international journal of hydrogen energy xxx (xxxx) xxx
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/he
Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles M. Virji a,*, G. Randolf b, M. Ewan c, R. Rocheleau c a
Hydrogen and Renewable Energy System Analysis (H2RESA), Pickering, Ontario, L1X 1R1, Canada GRandalytics, Honolulu, HI, 96822, USA c Hawaii Natural Energy Institute, University of Hawaii-Manoa, Honolulu, HI, 96822, USA b
highlights Hydrogen energy system (HES) is analyzed as a potential grid management tool. A simulation tool is presented and employed for grid management investigation. Framework for electrolyzer long-term degradation study under renewable cyclic load. Insight into critical fact when designing HES for grid management application. The impact of operating HES as a pure grid frequency stabilizer is also presented.
article info
abstract
Article history:
One of the objectives of the research project at Hawaii Natural Energy Institute (HNEI) is to
Received 7 October 2019
demonstrate long-term durability of the electrolyzer when operated under cyclic operation
Received in revised form
for frequency regulation on an Island grid system. In this paper, a Hydrogen Energy System
9 January 2020
with an electrolyzer is analyzed as a potential grid management tool. A simulation tool
Accepted 12 January 2020
developed with a validated model of the hydrogen energy system and Island of Hawaii grid
Available online xxx
model is presented and employed for this investigation. The simulation study uses realistic measured solar and wind power profiles to understand what optimal electrolyzer size
Keywords:
would be required to achieve the maximum level of grid frequency stabilization. The
Hydrogen
simulation results give insight into critical information when designing a hydrogen energy
Hydrogen energy system
system for grid management applications and the economic impact it has when operated
Electrolyzer
as a pure grid management scheme or as a limitless hydrogen production system.
Grid management
© 2020 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.
Renewable energy
Introduction The exploitation of renewable energy resources on the Hawaiian Isles are increasing rapidly as the cost of electricity from the imported conventional fossil fuel energy sources become expensive and unsustainable. The very ambitious
goal adopted by the Hawaii’s state legislature in 2015 for the Hawaii Clean Energy Initiative of having 100% renewable energy by 2045, has further magnified the development of renewables in Hawaii [1,3,4]. This growth in environmental friendly renewable energy resources has had and will continue to have significant impact on the performance and efficiency of the small electric grids on the Islands and has
* Corresponding author. E-mail address:
[email protected] (M. Virji). https://doi.org/10.1016/j.ijhydene.2020.01.070 0360-3199/© 2020 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved. Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
2
international journal of hydrogen energy xxx (xxxx) xxx
necessitated the use of grid balancing technology such as energy storage systems to address the instability of these confined and secluded electrical grid systems [2,3]. The integration of renewable energy into existing small island electrical power grids comes with its own challenges. Firstly, due to variability and uncertain nature of the renewable energy, especially solar and wind energy, the grid system has to be constantly balanced. Secondly, the grid system has to manage the peak and base loads during generation inadequacy and surplus situations [5]. Lastly but not the least, the renewable energy sources are not always evenly distributed or close to the consumer in order to maximize their integration and usage. Some of these challenges can be eliminated or minimized by use of an electrolyzer-based hydrogen energy system (HES) which can increase or curtail hydrogen production to level out renewable energy production fluctuations. A hydrogen energy system with an electrolyzer has many different applications in the chemical process and power generation industries and most recently greater numbers of these systems are being used for fueling stations to provide hydrogen for fuel cell vehicles. Hydrogen production by electrolysis for fueling station is sensible and economical if the electricity is produced from a renewable energy source [6]. In order to utilize these electrolyzers for grid management, they should have some specific dynamic characteristics. The electrolyzer should respond fast within faction of seconds to commanded production levels in order to balance energy availability and be durable and secure during these cyclic operations. They should also have the ability to safely startup (from cold to full power) and shutdown (from maximum power) within several minutes [5]. Many commercial electrolyzers are not designed to operate in the dynamic or load-following mode, but mostly in a constant demand mode until storage capacity is reached. However, research work at the National Renewable Energy Laboratory (NREL), has revealed that electrolyzers can respond sufficiently fast and are capable of operating for long duration to participate in grid management applications [5]. Initial research work at Hawaii Natural Energy Institute (HNEI) has also shown that with modified control algorithm, a commercial electrolyzer can respond within fraction of seconds to a given dynamic setpoint demand and can be operated effectively in a cyclic mode with a dynamic load profile [6]. Although, the electrolyzer system is capable of reacting to changes in the range of about 2 to 0.5 Hz, it is not as fast as a Battery Energy Storage System (BESS) which can balance instantaneous grid-wide mismatches between generation and load within 100 ms [7]. HNEI has a 65 kg/day electrolyzer hydrogen fueling station located at the Natural Energy Laboratory of Hawaii Authority (NELHA) on the Island of Hawaii [8], It is being used to supply 350 bar hydrogen for fuel-cell hybrid buses for the County of Hawaii Mass Transportation Agency (MTA) and Hawai’i Volcanoes National Park (HAVO) [6]. One of the objectives of this research project is to demonstrate long-term durability of the electrolyzer when operated under cyclic operation for frequency regulation on an Island grid system [8]. As most commercial electrolyzers are usually designed to operate at constant loads and consequently, long term durability studies of electrolyzers under cyclic load are marginal [9e11]. The degradation of electrolyzer stacks performance is more likely
to occur under intermittent power source where they will experience extremely fluctuating loads and frequent startup/ shutdowns. A recent 250 h durability test under a solar cycle with highly fluctuating loads and frequent shutdowns resulted in an average voltage degradation rate of 280 mV/h [12] which was significant to a commercial product of Proton OnSite has shown when it was run over 20,000 h [10]. Therefore, this research project also includes development of a Hardware-in-the-Loop (HiL) measurement system for assessing and integrating the long term degradation of the electrolyzer stacks (hardware) under the impacts of dynamic operating conditions. However, in this paper the framework for long term test using hardware-in-the-loop (HiL) technique which requires a simulation tool with an electrolyzer cell and system model is detailed [13,14]. In order to better understand the performance of the 65 kg/ day HES for grid management application, a model of the system was developed and tuned to match the measured dynamic characteristics of the plant. An electrical grid model of the Island of Hawaii was also developed to evaluate the operational limits of the HES for grid management under measured solar and wind power profiles. The scope of this paper is to describe simulation tool with a validated model of the hydrogen energy system and Island of Hawaii grid model as well as present the simulation grid management results under realistic measured renewable power profiles.
