- Email: [email protected]
Reliability considerations of a fuel cell backup power system for telecom applications
Journal of Power Sources 309 (2016) 66e75
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
Reliability considerations of a fuel cell backup power system for telecom applications Mustafa Fazil Serincan Department of Mechanical Engineering, Istanbul Bilgi University, Kazım Karabekir Cad. No: 2/13, Eyup, 34060 Istanbul, Turkey
h i g h l i g h t s A fuel cell backup power unit is tested in real-world conditions at a GSM base station. Reliability of the fuel cell system is found to be 98.5% following 260 cycles. Inverter in the base station is redesigned to attenuate the current harmonics. Reliability of the system under fault conditions is studied. Sizing of system components is critical for an isolated operation during faults.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 November 2015 Received in revised form 18 January 2016 Accepted 22 January 2016 Available online 5 February 2016
A commercial fuel cell backup power unit is tested in real life operating conditions at a base station of a Turkish telecom operator. The fuel cell system responds to 256 of 260 electric power outages successfully, providing the required power to the base station. Reliability of the fuel cell backup power unit is found to be 98.5% at the system level. On the other hand, a qualitative reliability analysis at the component level is carried out. Implications of the power management algorithm on reliability is discussed. Moreover, integration of the backup power unit to the base station ecosystem is reviewed in the context of reliability. Impact of inverter design on the stability of the output power is outlined. Significant current harmonics are encountered when a generic inverter is used. However, ripples are attenuated significantly when a custom design inverter is used. Further, fault conditions are considered for real world case studies such as running out of hydrogen, a malfunction in the system, or an unprecedented operating scheme. Some design guidelines are suggested for hybridization of the backup power unit for an uninterrupted operation. © 2016 Elsevier B.V. All rights reserved.
Keywords: Fuel cell Reliability Telecom Power management Backup power Fault
1. Introduction Fuel cells have emerged as promising alternatives to conventional power generation technologies because of their high power and energy densities and environmental friendly operations. Contrary to these positive implications about fuel cells, economic maturity of the technology is far from being competitive to the conventional counterparts. Indeed, the commercialization progress is expected to be sluggish until the mass production of the fuel cell systems can be realized. On the other hand, there are some sectors where fuel cells are shown to be commercially more feasible alternatives already, compared to the conventional technologies. Material handling and
E-mail addresses: [email protected], [email protected]. http://dx.doi.org/10.1016/j.jpowsour.2016.01.083 0378-7753/© 2016 Elsevier B.V. All rights reserved.
backup power applications can be given as examples to earlymarkets for fuel cell technology. As a matter of fact, fuel cell forklifts and fuel cell backup power generators have been increasingly used worldwide in commercial applications [1]. Fuel cell systems have been demonstrated as more viable business cases compared to alternative technologies such as diesel and battery power units [2e6]. Deployments of fuel cell backup power systems for telecom applications have been on the rise throughout the world [7e15]. While fuel cell technology has become cost competitive in early market applications recently, studies are still going on to improve system efficiency and reliability and reduce return of investment durations. Reliability for telecom equipment is conventionally defined as “the ability of an item to perform a required function under given conditions for a given time interval” [16]. This definition can also be
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
interpreted as the probability of the system functioning as desired for a specified duration. However, it has been discussed that lumping the reliability concept to a single parameter may have limited practicality in telecom applications [17] and new metrics for reliability focusing on fault analysis have been proposed [18]. Reliability is a very important concept in the context of energy management of the base station. Alsharif et al. [19] discuss that one of the main drawbacks of diesel generators is the very low reliability of the technology when used as an energy source for an offgrid telecom base station. Kanzumba and Vermaak [20] report that diesel generators are responsible for 65% of the loss of telecom service due to failures of these devices. In this regards, fuel cell backup power units which have superior reliability compared to both diesel generators and batteries are more advantageous in telecom base stations. Reliability is a main concern for the design and implementation of a hybrid fuel cell system. Sulaiman et al. [21] conclude that apart from the economical concerns, main issues and challenges in a fuel cell hybrid system are due to the peripheral system components; such as the battery lifetime, energy management system and power electronics interface. It is asserted that most of the research regarding validation of the energy management system has been limited to simulations while ensuring the reliability of the systems in real world applications is required. In this regards, hybridization of a fuel cell system is of a key concern for a reliable and efficient operation [22e26]. Xhan et al. [27] proposed an energy management system for a hybrid UPS system which alternates operating modes between the fuel cell, the battery and the supercapacitors for a reliable power output. Li et al. [28] present a power sharing strategy for a fuel cell-batterysupercapacitors hybrid tramway in which the proposed system distributes the power demand to each source appropriately to guarantee a safe operation and extend the lifetime of each source. Xie et al. [29] developed an energy management algorithm for a hybrid powertrain which stabilizes the DC bus voltage via a fuzzy logic controller to protect the fuel cells and the batteries. The controller is also responsible for power distribution during a dynamic load cycle. Vasallo et al. [30] implemented an energy management strategy to determine the sizing of a hybrid UPS system for optimal power sharing between the fuel cell and the batteries. As stated earlier most of these studies are based on simulations while validation of power management system in real world applications and its influence on reliability is still of great interest. On the other hand, integration of the fuel cell system to the ecosystem of the application is also very critical. Guilbert et al. [31] study the interaction between the fuel cell and DC/DC converter in case of a faulty operating scheme in an electric vehicle application. They address that malfunction in the system at open circuit not only increases the fuel cell current ripples but also induces stresses on the inductors. They emphasize on the selection of the DC/DC converter for a reliable and efficient operation. Fontes et al. [32] studied the effects of current harmonics on a fuel cell stack. They conclude that although a fuel cell stack can filter high frequency current harmonics due to its double layer capacitance, impact of this operating scheme on the durability of the fuel cell and the maximum RMS current that can be tolerated by the fuel cell stack need to be investigated in more detail. Faults encountered during the operation may affect either the fuel cell system or the radio equipment in the base station or both. Hence malfunctions in the fuel cell system or unprecedented operating schemes have an imminent impact on the reliability. There have been a number studies recently focusing on the detection and isolation of the faults in hybrid fuel cell systems [33e41]. Most of these studies propose the use of sophisticated electronics designs and control techniques to minimize the effects
67
of the faults on the operational reliability of the systems. While active fault prevention methods like these are proven to be effective, they require complex system architectures and indubitably exacerbate the economic feasibility of a commercial fuel cell system. Hence, passive fault isolation techniques would be much more desirable, while reliability concerns with them are yet to be addressed. This study focuses on the reliability analysis of a commercial fuel cell backup power unit installed in a base station run by a Turkish GSM operator, Turkcell. Apart from a lumped system level approach, reliability of the fuel cell backup power unit is considered in real-life case studies; e.g. integration issues of the unit with the base station ecosystem, running out of hydrogen, and inadvertent interruption of the fuel cell operation by the site personnel. Unlike the controlled laboratory experiments, on-field tests present a unique environment for evaluating the reliability of the fuel cell backup power unit that endure a wide range of environmental conditions and unprecedented operating schemes, dynamic site loads and different levels of grid quality. Operation of the backup power system is elucidated in the following sections. Load sharing between the fuel cell and the batteries is examined for various operating conditions. Effects of utilizing an inverter to generate alternative current are discussed in the context of integration of the fuel cell backup power unit to the existing base station ecosystem. Inverter design is found to be decisive to get a high quality and stable power output from the backup power unit. Moreover, system responses are examined for separate cases when fuel cell exhibits a failure mode and recovers from a failure. Load sharing during a fuel cell failure is investigated to see how an uninterrupted system operation is sustained without any power loss. Finally, effects of hydrogen depletion on the system operation is discussed in a real-life case study for which the fuel cell is forced to respond to a power outage while the hydrogen in the tank is completely depleted. 2. System configurations and test setup A fuel cell backup power system has been installed in a base station of Turkish mobile telecom operator Turkcell in Bursa. A schematic of the installation can be seen in Fig. 1. The fuel cell backup power unit is comprised of a fuel cell power module and a startup battery, which are connected in parallel. There is a boost converter after the fuel cell stack increasing the voltage to the nominal operating range of the base station. Fuel cell power module also includes peripheral equipment accounting for thermal management and hydrogen and air supply to the stack. Main specifications of the tested backup power unit are given in Table 1. Rated power of the system is 6 kW at a nominal DC output voltage of 48 V while the typical load demand for the base station is about 1.5 kW. Also, the system is equipped with an inverter to provide the AC power to the air conditioner inside the base station cabinet during the grid failures. Note that negative potential is an industry standard in telecommunications. However, to avoid any confusion we have chosen to report positive voltage values in this study. A low temperature PEM fuel cell stack is exploited in the backup power unit with an operating temperature around 60e65 C. Heat generated during the backup operation is rejected from the fuel cell stack via liquid cooling. There is an external liquid to air heat exchanger connected to the backup power unit supplying cold water to the system. Hydrogen consumption of the fuel cell is about 14 slpm kW1 at the rated conditions while this value increases at the partial loads. Hydrogen is generated at the test site with an alkali electrolyzer
68
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
Fig. 1. Schematics of the experimental setup.
Table 1 Specifications of fuel cell backup power systems. Property
Value
Rated power Nominal output voltage Fuel cell type Thermal management Power management Fuel consumption at rated power Hydrogen input pressure Hydrogen supply Hydrogen storage Backup time at 1.5 kW Additional energy storage On site energy storage
6 kW 48 VDC/220 VAC Low temperature PEM Liquid cooling Output regulated with a DC/DC converter 14 slpm kW1 2e5 barg On-field electrolyzer 6 cylinders (49 L, 150 bar) ~40 h 55 Ah AGM batteries 100 Ah AGM batteries
unit. Output pressure of the electrolyzer is boosted with a compressor. Hydrogen is stored in standard 49 L industrial hydrogen cylinders at 150 bar. A bundle of 6 cylinders is used to guarantee at least 40 h of autonomy while additional cylinders could be added to increase the backup time. Hydrogen pressure is reduced to 2e5 barg at the inlet of the power module where it is delivered to the fuel cell through an internal buffer tank. Backup power unit is installed parallel to the existing site batteries in the base station. This redundant configuration is requested by Turkcell to guarantee fail-safe operation during the test period. However, some of the battery groups in the original configuration are deactivated to utilize the fuel cell unit more for backup power generation. In the final configuration site batteries correspond to 100 Ah while the startup batteries in the unit provide 55 Ah of additional storage. The main role of the batteries in this installation is to provide the backup power instantly when the fuel cell system kicks off from a dormant state. A fuel cell at start-up cannot generate its rated power until the temperature and the humidity in the stack reach steady-state operating conditions. While the response of a fuel cell strongly depends on the design of the stack, typical cold start-up times can be assumed in the order of 1 min [42]. In this transient phase, batteries provide power demanded by the base station. Output of the backup power unit is connected to the base station
DC bus. Current and voltage measurements are carried out at the fuel cell, battery and the load terminals as schematically depicted in Fig. 1. Current measurements are done via shunts placed on the power cables. The relation between the current measured at different shunts is
iFC þ ibatt ¼ iload
(1)
where negative value implies current flowing out, e.g. battery discharging while positive value implies current flowing in, e.g. battery charging. On the other hand,
VFC ¼ Vbatt ¼ Vload
(2)
because as it can be seen in Fig. 1 the fuel cell, battery and the DC bus are connected in parallel. Current and voltage measurements for the fuel cell reflect the values after the DC/DC converter. Note that load here accounts for anything connected to the DC bus beyond the measurement box. In other words, site batteries and the inverter also contribute to the load measured by the shunt apart from the telecom equipment. Voltage and current measurements are recorded by an industrial PC located inside the base station cabinet. Signals are first transmitted to a signal conditioning board where they are converted to digital inputs then sent to the I/O card found in the PC. Ambient temperature is also measured by a thermocouple, which is directly connected to the PC via the USB port. A standard test software written in Labview by the other project partners [44], is used to monitor and control the operation of the backup power unit. The PC is connected to the Internet via a 3G-modem enabling tests from a remote location. Both the industrial PC and the 3Gmodem are powered from the DC bus. Moreover, backup power unit has its own data logger containing internal measurements such as fuel cell anode pressure, stack temperature and minimum and maximum cell voltages. This internal dataset is transferred to the test PC via Ethernet. As the control output of the system, I/O card generates a signal and sends it to the AC switch to turn the electricity grid on and off remotely. For this purpose, a relay module is used in the signalconditioning box which generates a 5 A/30 VDC analogue signal.
