An advanced real time energy management system for microgrids

Energy 114 (2016) 742e752

Contents lists available at ScienceDirect

Energy journal homepage: www.elsevier.com/locate/energy

An advanced real time energy management system for microgrids Moataz Elsied a, c, Amrane Oukaour a, *, Tarek Youssef b, Hamid Gualous a, Osama Mohammed b a

University of Caen Basse Normandie, LUSAC Laboratory, Caen, France Electrical and Computer Engineering, Florida International University FIU, USA c Faculty of Engineering, Helwan University, Egypt b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 11 July 2015 Received in revised form 1 August 2016 Accepted 15 August 2016

This paper presents an advanced Real-Time Energy Management System (RT-EMS) for Microgrid (MG) systems. The proposed strategy of RT-EMS capitalizes on the power of Genetic Algorithms (GAs) to minimize the energy cost and carbon dioxide emissions while maximizing the power of the available renewable energy resources. MATLAB-dSPACE Real-Time Interface Libraries (MLIB/MTRACE) are used as new tools to run the optimization code in Real-Time Operation (RTO). The communication system is developed based on ZigBee communication network which is designed to work in harsh radio environment where the control system is developed based on Advanced Lead-Lag Compensator (ALLC) which its parameters are tuned online to achieve fast convergence and good tracking response. The proposed RT-EMS along with its control and communication systems is experimentally tested to validate the results obtained from the optimization algorithm in a real MG testbed. The simulation and experimental results using real-world data highlight the effectiveness of the proposed RT-EMS for MGs applications. © 2016 Elsevier Ltd. All rights reserved.

Index terms: Distributed Energy Resources (DERs) MicroGrid (MG) Energy Management System (EMS) Genetic Algorithm (GA)

1. Introduction The concept of MGs has received an increasing interest from the scientific community. This is mainly due to their flexible and intelligent characteristics bringing a significant potential to promote and integrate renewable energy resources. Moreover, MGs are regarded as a way to improve overall system reliability, efficiency, and resiliency. They can be used as an autonomous power source (islanding mode) in parallel with a main grid (grid connected mode), as well as during transition to islanding mode (on/off grid mode) [1e3]. The power electronics interface circuits (PEICs) bridge different MG components making them into two main configurations: AC and DC microgrids [4,5]. AC microgrids (ACMGs) are gaining in popularity due to their variety of applications. On the other hand, DC microgrids (DCMGs) provide DC power and are usually connected to a DC bus line [6]. In this case, an inverter (DC/AC) and a rectifier (AC/DC) are both required for bidirectional connection with ACMGs.

* Corresponding author. University of Caen Basse Normandie, LUSAC laboratory, BP78 Rue Louis Aragon, 50130 Cherbourg, France. E-mail address: [email protected] (A. Oukaour). http://dx.doi.org/10.1016/j.energy.2016.08.048 0360-5442/© 2016 Elsevier Ltd. All rights reserved.

ACMGs and DCMGs along with their respective PEICs still face numerous challenges such as efficient energy control and management [7e11]. Therefore, a Microgrid Energy Management System (MG-EMS) is required to identify operation costs and emission limits while taking into account the consumer power demand for each Distributed Energy Resource (DER) and Energy storage system (ESS) unit. As such, efficient algorithms are developed to optimize the use of individual DERs by minimizing an objective function while considering the system constraints such as load power balance, fuel cost, performance, specifications, limitations due to safety, fuel supply limitation, restrictions on noise/pollutant emissions, and so on [12e17]. Various optimization algorithms have been considered for this problem. In Refs. [18], the evolutionary optimization algorithm is proposed to minimize the sum of the total capital, operational and maintenance cost of DERs subject to constraints such as energy and emission limits of each DER and Loss of Power Supply Probability (LPSP) of the microgrid. Ant Colony Optimization (ACO) is used in Ref. [19] to solve the economic and environmental dispatch of MGs containing different types of generation systems. Other than that, an advanced EMS in a typical MG working in grid and island operating mode is introduced in Ref. [20] based on Advanced Integrated Multidimensional Modeling Software (AIMMS) to

M. Elsied et al. / Energy 114 (2016) 742e752

determine the optimal operating strategies to minimize the energy costs and pollutant emissions hence maximizing the output of the available renewable energy resources. However, these methods are often used offline, which restrict their use for real-time applications. Real-Time Energy Management Systems (RT-EMSs) is described as one which controls an environment by receiving data, processing them, and returning the results sufficiently quickly to affect the environment at that time [21,22]. To improve the performance of RT-EMSs, a robust and reliable communication network is required to exchange the information and decision commands between the EMS, DERs, ESSs, local controllers, and PEIC in real time. The communication network should be flexible and expandable to provide a link with all the nodes within the MG location. For customer privacy and system security, a strong encryption algorithm should be used to secure the communication between different nodes and MG-EMS [23,24]. An advanced RE-EMS is proposed for MG application based on wired communication network in Ref. [25]. Wired communication systems do not have interference problems and their functions are not dependent on batteries, as wireless solutions often do. On the other hand, wireless communications have some advantages over wired technologies, such as low-cost infrastructure and ease of connection to difficult or unreachable areas. However, the nature of the transmission path may cause the signal to attenuate. Recently, different wireless communication technologies are available and utilized for smart MGs to meet certain requirements such as low cost, simple deployment, working in noisy and harsh environment and support large number of nodes to accommodate for all smart appliances, distributed energy resources and controllers [26]. Among these technologies, ZigBee communication is widely accepted as the most suitable standards for MGs residential