Hydrogen energy system specifications The HES consists of a Proton C30 electrolyzer unit with two cooling chillers to produce hydrogen at 30 bar at a rate of 65 kg/day. The hydrogen is stored in a small buffer tank and is further compressed to 450 bar by a HydroPac compressor at a rate of 120 kg/day. The compressed hydrogen is then stored in 12 tanks inside a transport trailer which is connected via a fueling post for onsite dispensing or transported to other dispensing site on the Island. There are 3 hydrogen transport trailers (HTT) which can store up to 100 kg of hydrogen each. Fig. 1 shows the overview of the main components of the HES. The hydrogen generator in the HES is a Proton Exchange Membrane (PEM) water electrolysis system configured with three PEM stacks each operating at maximum current of 410 A to produce 30 Nm3/h. of hydrogen. The unit has a deionizedwater (di-water) production and circulating system to supply di-water to the stacks for generating hydrogen and oxygen, and for removing excess heat. The oxygen is safely vented out to the atmosphere while the hydrogen is separated from water and dried to deliver high purity (99.999%) product gas at 30 bar. The unit has two liquid cooling systems to maintain stack and power supply temperatures of 50 C and 40 C respectively. Supervisory controls enable safe performance and automatic operation of the electrolyzer. The C30 Proton electrolyzer is a fully automated system and like many of these commercial units, it regulates the hydrogen production rate to maintain the set buffer tank pressure. However, this particular electrolyzer unit has a mass flow controller to govern its operation externally via a 4e20 mA signal. This allows for controlling the electrolyzer and its hydrogen production, for example, by a grid management algorithm.
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
international journal of hydrogen energy xxx (xxxx) xxx
The product hydrogen is stored in a buffer tank before its compressed to 450 bar by a HydroPac compressor system. The compressor only operates when the HTT is connected to the fueling post and the cylinders are below a certain pressure. The compressor starts automatically when the buffer tank pressure is greater than 29.5 bar and runs until the buffer tank pressure drops below 16.8 bar. One compressor cycle takes about 20 min. The operation of the compressor and the monitoring of all critical pressures and temperatures of the system are done by the main programmable logic controller (PLC). The PLC is also responsible for the smooth and safe operation of all the main and ancillary components of the fueling station which includes the H2 generator, compressor, three chillers, fueling posts and hydrogen transport trailer storage. The dynamic operation of the electrolyzer via a 4e20 mA control signal is also implemented into the PLC and available through the Human Machine Interface (HMI). Normally the hydrogen energy system is operated at a constant maximum production rate from the HMI to generate hydrogen on demand to refill the storage capacity of the fueling station. One of the primary objectives of this project is to evaluate the operation of the electrolyzer in a dynamic mode by the grid management control algorithm in order to stabilize frequency of the Island of Hawaii electrical grid which fluctuates due influx of intermittent solar and wind renewable power sources. Utilizing this external control capability of the electrolyzer, a PLC was employed to run various test patterns including step functions and frequency sweeps at various amplitudes and offsets to validate the model, refine parameters and obtain an envelope of operations. Fig. 2 shows the front-end of the PLC that was employed. Through this PLC front-end, it was also possible to upload measured BESS power data in order to investigate the electrolyzer’s behavior with actual power demand data. Different filter algorithms and cut-off limits could be applied to the data to find how much a grid managing battery could be reduced in size by getting support from an electrolyzer [6,15]. A feasibility study was performed to understand the level of grid improvement with the current size of electrolyzer and
3
what optimal electrolyzer size would be required to achieve the maximum level of grid frequency stabilization. The impact of operating the electrolyzer in dynamic mode on the production of the hydrogen for the fueling station was also analyzed using the PLC (Fig. 2). A dynamic model of electrolyzer system and the Island of Hawaii grid were developed to be used in the HES Simulation Tool for this feasibility study under different operational strategies and renewable power profiles.
Modeling of the hydrogen energy system Models of main components of the HES were developed and validated against measured data logged during the precommissioning and testing of the system in March 2015 [6].