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
The signal conditioning box itself is powered by 24 VDC through the DC bus. Control signal is then sent to the contactor switch installed on the AC line at the entrance of the base station. A grid failure simulation is initiated by controlling the switch remotely in the Labview program. After the test is completed, software drives the switch back to its normal position restoring the grid in the base station. Tests have been implemented following a standardized test procedure developed in the project as a collaborative effort [43,44]. Test procedure is comprised of short and long cycles corresponding to 15 min and 240 min of power outages respectively. Each cycle includes a waiting period before switching off the AC grid. Waiting period is either 1 min or 1 h (minimum) following the previous cycle to simulate hot or cold start-up respectively. As a performance metric, reliability of the backup power unit has been calculated quantitatively following the conventional definition given in the previous section as [16];
R¼
number of cycles number of system failures 100% number of cycles
(3)
where number of cycles is the total power outage simulations, system failures represent the instances when the system is unable to provide the power demanded by the base station. In other words, reliability considered here is the frequency of the backup power system conforming to the normal operating scheme during the test period. This gives a single indication for the reliability considered at the system level. Normal operation is determined based on the service voltage range from 40.5 VDC to 57.0 VDC, which is specified by the European Telecommunications Standards Institute (ETSI) [45]. Operating voltage being beyond this range for more than 20 ms is regarded as a failure. On the other hand, a malfunction in the fuel cell stack may be masked by the start-up battery since the response of the battery is instantaneous and the DC/DC converter set voltage guarantees the ETSI operating range. However, as the battery is being depleted, operating voltage will eventually drop below the minimum allowable limit. In the test setup this scenario is almost always avoided because of the additional capacity provided by the site batteries. In order to account for all the pitfalls in the fuel cell system, a failure is categorized as the likelihood of the system going beyond the ETSI voltage limits. 3. Results and discussion 3.1. Performance of the fuel cell backup power unit Fuel cell backup power unit installed in Turkcell site has followed a standard test protocol to simulate short and long electricity blackouts. A total of 260 power outage simulations were imposed on the system consisting of 228 short (15 min) and 18 long (4 h) standard test cycles with additional 14 custom cycles. Tests started on July 16th, 2012 and continued until April 6th, 2014. In the course of the tests fuel cell system produced 139.1 kWh of energy for the base station equipment for a cumulative 130 h of backup operation. This corresponds to an average power of 1 kW. Fuel cell system is found to be in a healthy state in 256 of the 260 test cycles, responding to load requests successfully by providing the demanded backup power to the telecom equipment. For the remaining 4 times fuel cell system unexpectedly failed to start. On the other hand, batteries installed in the unit have supplemented the load demand to sustain an uninterrupted backup operation. Although start-up batteries provide a buffer to prevent failure from happening immediately, working in this condition for an extended amount of time would eventually lead the system voltage to fall
69
down below 40.5 V. Therefore, these instances are considered as failures although the base station operation has never been interrupted in the course of the tests. The possible causes of the failures are discussed in the following sections. Moreover, there have been some instances that the backup power operation is ceased due to secondary issues other than the backup power unit itself. Some complications have been experienced with the on-site hydrogen generation equipment, which forced the tests to be halted for a while. Considering only the technical issues related to the backup power unit, reliability of the fuel cell system is calculated as 98.5% using Eq. (3):
R¼
260 256 100% ¼ 98:5% 260
(4)
This is a notable result for a fuel cell system tested comprehensively in a real-world application, which justifies technology as a commercial alternative to the existing backup power solutions. At the end of the tests, fuel cell backup power system is still capable to deliver its rated power. In real life conditions total number of cycles carried out would correspond to 5e10 years of commercial operation depending on the duration and the frequency of the electricity grid failures. Based on these observations, lifetime of the fuel cell system is proven to be commercially viable. Details on the test procedure and the performance results can be found in a recently submitted study [46]. Interpretation of reliability exploited up to now is reasonable only in a system level approach. On the other hand, each component in the system has its own characteristics that affect the performance and stability of the operation. Apart from considering reliability as a measure of capability to provide the requested power; how system reacts to different operating regimes and how individual components work with each other are also discussed. Hence, the context of reliability analysis will be extended further to incorporate investigation of individual components, especially the fuel cell stack. In the following sections, reliability of the backup power operation will be considered for a number of real-life case studies experienced in the course of field tests. 3.2. Effects of power management Two different power sources in the unit, fuel cell and battery can individually provide current to the system. In case of a cold startup, when fuel cell operating variables are far from their steady state values, the batteries provide most of the load demand. On the other hand, share of the total power supply provided by the fuel cell increases as the fuel cell warms up. This can be characterized by the ratio of the fuel cell current to the total current coming out of the backup power unit as;
SFC ¼
IFC 100 IFC þ Ibatt
(5)
Recall that Ibatt measurement represents only the start-up batteries as depicted in Fig. 1. Consistent with the notation used in Eq. (1) fuel cell and battery currents can have opposite signs. Such case can be attained at the start-up when some power is needed for the fuel cell to run the auxiliary equipment. In this case, SFC will be negative. A typical response of the backup power unit can be seen in Fig. 2, which shows the DC bus voltage and the load current. This test imposes three consecutive power outages on the base station each lasting 15 min. Test cycles are distinguished with the vertical dashed lines in the figure. First cycle invokes the system with a cold start-up, followed by two new cycles with 1-min waiting period in
70
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
Fig. 2. Response of the DC bus voltage (top), response of the load current (bottom) to 3 consecutive power outages of 15 min on 18 March 2014.
between power outages. During this period, electricity grid is restored and the fuel cell system shuts down. However, temperature of the system changes little in this period so the fuel cell exhibits warm start-up in the consecutive cycle. Before the grid failure, DC bus voltage reads 54.5 V, which is the set value of the base station rectifier. Until the fuel cell kicks in, only batteries deliver power to the base station. As the batteries are discharged, bus voltage drops. Power management algorithm of the backup power unit dictates that, fuel cell starts when the DC bus voltage falls below 50.5 V. Since a considerable number of the site batteries are deactivated during the tests, initial voltage drop is abrupt. Also, it can be seen from Fig. 2 that voltage further drops even after the fuel cell starts because the fuel cell takes some time to reach stable operating conditions. Nonetheless, once the fuel cell passes this transient operating regime, it delivers greater power, which results in a quick recovery of the bus voltage. Power management algorithm in the backup power unit prevents the batteries to be overcharged by the fuel cell. This is done via the DC/DC converter by increasing and decreasing the fuel cell output current alternatingly. This way, DC bus voltage is maintained around a nominal operating voltage. As can be seen in Fig. 2, bus voltage periodically varies between 51 and 51.5 V for the rest of the backup operation. Current output of the backup power unit can also be seen in Fig 2. When the grid goes off, batteries initially govern the load. Site batteries dominate in this stage, as they constitute a larger capacity compared to the start-up batteries installed in the unit. Batteries are discharged while they provide energy to the base station. As
discussed before, discharged site batteries also contribute to the load. It should be noted that load due to the other base station operation is relatively constant at around 30 A. After the fuel cell starts, it provides energy not only for the base station equipment but also to charge the batteries. Therefore, the load current goes up to 45 A. In the steady operation when the DC/DC converter regulates the system voltage, load current fluctuates between 25 and 35 A. It can be further estimated from this operating regime that when backup power unit gives 25 A, site batteries provide an additional 5 A to satisfy the telecom equipment demand. On the other hand, when system voltage is ramped up by the converter, fuel cell unit provides 35 A of which 5 A is dedicated for charging the batteries. Fig. 3 shows how the fuel cell and the unit batteries share the load (recall the sign convention in Eq. (1)). At the beginning of the cycle fuel cell module draws energy from the unit battery for about half a minute to power its peripherals. After getting the initial impetus, fuel cell output current gradually increases. 1 min after its start-up, fuel cell reaches steady operation and overtakes the startup battery in the unit. At around 90 s into the grid failure, battery current becomes negative meaning it is being charged. This is when the fuel cell is capable of fulfilling the entire load demand. At about 100 s into the grid failure fuel cell current reaches about 50 A. Assuming the base station load demand is approximately 30 A, it can be estimated that fuel cell provides 20 A to charge the batteries. The unit battery consumes about 5 A, so the remaining 15 A is dedicated for charging the site batteries. As the batteries are being charged, fuel cell current decreases gradually to its nominal operating values. During its regular operation, fuel cell current increases and decreases alternatingly to charge and discharge the batteries as modulated by the DC/DC converter. It can be suggested from Figs. 2 and 3 that batteries are put to charge mode for 2 min, while they are discharged for 3 min, forming a limit cycle every 5 min in a regular steady operating regime. 3.3. Effects of inverter design Telecom equipment standalone constitutes a load of 20 A, which corresponds to roughly 1 kW. Recalling the fuel cell unit is rated 6 kW, it is apparent that fuel cell operates in a sub-optimal region. As a matter of fact, in the early stages of the tests when the site load was only comprised of the telecom equipment, only 10e15% of fuel cell capacity was essentially utilized. Beyond working in a nonoptimal region, this would cause instable operation and decrease
Fig. 3. Load, fuel cell and battery currents during cold start-up cycle; the first cycle in Fig. 2.
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
the reliability of the entire system. For this reason, site load is increased by coupling the air conditioner to the DC bus with an inverter. Because of the intensive energy requirement of the air conditioners, site batteries are depleted rapidly during a grid failure. Therefore, typically when backup power is managed by the batteries, it is a common practice to maintain the temperature of the base station cabinet by free cooling. On the other hand free cooling has significant limitations and can provide cooling only as much as the ambient temperature allows. Therefore, air conditioner emerges as a necessity in warmer locations. A fuel cell system can be used in a base station where the cabinet is cooled with an air conditioner. For this purpose, DC output of the back up power unit is converted to AC via an inverter. On the other hand, design of the inverter is reported to be critical for the fuel cell operation [47e49] while an inverter that is suitable to be used with the batteries may not work impeccably with a fuel cell system. Also, fuel cell manufacturers would not have the information about the designs of the inverters found in the base station. Yet, the backup power unit is expected to have a seamless integration with the existing base station ecosystem. In this regard, field tests of fuel cell backup power units give an opportunity to see the performance of the systems for the use with a third party inverter. Initially an off-the shelf commercial inverter is integrated to the DC bus to power the air conditioner in the base station during the backup operation. However, the system exhibits a poor response with high frequency ripples in current and voltage. Fig. 4 shows the DC bus voltage and the load current responses to a 4-h test cycle. The insets in the figure gives a more detailed picture of the voltage and current harmonics over a duration of 5 min. It can be observed
Fig. 4. Response of the DC bus voltage (top) response of the load current (bottom) to a 4-h grid failure on 17 July 2012. Insets elaborates the voltage and current ripples caused by the inverter dynamics.