743

network domain by the U.S. National Institute for Standards and Technology (NIST) [27]. ZigBee is an ultra-low power wireless networking technology which makes it suitable to embed in wide range of devices and application. ZigBee is chosen as a good option for MGs implementations according to its simplicity, mobility, robustness, low bandwidth requirements, low cost of deployment and easy network implementation. In addition, it has the ability to operate within an unlicensed spectrum [28]. In this regard, this paper presents an advanced RT-EMS suitable for MGs applications. The optimization problem is solved using GA to satisfy the load demand requirements during 24 h operation with the lowest utility cost by finding an hourly optimal allocation for each DER unit. In order to improve the performance of RT-EMS, the communication network is designed based on ZigBee technology which has the ability to work in harsh radio environment. The proposed RT-EMS along with its communication system is experimentally tested to validate results obtained from the optimization algorithm on a real-time MG testbed. The simulation and experimental results highlight the effectiveness of the proposed RT-EMS for MGs. The paper is organized as follows: firstly, the MG system description is illustrated in Section 2. In Section 3, the optimization model and its algorithm are introduced and analyzed. In Section 4, the experimental setup using the proposed RT-EMS with its communication system are presented and discussed. Finally, the experimental results are discussed in Section 5. 2. MG system description The schematic diagram of the MG-EMS under consideration is depicted in Fig. 1. There are three DERs: Micro-turbine (MT), wind,

Fig. 1. The schematic diagram of the MGEMS under consideration.

744

M. Elsied et al. / Energy 114 (2016) 742e752

and Fuel cell (FC). The integration of DERs and ESS is interfaced with the load and the main grid by power electronic converters. Using various energy sources, different power converters topologies are necessary for power flow control to fulfil load power demand requirements by receiving control signals from the MG-EMS. The MGEMS based on GA is introduced to provide each power source by its reference value. It is important to note that the GA takes into account the system variables, constraints, parameters, cost and emission objective functions. Moreover, it has decision making capabilities for the operation of the MG in grid-connected or islanding mode. In grid-connected mode, bidirectional power is allowed based on market policies. Also, the MG is able to disconnect itself from the main grid if low power quality events occur on the main grid. All local controllers for FC, wind, ESS, and MT are designed to work in current control mode to control the power flow between the DERs, ESS, and load demand. Unlike these controllers, the local controller of the main utility grid is worked in voltage and current controllers to regulate the DC-bus voltage at a constant DC-voltage. The optimal power for each DER unit delivered by GA is divided by the DC-link voltage then subtracted from the measured output current. The error signal is then fed to the online current controller to generate the required PWM pulses to the boost converters. To exchange the information and decision commands between the EMS, DERs, ESSs, local controllers, and PEIC in real time, the communication system is developed based on ZigBee communication networks. The Zigbee network utilizes an AES 256 symmetric-key encryption algorithm for data encryption which has been approved by national security agency (NSA) [26]. The transparent over air encryption mechanism is chosen to provide the security for wireless signal while require no modifications on the applications and controller level, this is achieved by performing encryption and decryption processes by the communication processor. The transparent encryption is important to provide the security while ensure the interoperability with controllers, resources and application. The symmetric encryption is chosen to provide strong encryption with minimal processing efforts compared to asymmetric encryption, this is important to achieve required data rate without significantly increasing the cost required for higher speed processor. RT-EMS along with its communication networks presented in this paper support more efficient network operation and allow the communication of frequent price updates. In addition, it allows a more dynamic, reactive pricing mechanism required to take into account real-time availability of fluctuating RES and follow the evolution of balance between supply and demand in real time. 3. MG-EMS methodology In this section, an optimization model of the MG-EMS is presented. The decision variables include electricity generation of DERs and the main utility grid to allocate each source by its optimal power generation set point. As such, each power source contributes to the load by its optimized rate provided by the MG-EMS.

Additionally, energy sources can also be put in ON or OFF state to reach optimal operation of the overall system while satisfying its constraints. The mathematical objective functions are described as follows. 3.1. Proposed objective fitness function The choice of the fitness function is a key for optimal operation. With the fitness function proposed in Refs. [29,30], the grid utility, start-up and operating costs of each DER are considered. Since the aim of the fitness function is to fulfill load demand requirements for a given day in an economical manner, a modified fitness function is expressed as follows:

Min f1 ðxÞ ¼ Min

T X

cost t

t¼1

80 ug T < X X @ ¼ Min DDERi ðtÞPDERi ðtÞCDERi ðtÞ : t¼1 i¼1 0 us X  þ DDERi ðtÞSUCDERi ðtÞÞ þ @ DESSj ðtÞPESSj ðtÞCESSj ðtÞ j¼1

19 = þ ðPgr ðtÞCgr ðtÞ  Pgs ðtÞCgs ðtÞA ; (1) where CDERi(t),CESSj(t) are the costs of output active power of the ith DERs and jth ESS at hour t. Cgr(t),Cgs(t) are the costs of receiving (buying)/sending (selling) active power from/to the main grid at hour t. Pgr(t),Pgs(t) are the receiving/sending active power from/to main grid at hour t, respectively. In the proposed cost function, x (1, 2  u*T) is considered as the decision variable vector with one row and columns (2  u*T), consisting of the output power from all DER units and ESS, the amount of exchange power with main utility grid, and on/off mode in a vision planned for the day ahead. This vector can be expressed as follows:

x ¼ ½PG ; DG ð1;