PEM electrolyzer cell model The C30 Proton electrolyzer unit uses three Proton Exchange Membrane (PEM) electrolyzer stacks to produce hydrogen. Many PEM cells are connected in series to make one electrolyzer stack. Each cell consists of anode and cathode electrodes which are separated physically by a solid proton exchange membrane. The hydrogen is produced in each cell by water electrolysis, an electrochemical reaction, whereby the water is split into hydrogen and oxygen when voltage is applied across the anode and cathode electrodes of the cell. The electrochemical reactions for hydrogen production of a PEM electrolyzer cell are represented in the following equations: Anode: H2 OðIÞ þ DC Electricity/2Hþ þ ½O2ðgÞ þ 2e Cathode: 2Hþ þ 2e /H2ðgÞ Overall: H2 OðIÞ þ DC Electricity/H2ðgÞ þ ½O2ðgÞ
(1)
The water fed into the anode side is split into hydrogen ions, electrons and oxygen gas by the higher potential (voltage) provided by the DC electrical power source. The resultant hydrogen ion (Hþ) and some water molecules are transported across the PEM to the cathode side while the un-
Fig. 1 e Overview of the main components of the HES. Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
4
international journal of hydrogen energy xxx (xxxx) xxx
Fig. 2 e The Front-end of the PLC used to evaluate the dynamic operational envelope.
reacted water and product O2 gas exit the cell at the anode outlet. The hydrogen ions are reduced to H2 gas in the cathode side by the electrons flowing through the external circuit. The cells are usually pressurized to a nominal pressure so that the hydrogen product gas exits at desired pressure. The product gas is further treated in a dryer unit to achieve high purity (99.999%) pressurized hydrogen [16]. The primary aim of developing a PEM electrolyzer cell model is to estimate the cell voltage as a function of operating current, temperature and pressure in order to determine the power consumption and efficiency of the electrolyzer system. These operating parameters as well as others, such as the electrodes and membrane properties, are varied to define a polarization curve which is validated with a measured voltage-current (VI) characteristic curve. The standard cell model used in many research papers to estimate the PEM electrolyzer cell voltage is given by equation (2) [17e20]. Vcell ¼ Vo þ hact; ao þ hact; ca þ hohmic þ hcon
(2)
where, VO is the equilibrium voltage hact, ao &hact, ca are activation over-potentials at anode and cathode, respectively hohmic is the ohmic over-potential across the proton exchange membrane hcon is the concentration over-potential. In a practical PEM electrolyzer, the cell voltage (Vcell) required for the electrochemical reactions for hydrogen production is always larger than the equilibrium voltage (Vo) or the theoretical open circuit voltage (OCV) due to the activation, ohmic and concentration over-potentials which represent the irreversible losses within the cell [17,19]. These three major irreversible losses occur at different current densities and can be estimated by different kinetic loss mechanisms in the cell. The equations which govern the equilibrium voltage and the kinetic loss mechanisms for the three major irreversible losses are defined in the next section. Validation of these models with measured data or
parameters used in the research literature is also described in the next section.
Equilibrium voltage (Vo) The equilibrium voltage or open circuit voltage for PEM water electrolysis as characterized in equation (1) can be derived from the Nernst equation and is given by the following equation (3) [17,19,21]: pffiffiffiffiffiffiffi DS R:TEzer PH2: PO2 TEzer Tref þ ln n:F n:F PH2O
Vo ¼ Eo
(3)
where, Eo - Standard cell potential at 298 K and 1atm DS - Molar change in entropy of water n - Number of electrons exchanged in a chemical reaction F - Faraday’s constant TEzer - Cell operating temperature (K) Tref - Reference temperature (K) R - Universal gas constant ¼ 8.314 kJ kmol1 K1 Pp - Partial pressures of species, p ¼ H2, O2 and H2O (Pa). The water electrolysis process requires both, electrical and thermal energy for the production of hydrogen. The derived Nernst equation (3) accounts for the thermal energy required for the electrochemical reaction by the entropy term (TDS). Since no external heat source is present, this thermal energy is supplied by electrical energy. Therefore, a much higher DC voltage is required to initiate the reaction [19] and this voltage is determined by the enthalpy (DH) of the reaction and is called the thermoneutral voltage, Vth ¼ DH/n. F ¼ 1.481 V.
Activation losses Activation losses at the electrode surfaces are caused due to slow reaction kinetics between the electrode and the electrolyte for the oxygen and hydrogen evolution reactions at the anode and cathode electrodes respectively. These irreversible processes result in the anode and cathode activation overpotentials which combine together to form the total activation losses of the PEM electrolyzer cell. It is found in many
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
international journal of hydrogen energy xxx (xxxx) xxx
published literature that activation losses on the anode side are more dominant especially at lower current densities [19,21]. The anode and cathode activation losses are determined by the following equation (4) [17], based on the ButlereVolmer equation: R:TEzer R:TEzer 1 jEzer sinh þ hact ¼hact; ao þhact; ca ¼ F 2:Jao F j Ezer 1 sinh 2:Jca
(4)
where, jEzer - Current density (A/m2) Eact; ao;ca ref - Exchange current density (A/m2) Jao;ca ¼ ðJao;ca Þ:exp R:T Ezer ref Jao;ca - Pre-exponential factor Eact;ao;ca - Activation energy, (J/mol)
Ohmic losses The ohmic losses are due to resistance of the electrons and protons flowing through the cell components and membrane respectively. Based on Ohm’s law, the ionic resistance is proportional to the current flowing through the cell. Ionic resistance is a function of the dryness, thickness and temperature of the membrane [17,21]. The ohmic losses in the PEM electrolyzer cell, as published in the research papers, are determined by the following equation (5) [17]. hohmic ¼ jEzer
Tmem smem
(5)