71
that ripples occur with an average period of 16 s and amplitude of 0.15 V. Ripples get even worse when voltage fluctuations reach up to 0.4 V at around 13:15. An identical trend in current ripples is observed with the same frequency. However, the effect is more pronounced as the amplitude of the fluctuations reaches 10 A when voltage ripples are at their highest. Current response also hints about the operating regime of the air conditioner. When air conditioner is off at the initial and final stages of the test, the load current reads 20 A. Operation with the air conditioner requires an additional 10 A on average. Yet, instantaneous current demand can go up to 50 A. Current response is determined by the changes in the air conditioner operating profile. As the compressor in the air conditioner ramps up and down periodically, current demand from the backup power unit varies correspondingly. Note that when the air conditioner is not working, ripples are attenuated but do not vanish completely. Harmonics seen in the load current will be identically reflected on both fuel cell and battery currents. This can be verified in Fig. 3 where all forms of current profiles exhibit identical fluctuations. Ripple currents are proven to shorten the lifetimes of both fuel cells and batteries [50e52]. Gemmen suggests the current ripples should be less than 10% for a reliable operation of the fuel cell system [53]. Also for the lead acid batteries it is recommended that the ripple voltage is less than 0.5% of the DC voltage [54]. Ripples experienced in the tests are well beyond these limitations. The specifications of the inverter states a voltage total distortion harmonics (TDH) value as much as 3%. Also, the output of the inverter is a modified sine which is a typical practice for most telecom applications. However, inverters with modified sine output must be tested and rated to be used with a particular equipment. These two are considered to be the main sources of high voltage and current ripples in the DC bus. After these observations, the inverter is sent back to the manufacturer and replaced with a new one having a custom design. Different strategies for elimination and mitigation of inverter ripples in fuel cell applications are outlined in the literature [55e57]. However, implementations of these designs are beyond the scope of our study. The inverter manufacturer applied its own methods to decrease the ripples practically. First modification is that the inverter gives a true sine output. Also, a choke coil is employed in the inverter to decrease voltage TDH. Although the new design may not be the most optimum for a fuel cell application, yet a significant reduction in ripples is achieved. Fig. 5 compares the voltage responses of the system for three different cases tested: i) there is no inverter installed (top), ii) a generic inverter is installed (middle), iii) a custom design inverter is installed (bottom). With the custom design inverter, voltage ripples are attenuated considerably. Voltage fluctuates in a band of 0.01 V on average, which has a negligible effect on the system performance. Rest of the tests has been implemented with this inverter. Likewise, ripples in current response can be observed with an average amplitude of 1 A. Following these discussions, it can be concluded that inverter design is very critical for the reliability of the fuel cell backup power unit. Generic inverter designs may cause instabilities in the system and may even damage the telecom equipment. It can even risk the fuel cell operation, as it will induce pressure variations inside the stack. Effect of these variations is yet to be studied. Issues with the inverter are experienced during the commissioning of the fuel cell backup system in the base station. On the other hand, two of the failures experienced with the power module exist shortly after the fuel cell is exposed to large current and voltage ripples, while the other two are experienced in the later stages of the field tests. This observation might suggest a
72
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
be sized carefully for extra backup times in order to sustain a failsafe operation in case of hydrogen depletion in the tank, hence to increase the reliability of the system. Four distinct failures experienced with the fuel cell system can hardly be linked to operation on low hydrogen levels, because the system has not gone through starvation. Two system failures happen before this event. On the other hand, after hydrogen depletion incident the fuel cell system exhibits normal operating regimes for a long time until the next two failures occur. Hence, these failures are more likely due to some other implications of long term tests. 3.5. Operation in failure and recovery
Fig. 5. Comparison of voltage responses when there is no inverter installed (top), a generic inverter is installed (middle), a custom design inverter is installed (bottom) in the DC bus.
correlation between the earlier failures with the ripples. Yet, there is not enough evidence to associate them certainly, as early failures might well be due to some other issues pertaining to the system design. 3.4. Hydrogen depletion Hydrogen supply is one of the main considerations when working with fuel cell systems. Backup duration is determined with the hydrogen storage on-site and can be extended by installing additional cylinders. As the system generates energy for the base station, hydrogen is being depleted continuously. Replacements cylinders must be provided when the hydrogen content in the system is diminishing below critical levels. In this test setup hydrogen is supplied from the on-site electrolyzer. However, due to a fault in the hydrogen generation equipment, electrolyzer stopped working. As a result hydrogen in the tank is depleted completely as the tests continue without being aware of this situation. Fig. 6 plots hydrogen pressure in the tank and in the fuel cell anode during a standard 4-h test cycle (plots on the left), and corresponding dynamics of fuel cell current and DC bus voltage (plots on the right). At the beginning of the test, hydrogen in the tank has a pressure of 8 bar but it decreases linearly as the fuel cell converts chemical energy stored in hydrogen to electric energy. Hydrogen pressure vanishes completely after 1 h. On the other hand, pressure at the stack is observed to be relatively constant at 0.3 bar. This is related to the additional capacity provided by the piping and internal components of the unit. Indeed, it is highly probable that a smaller buffer tank exists in the fuel cell system that lets the fuel cell operate even when hydrogen is entirely deprived in the main storage tanks. Corresponding system response is given in the plots on the right hand side. Both fuel cell current and voltage profiles are observed to be normal throughout the 4-h test cycle. On the other hand, once the internal buffer is also consumed fuel cell will fail and the batteries will deliver the backup power. An internal buffer tank must
Reliability of the system in case of a failure is another important consideration. Failure of the fuel cell may occur either before the system starts or in the middle of the operation. The former is a relatively simpler management task which requires incorporation of a temporary backup power supply until the fuel cell maintenance takes place. However, the latter scenario happens unexpectedly and does not allow a lead-time to take a corrective action. In this case, the system must sustain the backup power demanded by the base station until the site personnel arrive at the site and install an alternative backup source. Fig. 7 shows current and voltage response of the system when the fuel cell unprecedentedly fails soon after it is started. It can be observed from the figure that the fuel cell is activated at 11:18 when the DC bus voltage falls below 50.5 V. After the auxiliaries in the fuel cell module are initiated, fuel cell starts providing current. Fuel cell gives 20 A at its first attempt but it is followed by an abrupt drop in current resulting in loss of power completely. In the meantime batteries kick in immediately and compensate for the fuel cell power loss. As a result backup operation is recovered very quickly and voltage drop due to fuel cell failure is limited to only 0.5 V. Share of the load is taken completely by the batteries for the rest of the test. For a grid failure of 15 min batteries maintain the demanded power without any significant voltage drop. Note that battery current plotted in Fig. 7 is for the start-up batteries in the unit only. Assuming base station load is 30 A on average, it can be estimated that 1/3 of the load is covered by the start-up batteries while 2/3 is covered by the site batteries. On the other hand, batteries will be depleted and bus voltage will eventually drop below ETSI limits for extended backup durations. Also when the site batteries are completely decoupled, an ultimate system failure will be encountered sooner. In this regard batteries directly affect system reliability. Therefore, it is crucial to size and select the batteries in the unit carefully for sustaining backup operation even in case of a fuel cell failure, hence achieving higher system reliability. Although it may not be as critical as the foregoing scenario, it is also interesting to examine how the system responds in case fuel cell recovers from a failure when the system is already running on the batteries. Fig. 8 depicts current and voltage responses in a similar situation when the fuel cell is manually started at the final stages of the test. Note that this test is implemented with the entire site batteries coupled to the DC bus and the inverter is deactivated. This explains relatively slow voltage drop. It can be observed that fuel cell does not start when the voltage falls far below 50.5 V. After the site personnel check the system status, error messages are found on the fuel cell interface. Once the error messages are cleared, fuel cell is reset and started manually at 11:58 p.m. Following its regular start-up routine fuel cell begins delivering current to the load, however, it cannot undertake the entire load share right away. Tests are discontinued shortly after, to
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
73
Fig. 6. Change in hydrogen tank pressure (top left) and anode pressure (bottom left) during a 4-h grid failure on 16 October 2012. Responses in fuel cell current (top right) and DC bus voltage (bottom right) during the same failure.
Fig. 7. Current and voltage responses of the system to a 15 min power outage on 24 December 2013. An unexpected fuel cell failure is experienced at 11:19 a.m.
implement a standard cycle. Here the importance of the batteries in the system is once more emphasized because even when the fuel cell recovers from failure it may take some additional time to reach its nominal operating regime. 4. Conclusions A fuel cell backup power system has been tested under real life conditions at a telecom base station operated by Turkcell. A test procedure consisting of 260 cycles is implemented to estimate the long-term reliability of the system in case of real life grid failure scenarios. The entire system responded 256 of the 260 test cycles successfully. With a system level approach, reliability of the system is calculated as 98.5%. For the remaining 4 times fuel cell failed to provide the requested current. However, with the start-up batteries backup operation is sustained without any disruption. Also, the system is investigated at component level to have a more comprehensive reliability analysis. It is shown that power
Fig. 8. Current and voltage responses of the system to a 15 min power outage on 17 July 2012. Fuel cell recovers from failure at 11:58 a.m.
management of the unit is decisive in load sharing between the fuel cell and the batteries. The batteries govern system response in the transient start-up period. Merely 1 min into the backup operation, fuel cell is warmed up and undertakes almost the entire load demand. Moreover, inverter design is found to have a great impact on the reliability of the system. It has been shown that a third party generic design inverter used by the operator results in large ripples in system response. While the voltage fluctuates as much as 0.4 V, current ripples are more pronounced with fluctuations reaching 10 A. With a custom design inverter, ripples attenuated considerably by 90%. Batteries installed in the unit are shown to be very decisive in improving the reliability of the unit. In case of a fuel cell failure in the middle of a backup operation, batteries account for the entire load demand instantly, without letting the system voltage drop below the limits. Also batteries are shown to support fuel cell recovering from a failure. It is imperative to select and size start-up
74
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
batteries in the unit to increase reliability of the operation. Further, the system is tested when the fuel cell runs out of hydrogen in the middle of a backup operation. It has been observed that even when the hydrogen is completely depleted in the main storage; hydrogen content inside the unit is sufficient enough to sustain the backup operation for another 3 h. Capacity of the pipes and other system components provide a safety margin for the operation. In this regard, design of an internal buffer tank is very important to increase the reliability of the system amid the risk of hydrogen starvation. Tests at real-life conditions have proven the reliability of the fuel cell backup power system at both system and component levels. Yet, to guarantee the utmost reliability of the operation and perfect compliance with the base station ecosystem, improvements are needed not only on the fuel cell technology but on the peripheral system components as well. Acknowledgement The research leading to these results has received funding from the European Union's Seventh Framework Programme (FP7/20072013) for the Fuel Cells and Hydrogen Joint Technology Initiative under grant agreement no [256766]. Matching funds have been received from Republic of Turkey, Ministry of Energy and Natural Resources under the “International Centre for Hydrogen Energy Technologies” project. Turkcell Communication Services Inc. further provided a great amount of in kind support throughout the study. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.jpowsour.2016.01.083. References [1] Fuel Cell Industry Review 2013, Johnson Matthey PLC, Herts, UK, 2013. [2] DOE, Hydrogen Program e Early Markets: Fuel Cells for Backup Power, 2009. [3] APCO, Annual Conference and Expo: Fuel Cells for Critical Communications Backup Power, 2009. [4] J. Kurtz, G. Saur, S. Sprik, C. Ainscough, Backup Power Cost of Ownership Analysis and Incumbent Technology Comparison, NREL/TP-5400-60732, 2014. [5] Warid Pakistan Expects to Save More than US$ 6 Million per Year After Trialling Energy Efficiency Solutions, Mobile Energy Efficiency Optimization Case Study, first ed., GSMA, London, 2014. [6] Pakistan telecom trial validates ElectraGen backup power benefits, Fuel Cells Bull. 2014 (4) (April 2014) 3. [7] DOE, Hydrogen and Fuel Cells Program Record #14009. [8] 2011 Fuel Cell Technologies Market Report, US Department of Energy, Energy Efficiency and Renewable Energy, 2012. [9] Ballard Power Systems Inc, Annual Information Form, February 26, 2014. [10] Progress in FCH deployment in Europe and the world, in: Ballard Company Presentation by John Sheridan, FCH JU 6th Stakeholder General Assembly, 13.11.2013. [11] 2013 Fuel Cell Technologies Market Report, US DOE Energy Efficiency and Renewable Energy, November 2014. [12] Ballard deploys initial units at telecom sites in Manila, new CEO, Fuel Cells Bull. 2014 (9) (2014) 3. [13] 2010 Fuel Cell Technologies Market Report, US Department of Energy, Energy Efficiency and Renewable Energy, 2011. [14] 2012 Fuel Cell Technologies Market Report, U.S. Department of Energy, Efficiency and Renewable Energy, 2013. [15] Cascadiant wins two clean energy telecom contracts in Indonesia, Fuel Cells Bull. 2013 (2) (2013) 5. [16] ITU-T Recommendation E.800 08/94 p.17, Switzerland, Terms and Definitions Related to Quality of Service and Network Performance Including Dependability, 1994. [17] M. Ayers, Telecommunications System Reliability Engineering, Theory, and Practice, Wiley, Hoboken, N.J, 2012. [18] H. Kehl, Fault Analysis and Fault Tolerance of a Base Station System for Mobile Communications, PhD Dissertation in Technische Universitat Munchen, 2000. [19] M. Alsharif, R. Nordin, M. Ismail, Energy optimisation of hybrid off-grid system for remote telecommunication base station deployment in Malaysia, J. Wirel. Com. Netw. 64 (2015), http://dx.doi.org/10.1186/s13638-015-0284-7.