2*u*TÞ

(2)

PG is the output active power from all DERs and ESS units where DG is the state vector denoting the ON or OFF states for all DERs and ESS units during each hour of the day. T represents the total number of time intervals which is equal to 24 h for one day of operation. The total number of variables (u) can be expressed using the following equation:

u ¼ ug þ us þ 1

(3)

where ug, us are the total number of all DERs and ESS units and the extra number 1 indicates the value of the market production. Each variable in Eq. (2) can be defined as follows:

PG ¼ ½PDER i h ; PESS  PDER ¼ PDER1 ; PDER2 ; ………………………::; PDERðugþ1Þ PDERi ¼h½PDERi ð1Þ; PDERi ð2Þ; …:PDERi ðtÞ; ……; PDERi ðTÞ; i i ¼ 1; 2; …::; ðug þ 1Þ PESS ¼ PESS1 ; PESS2 ; ……………………………::; PESSðusÞ   PESSj ¼ PESSj ð1Þ; PESSj ð2Þ; …:::PESSj ðtÞ; ……; PESSj ðTÞ ; j ¼ 1; 2; ……; us

(4)

M. Elsied et al. / Energy 114 (2016) 742e752

where PDERi(t),PESSj(t) are the output active power of the ith DERs and jth ESS units at hour t.

745

each year [20]. Without limit emissions of air pollutants, ocean acidification, sea level rise and increased storm surge, harm to agriculture and forests, species extinctions and ecosystem damage.

DG ¼ ½DhDER ; DESS  i DDER ¼ DDER1 ; DDER2 ; ………………:…::::; DDER;ðugþ1Þ DDERi ¼ DDERi ð1Þ; DDERi ð2Þ; …; DDERi ðtÞ…::DDERi ðTÞ; i ¼ 1; 2; ……; ðug þ 1Þ DDERi 2f0; i h 1g DESS ¼ DESS1 ; DESS2 ; ……………………………:; DESSðusÞ   DESSj ¼ DESSj ð1Þ; DESSj ð2Þ; …; DESSj ðtÞ…:…DESSj ðTÞ ; j ¼ 1; 2; ……::…; us DESSi 2f0; 1g

(5)

Such constraints can be written as follows: Where DDERi(t),DESSj(t) are the state vectors which demonstrate the ON or OFF states of all DERs and ESS units at hour t, respectively. The DER startup cost(SUCDERi) depends on the time the unit has been off prior to the new startup. The mathematical expression of this cost can be written as the following equation:

  toff i SUCDERi ðtÞ ¼ bi 1  e t

(6)

where bi is the cold startup cost, t is the unit cooling time constant and toff i is the time a unit has been off prior to the startup for each DER [29]. 3.2. MG system constraints Under the consideration of DERs and ESS characteristics, MGEMS should ensure the steady state security and reserve capacity margin for MG when participating in the power market. Accordingly, the following constraints are defined and applied to the optimization model of MG. 3.2.1. Load power balance This constraint states that the summation of load demand power Pl(t) must be equal to the summation of DERs power PDERi(t), ESS power PESSj(t) and the main grid net power (Pgs(t)Pgr(t)) at hour t. The mathematical expression for such constraint can be written as follows:

PDERi ðtÞ or ECO2

ECO2

DERi

ENOX

PDERi ðtÞ or ENOX

ESO2

PDERi ðtÞ or ESO2

DERi

DERi

l¼1

Pl ðtÞ 

ug X i¼1

PDERi ðtÞ 

us X

PESSj ðtÞ  Pgr ðtÞ þ Pgs ðtÞ ¼ 0

(8)

PESSj ðtÞ or ENOXg ðtÞPg ðtÞ  Emax NOX ðtÞ

(9)

PESSj ðtÞ or ESO2g ðtÞPg ðtÞ  Emax SOX ðtÞ

Pg ðtÞ ¼ Pgs ðtÞ  Pgr ðtÞ Emax CO2 ,

Emax NOX

(10) (11)

Emax SO2

and represent the maximum emission rates of CO2, NOX and SO2 respectively for each unit in kg at time t. ECO2 ; ENOX ; and ESO2 are the emission factors of carbon diDERi DERi DERi oxide, nitrogen oxides and sulfur dioxide emission delivered by each DER in kg/kWh. ECO2 ; ENOX ; and ESO2 are the emission ESSj ESSj ESSj factors of carbon dioxide, nitrogen oxides and sulfur dioxide emissions delivered by each ESS in kg/kWh, and ECO2g ; ENOXg ; and ESO2g are the emission factors of carbon dioxide, nitrogen oxides and sulfur dioxide emission delivered by the utility grid in kg/kWh respectively [16] [17], and [25]. The rest of the above parameters are defined in Section 3.1. The mathematical formulation of the net emission of carbon dioxide, nitrogen oxides, and sulfur dioxide during 24 h of operation are expressed in Eq. (12), Eq. (13), and Eq. (14) respectively.

f2 ðxÞ ¼

(7)

8 ug T
þ

j¼1

:

DDERi ðtÞPDERi ðtÞECO2

us X

DESSj ðtÞPESSj ðtÞECO2

f3 ðxÞ ¼

8 ug T
þ

:

þ

:

DDERi ðtÞPDERi ðtÞENOX

DESSj ðtÞPESSj ðtÞENOX

ESSj

9 = ;

(12)

ðtÞ

ðtÞ þ Pg ðtÞESO2g ðtÞ

i¼1

us X j¼1

DERi

ESSj

8 ug T
DDERi ðtÞPDERi ðtÞESO2

DESSj ðtÞPESSj ðtÞESO2

j¼1

f4 ðxÞ ¼

ðtÞ þ Pg ðtÞECO2g ðtÞ

ESSj

i¼1

us X

ðtÞ

DERi

i¼1

j¼1

where ul is the total number of load levels. 3.2.2. Emission constraint To improve the public image, DERs, ESS and the main grid should work within acceptable emission limits which depend on the emission standards reference guide of each country. This part of the work presents the constraint for the emissions caused by fossilfueled thermal units. Three of the most important emissions are considered in the current work: carbon dixode (CO2), sulfur dioxide (SO2) and nitrogen oxides (NOx). It is assumed that each unit should not exceed the maximum significant emission rate for each gas in kg at hour t. These limits are designed to regulate air emissions produced by MG and utility grid units. The emissions reductions have led to dramatic improvements in the quality of the air, preventing hundreds of thousands of cases of serious health effects

PESSj ðtÞ or ECO2g ðtÞPg ðtÞ  Emax CO2 ðtÞ

ESSj

ESSj

t¼1 ul X

ESSj

DERi

9 = ;

(13)

ðtÞ

ðtÞ þ Pg ðtÞENOXg ðtÞ

9 = ;

(14)

746

M. Elsied et al. / Energy 114 (2016) 742e752

The mathematical description of the carbon dioxide emission factors can be stated as follow:

E CO2DERi ðtÞ ¼ ½E CO2DERi ð1Þ; E CO2DERi ð2Þ; ……::E CO2DERi ðTÞ (15) E NOxDERi ðtÞ ¼ ½E NOxDERi ð1Þ; E NOxDERi ð2Þ; ……E NOxDERi ðTÞ (16) E SO2DERi ðtÞ ¼ ½E SO2DERi ð1Þ; E SO2DERi ð2Þ; ……::E SO2DERi ðTÞ (17)

Fig. 3. Load demand profile.

3.2.3. Supply constraints For stable operation, each DER and ESS unit is subject to the technical limits which include the upper and lower bounds as follows: min max PDERi ðtÞ  PDERi ðtÞ  PDERi ðtÞ min max ðtÞ  PESSj ðtÞ  PESSj ðtÞ PESSj

i ¼ 1; 2; 3…ug þ 1

(18)

j ¼ 1; 2; 3…us

(19)

min ðtÞ; P max ðtÞ are the minimum and maximum operating Where PDERi DERi powers of each DER at time t.

3.2.4. Battery storage constraints According to a restriction which may exist on the charge and discharge rate of the storage devices, the following equation should be satisfied in each time interval:

WESSðtÞ ¼ WESSðt1Þ þ Pcharging ðDtÞhch 

Pdischarging

hdisch

ðDtÞ

(20)

where WESS(t) ¼ PESSj * t and can be defined as WESS(t1),WESS(t) are the amounts of energy storage inside the battery at time(t1) and (t) respectively, Pcharging,Pdischarging are the permitted rates of charge and discharge through a definite period of time Dt. hch is the battery efficiency during charging and hdisch during discharging and their values are considered in this work as 0.9 and 0.85 respectively. The minimum and maximum storage levels are constrained to lengthen service lifetime of the battery by avoiding deep discharging and over charging. These constraints can be

Fig. 2. MG system variables.

Fig. 4. Grid selling and buying tariff.

written as follows:

WESS

min

 WESSðtÞ  WESS

Pcharging  Pcharging

max

(21)

max

Pdischarging  Pdischarging

max

(22)

Where WESS_min, WESS_max are the minimum and maximum limits of the battery energy storage and Pcharging_max, Pdischarging_max are the maximum rates of charge and discharge during each interval Dt respectively.

3.3. System variables (system inputs) This section presents a full description for system variables such as: load profiles, grid tariff, and wind power profiles in MG system. It is important to note that all variables shown in Fig. 2 and used in this section are formulated based on practical knowledge. The forecasting model used to predict the load and wind power in this paper is regenerated from the model derived in Ref. [31] using the historical load and wind data described the French transmission network of electricity (RTE) [32].

Fig. 5. Wind power profile.

M. Elsied et al. / Energy 114 (2016) 742e752

747

Table 1 Power capacity, tariff and emission for each source. MG system

MT

FC

Wind

ESS

Tariff (Euro/kWh) CO2 (kg/kWh) SO2 (kg/kWh) NOX (kg/kWh) Min power (W) Max power (W)