where, Tmem e Membrane thickness, (m) smem e Membrane conductivity, (S/m) 1
1 TEzer smem ¼ ð0:5139:l 0:326Þ:exp 1268: 303 l e Nafion Membrane hydration factor (value ranges from 14 to 25).
Table 1 e Parameters employed in the PEM electrolyzer cell model. Parameters Cell Temperature, TEzer Cell Pressure, PEzer Standard Cell Potential,Eo Entropy of Water, △S No. of Electrons, n Faraday’s Constant, F Universal Gas Constant, R Reference Temperature, Tref Current Density (CD), jEzer Anode Exchange CD, Jao Cathode Exchange CD, Jca Anode Activation Energy, Eact; ao Cathode Activation Energy, Eact; ca ref Pre-exponential Factors, Jao;ca Membrane Thickness, Tmem Membrane Hydration Factor, l Charge Transfer Coefficient, a Limiting Current Density, jL
Units
References
330.15 30.00 1.229 163.30 2 96,485 8.314 298.15 Variable 0.0167 0.654 76x103
Values
K bar V Jmol1K1 # C mol1 Jmol1K1 K A/m2 A/m2 A/m2 Jmol1
Measured Measured [18e20] [19,20] [18e20] [18e20] [19,20] [19,20] Measured [17] [17] [17]
18x103
Jmol1
[17]
Calculated 50x106 20 2 20000
A/m2 m # # A/m2
[17] [19] [19] [22] [22]
5
Concentration losses The concentration losses occur mainly at high current densities when Oxygen bubbles at the anode block the contact between the electrode and the reactant (H2O) which increases the overpotential exponentially due to high local current density [17,22]. The concentration losses are determined by the following equation (6) [13,22]. hcon ¼
RTEzer jEzer ln 1 fnF jL
(6)
where, a e Charge transfer coefficient, jL e Limiting current density (A/m2)
PEM electrolyzer cell model validation The PEM electrolyzer cell model encompasses equations (1 e 6) and the model uses these equations to calculate the cell voltage as a function of the operating parameters of the electrolyzer cell, anode and cathode electrodes and membrane properties. The PEM electrolyzer model was programmed in Matlab®Simulink® software and model parameters of all the six equations were varied to predict a polarization (VI) curve which best matches a measured Proton C30 electrolyzer VI curve. Table 1 shows measured data and data found in the literature, for an electrolyzer with Nafion polymer membrane, which were used for the model validation. These parameters in Table 1 were inserted into the model to generate a VoltageCurrent (VI) curve for the PEM electrolyzer cell. Some of the important parameter values were varied to study the effect on the performance of the cell as compared to the measured data. The parameters values which generated the very similar VI curve to the measured VI curve were chosen to generate the final validated curve. The results in Fig. 3A-F show that at high current (200 A) the model predicts the performance of the cell more accurately than at low current. This was further investigated by changing the parameters to beyond its norm values, but the model data could not converge to measured data accurately at low current. One reason for this could be the quality of the data collected from the Proton C30 electrolyzer stack at low current, which show a larger error due to system interference. More accurate data, measured as closely to the stacks as possible via an external current sensor would be required for the validation process at low current. However, the model is adequate to predict the performance of a Proton C30 single cell as it mainly operates in the high current region (200e400 A). The best typical parameter values selected for the model are shown in Table 1 and were employed in the cell model to generate a VI curve similar to those shown in Fig. 2. This cell polarization characteristic curve was used to estimate the performance of the stacks, the total hydrogen production and the overall electrolyzer system power consumption in the HES Model.
PEM electrolyzer system model A PEM electrolyzer system model was developed around the single cell model to integrate the control and dynamic characteristics of the Proton C30 electrolyzer system during the
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
6
international journal of hydrogen energy xxx (xxxx) xxx
Fig. 3 e Comparison of Simulated VI Curves for varying parameters with measured data.
startup (powering up), safety check, venting and normal operational processes. The temperature, current and pressure controllers and time constants were designed and tuned to simulate the dynamic ramping characteristics of the electrolyzer during the production of hydrogen. Fig. 4 shows the PEM electrolyzer system model implemented in the Matlab®Simulink® environment. The electrolyzer model estimates and controls the stack current and temperature for the production of pressurized hydrogen gas, which impacts the pressure within the system, and the product pressure in the buffer tank.
In the PEM Electrolyzer system model, it is assumed that all the cells in each of the three stacks have uniform temperature, pressure, current and reactant distribution to perform according to the characteristic polarization curve. For a given supply current, the voltage of each cell is determined and multiplied by the total number of the cells to evaluate the overall DC power consumed by the three stacks. The pressurized wet hydrogen produced at the cathode side builds up the system pressure in the pipes and small volumes inside the electrolyzer unit and is then fed into a pressure swing dryer unit. The dry hydrogen is then supplied as product gas to the buffer tank.
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
international journal of hydrogen energy xxx (xxxx) xxx
7
Fig. 4 e PEM electrolyzer system model in the Matlab®-Simulink®Environment.
Fig. 5 e Response of the electrolyzer system when operated at different MFC demands.
The system pressure sensor is used to control the electrolyzer current. A mechanical back pressure regulator (BPR) aims to maintain constant system pressure. In the actual Proton C30 Electrolyzer, the BPR is set to a pressure of 29.2 bar while the maximum system pressure is defined as to 31.5 bar. When the buffer tank gets full, its pressure and the system pressure increase and the BPR opens in order to maintain constant system pressure, until it is fully open. If the pressure increases further, the system pressure transducer governs the electrolyzer controller to slowly ramp down the current. If this is not enough to reduce the pressure below a safe level it quickly
drops the current. The control, the buildup of the system pressure and the operational logic for BPR were all modelled in H2 generation and system pressure block shown in Fig. 3. Similar to the Proton C30 electrolyzer model, a mass flow controller (MFC) model was developed, calibrated and inserted into the H2 generation and system pressure block to operate the electrolyzer at different flow rates (or currents) between 0 and 100%. This gives the electrolyzer system the flexibility to operate in the normal mode at 100% capacity or in grid management mode (90% nominal ± 10% dynamic range). Surplus power from renewable sources such as solar or wind can be
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
8
international journal of hydrogen energy xxx (xxxx) xxx
Fig. 6 e Comparison of the simulated step-down test results with the measured results obtain from the Proton C30 Electrolyzer.