[20] M. Crouch, Fuel Cell Systems for Base Stations: Deep Dive Study, first ed., GSMA Green Power for Mobile White Paper, London (, 2011. [21] N. Sulaiman, M. Hannan, A. Mohamed, E. Majlan, W. Wan Daud, A review on energy management system for fuel cell hybrid electric vehicle: issues and challenges, Renew. Sust. Energy Rev. 52 (2015) 802e814, http://dx.doi.org/ 10.1016/j.rser.2015.07.132. [22] C. Chao, J. Shieh, A new control strategy for hybrid fuel cell-battery power systems with improved efficiency, Int. J. Hydrogen Energy 37 (17) (2012) 13141e13146, http://dx.doi.org/10.1016/j.ijhydene.2012.03.143. [23] F. Odeim, J. Roes, L. Wülbeck, A. Heinzel, Power management optimization of fuel cell/battery hybrid vehicles with experimental validation, J. Power Sources 252 (2014) 333e343, http://dx.doi.org/10.1016/ j.jpowsour.2013.12.012. [24] J. Hwang, W. Chang, Characteristic study on fuel cell/battery hybrid power system on a light electric vehicle, J. Power Sources 207 (2012) 111e119, http://dx.doi.org/10.1016/j.jpowsour.2012.02.008. [25] M. Abu Mallouh, E. Abdelhafez, M. Salah, M. Hamdan, B. Surgenor, M. Youssef, Model development and analysis of a mid-sized hybrid fuel cell/battery vehicle with a representative driving cycle, J. Power Sources 260 (2014) 62e71, http://dx.doi.org/10.1016/j.jpowsour.2014.02.104. [26] J. Bernard, M. Hofer, U. Hannesen, A. Toth, A. Tsukada, F.N. Büchi, P. Dietrich, Fuel cell/battery passive hybrid power source for electric powertrains, J. Power Sources 196 (14) (2011) 5867e5872, http://dx.doi.org/10.1016/ j.jpowsour.2011.03.015. [27] Y. Zhan, Y. Guo, J. Zhu, L. Li, Power and energy management of grid/PEMFC/ battery/supercapacitor hybrid power sources for UPS applications, Int. J. Electr. Power 67 (2015) 598e612, http://dx.doi.org/10.1016/ j.ijepes.2014.12.044. [28] Q. Li, W. Chen, Z. Liu, M. Li, L. Ma, Development of energy management system based on a power sharing strategy for a fuel cell-battery-supercapacitor hybrid tramway, J. Power Sources 279 (2015) 267e280, http://dx.doi.org/ 10.1016/j.jpowsour.2014.12.042. [29] C. Xie, X. Xu, P. Bujlo, D. Shen, H. Zhao, S. Quan, Fuel cell and lithium iron phosphate battery hybrid powertrain with an ultracapacitor bank using direct parallel structure, J. Power Sources 279 (2015) 487e494, http://dx.doi.org/ 10.1016/j.jpowsour.2015.01.029. [30] M. Vasallo, J. Bravo, J. Andújar, Optimal sizing for UPS systems based on batteries and/or fuel cell, Appl. Energy 105 (2013) 170e181, http://dx.doi.org/ 10.1016/j.apenergy.2012.12.058. [31] D. Guilbert, A. Mohammadi, A. Gaillard, A. N'Diaye, A. Djerdir, Interactions between fuel cell and DC/DC converter for fuel cell electric vehicle applications: influence of faults, in: IECON 2013-39th Annual Conference of the IEEE Industrial Electronics Society, 2013, http://dx.doi.org/10.1109/ iecon.2013.6699255. [32] G. Fontes, C. Turpin, S. Astier, T. Meynard, Interactions between fuel cells and power converters: influence of current harmonics on a fuel cell stack, IEEE Trans. Power Electr. 22 (2) (2007) 670e678, http://dx.doi.org/10.1109/ tpel.2006.890008. [33] H. Aouzellag, K. Ghedamsi, D. Aouzellag, Energy management and fault tolerant control strategies for fuel cell/ultra-capacitor hybrid electric vehicles to enhance autonomy, efficiency and life time of the fuel cell system, Int. J. Hydrogen Energy 40 (22) (2015) 7204e7213, http://dx.doi.org/10.1016/ j.ijhydene.2015.03.132. [34] A. Rosich, R. Sarrate, F. Nejjari, On-line model-based fault detection and isolation for PEM fuel cell stack systems, Appl. Math. Model. 38 (11/12) (2014) 2744e2757, http://dx.doi.org/10.1016/j.apm.2013.10.065. [35] Z. Li, R. Outbib, S. Giurgea, D. Hissel, Y. Li, Fault detection and isolation for polymer electrolyte membrane fuel cell systems by analyzing cell voltage generated space, Appl. Energy 148 (2015) 260e272, http://dx.doi.org/ 10.1016/j.apenergy.2015.03.076. [36] D. Rotondo, V. Puig, F. Nejjari, Fault tolerant control of a pem fuel cell using qLPV virtual actuators, IFAC-PapersOnLine 48 (21) (2015) 271e276, http:// dx.doi.org/10.1016/j.ifacol.2015.09.539. [37] D. Guilbert, A. Gaillard, A. Mohammadi, A. N'Diaye, A. Djerdir, Investigation of the interactions between proton exchange membrane fuel cell and interleaved DC/DC boost converter in case of power switch faults, Int. J. Hydrogen Energy 40 (1) (2015) 519e537, http://dx.doi.org/10.1016/ j.ijhydene.2014.10.072. [38] M. Kamal, D. Yu, D. Yu, Fault detection and isolation for PEM fuel cell stack with independent RBF model, Eng. Appl. Artif. Intel. 28 (2014) 52e63, http:// dx.doi.org/10.1016/j.engappai.2013.10.002. [39] L. Riascos, M. Simoes, P. Miyagi, On-line fault diagnostic system for proton exchange membrane fuel cells, J. Power Sources 175 (1) (2008) 419e429, http://dx.doi.org/10.1016/j.jpowsour.2007.09.010. [40] A. De Bernardinis, D. Candusso, F. Harel, G. Coauerv, Power electronics interface for an hybrid PEMFC generating system with fault management strategies for transportation, in: Proceedings of 13th European Conference on Power Electronics and Applications, Barcelona, 2009, pp. 1e10. [41] D. Guilbert, M. Guarisco, A. Gaillard, A. N'Diaye, A. Djerdir, FPGA based faulttolerant control on an interleaved DC/DC boost converter for fuel cell electric vehicle applications, Int. J. Hydrogen Energy 40 (45) (2015) 15815e15822, http://dx.doi.org/10.1016/j.ijhydene.2015.03.124. [42] FCgen-1020ACS Fuel Cell Stack, Product Manual, Ballard, 2010. [43] submitted to FCH-JU, Development of Test Protocol, Fitup Project Deliverable, 2011.
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75 [44] U. Trachte, P. Sollberger, B. Wellig, Fuel cell field test demonstration of economic and environmental viability for portable generators, backup and UPS power system applications, in: Proceedings of 4th European PEFC and H2 Forum 2013, Lucerne, 2013, p. B0404. [45] ETSI EN 300 132-2, France, Environmental Engineering (EE); Power Supply Interface at the Input to Telecommunications and Datacom (ICT) Equipment; Part 2: Operated by -48 V Direct Current (dc), 2011, p. 9. [46] M.F. Serincan, Validation of fuel cell hybridization methodologies in realworld telecom backup power applications, submitted to Int. J. Hydrogen Energy. (unpublished results). [47] N. Bizon, A new topology of fuel cell hybrid power source for efficient operation and high reliability, J. Power Sources 196 (6) (2011) 3260e3270, http:// dx.doi.org/10.1016/j.jpowsour.2010.11.049. [48] W. Shireen, R.A. Kulkarni, M. Arefeen, Analysis and minimization of input ripple current in PWM inverters for designing reliable fuel cell power systems, J. Power Sources 156 (2) (2006) 448e454, http://dx.doi.org/10.1016/ j.jpowsour.2005.06.012. [49] A. Kirubakaran, S. Jain, R.K. Nema, A review on fuel cell technologies and power electronic interface, Renew. Sust. Energy Rev. 13 (9) (2009) 2430e2440, http://dx.doi.org/10.1016/j.rser.2009.04.004. [50] J. Kim, I. Lee, Y. Tak, B. Cho, Impedance-based diagnosis of polymer electrolyte membrane fuel cell failures associated with a low frequency ripple current, Renew. Energy 51 (2013) 302e309, http://dx.doi.org/10.1016/ j.renene.2012.09.053. n, A. Jimenez, E. Moreno, Power [51] J.J. Brey, C.R. Bordallo, J.M. Carrasco, E. Galva
[52]
[53]
[54]
[55]
[56]
[57]
75
conditioning of fuel cell systems in portable applications, Int. J. Hydrogen Energy 32 (2007) 1559e1566, http://dx.doi.org/10.1016/ j.ijhydene.2006.10.033. W. Lee, B. Han, H. Cha, Battery ripple current reduction in a three-phase interleaved dc-dc converter for 5kW battery charger, in: 2011 IEEE Energy Conversion Congress and Exposition, Phoenix, 2011, pp. 3535e3540, http:// dx.doi.org/10.1109/ECCE.2011.6064247. R.S. Gemmen, Analysis for the effect of inverter ripple current on fuel cell operating condition, J. Fluids Eng. 125 (3) (2003) 576e585, http://dx.doi.org/ 10.1115/1.1567307. Effects of AC Ripple Current on VRLA Battery Life, Emerson Network Power white paper, Columbus OH, http://www.emersonnetworkpower.com/ documentation/en-us/brands/liebert/documents/white%20papers/effects% 20of%20ac%20ripple%20current%20on%20vrla%20battery%20life.pdf, (accessed 12.01.16). G.R. Zhu, S.C. Tan, Y. Chen, C.K. Tse, Mitigation of low-frequency current ripple in fuel-cell inverter systems through waveform control, IEEE Trans. Power Electr. 28 (2) (2013) 779e792, http://dx.doi.org/10.1109/TPEL.2012.2205407. S.K. Mazumder, R.K. Burra, K. Acharya, A ripple-mitigating and energyefficient fuel cell power-conditioning system, IEEE Trans. Power Electr. 22 (4) (2007) 1437e1452, http://dx.doi.org/10.1109/TPEL.2007.900598. Y. Zhan, Y. Guo, J. Zhu, L. Li, Performance comparison of input current ripple reduction methods in UPS applications with hybrid PEM fuel cell/supercapacitor power sources, Int. J. Electr. Power 64 (2015) 96e103, http:// dx.doi.org/10.1016/j.ijepes.2014.07.020.
