Energy management and economic analysis of multiple energy storage systems in solar PV/PEMFC hybrid power systems

This study proposes an energy management system (EMS) to manage a standalone hybrid power system (HPS) comprising solar photovoltaic (PV), proton exchange membrane fuel cell (PEMFC), and a battery energy storage. The battery and a hydrogen storage system in PEMFC provide short- and long-term electricity storage, respectively. The EMS ensures electrical efﬁciency during normal/abnormal operation of the HPS and related limitations, namely unexpected variations in solar irradiance and loads, PEMFC cold start, and battery charging status. The proposed EMS is evaluated using mathematical modelling, simulation studies, and real-time digital simulation. The results show that the proposed EMS is resilient to complex solar PV/load power changes and keeps the dc-link voltage in place. The viability of the proposed HPS is analysed using daily average solar insolation and load proﬁles. This analysis is useful to schedule problems of load/generation. The economic analysis of the proposed EMS, done using Homer pro, ensures that the HPS operates cost-effectively.


INTRODUCTION
Renewable energy sources (RESs) are used in power grids to deliver clean energy to consumers. The RESs are uncontrollable, as they are nature-dependent to generate electricity [1]. Thus, it becomes difficult to plan and use RESs. Energy storage system (ESSs) such as fuel cells and batteries can help solve the aforementioned issues by injecting the required energy into the network grid, thereby making the process controllable. RESs and ESSs are correlated in various ways [2,3], depending on the availability and cost. An energy management system (EMS) is required to manage energy sources to optimally fulfil the load demand. The objectives of EMSs differ depending on RESs or algorithms [4][5][6], such as load fulfilment and cost minimization. EMS algorithms such as fuzzy logic control [7], heuristic algorithm [8], and other soft computing techniques [9] have been used. However, these algorithms are complex and computationally intensive. In addition, most combinations of RESs cannot detect the battery state of charge (SOC bat ) and the slow ramp rate of proton exchange membrane fuel cell (PEMFC; in planning).
This study proposes an EMS, applied to a standalone hybrid power system (HPS), that comprises solar photovoltaic (PV), PEMFC, and lead-acid battery. Solar PV is a commercial low-cost matured RES [1]. PEMFC is a clean energy source, allowing bi-directional energy flow and storing and generating electricity. The generation occurs owing to the energy released caused by chemical reactions [10]. Hydrogen, fuel for PEMFC, is stored in the hydrogen storage system (HSS). High-purity hydrogen is an important requirement for energy generation in PEMFC. It is produced by the electrolysis of water, an eco-friendly carbon-neutral method [11]. Electrolyzers require 48-55 kWh of electricity to supply 1 kg of hydrogen at an efficiency of 65-70%, although the overall production of industrial hydrogen using electrolysis is 4% worldwide [12] owing to the economics involved. However, PEMFC cannot feed dynamic loads owing to difficulties in cold, slow ramp rates, and fluctuations in output voltage (V FC ) [10]. In addition, the current produced by PEMFC (i fc ) contains high low-frequency ripples, rendering the system unstable, especially in a grid-connected mode [13]. While PEMFC can store energy itself, its integration with ESS can help feed dynamic loads [14]. Moreover, fluctuations in the solar PV output caused by weather can stress or damage the inverter [15]. The ESS helps overcome power fluctuations at the inverter input and release stress, making solar PV a reliable energy source [16]. However, the high price of ESSs increases the overall system cost [17]. In this study, the lead-acid battery acts as an ESS.
As an HSS in a PEMFC serves as energy storage, in this study, the combination of lead-acid battery and HSS is called multi-ESS (MESS). The proposed EMS uses the voltage and current parameters of the solar PV, ESS, and DC and AC buses to share power among energy sources, MESS, and loads. The EMS controls the power flow to qualitatively and quantitatively maintain equilibrium between generation and consumption. Many studies mention several control strategies for the charging and discharging of batteries [18]. In this study, the maximum selfconsumption of the power generated from solar PV [19] can charge the battery. Previously, the excess P pv was used to diminish in dump loads [20]. This study proposes to store the excess P pv in HSS as hydrogen. Moreover, the PEMFC and battery operation can help overcome load fluctuations. The proposed EMS algorithm, considering standalone HPS, is evaluated in the following scenarios: i. Scenario 1: Variable P pv and constant P Load ii. Scenario 2: Variable P pv and variable P Load iii. Scenario 3: Effect of fault on the EMS iv. Scenario 4: HPS analysis using daily average solar insolation profiles The proposed EMS is evaluated in Scenarios 1 and 2 under normal conditions and in Scenario 3 under dynamic operating conditions, whereas the long-term system operation is observed in Scenario 4.
The EMS estimates the availability of the communication system to transmit and collect data from each source and method. The proposed research states the following findings: i. The proposed design removes energy wastage in dump loads and allows adequate energy storage in HSS. ii. The ESS integration helps overcome variations at the input terminal of the inverter. iii. The proposed EMS is simple, fast, and efficient in computation. iv. The EMS stabilizes V dc_act regardless of source power variations. v. According to SOC bat and P pv , the EMS switches among the battery, PEMFC, and electrolyzer.
The rest of the paper is structured as follows: Section 2 details various components of the proposed system. Section 3 analyses the economic feasibility of the system using Homer Pro and compares it with the existing literature. Section 4 discusses the specifications and control methods for the power device. Section 5 deals with the planned EMS. Section 6 analyses the results, and Section 7 concludes the paper. Figure 1 shows the considered off-grid HPS. The solar PV array and PEMFC are connected to a common DC-link using a boost converter and a bi-directional DC-DC converter, respectively. Moreover, the bi-directional DC-DC converter (BDCC) connects the battery to the DC-link, whereas HSS is a part of PEMFC. The DC-link is a 5000-µF capacitor with V dc _ ref of 700 V. The mathematical models of the solar PV and the MESS