utilized to generate hydrogen which can be stored in the fueling station’s storage tanks for fueling fuel cell hybrid vehicles. Fig. 5 shows the response of the electrolyzer when operated at different MFC demands. PID controller and time constants were implemented in the PEM electrolyzer system model. These were tuned to have very similar initial, transient and steady state operational characteristics as the Proton C30 electrolyzer system during startup and at normal operation. The PID controller and the time constants were tuned using measured data acquired from the electrolyzer’s ramping up and ramping down cycles during the commissioning tests at Powertech. The time delays observed during these ramping tests were also included in the control system of the electrolyzer. The same step-up and stepdown cycles performed on the Proton C30 electrolyzer system during the commissioning tests were repeated by the PEM electrolyzer system simulation in order to tune and validate the control system parameters. Fig. 6 show the simulated
result of the step-down tests compared to the measured results obtained under the same tests profiles respectively. The simulation results show very similar current step-up and step-down characteristics and delay times compared to those seen in the real electrolyzer under the same step-up and step-down tests. However, the simulation does not show three (Fig. 6) pronounced points of support as it ramps down. Further investigations are required to fully understand this behavior of the controller when ramping the electrolyzer demand up or down quickly.
Buffer tank and compressor models The buffer tank block in Fig. 3 has the buffer tank model to estimate the pressure of the tank from the flow rate of the dry hydrogen supplied from the electrolyzer unit. The pressure builds up in the tank until the upper limit is reached after which either the compressor is switched on to consume the
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
international journal of hydrogen energy xxx (xxxx) xxx
9
Fig. 7 e Simulated results for different operating states compared with measured data from the Proton C30 Electrolyzer Unit.
Fig. 8 e Island of Hawaii grid model in the Matlab®-Simulink®Environment.
stored hydrogen or hydrogen production is stopped. The hydrogen compressor is modelled by a rate algorithm which allows the compressor to switches ON and OFF when the buffer tank is full and empty respectively. The compressor switches ON when the buffer pressure is greater than 29.5 bar and runs until the buffer tank pressure drops below 16.8 bar. These upper and lower buffer tank pressure limits results in a hydrogen compressing time of about 20 min and a buffer tank filling time of approximately 13 min, as measured in the actual HES system.
PEM electrolyzer and HES system model validations The PEM Electrolyzer System Model was validated with measured data obtained during performance testing at Powertech in March 2015. The electrolyzer system model was setup to run in same way as the actual Proton C30 electrolyzer unit from the “Power-up and Prestart State” where all the safety checks are done to “Generate and Venting State” where the stacks’ operation and hydrogen quality are verified to “Normal Pressurized Operation State” for generating the
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
10
international journal of hydrogen energy xxx (xxxx) xxx
hydrogen as product gas. Fig. 7 shows the simulated results compared to the measured data for the H2 flow rate, system and product pressure, and one stack current during the different operational states. This simulated results show that the control parameters and the time constants in the model are well tuned to accurately estimate the operational characteristics of the actual Proton C30 electrolyzer system.
Hawaii Island electrical grid model To analyze the effects of the hydrogen electrolyzer system on the electrical grid, models of the Island of Hawaii electrical grid and an electrolyzer grid interface were developed. The conventional model of grid frequency dynamics based on the swing equation [7,23] was modified to develop the Island of Hawaii grid model shown in Fig. 8. The grid model includes a system inertia algorithm which characterizes the rotational inertia of the generators and the frequency dependent loads while the “governor” transfer function with time delay represents all the turbine generators in the Island of Hawaii power plant. The model parameters of the system inertia (M, D) and governor (Kp, Ki, time constants) algorithm specifically for the Island of Hawaii grid were obtain from Hawaii Natural Energy Institute (HNEI) study of a grid-scale Battery Energy Storage System (BESS) on the Island of Hawaii [7]. These algorithms were developed and validated during this study utilizing a fast-acting 1 MW lithium-ion (Li-ion) BESS on the Hawaii Island grid. The Island of Hawaii grid model determines the grid frequency depending on the total number of loads on the grid (i.e. daily residential, industrial and/or electrolyzer demand) compared to the total number of power generators connected on the grid (i.e. fossil fuel turbines, BESS, and/or renewable power sources). An electrolyzer grid controller algorithm was developed to ramp the electrolyzer system between 80% and 100% of its nominal hydrogen production capacity as a function of grid frequency. The electrolyzer normally operates at 90% of its production rate to generate hydrogen for the fueling stations, while consuming power from the grid. The electrolyzer grid controller, when activated, allows the electrolyzer to ramp up from 90% to 100% when grid frequency is greater than 60 Hz and ramp down to 80% when the frequency is less than 60 Hz. Fig. 3 shows the electrolyzer grid controller implemented in
Table 2 e Parameters employed in the grid management simulation tool. Parameters
Values
Simulation Setup Parameters Simulation Time Variable Simulation Sample Time 0.01 Island of Hawaii Grid Model Parameters M 1.5 D 26.7 3.3 Kp 0.72 Ki Time Delay, Tg 0.4 Nominal Grid Frequency, fo 60
Units
References
s s
N/A N/A
MW/Hz/s MW/Hz MW/Hz MW/s/Hz s Hz
[7] [7] [7] [7] [7] [7]
the PEM Electrolyzer System block in the Matlab®-Simulink® Environment.