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
Reliability considerations of a fuel cell backup power system for telecom applications Mustafa Fazil Serincan Department of Mechanical Engineering, Istanbul Bilgi University, Kazım Karabekir Cad. No: 2/13, Eyup, 34060 Istanbul, Turkey
h i g h l i g h t s A fuel cell backup power unit is tested in real-world conditions at a GSM base station. Reliability of the fuel cell system is found to be 98.5% following 260 cycles. Inverter in the base station is redesigned to attenuate the current harmonics. Reliability of the system under fault conditions is studied. Sizing of system components is critical for an isolated operation during faults.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 November 2015 Received in revised form 18 January 2016 Accepted 22 January 2016 Available online 5 February 2016
A commercial fuel cell backup power unit is tested in real life operating conditions at a base station of a Turkish telecom operator. The fuel cell system responds to 256 of 260 electric power outages successfully, providing the required power to the base station. Reliability of the fuel cell backup power unit is found to be 98.5% at the system level. On the other hand, a qualitative reliability analysis at the component level is carried out. Implications of the power management algorithm on reliability is discussed. Moreover, integration of the backup power unit to the base station ecosystem is reviewed in the context of reliability. Impact of inverter design on the stability of the output power is outlined. Significant current harmonics are encountered when a generic inverter is used. However, ripples are attenuated significantly when a custom design inverter is used. Further, fault conditions are considered for real world case studies such as running out of hydrogen, a malfunction in the system, or an unprecedented operating scheme. Some design guidelines are suggested for hybridization of the backup power unit for an uninterrupted operation. © 2016 Elsevier B.V. All rights reserved.
Keywords: Fuel cell Reliability Telecom Power management Backup power Fault
1. Introduction Fuel cells have emerged as promising alternatives to conventional power generation technologies because of their high power and energy densities and environmental friendly operations. Contrary to these positive implications about fuel cells, economic maturity of the technology is far from being competitive to the conventional counterparts. Indeed, the commercialization progress is expected to be sluggish until the mass production of the fuel cell systems can be realized. On the other hand, there are some sectors where fuel cells are shown to be commercially more feasible alternatives already, compared to the conventional technologies. Material handling and
E-mail addresses: [email protected], [email protected]. http://dx.doi.org/10.1016/j.jpowsour.2016.01.083 0378-7753/© 2016 Elsevier B.V. All rights reserved.
backup power applications can be given as examples to earlymarkets for fuel cell technology. As a matter of fact, fuel cell forklifts and fuel cell backup power generators have been increasingly used worldwide in commercial applications [1]. Fuel cell systems have been demonstrated as more viable business cases compared to alternative technologies such as diesel and battery power units [2e6]. Deployments of fuel cell backup power systems for telecom applications have been on the rise throughout the world [7e15]. While fuel cell technology has become cost competitive in early market applications recently, studies are still going on to improve system efficiency and reliability and reduce return of investment durations. Reliability for telecom equipment is conventionally defined as “the ability of an item to perform a required function under given conditions for a given time interval” [16]. This definition can also be
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
interpreted as the probability of the system functioning as desired for a specified duration. However, it has been discussed that lumping the reliability concept to a single parameter may have limited practicality in telecom applications [17] and new metrics for reliability focusing on fault analysis have been proposed [18]. Reliability is a very important concept in the context of energy management of the base station. Alsharif et al. [19] discuss that one of the main drawbacks of diesel generators is the very low reliability of the technology when used as an energy source for an offgrid telecom base station. Kanzumba and Vermaak [20] report that diesel generators are responsible for 65% of the loss of telecom service due to failures of these devices. In this regards, fuel cell backup power units which have superior reliability compared to both diesel generators and batteries are more advantageous in telecom base stations. Reliability is a main concern for the design and implementation of a hybrid fuel cell system. Sulaiman et al. [21] conclude that apart from the economical concerns, main issues and challenges in a fuel cell hybrid system are due to the peripheral system components; such as the battery lifetime, energy management system and power electronics interface. It is asserted that most of the research regarding validation of the energy management system has been limited to simulations while ensuring the reliability of the systems in real world applications is required. In this regards, hybridization of a fuel cell system is of a key concern for a reliable and efficient operation [22e26]. Xhan et al. [27] proposed an energy management system for a hybrid UPS system which alternates operating modes between the fuel cell, the battery and the supercapacitors for a reliable power output. Li et al. [28] present a power sharing strategy for a fuel cell-batterysupercapacitors hybrid tramway in which the proposed system distributes the power demand to each source appropriately to guarantee a safe operation and extend the lifetime of each source. Xie et al. [29] developed an energy management algorithm for a hybrid powertrain which stabilizes the DC bus voltage via a fuzzy logic controller to protect the fuel cells and the batteries. The controller is also responsible for power distribution during a dynamic load cycle. Vasallo et al. [30] implemented an energy management strategy to determine the sizing of a hybrid UPS system for optimal power sharing between the fuel cell and the batteries. As stated earlier most of these studies are based on simulations while validation of power management system in real world applications and its influence on reliability is still of great interest. On the other hand, integration of the fuel cell system to the ecosystem of the application is also very critical. Guilbert et al. [31] study the interaction between the fuel cell and DC/DC converter in case of a faulty operating scheme in an electric vehicle application. They address that malfunction in the system at open circuit not only increases the fuel cell current ripples but also induces stresses on the inductors. They emphasize on the selection of the DC/DC converter for a reliable and efficient operation. Fontes et al. [32] studied the effects of current harmonics on a fuel cell stack. They conclude that although a fuel cell stack can filter high frequency current harmonics due to its double layer capacitance, impact of this operating scheme on the durability of the fuel cell and the maximum RMS current that can be tolerated by the fuel cell stack need to be investigated in more detail. Faults encountered during the operation may affect either the fuel cell system or the radio equipment in the base station or both. Hence malfunctions in the fuel cell system or unprecedented operating schemes have an imminent impact on the reliability. There have been a number studies recently focusing on the detection and isolation of the faults in hybrid fuel cell systems [33e41]. Most of these studies propose the use of sophisticated electronics designs and control techniques to minimize the effects
67
of the faults on the operational reliability of the systems. While active fault prevention methods like these are proven to be effective, they require complex system architectures and indubitably exacerbate the economic feasibility of a commercial fuel cell system. Hence, passive fault isolation techniques would be much more desirable, while reliability concerns with them are yet to be addressed. This study focuses on the reliability analysis of a commercial fuel cell backup power unit installed in a base station run by a Turkish GSM operator, Turkcell. Apart from a lumped system level approach, reliability of the fuel cell backup power unit is considered in real-life case studies; e.g. integration issues of the unit with the base station ecosystem, running out of hydrogen, and inadvertent interruption of the fuel cell operation by the site personnel. Unlike the controlled laboratory experiments, on-field tests present a unique environment for evaluating the reliability of the fuel cell backup power unit that endure a wide range of environmental conditions and unprecedented operating schemes, dynamic site loads and different levels of grid quality. Operation of the backup power system is elucidated in the following sections. Load sharing between the fuel cell and the batteries is examined for various operating conditions. Effects of utilizing an inverter to generate alternative current are discussed in the context of integration of the fuel cell backup power unit to the existing base station ecosystem. Inverter design is found to be decisive to get a high quality and stable power output from the backup power unit. Moreover, system responses are examined for separate cases when fuel cell exhibits a failure mode and recovers from a failure. Load sharing during a fuel cell failure is investigated to see how an uninterrupted system operation is sustained without any power loss. Finally, effects of hydrogen depletion on the system operation is discussed in a real-life case study for which the fuel cell is forced to respond to a power outage while the hydrogen in the tank is completely depleted. 2. System configurations and test setup A fuel cell backup power system has been installed in a base station of Turkish mobile telecom operator Turkcell in Bursa. A schematic of the installation can be seen in Fig. 1. The fuel cell backup power unit is comprised of a fuel cell power module and a startup battery, which are connected in parallel. There is a boost converter after the fuel cell stack increasing the voltage to the nominal operating range of the base station. Fuel cell power module also includes peripheral equipment accounting for thermal management and hydrogen and air supply to the stack. Main specifications of the tested backup power unit are given in Table 1. Rated power of the system is 6 kW at a nominal DC output voltage of 48 V while the typical load demand for the base station is about 1.5 kW. Also, the system is equipped with an inverter to provide the AC power to the air conditioner inside the base station cabinet during the grid failures. Note that negative potential is an industry standard in telecommunications. However, to avoid any confusion we have chosen to report positive voltage values in this study. A low temperature PEM fuel cell stack is exploited in the backup power unit with an operating temperature around 60e65 C. Heat generated during the backup operation is rejected from the fuel cell stack via liquid cooling. There is an external liquid to air heat exchanger connected to the backup power unit supplying cold water to the system. Hydrogen consumption of the fuel cell is about 14 slpm kW1 at the rated conditions while this value increases at the partial loads. Hydrogen is generated at the test site with an alkali electrolyzer
68
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
Fig. 1. Schematics of the experimental setup.
Table 1 Specifications of fuel cell backup power systems. Property
Value
Rated power Nominal output voltage Fuel cell type Thermal management Power management Fuel consumption at rated power Hydrogen input pressure Hydrogen supply Hydrogen storage Backup time at 1.5 kW Additional energy storage On site energy storage
6 kW 48 VDC/220 VAC Low temperature PEM Liquid cooling Output regulated with a DC/DC converter 14 slpm kW1 2e5 barg On-field electrolyzer 6 cylinders (49 L, 150 bar) ~40 h 55 Ah AGM batteries 100 Ah AGM batteries
unit. Output pressure of the electrolyzer is boosted with a compressor. Hydrogen is stored in standard 49 L industrial hydrogen cylinders at 150 bar. A bundle of 6 cylinders is used to guarantee at least 40 h of autonomy while additional cylinders could be added to increase the backup time. Hydrogen pressure is reduced to 2e5 barg at the inlet of the power module where it is delivered to the fuel cell through an internal buffer tank. Backup power unit is installed parallel to the existing site batteries in the base station. This redundant configuration is requested by Turkcell to guarantee fail-safe operation during the test period. However, some of the battery groups in the original configuration are deactivated to utilize the fuel cell unit more for backup power generation. In the final configuration site batteries correspond to 100 Ah while the startup batteries in the unit provide 55 Ah of additional storage. The main role of the batteries in this installation is to provide the backup power instantly when the fuel cell system kicks off from a dormant state. A fuel cell at start-up cannot generate its rated power until the temperature and the humidity in the stack reach steady-state operating conditions. While the response of a fuel cell strongly depends on the design of the stack, typical cold start-up times can be assumed in the order of 1 min [42]. In this transient phase, batteries provide power demanded by the base station. Output of the backup power unit is connected to the base station
DC bus. Current and voltage measurements are carried out at the fuel cell, battery and the load terminals as schematically depicted in Fig. 1. Current measurements are done via shunts placed on the power cables. The relation between the current measured at different shunts is
iFC þ ibatt ¼ iload
(1)
where negative value implies current flowing out, e.g. battery discharging while positive value implies current flowing in, e.g. battery charging. On the other hand,
VFC ¼ Vbatt ¼ Vload
(2)
because as it can be seen in Fig. 1 the fuel cell, battery and the DC bus are connected in parallel. Current and voltage measurements for the fuel cell reflect the values after the DC/DC converter. Note that load here accounts for anything connected to the DC bus beyond the measurement box. In other words, site batteries and the inverter also contribute to the load measured by the shunt apart from the telecom equipment. Voltage and current measurements are recorded by an industrial PC located inside the base station cabinet. Signals are first transmitted to a signal conditioning board where they are converted to digital inputs then sent to the I/O card found in the PC. Ambient temperature is also measured by a thermocouple, which is directly connected to the PC via the USB port. A standard test software written in Labview by the other project partners [44], is used to monitor and control the operation of the backup power unit. The PC is connected to the Internet via a 3G-modem enabling tests from a remote location. Both the industrial PC and the 3Gmodem are powered from the DC bus. Moreover, backup power unit has its own data logger containing internal measurements such as fuel cell anode pressure, stack temperature and minimum and maximum cell voltages. This internal dataset is transferred to the test PC via Ethernet. As the control output of the system, I/O card generates a signal and sends it to the AC switch to turn the electricity grid on and off remotely. For this purpose, a relay module is used in the signalconditioning box which generates a 5 A/30 VDC analogue signal.