0.46 0.37 3e-6 9e-5 50 1200

0.31 0.23 2.4e-6 6.5e-6 50 1200

0.7 0 0 0 0 200

0.31 0 0 0 e e

3.3.1. Hour forecasting of load demand profile The data forecasting process was based on load data collected over 4 years on an hourly basis from RTE. It is well known that the electricity demand varies according to some factors such as the time of day, time of year, geographical location and climate. Fig. 3 shows an average daily electricity load profile during one day in summer season and illustrates the extent of daily variations in electricity demand. 3.3.2. Grid tariff Demand variation is reflected in main grid market fare and/or supply contracts that ensure more generation when demand is high and less when it is low. Fig. 4 shows the used power fare (buyingselling) during 24 h. This price is for general residential usage and may be suitable for customers who use most of their electricity during off-peak times. The authors assumed in the studied scenario that the selling prices are always higher than the buying prices to increase the revenue of the MG system in grid-connected mode during day ahead market (DAM). However, in some other scenarios, the selling prices are not necessarily higher than buying prices. Moreover, it is assumed that energy prices during peak hours are abit higher than usual [33] to show the power exchange between the MG and the utility in grid-connected mode. 3.3.3. Wind power availability Fig. 5 shows the hourly assumed wind power profile considering the time of day, time of year and climate change. This power profile is obtained from RTE- France during one day in summer [32]. 3.4. System parameters This section presents supply power limits, operating cost, and emission factors for each DER during operating time. 3.4.1. Supply capacity power limits Each generator has maximum limits for the output power during operating mode as listed in Table 1. MT and FC are working all the time with a power range between minimum and maximum limits. 3.4.2. Supply cost and emission factors In DERs units, the total cost per produced kWh (unit cost) is traditionally calculated by discounting the investment cost and operating/maintenance costs over the lifetime divided by the annual electricity production. The first row in Table 1 shows that the unit cost of generation is calculated as an average cost over the lifetime. In fact, actual costs are usually lower than the calculated average ones at the beginning of life due to low operating/maintenance costs increasing over DERs usage time. The startup/shut down costs are not considered for operation scheduling in this paper according to their values are very small in small-scale DERs. The second, third, and fourth row of the table present the total CO2,SO2, and NOx emissions of each DER unit. It is proposed in this study that each generator unit should not exceed the maximum

Fig. 6. GA flowchart.

emission rate for each gas which considered as 17 kg of CO2, 7 g of SO2, and 5 g of NOX emissions at hour t. These maximum rates vary from country to country according based on the emission standards reference guide of each country. On the other hand, it is assumed that 70% of the utility grid power production is coming from nuclear generation system (0 emission) while 30% is coming from a coal generation system (0.95 kg/kWh of CO2, 0.35 g/kWh of SO2, and 0.2 g/kWh of NOx). The minimum state of charge (SOC) of the battery is chosen 20% where the maximum is limited to 80%. 3.5. Genetic algorithm (GA) This section presents the optimization algorithm for the produced power in the MG system. The optimization problem for MG operation has been analyzed using a GA. The genetic algorithm (GA) is a search heuristic that is routinely used to generate useful solutions to optimization and search problems. It generates solutions to

748

M. Elsied et al. / Energy 114 (2016) 742e752

Fig. 7. The optimal power allocation for the MG system.

optimization problems using techniques inspired by natural evolution such as inheritance, mutation, selection, and crossover [34,35]. Moreover, GA is applicable to both continuous and discrete optimization problems. In MG-EMS optimization scenarios, GA manifested the ability to evolve solutions with multiple objective criteria like cost and emission functions. In the paper work, the GA is implemented to determine the optimal settings by minimizing the cost functions (1) subjected to the given constraints (7)e(23), where the Population Size ¼ 70, Crossover Fraction ¼ 0.8, mutation rate ¼ 0.2, and Generations ¼ 200. The selection function was the Roulette and the crossover function was the arithmetic one. The mutation function was selected to be the Gaussian and stop criteria is the number of specified generations. The operation cost model of the MG is continuous and comprises many constraints. The step technique employed for determining the best solutions of the optimization problem are summarized graphically as shown in Fig. 6.

3.6. Optimization results In this section, optimization results of the MG system are evaluated using GA. The optimization of the installed capacity of DERs and the main grid is investigated during 24 h of operation. The optimization model takes into account the variation of the load demand profile, main grid tariff, and wind power variation during the day. Furthermore, it is considered that all generators are either in the MG or in the main grid in operating mode and the algorithm is intended to choose the optimum mode according to the cost and emission objective functions. Fig. 7 shows the optimal power allocation for the MG generators and the main grid. It is clear from the results that all DERs in MG

system and main grid are aimed to verify load profile requirements. As observed from the results, in the 6 h of the day, a major part of the load is supplied by FC and the main grid because the energy price of their units are lower compared to the others DERs. Over the next 6 h of the day (midday), the DER output power is increased due to the growth of the load demand as well as the increase of main grid energy price. During this period of the day, the MG can sell some power to the main grid if the generated power exceeds the load demand. It should be also noted from the results that the charging process of the ESS during the operating hours is done when the load demand and energy price are low, but the discharge action is postponed when the load demand reaches its peak values.

4. Hardware implementation In order to implement the GA results in real-time with its control strategy, an experimental test bench, shown in Fig. 8, has been designed in our laboratory. Four DC programmable power supplies, i.e., EA-SI8080 and GEC5000, are used to emulate the typical power profiles of FC, MT, Wind, and the main grid sending power. Two DC programmable electronic loads are used to emulate the load demand profile and main grid during receiving power from MG system. The Li-Ion battery is used as an ESS for the MG system. The open circuit voltage of the battery module is equal to 19.6 V. Five modules of DC/DC boost converters are connected in parallel in order to integrate all sources to common DC-bus, where its voltage is regulated at 50 V. The parameters of the boost converters and the battery module are listed in Table 2. MTX1032 and FLUKE 80i-110s are used respectively as differential voltage and current probes to measure the voltage and current signal. The Semikron SKM121AR is used as a switching device for the DC/DC boost converters modules

Fig. 8. The experimental test bench.