Analyses of HES as a grid management tool A grid management simulation tool was developed in the Matlab®-Simulink® environment to analyses HES as a potential grid contrivance for a smaller electrical grid like the one of the Island of Hawaii. The grid management simulation tool consists of a PEM electrolyzer system model (Fig. 4), Island of Hawaii electrical grid model (Fig. 8) and source of power, load and grid frequency input signals to drive the simulation. For most of the analyses the setup parameters shown in Table 2 were used by the simulation tool. For all grid management analysis with the Island of Hawaii grid model, it is assumed that all other loads and electrical power sources remain constant beside those which are varied in the investigation. The regulating frequency, ƒc, of the electrolyzer grid controller (Fig. 3) determines how sensitive the grid controller ramps the electrolyzer up or down. When ƒc is set to a window value of [60.00, 60.00] Hz, then the smallest deviation between grid frequency (ƒgrid) and regulating (ƒc) frequency, ƒgrid-ƒc, would cause the electrolyzer to ramp up or down. In contrast, when the ƒc is set to a wide window of [59.90, 60.10], (i.e. ±0.1 Hz of nominal grid frequency, ƒo) then the electrolyzer ramping up and down would be very minimal since the deviation of the grid frequency needs to exceed ±0.1 Hz. Fig. 9 shows the response of the electrolyzer with different sets of grid controller regulating frequencies, ƒc. These results demonstrate the sensitive of the controller as mentioned above, from most cyclic operation at ƒc ¼ [60.00, 60.00] Hz to constant operation at ƒc ¼ [59.90, 60.10] Hz. Hence, for a particular grid system, an optimal window value of the ƒc has to be determined to give a smooth and efficient grid management operation by the electrolyzer.
Grid management of the HES with measured grid frequency profile In this study, the functionality of the electrolyzer grid controller (Fig. 3) algorithm was tested with a grid frequency profile measured on the Island of Hawaii during the BESS investigation [7]. The deviation of the grid frequency from the nominal grid frequency of 60 Hz was used to regulate the electrolyzer between 80% and 100% of its production rate for the grid management application. During the grid management operation, after the startup period of 200 s, the electrolyzer operates mostly at 90% and it then regulates between 80 and 100% depending on the grid condition. When surplus power is available on the grid (grid frequency > 60.00 Hz), e.g. from the renewable energy sources, the controller ramps up the hydrogen production rate towards 100% and when there is an increase in power demand (grid frequency < 60.00 Hz), the controller ramps down the hydrogen production rate towards 80%, thereby reducing the electrolyzer power demand on the grid. For this analysis, the window value of ƒc for the electrolyzer grid controller was selected to be [60.00, 60.00] Hz. Fig. 9 (top
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
international journal of hydrogen energy xxx (xxxx) xxx
11
Fig. 9 e Results of sensitivity study of the grid controller regulating frequency (ƒc).
Fig. 10 e Effects of increasing solar power on the grid frequency and on the C30 Proton Electrolyzer performance.
left) shows the performance of the Proton C30 electrolyzer system under the measured Island of Hawaii grid frequency variability. Due to inherent time constants and time delays of the electrolyzer system (section PEM Electrolyzer System Model), the electrolyzer does not respond to faster transients seen in the measured grid frequency plot and therefor the ramping up or down of the electrolyzer is much smoother compared to the very cyclic nature of the grid frequency.
Grid management of the electrolyzer system with measured solar power profile A solar power profile measured on the Island of Oahu in the month of July with average and maximum power of 10 kW and 23 kW respectively was used to investigate the influence of a solar power source on grid frequency and the grid management performance of the electrolyzer.
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
12
international journal of hydrogen energy xxx (xxxx) xxx
Fig. 11 e Performance of different sizes of electrolyzer systems and grid frequency with measured wind power profile.
The Island of Hawaii grid model was used to study the impact of the solar power on the grid frequency. The impact of the solar power (kW) feeding into a MW grid system was minimal and the frequency deviation was in the mHz range which did not trigger any ramping of the electrolyzer with ƒc set to [59.999, 60.001], thus the electrolyzer mostly operated at a 90% production rate. To avoid continuous ramping of the electrolyzer caused by mHz frequency deviation, the regulating frequency window, ƒc, of [59.999, 60.001] was
select, limiting ramping of the electrolyzer to frequency deviation greater than ±1 m Hz. However, if the solar plant was scaled up and the power increased to about a megawatt, then the impact of the solar power on the grid frequency and the operation of the electrolyzer is much more prominent. The solar power was increased to 1.15 MW (i.e. 50 times the measured profile) to study the effect of solar power on the performance of the electrolyzer as a grid management tool. Fig. 10 shows the effect of increasing
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
international journal of hydrogen energy xxx (xxxx) xxx
solar power to 1.15 MW on the grid frequency and performance of the electrolyzer. The results of this study show, as expected, that the large renewable solar plant with its surplus power had a greater influence on the grid frequency and therefore, on the grid management by the C30 Proton electrolyzer. The surplus power from a 1.15 MW solar plant caused the electrolyzer to operate above 90% capacity for at least 48% of the time.