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
The signal conditioning box itself is powered by 24 VDC through the DC bus. Control signal is then sent to the contactor switch installed on the AC line at the entrance of the base station. A grid failure simulation is initiated by controlling the switch remotely in the Labview program. After the test is completed, software drives the switch back to its normal position restoring the grid in the base station. Tests have been implemented following a standardized test procedure developed in the project as a collaborative effort [43,44]. Test procedure is comprised of short and long cycles corresponding to 15 min and 240 min of power outages respectively. Each cycle includes a waiting period before switching off the AC grid. Waiting period is either 1 min or 1 h (minimum) following the previous cycle to simulate hot or cold start-up respectively. As a performance metric, reliability of the backup power unit has been calculated quantitatively following the conventional definition given in the previous section as [16];
R¼
number of cycles number of system failures 100% number of cycles
(3)
where number of cycles is the total power outage simulations, system failures represent the instances when the system is unable to provide the power demanded by the base station. In other words, reliability considered here is the frequency of the backup power system conforming to the normal operating scheme during the test period. This gives a single indication for the reliability considered at the system level. Normal operation is determined based on the service voltage range from 40.5 VDC to 57.0 VDC, which is specified by the European Telecommunications Standards Institute (ETSI) [45]. Operating voltage being beyond this range for more than 20 ms is regarded as a failure. On the other hand, a malfunction in the fuel cell stack may be masked by the start-up battery since the response of the battery is instantaneous and the DC/DC converter set voltage guarantees the ETSI operating range. However, as the battery is being depleted, operating voltage will eventually drop below the minimum allowable limit. In the test setup this scenario is almost always avoided because of the additional capacity provided by the site batteries. In order to account for all the pitfalls in the fuel cell system, a failure is categorized as the likelihood of the system going beyond the ETSI voltage limits. 3. Results and discussion 3.1. Performance of the fuel cell backup power unit Fuel cell backup power unit installed in Turkcell site has followed a standard test protocol to simulate short and long electricity blackouts. A total of 260 power outage simulations were imposed on the system consisting of 228 short (15 min) and 18 long (4 h) standard test cycles with additional 14 custom cycles. Tests started on July 16th, 2012 and continued until April 6th, 2014. In the course of the tests fuel cell system produced 139.1 kWh of energy for the base station equipment for a cumulative 130 h of backup operation. This corresponds to an average power of 1 kW. Fuel cell system is found to be in a healthy state in 256 of the 260 test cycles, responding to load requests successfully by providing the demanded backup power to the telecom equipment. For the remaining 4 times fuel cell system unexpectedly failed to start. On the other hand, batteries installed in the unit have supplemented the load demand to sustain an uninterrupted backup operation. Although start-up batteries provide a buffer to prevent failure from happening immediately, working in this condition for an extended amount of time would eventually lead the system voltage to fall
69
down below 40.5 V. Therefore, these instances are considered as failures although the base station operation has never been interrupted in the course of the tests. The possible causes of the failures are discussed in the following sections. Moreover, there have been some instances that the backup power operation is ceased due to secondary issues other than the backup power unit itself. Some complications have been experienced with the on-site hydrogen generation equipment, which forced the tests to be halted for a while. Considering only the technical issues related to the backup power unit, reliability of the fuel cell system is calculated as 98.5% using Eq. (3):
R¼
260 256 100% ¼ 98:5% 260
(4)
This is a notable result for a fuel cell system tested comprehensively in a real-world application, which justifies technology as a commercial alternative to the existing backup power solutions. At the end of the tests, fuel cell backup power system is still capable to deliver its rated power. In real life conditions total number of cycles carried out would correspond to 5e10 years of commercial operation depending on the duration and the frequency of the electricity grid failures. Based on these observations, lifetime of the fuel cell system is proven to be commercially viable. Details on the test procedure and the performance results can be found in a recently submitted study [46]. Interpretation of reliability exploited up to now is reasonable only in a system level approach. On the other hand, each component in the system has its own characteristics that affect the performance and stability of the operation. Apart from considering reliability as a measure of capability to provide the requested power; how system reacts to different operating regimes and how individual components work with each other are also discussed. Hence, the context of reliability analysis will be extended further to incorporate investigation of individual components, especially the fuel cell stack. In the following sections, reliability of the backup power operation will be considered for a number of real-life case studies experienced in the course of field tests. 3.2. Effects of power management Two different power sources in the unit, fuel cell and battery can individually provide current to the system. In case of a cold startup, when fuel cell operating variables are far from their steady state values, the batteries provide most of the load demand. On the other hand, share of the total power supply provided by the fuel cell increases as the fuel cell warms up. This can be characterized by the ratio of the fuel cell current to the total current coming out of the backup power unit as;
SFC ¼
IFC 100 IFC þ Ibatt
(5)
Recall that Ibatt measurement represents only the start-up batteries as depicted in Fig. 1. Consistent with the notation used in Eq. (1) fuel cell and battery currents can have opposite signs. Such case can be attained at the start-up when some power is needed for the fuel cell to run the auxiliary equipment. In this case, SFC will be negative. A typical response of the backup power unit can be seen in Fig. 2, which shows the DC bus voltage and the load current. This test imposes three consecutive power outages on the base station each lasting 15 min. Test cycles are distinguished with the vertical dashed lines in the figure. First cycle invokes the system with a cold start-up, followed by two new cycles with 1-min waiting period in
70
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
Fig. 2. Response of the DC bus voltage (top), response of the load current (bottom) to 3 consecutive power outages of 15 min on 18 March 2014.
between power outages. During this period, electricity grid is restored and the fuel cell system shuts down. However, temperature of the system changes little in this period so the fuel cell exhibits warm start-up in the consecutive cycle. Before the grid failure, DC bus voltage reads 54.5 V, which is the set value of the base station rectifier. Until the fuel cell kicks in, only batteries deliver power to the base station. As the batteries are discharged, bus voltage drops. Power management algorithm of the backup power unit dictates that, fuel cell starts when the DC bus voltage falls below 50.5 V. Since a considerable number of the site batteries are deactivated during the tests, initial voltage drop is abrupt. Also, it can be seen from Fig. 2 that voltage further drops even after the fuel cell starts because the fuel cell takes some time to reach stable operating conditions. Nonetheless, once the fuel cell passes this transient operating regime, it delivers greater power, which results in a quick recovery of the bus voltage. Power management algorithm in the backup power unit prevents the batteries to be overcharged by the fuel cell. This is done via the DC/DC converter by increasing and decreasing the fuel cell output current alternatingly. This way, DC bus voltage is maintained around a nominal operating voltage. As can be seen in Fig. 2, bus voltage periodically varies between 51 and 51.5 V for the rest of the backup operation. Current output of the backup power unit can also be seen in Fig 2. When the grid goes off, batteries initially govern the load. Site batteries dominate in this stage, as they constitute a larger capacity compared to the start-up batteries installed in the unit. Batteries are discharged while they provide energy to the base station. As
discussed before, discharged site batteries also contribute to the load. It should be noted that load due to the other base station operation is relatively constant at around 30 A. After the fuel cell starts, it provides energy not only for the base station equipment but also to charge the batteries. Therefore, the load current goes up to 45 A. In the steady operation when the DC/DC converter regulates the system voltage, load current fluctuates between 25 and 35 A. It can be further estimated from this operating regime that when backup power unit gives 25 A, site batteries provide an additional 5 A to satisfy the telecom equipment demand. On the other hand, when system voltage is ramped up by the converter, fuel cell unit provides 35 A of which 5 A is dedicated for charging the batteries. Fig. 3 shows how the fuel cell and the unit batteries share the load (recall the sign convention in Eq. (1)). At the beginning of the cycle fuel cell module draws energy from the unit battery for about half a minute to power its peripherals. After getting the initial impetus, fuel cell output current gradually increases. 1 min after its start-up, fuel cell reaches steady operation and overtakes the startup battery in the unit. At around 90 s into the grid failure, battery current becomes negative meaning it is being charged. This is when the fuel cell is capable of fulfilling the entire load demand. At about 100 s into the grid failure fuel cell current reaches about 50 A. Assuming the base station load demand is approximately 30 A, it can be estimated that fuel cell provides 20 A to charge the batteries. The unit battery consumes about 5 A, so the remaining 15 A is dedicated for charging the site batteries. As the batteries are being charged, fuel cell current decreases gradually to its nominal operating values. During its regular operation, fuel cell current increases and decreases alternatingly to charge and discharge the batteries as modulated by the DC/DC converter. It can be suggested from Figs. 2 and 3 that batteries are put to charge mode for 2 min, while they are discharged for 3 min, forming a limit cycle every 5 min in a regular steady operating regime. 3.3. Effects of inverter design Telecom equipment standalone constitutes a load of 20 A, which corresponds to roughly 1 kW. Recalling the fuel cell unit is rated 6 kW, it is apparent that fuel cell operates in a sub-optimal region. As a matter of fact, in the early stages of the tests when the site load was only comprised of the telecom equipment, only 10e15% of fuel cell capacity was essentially utilized. Beyond working in a nonoptimal region, this would cause instable operation and decrease
Fig. 3. Load, fuel cell and battery currents during cold start-up cycle; the first cycle in Fig. 2.
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
the reliability of the entire system. For this reason, site load is increased by coupling the air conditioner to the DC bus with an inverter. Because of the intensive energy requirement of the air conditioners, site batteries are depleted rapidly during a grid failure. Therefore, typically when backup power is managed by the batteries, it is a common practice to maintain the temperature of the base station cabinet by free cooling. On the other hand free cooling has significant limitations and can provide cooling only as much as the ambient temperature allows. Therefore, air conditioner emerges as a necessity in warmer locations. A fuel cell system can be used in a base station where the cabinet is cooled with an air conditioner. For this purpose, DC output of the back up power unit is converted to AC via an inverter. On the other hand, design of the inverter is reported to be critical for the fuel cell operation [47e49] while an inverter that is suitable to be used with the batteries may not work impeccably with a fuel cell system. Also, fuel cell manufacturers would not have the information about the designs of the inverters found in the base station. Yet, the backup power unit is expected to have a seamless integration with the existing base station ecosystem. In this regard, field tests of fuel cell backup power units give an opportunity to see the performance of the systems for the use with a third party inverter. Initially an off-the shelf commercial inverter is integrated to the DC bus to power the air conditioner in the base station during the backup operation. However, the system exhibits a poor response with high frequency ripples in current and voltage. Fig. 4 shows the DC bus voltage and the load current responses to a 4-h test cycle. The insets in the figure gives a more detailed picture of the voltage and current harmonics over a duration of 5 min. It can be observed
Fig. 4. Response of the DC bus voltage (top) response of the load current (bottom) to a 4-h grid failure on 17 July 2012. Insets elaborates the voltage and current ripples caused by the inverter dynamics.