M. Elsied et al. / Energy 114 (2016) 742e752

749

Table 2 Boost converters and Battery Module Parameters. Items

Parameters

Boost converter Parameters

Inductance Inductor resistance (Rl) Capacitance (C) Capacitor Resistance(Rc) Switching Frequency (fs) Capacity Initial SoC Nominal Voltage Max Output Voltage DC Internal Resistance

Li-ion Battery Parameters

2 mH 40 mU 6800 mF 24 mU 10 kHz 65 Ah 80% 19.2 V 21.9 V 10 mU

to achieve small switching and conduction losses. 4.1. Control design and implementation The control system is divided based on the speed and functions on low and high level control layers. 4.1.1. Low level control design The low level control layer generates the necessary PWM signals to the DC-DC converters based on ALLC, which is implemented on the ARM Cortex M4 processor. This control layer performs fast computation for the control action and protection function which requires high bandwidth with a fast and predictable time response. The DSP extension for ARM Cortex M4 assists in fast computation of the control output. The built in dedicated PWM module and analog to digital converters with direct memory access make it possible to acquire analog feedback signals with a fast sampling rate while operating on fast switching frequencies. The STM32f407vgt6 processor running at 160 MHz was used for the embedded implementation of the control layer. In this implementation, the voltage and current ALLC are employed during transient and steady-state conditions for controlling the main utility grid boost converter as shown in shown in Fig. 9. In addition, the current ALLC shown in Fig. 10 is employed for MT, wind, ESS, and FC. The controllers are designed based on the Small Signal Model (SSM) using frequency response techniques. The SSM for the boost converters is derived in details in Refs. [36,37]. The equation of current and voltage control can be written as follow:

Ci ðsÞ ¼ kci

(23)

ðs þ zvÞ sðs þ pvÞ

(24)

Fig. 9. Main control scheme for the main utility grid.

Where, Cv(s) is the voltage compensator and Ci(s) is the current compensator that assures cancellation of the static error and high bandwidth. A pole at the origin is considered as an integral action and provides a very high gain at low frequencies. Moreover, the polezero pairs (pi,zi) for current controller and (pv,zv) for voltage controller aim to reduce the phase shift between the frequency of the two plant zeros and the frequency of two plant poles.

zi ¼ wcni

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1  sin∅mdi ; 1 þ sin∅mdi

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1  sin∅mdv zv ¼ wcnv ; 1 þ sin∅mdv

zi pi ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

(25)

zv pv ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

(26)

1sin∅mdi 1þsin∅mdi

1sin∅mdv 1þsin∅mdv

Where, wcni and wcnv are the new crossover frequencies of the current inner loop and the voltage outer loop, respectively. The compensator phase margin of the current and the voltage controller are Abbreviated as ∅mdi and ∅mdv ; respectively. kci is the current controller gain where kvi is the voltage controller gain which can be calculated as the following:

   1   at w ¼ wcni kci ¼  TiðSÞ

(27)

   1   at w ¼ wcnv kvi ¼  TvðSÞ

(28)

Where Ti(s),Tv(s) are the open loop transfer functions for the inner and outer loop, respectively.

Ti ðsÞ ¼ Ci ðsÞHi ðsÞ Tv ðsÞ ¼

ðs þ ziÞ sðs þ piÞ

Cv ðsÞ ¼ kcv

Fig. 10. Main control scheme for the others types of DERs.

for the inner loop

Cv ðsÞCi ðsÞHv ðsÞ 1 þ Ci ðsÞHi ðsÞ

for the outer loop

(29) (30)

Where Hi(s),Hv(s) are the current and voltage transfer functions of DC/DC boost converters. For the main utility grid local controller, the crossover frequency of the current-loop is selected to be 1.2 times of the plant crossover frequency, with a phase margin equal to 50 . To avoid the interaction between subsystems, low bandwidth control is used for the voltage outer loop which has a crossover frequency 1.1 times of crossover frequency one and a phase margin equal to 65 . For the remaining local controllers, the crossover frequency of the currentloop is selected to be 1.2 times of the plant crossover frequency. 4.1.2. High level control design based on MLIB/MTRACE The high level control layer is the optimization layer which generates the power reference for the distributed resources to optimize the cost and emissions. To guarantee real-time performance, the optimization code is running in a dSpace 1104 embedded controller. This control layers receive measurements from the low level controllers, load demand variation and calculate

750

M. Elsied et al. / Energy 114 (2016) 742e752

Fig. 11. Communication network topology.

the power reference for each source taking into account the economical and environmental impacts. The power reference is sent back to the low level controller to control the DC-DC converters through the communication network. The forecasted data for load and available energy is exchanging with the MG-EMS through the dSPACE Real-time MLIB/MTRACE libraries as shown in Fig. 11. The MLIB/MTRACE are considered as new tools to be used in real time energy management systems. The function of these two libraries is to provide direct access from MATLAB scripts to runtime data of the application running on a dSPACE board in real time, without interrupting the process [38]. MTRACE provides real-time data capture for the system variables such as load demand, available wind power, and grid tariff for logging and visualization, while MLIB writes the GA's output values back to the dSPACE processor memory without interrupting the operation. In this study, GA worked to run the optimization code and to take decision every 0.5 s where the sampling time of the load variation and power profile of non dispatchable was set to 1s. By this way, the optimization decision will include all operating points. In addition, it is assumed that each 1s during operating test is equivalent to 1hr of real operation.