a significant reduction in grid frequency variability observed especially when comparing the grid frequency fluctuations of a 9.0 MW system with that of just the wind power without an electrolyzer. Furthermore, we observe that the electrolyzer system with 3.1 MW nominal power consumption which is closer to the maximum power provided by the renewable wind power, achieves a reasonable reduction in grid frequency variability. The results also show that during the most significant change in wind power around 1 h 50 min mark (circled), the grid controller does ramp down the electrolyzer power consumption as the wind power drops suddenly. However, at the same time mark, there is much less change in the grid frequency for electrolyzer size 2.7 MW compared to when the electrolyzer was not operating (0.0 MW), where much more downward (<60 Hz) fluctuations on the grid frequency are observed. By contrast, in the 9.0 MW electrolyzer, the grid frequency drop is very minimal compared to the 2.7 MW or 3.1 MW system at the same time mark. This implies that there is a certain optimal electrolyzer size for a particular wind power profile where the grid stabilization by an electrolyzer system has maximum effect. To quantify the variability of the grid frequency due to the operation of the electrolyzer system on the Big Island grid system, a Variability Factor is defined to measure the reduction of the grid frequency variability caused by the electrolyzer system. The Variability Factor accounts for improvement in grid frequency as a percentage of total number of frequency data points. However, only the improvement in grid frequency (ƒnew -ƒoriginal) which is greater than 1 Millihertz was considered for frequency > 60 Hz (i.e. negative change e.g. 60e60.001 ¼ 0.001) and frequency <60 Hz (i.e. positive change e.g. 60e59.991 ¼ 0.001). The equation below shows how the Variability Factor is calculated:
Grid management of the electrolyzer system with measured wind power profile A wind power profile measured on the Island of Hawaii with average and maximum power of about 1.6 MW and 3 MW respectively was used to investigate the influence of wind power on grid frequency and grid management by the electrolyzer [7]. For this investigation, it is assumed that all other loads and power sources on the Island of Hawaii grid model remain constant beside those which are varied in the investigation. A two-hour long measured wind power profile was used by the HES simulation tool to investigate the impact of the wind power on the grid frequency and the potential of grid management by the electrolyzer system. For this study, the regulating frequency, ƒc was set to [59.999, 60.001]. Fig. 10 (0.18 MW electrolyzer plot) shows the grid frequency fluctuations and electrolyzer performance under the wind power profile. The grid frequency plot shows constant deviations from nominal frequency, ƒ0, of 60.00 Hz due to the unsteadiness of the wind power. This resulted in the electrolyzer operating at 68% of time above 0.16 MW (>90% power consumption) using the excess wind power. Compared to a 0.18 MW electrolyzer system, the wind power profile provides 15 times more power to the Island
2 6 6 6 Variability Factorð%Þ ¼ 6 6 4
P P
! fnew foriginal < 0:001 Hz
ðNo: of frequency data pointsÞ þ ! foriginal > 0:001 Hz
P fnew P ðNo: of frequency data pointsÞ
13
! 100
3
7 7 7 7 7 ! 5
100
of Hawaii grid system which was good for production of hydrogen under the grid management mode of operation. However, what would be the impact of a megawatt (MW) electrolyzer system on the grid frequency stabilization and performance as a grid management tool when operated with the Island of Hawaii megawatt wind power profile. The two-hour wind power profile with an average power of 2 MW was used again to study the effect of scaling up the electrolyzer system in small increments from its original size of 0.18 MWe9 MW, which is 50 times the size of the original Proton C30 electrolyzer unit. Fig. 11 shows the impact of selected sizes of the electrolyzer system on the grid frequency under the wind power profile. These results show that there is
This calculation was applied to the results of the wind power investigation with different sizes of electrolyzers to determine if there was any trend and an optimal size of the electrolyzer to have maximum impact on the grid frequency. Fig. 12 shows the results of variability factors as function of different sizes of electrolyzers. The figure also shows the average and maximum power of the wind power profile used in this investigation. This result shows a clear trend. As electrolyzer size increases, the reduction in grid frequency variability factor also increases up to a maximum value of 51%, thereafter the increase in the electrolyzer sizes has minimal effect. The most interesting conclusion from this result was that the optimal size of the electrolyzer (3.1 MW) to achieve the large change in
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
14
international journal of hydrogen energy xxx (xxxx) xxx
electrolyzer power demand on the grid is much lower than the larger system. For grid frequency stabilization, the electrolyzer system operating within wider power limits of 50%e75%-100% would be very valuable especially with a smaller size of electrolyzer systems, but for production of hydrogen for the fueling station this limit would be very uneconomical. With these operational limits (50%-75%-100%) there was approximately 22% reduction in hydrogen production over the two hour period. With this decrease in hydrogen production and the lower nominal point of operation of 75%, when renewable power is not available, there could be insufficient hydrogen to meet the daily fuel demand of the fueling station and this would cause deficit in revenue to cover the daily operational expenses of the fueling station. Hence, there is always a fine economical tradeoff between operating a hydrogen energy system as a pure grid management scheme or as a limitless hydrogen production system.
Conclusions
Fig. 12 e Effect of electrolyzer size on grid frequency variability.