71
that ripples occur with an average period of 16 s and amplitude of 0.15 V. Ripples get even worse when voltage fluctuations reach up to 0.4 V at around 13:15. An identical trend in current ripples is observed with the same frequency. However, the effect is more pronounced as the amplitude of the fluctuations reaches 10 A when voltage ripples are at their highest. Current response also hints about the operating regime of the air conditioner. When air conditioner is off at the initial and final stages of the test, the load current reads 20 A. Operation with the air conditioner requires an additional 10 A on average. Yet, instantaneous current demand can go up to 50 A. Current response is determined by the changes in the air conditioner operating profile. As the compressor in the air conditioner ramps up and down periodically, current demand from the backup power unit varies correspondingly. Note that when the air conditioner is not working, ripples are attenuated but do not vanish completely. Harmonics seen in the load current will be identically reflected on both fuel cell and battery currents. This can be verified in Fig. 3 where all forms of current profiles exhibit identical fluctuations. Ripple currents are proven to shorten the lifetimes of both fuel cells and batteries [50e52]. Gemmen suggests the current ripples should be less than 10% for a reliable operation of the fuel cell system [53]. Also for the lead acid batteries it is recommended that the ripple voltage is less than 0.5% of the DC voltage [54]. Ripples experienced in the tests are well beyond these limitations. The specifications of the inverter states a voltage total distortion harmonics (TDH) value as much as 3%. Also, the output of the inverter is a modified sine which is a typical practice for most telecom applications. However, inverters with modified sine output must be tested and rated to be used with a particular equipment. These two are considered to be the main sources of high voltage and current ripples in the DC bus. After these observations, the inverter is sent back to the manufacturer and replaced with a new one having a custom design. Different strategies for elimination and mitigation of inverter ripples in fuel cell applications are outlined in the literature [55e57]. However, implementations of these designs are beyond the scope of our study. The inverter manufacturer applied its own methods to decrease the ripples practically. First modification is that the inverter gives a true sine output. Also, a choke coil is employed in the inverter to decrease voltage TDH. Although the new design may not be the most optimum for a fuel cell application, yet a significant reduction in ripples is achieved. Fig. 5 compares the voltage responses of the system for three different cases tested: i) there is no inverter installed (top), ii) a generic inverter is installed (middle), iii) a custom design inverter is installed (bottom). With the custom design inverter, voltage ripples are attenuated considerably. Voltage fluctuates in a band of 0.01 V on average, which has a negligible effect on the system performance. Rest of the tests has been implemented with this inverter. Likewise, ripples in current response can be observed with an average amplitude of 1 A. Following these discussions, it can be concluded that inverter design is very critical for the reliability of the fuel cell backup power unit. Generic inverter designs may cause instabilities in the system and may even damage the telecom equipment. It can even risk the fuel cell operation, as it will induce pressure variations inside the stack. Effect of these variations is yet to be studied. Issues with the inverter are experienced during the commissioning of the fuel cell backup system in the base station. On the other hand, two of the failures experienced with the power module exist shortly after the fuel cell is exposed to large current and voltage ripples, while the other two are experienced in the later stages of the field tests. This observation might suggest a
72
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
be sized carefully for extra backup times in order to sustain a failsafe operation in case of hydrogen depletion in the tank, hence to increase the reliability of the system. Four distinct failures experienced with the fuel cell system can hardly be linked to operation on low hydrogen levels, because the system has not gone through starvation. Two system failures happen before this event. On the other hand, after hydrogen depletion incident the fuel cell system exhibits normal operating regimes for a long time until the next two failures occur. Hence, these failures are more likely due to some other implications of long term tests. 3.5. Operation in failure and recovery
Fig. 5. Comparison of voltage responses when there is no inverter installed (top), a generic inverter is installed (middle), a custom design inverter is installed (bottom) in the DC bus.
correlation between the earlier failures with the ripples. Yet, there is not enough evidence to associate them certainly, as early failures might well be due to some other issues pertaining to the system design. 3.4. Hydrogen depletion Hydrogen supply is one of the main considerations when working with fuel cell systems. Backup duration is determined with the hydrogen storage on-site and can be extended by installing additional cylinders. As the system generates energy for the base station, hydrogen is being depleted continuously. Replacements cylinders must be provided when the hydrogen content in the system is diminishing below critical levels. In this test setup hydrogen is supplied from the on-site electrolyzer. However, due to a fault in the hydrogen generation equipment, electrolyzer stopped working. As a result hydrogen in the tank is depleted completely as the tests continue without being aware of this situation. Fig. 6 plots hydrogen pressure in the tank and in the fuel cell anode during a standard 4-h test cycle (plots on the left), and corresponding dynamics of fuel cell current and DC bus voltage (plots on the right). At the beginning of the test, hydrogen in the tank has a pressure of 8 bar but it decreases linearly as the fuel cell converts chemical energy stored in hydrogen to electric energy. Hydrogen pressure vanishes completely after 1 h. On the other hand, pressure at the stack is observed to be relatively constant at 0.3 bar. This is related to the additional capacity provided by the piping and internal components of the unit. Indeed, it is highly probable that a smaller buffer tank exists in the fuel cell system that lets the fuel cell operate even when hydrogen is entirely deprived in the main storage tanks. Corresponding system response is given in the plots on the right hand side. Both fuel cell current and voltage profiles are observed to be normal throughout the 4-h test cycle. On the other hand, once the internal buffer is also consumed fuel cell will fail and the batteries will deliver the backup power. An internal buffer tank must
Reliability of the system in case of a failure is another important consideration. Failure of the fuel cell may occur either before the system starts or in the middle of the operation. The former is a relatively simpler management task which requires incorporation of a temporary backup power supply until the fuel cell maintenance takes place. However, the latter scenario happens unexpectedly and does not allow a lead-time to take a corrective action. In this case, the system must sustain the backup power demanded by the base station until the site personnel arrive at the site and install an alternative backup source. Fig. 7 shows current and voltage response of the system when the fuel cell unprecedentedly fails soon after it is started. It can be observed from the figure that the fuel cell is activated at 11:18 when the DC bus voltage falls below 50.5 V. After the auxiliaries in the fuel cell module are initiated, fuel cell starts providing current. Fuel cell gives 20 A at its first attempt but it is followed by an abrupt drop in current resulting in loss of power completely. In the meantime batteries kick in immediately and compensate for the fuel cell power loss. As a result backup operation is recovered very quickly and voltage drop due to fuel cell failure is limited to only 0.5 V. Share of the load is taken completely by the batteries for the rest of the test. For a grid failure of 15 min batteries maintain the demanded power without any significant voltage drop. Note that battery current plotted in Fig. 7 is for the start-up batteries in the unit only. Assuming base station load is 30 A on average, it can be estimated that 1/3 of the load is covered by the start-up batteries while 2/3 is covered by the site batteries. On the other hand, batteries will be depleted and bus voltage will eventually drop below ETSI limits for extended backup durations. Also when the site batteries are completely decoupled, an ultimate system failure will be encountered sooner. In this regard batteries directly affect system reliability. Therefore, it is crucial to size and select the batteries in the unit carefully for sustaining backup operation even in case of a fuel cell failure, hence achieving higher system reliability. Although it may not be as critical as the foregoing scenario, it is also interesting to examine how the system responds in case fuel cell recovers from a failure when the system is already running on the batteries. Fig. 8 depicts current and voltage responses in a similar situation when the fuel cell is manually started at the final stages of the test. Note that this test is implemented with the entire site batteries coupled to the DC bus and the inverter is deactivated. This explains relatively slow voltage drop. It can be observed that fuel cell does not start when the voltage falls far below 50.5 V. After the site personnel check the system status, error messages are found on the fuel cell interface. Once the error messages are cleared, fuel cell is reset and started manually at 11:58 p.m. Following its regular start-up routine fuel cell begins delivering current to the load, however, it cannot undertake the entire load share right away. Tests are discontinued shortly after, to
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
73
Fig. 6. Change in hydrogen tank pressure (top left) and anode pressure (bottom left) during a 4-h grid failure on 16 October 2012. Responses in fuel cell current (top right) and DC bus voltage (bottom right) during the same failure.
Fig. 7. Current and voltage responses of the system to a 15 min power outage on 24 December 2013. An unexpected fuel cell failure is experienced at 11:19 a.m.
implement a standard cycle. Here the importance of the batteries in the system is once more emphasized because even when the fuel cell recovers from failure it may take some additional time to reach its nominal operating regime. 4. Conclusions A fuel cell backup power system has been tested under real life conditions at a telecom base station operated by Turkcell. A test procedure consisting of 260 cycles is implemented to estimate the long-term reliability of the system in case of real life grid failure scenarios. The entire system responded 256 of the 260 test cycles successfully. With a system level approach, reliability of the system is calculated as 98.5%. For the remaining 4 times fuel cell failed to provide the requested current. However, with the start-up batteries backup operation is sustained without any disruption. Also, the system is investigated at component level to have a more comprehensive reliability analysis. It is shown that power
Fig. 8. Current and voltage responses of the system to a 15 min power outage on 17 July 2012. Fuel cell recovers from failure at 11:58 a.m.
management of the unit is decisive in load sharing between the fuel cell and the batteries. The batteries govern system response in the transient start-up period. Merely 1 min into the backup operation, fuel cell is warmed up and undertakes almost the entire load demand. Moreover, inverter design is found to have a great impact on the reliability of the system. It has been shown that a third party generic design inverter used by the operator results in large ripples in system response. While the voltage fluctuates as much as 0.4 V, current ripples are more pronounced with fluctuations reaching 10 A. With a custom design inverter, ripples attenuated considerably by 90%. Batteries installed in the unit are shown to be very decisive in improving the reliability of the unit. In case of a fuel cell failure in the middle of a backup operation, batteries account for the entire load demand instantly, without letting the system voltage drop below the limits. Also batteries are shown to support fuel cell recovering from a failure. It is imperative to select and size start-up
74
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75
batteries in the unit to increase reliability of the operation. Further, the system is tested when the fuel cell runs out of hydrogen in the middle of a backup operation. It has been observed that even when the hydrogen is completely depleted in the main storage; hydrogen content inside the unit is sufficient enough to sustain the backup operation for another 3 h. Capacity of the pipes and other system components provide a safety margin for the operation. In this regard, design of an internal buffer tank is very important to increase the reliability of the system amid the risk of hydrogen starvation. Tests at real-life conditions have proven the reliability of the fuel cell backup power system at both system and component levels. Yet, to guarantee the utmost reliability of the operation and perfect compliance with the base station ecosystem, improvements are needed not only on the fuel cell technology but on the peripheral system components as well. Acknowledgement The research leading to these results has received funding from the European Union's Seventh Framework Programme (FP7/20072013) for the Fuel Cells and Hydrogen Joint Technology Initiative under grant agreement no [256766]. Matching funds have been received from Republic of Turkey, Ministry of Energy and Natural Resources under the “International Centre for Hydrogen Energy Technologies” project. Turkcell Communication Services Inc. further provided a great amount of in kind support throughout the study. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.jpowsour.2016.01.083. References [1] Fuel Cell Industry Review 2013, Johnson Matthey PLC, Herts, UK, 2013. [2] DOE, Hydrogen Program e Early Markets: Fuel Cells for Backup Power, 2009. [3] APCO, Annual Conference and Expo: Fuel Cells for Critical Communications Backup Power, 2009. [4] J. Kurtz, G. Saur, S. Sprik, C. Ainscough, Backup Power Cost of Ownership Analysis and Incumbent Technology Comparison, NREL/TP-5400-60732, 2014. [5] Warid Pakistan Expects to Save More than US$ 6 Million per Year After Trialling Energy Efficiency Solutions, Mobile Energy Efficiency Optimization Case Study, first ed., GSMA, London, 2014. [6] Pakistan telecom trial validates ElectraGen backup power benefits, Fuel Cells Bull. 2014 (4) (April 2014) 3. [7] DOE, Hydrogen and Fuel Cells Program Record #14009. [8] 2011 Fuel Cell Technologies Market Report, US Department of Energy, Energy Efficiency and Renewable Energy, 2012. [9] Ballard Power Systems Inc, Annual Information Form, February 26, 2014. [10] Progress in FCH deployment in Europe and the world, in: Ballard Company Presentation by John Sheridan, FCH JU 6th Stakeholder General Assembly, 13.11.2013. [11] 2013 Fuel Cell Technologies Market Report, US DOE Energy Efficiency and Renewable Energy, November 2014. [12] Ballard deploys initial units at telecom sites in Manila, new CEO, Fuel Cells Bull. 2014 (9) (2014) 3. [13] 2010 Fuel Cell Technologies Market Report, US Department of Energy, Energy Efficiency and Renewable Energy, 2011. [14] 2012 Fuel Cell Technologies Market Report, U.S. Department of Energy, Efficiency and Renewable Energy, 2013. [15] Cascadiant wins two clean energy telecom contracts in Indonesia, Fuel Cells Bull. 2013 (2) (2013) 5. [16] ITU-T Recommendation E.800 08/94 p.17, Switzerland, Terms and Definitions Related to Quality of Service and Network Performance Including Dependability, 1994. [17] M. Ayers, Telecommunications System Reliability Engineering, Theory, and Practice, Wiley, Hoboken, N.J, 2012. [18] H. Kehl, Fault Analysis and Fault Tolerance of a Base Station System for Mobile Communications, PhD Dissertation in Technische Universitat Munchen, 2000. [19] M. Alsharif, R. Nordin, M. Ismail, Energy optimisation of hybrid off-grid system for remote telecommunication base station deployment in Malaysia, J. Wirel. Com. Netw. 64 (2015), http://dx.doi.org/10.1186/s13638-015-0284-7.