Fig. 13. MT power profile.

4.2. Communication design and implementation Fig. 14. FC power profile.

The communication network is implemented using XBee

Fig. 15. Wind power profile. Fig. 12. Load demand power profile.

M. Elsied et al. / Energy 114 (2016) 742e752

Fig. 16. Battery power profile.

Fig. 17. Main grid power profile.

802.15.4 2.4 GHZ module which can provide data rate up to 250 Kbps. It can be configured as full function device, which means it can be configured as network coordinator, router or end device. The network topology is configured as a star topology as shown in Fig. 11 with the ZigBee module connected to the MG-EMS configured as network coordinator. The network coordinator initiates a Personal Area Network (PAN) by selecting a RF channel, a PAN ID, and allows routers and end-devices to join the PAN. The ZigBee modules connected to DER nodes are configured as end devices. The module can cover ranges of up to 90 m indoor and 1.6 km outdoor. For extended range for remote node, a ZigBee module can be configured as a router to repeat the signal from a remote node such as wind or renewable energy resources. XBee modules are connected to the DER controller and Dspace 1104 through a serial link with baud rate of 115.2 Kbps. 5. Experimental results This section presents the experimental results which are obtained based on the proposed RT-EMS and its communication and control systems. Fig. 12 shows the measured load power profile in the summer season and illustrates the extent of daily variations in electricity demand and how these vary by time of day, reflecting seasonal effects.

Fig. 18. FC reference power vs output power.

751

Figs. 13e17 show the DC/DC boost converters output power profile of MT, FC, Wind, the battery and main grid, respectively. As can be observed, the real output power produced by boost converter is tracking the RT-EMS references power profile. In the first hours [01 he06 h] of the day a major part of the load is supplied by the main grid because the energy price of their units are lower compared to the other DERs. In the next hours of the day (07hre24hr), the DERs output power is increased due to the growth of the load demand as well as the increase of main grid energy price. Therefore, the grid receives an active power which equals to the difference between DERs total power and load demand power. During this period of the day, the MG can sell some power to the main grid and can achieve good income. In addition, it should be also noted from Fig. 16 that the charging process of the ESS during the operating hours is done when the load demand and energy price are low, but the discharge action is postponed when the load demand reaches peak values. Fig. 18 shows the boost converter output power profile of the FC unit. It can be observed from the figure that the real output power produced by FC boost converter is tracking well by MG-EMS's FC reference power profile. This tracking indicates the fast response and efficient performance of the proposed control and communication systems. 6. Conclusion This paper presented an effective and advanced Real-Time Energy Management System (RT-EMS) based on an optimization model to investigate the optimal power allocation for DERs, ESS and the main grid. The optimization study takes into account the economical and environmental criteria. The proposed RT-EMS methodology and its behavior is illustrated and analyzed in detail through a daily load demand variation. The optimization problem of RT-EMS is solved using the power of a GA. To evaluate the performance of RT-EMS, a robust and reliable communication network based on the XBee 802.15.4 2.4 GHZ Module is used to exchange the information and decision commands between the EMS, DERs, ESSs, local controllers, and PEIC in real time. The results proved that the proposed RT-EMS is able to provide an economical and environmental solutions for all MG power systems which are working in grid connected mode. In the future work, we are planning to consider the participation of plug-in electric vehicles and load demand response programs in MG power systems. In such systems, RE-EMS and its communication network should work with a high level of flexibility in order to avoid the significant increase of consumption in peak periods. References [1] Lidula NWA, Rajapakse AD. Microgrids research: a review of experimental microgrids and test systems. Renew Sustain Energy Rev 2011;15:186e202. [2] Huang J. A review on distributed energy resources and MicroGrid. Renew Sustain Energy Rev 2008;12:2472e83. [3] Taha S. Recent developments in microgrids and example cases around the worlddA review. Renew Sustain Energy Rev 2011;15:4030e41. [4] Arindam C. Advancements in power electronics and drives in interface with growing renewable energy resources. Renew Sustain Energy Rev 2011;15: 1816e27. [5] Carrasco JM, Franquelo LG, Bialasiewicz JT, Galvan E. Power-electronic systems for the grid integration of renewable energy sources: a survey. IEEE transaction industrial Electron 2006;53(4):1002e16. [6] Justo Jackson John. AC-microgrids versus DC-microgrids with distributed energy resources: a review. Renew Sustain Energy Rev 2013;24:387e405. [7] Benefits of distribution-level power electronics for supporting distributed generation. IEEE transaction Power Deliv 2013;28(2):911e9. [8] Guerrero JM, Chiang Loh Poh, Lee Tzung-Lin, Chandorkar M. Advanced control architectures for intelligent MicrogridsdPart II: power quality, energy storage, and AC/DC microgrids. IEEE Transaction Industrial Electron 2012;60(4): 1263e70. [9] Etemadi AH, Davison EJ, Iravani R. A decentralized robust control strategy for

752

[10]

[11] [12] [13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]

[23]