grid frequency corresponded to the maximum wind power. This information is critical when designing a hydrogen energy system for grid management application in order to minimize the cost of the system and maximize its use as a renewable power grid frequency stabilizer. The hydrogen production of the 3.1 MW electrolyzer while operating in the grid management mode was also estimated over the two hour wind power profile and was compared to the electrolyzer when operating in normal mode at maximum capacity. The simulation results show that there is a 6.5% decrease in total hydrogen produced during the two hours when the electrolyzer is operated in grid management mode. However, this reduction in hydrogen production for the fueling station may be acceptable compared to the electricity cost the station would have to incur when operating the electrolyzer at maximum capacity on a non-renewable power source. The effect of changing the power levels for grid management of electrolyzers from 80% to 90%-100% to 50%-75%100% was also investigated for the 0.18 MW and 3.1 MW systems. When comparing these results with the 80%-90%100% level results, for the smaller Proton C30 electrolyzer system, the variability factor increases from 1.1% to 10%, but for the bigger system, (3.1 MW), the variability factor decreases from 51% to 49%. This indicates that when the electrolyzer is allowed to cycle within a wider power range, more renewable power is consumed at lower power limits (50%e75%) than at higher power limits of 80e90%. Hence, for the 0.18 MW electrolyzer there is an increase of nearly 9% in grid management performance since the impact of
Energy storage, such the hydrogen energy system, is on the verge of becoming an important building block of the electrical infrastructure of the future. The potential of a hydrogen energy system as a grid management device has been demonstrated by the simulation analysis under realistic renewable power profiles. Although the validated model of the electrolyzer and the HES are simple, they are well suited to make reasonable predictions of the system’s performance under a variety of operating conditions and control strategies in minimizing the grid frequency variability. These simulation results with a validated Island of Hawaii grid model and realistic renewable power profiles give insight into critical information when designing a hydrogen energy system for grid management applications in order to minimize the cost of the system and maximize its use as a renewable power grid frequency stabilizer. For wind power analysis results show that the optimal size of the electrolyzer, required to achieve maximum effect in grid frequency stabilization, corresponded to the maximum wind power of 3.1 MW. However, it was assumed that all other loads and power sources on the Island of Hawaii grid model remain constant beside those which are varied in the investigation. The simulation results also show that there was a 6.5% decrease in total hydrogen produced when the electrolyzer was operated in grid management mode over the two-hour wind power profile. Nevertheless, there is always a fine economical tradeoff between operating a hydrogen energy system as a pure grid management scheme or as a limitless hydrogen production system. Future work would involve improvement of the models with high quality measured data of the electrolyzer stacks and fine-tuning of the control algorithms of the hydrogen energy system for robust pressure and current regulation. The development of enhanced grid controller algorithms which would allow other storage systems such as BESS to be included in stabilization of the grid frequency will also be considered. These algorithms will have the capability of optimizing the balance between the longevity of the stacks
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070
international journal of hydrogen energy xxx (xxxx) xxx
and secured hydrogen production on the one side and reducing battery size on the other hand. The durability of the stacks is highly depended on its robustness to ramping and the impact of the individual parameters (cycle frequency, cycle amplitude, ramp rate, peak current, etc.) will have on the life-time of the stacks will also be investigated in depth. Furthermore, future gird management analyses will also include the effects of other fluctuating loads and power sources (both conventional and renewable) on the grid.
Acknowledgments In addition to original funding from the US DOE Fuel Cell Technology Office, other funding for this project was provided by the Office of Naval Research and the State of Hawaii. NELHA provided funding for the development of the site infrastructure design and engineering, and is taking on the task of managing the installation of the site improvements. US Hybrid and the Hawaii Center for Advanced Transportation Technologies (HCATT) provided cost sharing for the conversion of the County of Hawaii Mass Transit Agency bus by US Hybrid.
references
[1] Grimley M, Farrell J. Report, Institute for Local Self-Reliance (ILSR). 2015. https://ilsr.org/report-renewable-hawaii/. [2] NREL news report Part 1. April 2018. https://www.nrel.gov/ news/features/. [3] NREL news report Part 2. April 2018. https://www.nrel.gov/ news/features/.
15
[4] Hawaii legislative house bill 623. 2015. https://www.capitol. hawaii.gov/. [5] Eichman J, Harrison, and K, Peters M. National renewable energy laboratory, Technical Report. 2014. p. 1e35. NREL/TP5400-61758. [6] Ewan M, et al. ECS Transactions 2016;75(14):403e19. [7] Hawaii Natural Energy Institute. Technical report for DOE award DE-EE0003507, 1-65. https://www.hnei.hawaii.edu/ publications/; 2014. [8] Hawaii Natural Energy Institute. Projects. 2010e2015. https:// www.hnei.hawaii.edu/projects/havo-fuel-cell-buses. [9] Millet P, et al. Journal of Hydrogen Energy 2010;35:5043e52. [10] Ayers KE, et al. ECS Transactions 2010;33(1):3e15. [11] Sun S, et al. J Power Sources 2014;267:515e20. [12] Kokoh KB, et al. Int J Hydrogen Energy 2014;39:1924e31. [13] Moore R, et al. J Power Sources 2006;162(1):302e8. [14] Moore R, et al. Fuel cell seminar. 2007. https://www. fuelcellseminar.com/past/. [15] Mohanpurkar M, et al. Energies 2017;10(1836). http://www. mdpi.com/journal/energies. [16] Proton on site, technical specifications. https://www. protononsite.com/products-proton-site/c10-c20-c30/; 2010e2019. [17] Ni M, et al. World Hydrogen Energy Conference (WHEC) 2006;16:13e6. [18] Zhang H, et al. International Journal of Electrochemical Society 2012;7:4143e57. [19] Bessarabov D, et al. Handbook. CRC Press, Taylor & Francis Group; 2016. p. 1e392. [20] Atkins PW. Physical chemistry handbook. 3rd ed. Oxford University Press; 1986. p. 813. 259-282. [21] Garcia-Valverde R, et al. Int J Hydrogen Energy 2012;37:1927e38. [22] Al Ashkar H, et al. Hydrogen power theoretical and engineering solutions international symposium, vol. 13; 2015. p. 1e9. [23] Pena R, et al. Appl Sci 2018;8(12):2413. https://www.mdpi. com/2076-3417/8/12/2413/.
Please cite this article as: Virji M et al., Analyses of hydrogen energy system as a grid management tool for the Hawaiian Isles, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2020.01.070