[20] M. Crouch, Fuel Cell Systems for Base Stations: Deep Dive Study, first ed., GSMA Green Power for Mobile White Paper, London (, 2011. [21] N. Sulaiman, M. Hannan, A. Mohamed, E. Majlan, W. Wan Daud, A review on energy management system for fuel cell hybrid electric vehicle: issues and challenges, Renew. Sust. Energy Rev. 52 (2015) 802e814, http://dx.doi.org/ 10.1016/j.rser.2015.07.132. [22] C. Chao, J. Shieh, A new control strategy for hybrid fuel cell-battery power systems with improved efficiency, Int. J. Hydrogen Energy 37 (17) (2012) 13141e13146, http://dx.doi.org/10.1016/j.ijhydene.2012.03.143. [23] F. Odeim, J. Roes, L. Wülbeck, A. Heinzel, Power management optimization of fuel cell/battery hybrid vehicles with experimental validation, J. Power Sources 252 (2014) 333e343, http://dx.doi.org/10.1016/ j.jpowsour.2013.12.012. [24] J. Hwang, W. Chang, Characteristic study on fuel cell/battery hybrid power system on a light electric vehicle, J. Power Sources 207 (2012) 111e119, http://dx.doi.org/10.1016/j.jpowsour.2012.02.008. [25] M. Abu Mallouh, E. Abdelhafez, M. Salah, M. Hamdan, B. Surgenor, M. Youssef, Model development and analysis of a mid-sized hybrid fuel cell/battery vehicle with a representative driving cycle, J. Power Sources 260 (2014) 62e71, http://dx.doi.org/10.1016/j.jpowsour.2014.02.104. [26] J. Bernard, M. Hofer, U. Hannesen, A. Toth, A. Tsukada, F.N. Büchi, P. Dietrich, Fuel cell/battery passive hybrid power source for electric powertrains, J. Power Sources 196 (14) (2011) 5867e5872, http://dx.doi.org/10.1016/ j.jpowsour.2011.03.015. [27] Y. Zhan, Y. Guo, J. Zhu, L. Li, Power and energy management of grid/PEMFC/ battery/supercapacitor hybrid power sources for UPS applications, Int. J. Electr. Power 67 (2015) 598e612, http://dx.doi.org/10.1016/ j.ijepes.2014.12.044. [28] Q. Li, W. Chen, Z. Liu, M. Li, L. Ma, Development of energy management system based on a power sharing strategy for a fuel cell-battery-supercapacitor hybrid tramway, J. Power Sources 279 (2015) 267e280, http://dx.doi.org/ 10.1016/j.jpowsour.2014.12.042. [29] C. Xie, X. Xu, P. Bujlo, D. Shen, H. Zhao, S. Quan, Fuel cell and lithium iron phosphate battery hybrid powertrain with an ultracapacitor bank using direct parallel structure, J. Power Sources 279 (2015) 487e494, http://dx.doi.org/ 10.1016/j.jpowsour.2015.01.029. [30] M. Vasallo, J. Bravo, J. Andújar, Optimal sizing for UPS systems based on batteries and/or fuel cell, Appl. Energy 105 (2013) 170e181, http://dx.doi.org/ 10.1016/j.apenergy.2012.12.058. [31] D. Guilbert, A. Mohammadi, A. Gaillard, A. N'Diaye, A. Djerdir, Interactions between fuel cell and DC/DC converter for fuel cell electric vehicle applications: influence of faults, in: IECON 2013-39th Annual Conference of the IEEE Industrial Electronics Society, 2013, http://dx.doi.org/10.1109/ iecon.2013.6699255. [32] G. Fontes, C. Turpin, S. Astier, T. Meynard, Interactions between fuel cells and power converters: influence of current harmonics on a fuel cell stack, IEEE Trans. Power Electr. 22 (2) (2007) 670e678, http://dx.doi.org/10.1109/ tpel.2006.890008. [33] H. Aouzellag, K. Ghedamsi, D. Aouzellag, Energy management and fault tolerant control strategies for fuel cell/ultra-capacitor hybrid electric vehicles to enhance autonomy, efficiency and life time of the fuel cell system, Int. J. Hydrogen Energy 40 (22) (2015) 7204e7213, http://dx.doi.org/10.1016/ j.ijhydene.2015.03.132. [34] A. Rosich, R. Sarrate, F. Nejjari, On-line model-based fault detection and isolation for PEM fuel cell stack systems, Appl. Math. Model. 38 (11/12) (2014) 2744e2757, http://dx.doi.org/10.1016/j.apm.2013.10.065. [35] Z. Li, R. Outbib, S. Giurgea, D. Hissel, Y. Li, Fault detection and isolation for polymer electrolyte membrane fuel cell systems by analyzing cell voltage generated space, Appl. Energy 148 (2015) 260e272, http://dx.doi.org/ 10.1016/j.apenergy.2015.03.076. [36] D. Rotondo, V. Puig, F. Nejjari, Fault tolerant control of a pem fuel cell using qLPV virtual actuators, IFAC-PapersOnLine 48 (21) (2015) 271e276, http:// dx.doi.org/10.1016/j.ifacol.2015.09.539. [37] D. Guilbert, A. Gaillard, A. Mohammadi, A. N'Diaye, A. Djerdir, Investigation of the interactions between proton exchange membrane fuel cell and interleaved DC/DC boost converter in case of power switch faults, Int. J. Hydrogen Energy 40 (1) (2015) 519e537, http://dx.doi.org/10.1016/ j.ijhydene.2014.10.072. [38] M. Kamal, D. Yu, D. Yu, Fault detection and isolation for PEM fuel cell stack with independent RBF model, Eng. Appl. Artif. Intel. 28 (2014) 52e63, http:// dx.doi.org/10.1016/j.engappai.2013.10.002. [39] L. Riascos, M. Simoes, P. Miyagi, On-line fault diagnostic system for proton exchange membrane fuel cells, J. Power Sources 175 (1) (2008) 419e429, http://dx.doi.org/10.1016/j.jpowsour.2007.09.010. [40] A. De Bernardinis, D. Candusso, F. Harel, G. Coauerv, Power electronics interface for an hybrid PEMFC generating system with fault management strategies for transportation, in: Proceedings of 13th European Conference on Power Electronics and Applications, Barcelona, 2009, pp. 1e10. [41] D. Guilbert, M. Guarisco, A. Gaillard, A. N'Diaye, A. Djerdir, FPGA based faulttolerant control on an interleaved DC/DC boost converter for fuel cell electric vehicle applications, Int. J. Hydrogen Energy 40 (45) (2015) 15815e15822, http://dx.doi.org/10.1016/j.ijhydene.2015.03.124. [42] FCgen-1020ACS Fuel Cell Stack, Product Manual, Ballard, 2010. [43] submitted to FCH-JU, Development of Test Protocol, Fitup Project Deliverable, 2011.
M.F. Serincan / Journal of Power Sources 309 (2016) 66e75 [44] U. Trachte, P. Sollberger, B. Wellig, Fuel cell field test demonstration of economic and environmental viability for portable generators, backup and UPS power system applications, in: Proceedings of 4th European PEFC and H2 Forum 2013, Lucerne, 2013, p. B0404. [45] ETSI EN 300 132-2, France, Environmental Engineering (EE); Power Supply Interface at the Input to Telecommunications and Datacom (ICT) Equipment; Part 2: Operated by -48 V Direct Current (dc), 2011, p. 9. [46] M.F. Serincan, Validation of fuel cell hybridization methodologies in realworld telecom backup power applications, submitted to Int. J. Hydrogen Energy. (unpublished results). [47] N. Bizon, A new topology of fuel cell hybrid power source for efficient operation and high reliability, J. Power Sources 196 (6) (2011) 3260e3270, http:// dx.doi.org/10.1016/j.jpowsour.2010.11.049. [48] W. Shireen, R.A. Kulkarni, M. Arefeen, Analysis and minimization of input ripple current in PWM inverters for designing reliable fuel cell power systems, J. Power Sources 156 (2) (2006) 448e454, http://dx.doi.org/10.1016/ j.jpowsour.2005.06.012. [49] A. Kirubakaran, S. Jain, R.K. Nema, A review on fuel cell technologies and power electronic interface, Renew. Sust. Energy Rev. 13 (9) (2009) 2430e2440, http://dx.doi.org/10.1016/j.rser.2009.04.004. [50] J. Kim, I. Lee, Y. Tak, B. Cho, Impedance-based diagnosis of polymer electrolyte membrane fuel cell failures associated with a low frequency ripple current, Renew. Energy 51 (2013) 302e309, http://dx.doi.org/10.1016/ j.renene.2012.09.053. n, A. Jimenez, E. Moreno, Power [51] J.J. Brey, C.R. Bordallo, J.M. Carrasco, E. Galva
[52]
[53]
[54]
[55]
[56]
[57]
75
conditioning of fuel cell systems in portable applications, Int. J. Hydrogen Energy 32 (2007) 1559e1566, http://dx.doi.org/10.1016/ j.ijhydene.2006.10.033. W. Lee, B. Han, H. Cha, Battery ripple current reduction in a three-phase interleaved dc-dc converter for 5kW battery charger, in: 2011 IEEE Energy Conversion Congress and Exposition, Phoenix, 2011, pp. 3535e3540, http:// dx.doi.org/10.1109/ECCE.2011.6064247. R.S. Gemmen, Analysis for the effect of inverter ripple current on fuel cell operating condition, J. Fluids Eng. 125 (3) (2003) 576e585, http://dx.doi.org/ 10.1115/1.1567307. Effects of AC Ripple Current on VRLA Battery Life, Emerson Network Power white paper, Columbus OH, http://www.emersonnetworkpower.com/ documentation/en-us/brands/liebert/documents/white%20papers/effects% 20of%20ac%20ripple%20current%20on%20vrla%20battery%20life.pdf, (accessed 12.01.16). G.R. Zhu, S.C. Tan, Y. Chen, C.K. Tse, Mitigation of low-frequency current ripple in fuel-cell inverter systems through waveform control, IEEE Trans. Power Electr. 28 (2) (2013) 779e792, http://dx.doi.org/10.1109/TPEL.2012.2205407. S.K. Mazumder, R.K. Burra, K. Acharya, A ripple-mitigating and energyefficient fuel cell power-conditioning system, IEEE Trans. Power Electr. 22 (4) (2007) 1437e1452, http://dx.doi.org/10.1109/TPEL.2007.900598. Y. Zhan, Y. Guo, J. Zhu, L. Li, Performance comparison of input current ripple reduction methods in UPS applications with hybrid PEM fuel cell/supercapacitor power sources, Int. J. Electr. Power 64 (2015) 96e103, http:// dx.doi.org/10.1016/j.ijepes.2014.07.020.