M. Elsied et al. / Energy 114 (2016) 742e752 multi-DER MicrogridsdPart II: performance evaluation. IEEE Power Deliv 2012;27(4):1854e61. Vandoorn T, Vasquez J, De Kooning J, Guerrero J, Vandevelde L. Microgrids: hierarchical control and an overview of the control and reserve management strategies. IEEE Transaction Industrial Electron 2013;7(4):42e55. Huang Wei. Survey on microgrid control strategies. Energy Procedia 2011;12: 206e12. Chen Yen-Haw. Economic analysis and optimal energy management models for microgrid systems: a case study in Taiwan. Appl Energy 2013;103:145e54. Niknam T, Golestaneh F, Shafiei M. Probabilistic energy management of a renewable microgrid with hydrogen storage using self-adaptive charge search algorithm. Energy 2013;49:252e67. Niknam T, Meymand HZ, Mojarrad HD. An efficient algorithm for multiobjective optimal operation management of distribution network considering fuel cell power plants. Energy 2011;36:119e32. Ippolito MG, Silvestre Di, Riva Sanseverinoa E, Zizzo G. Multi-objective optimized management of electrical energy storage systems in an islanded network with renewable energy sources under different design scenarios. Energy 2014;64:648e62. nergie et Elsied M, Oukaour A, Gualous H, Hassan R, Amin Amr. Gestion de l'e me multisources base e sur l'algorithme ge  ne tique. SGE; optimisation du syste 14, July, 2014. Elsied M, Oukaour A, Gualous H, Hassan R, Amin Amr. An advanced energy management of microgrid system based on genetic algorithm. Istanbul: ISIE IEEE; 2014. Logenthiran T, Srinivasan D, Khambadkone AM, Raj TS. Optimal sizing of an islanded microgrid using Evolutionary Strategy. In: International conference on probabilistic methods applied to power systems (PMAPS). Singapore: IEEE; 2010. Colson CM, Nehrir MH, Wang C. Ant colony optimization for microgrid multiobjective power management. In: Power systems conference and exposition. PSCE '09, WA, Seattle; 2009. Elsied M, Oukaour A, Gualous H, Hassan R. Energy management and optimization in microgrid system based on green energy. Energy J May 2015;84: 139e51. Marzband M, Sumper A, Ruiz Alvarez A, Domnguez Garca JL, Tomoiaga B. Experimental evaluation of a real time energy management system for standalone microgrids in day-ahead markets. Appl Energy 2013;106:365e76. Marzband Mousa, Ghadimi Majid, Sumper Andrea, Luis Jose. Experimental validation of a real time energy management system using multi-period gravitional search algorithm for microgrid in islanded mode. Appl energy 2014;128:164e74. Wang W, Xu Y, Khanna M. A survey on the communication architectures in

smart grid. Comput Netw 2011;55:3604e29. [24] Usman A, Shami SH. Evolution of communication technologies for smart grid applications. Renew Sustain Energy Rev 2013;19:191e9. [25] Elsied M, Oukaour A, Gualous H, Lo Brutto Ottavio A. Optimal economic and environment operation of micro-grid power systems. Energy Convers Manag May 2016;122:182e94. [26] Gungor VC, Sahin D, Kocak T, Ergut S, Buccella C, Cecati C, et al. Smart grid technologies: communication technologies and standards. IEEE Trans Industrial Inf 2011;7:529e39. [27] Gezer C, Buratti C. A ZigBee smart energy implementation for energy efficient buildings. In: Proc. IEEE 73rd Veh. Technol. Conf. (VTC Spring); May 2011. p. 1e5. [28] Peizhong Y, Iwayemi A, Zhou C. Developing ZigBee deployment guideline under WiFi interference for smart grid applications. IEEE Trans Smart Grid Mar. 2011;2(1):110e20. [29] Mohamed Faisal A. Online management genetic algorithms of microgrid for residential application. Energy Convers Manag 2012;64:562e8. [30] Moghaddam Amjad Anvari. Multi-objective operation management of a renewable MG (micro-grid) with back-up micro-turbine/fuel cell/battery hybrid power source. Energy 2011;36:6490e507. [31] Mohamed Ahmed, Mohammed Osama. Real time energy management scheme for hybrid renewable energy systems in smart grid applications. Electr power Syst Res 2013;96:133e43. [32] RTE Electrical in France [online]. Available from: www.rte-france.com/en/ sustainable. [33] Eurostat, Electricity prices by type of user, [online]. Available from: http:// epp.eurostat.ec.europa.eu/tgm/table.do? tab¼table&init¼1&language¼en&pcode¼ten00117&plugin¼1. [34] Davis. The handbook of genetic algorithms. New York: Van Nostrand Reinhold; 1991. p. 61e71. [35] Vasconcelos JA. Genetic-algorithm-based optimization approach for energy management. IEEE Transaction Power Deliv 2013;28(1):162e70. [36] Elsied M, Oukaour A, Gualous H, hassan R, Amin A. An advanced dc/dc boost converter for hybrid electric vehicles applications based on Proton Exchange Membrane Fuel Cell (PEMFC). In: The 6th International Conference of fundamentals and development of fuel cells, France, Toulouse; 3e5 Feb 2015. [37] Ayoubi Y, Elsied M, Oukaour A, Chaoui H, Slamania Y, Gualous H. Four-phase interleaved DC/DC boost converter interfaces for super-capacitors in electric vehicle application based on advanced sliding mode control design. Electr Power Syst Res Feb 2016;96:134e86. [38] dSPACE User’s Guide Reference Guide. RTI 31: real-time interface to SIMULINK. dSPACE GmbH; 2012.