Solar PV array model
The solar PV array is a combination of series (N S ) and parallel (N P ) connected PV modules [21]. According to semiconductor technology, a solar cell generates 0.7-1.2 V [2]. Here, each solar PV cell is modelled as a single diode circuit equivalent with a parallel current source (Figure 2(a)). The relationship between V pv and I pv is provided in (1). where

Battery model
The equivalent circuit of the battery is shown in Figure 2(b) [22]. v bat and SOC bat are calculated in (2) and (3), respectively.

PEMFC model
PEMFC is an electrochemical component that generates electricity using chemical reactions [24]. V FC is calculated as shown below: The equivalent PEMFC circuit is shown in Figure 3(a), where different losses are described as resistance equivalent, R act , R con , and R ohm . The boost mode of BDCC increases V FC to a V dc_ref level. The PEMFC is presumed to be in "on" condition in the present function to avoid cold start.

Electrolyzer and HSS
The electrolyzer electrolyses the water to generate hydrogen at a rate defined in (5) [25,26]. The hydrogen is stored in HSS at a pressure defined by (6) to be utilized by PEMFC as per the load requirement. The equivalent circuit of the electrolyzer is shown in Figure 3(b). The source represents the reversible voltage obtained from Gibbs free energy caused by chemical reactions. Figure 3(b) shows the ESR of the electrolyzer cell [25].
The capacity of HSS is based on the maximum amount of hydrogen produced during electrolysis.

ECONOMIC ANALYSIS
The economic analysis of the standalone HPS is performed using Homer pro [27]. The following two parameters have been considered: net present cost (NPC) and cost of energy (COE). NPC is the present installation value and lifetime operation cost of the system. It is calculated based on the total annualized cost  ($/year) and capital recovery factor. COE is the "average cost per kWh" of usable electricity produced by the system. The COE is expressed as follows: The economic feasibility of the proposed HPS has been evaluated considering two schemes: i. Scheme 1: Total load supplied by the utility ii. Scheme 2: Total load supplied by standalone HPS Each scheme was evaluated to determine the most suitable power delivery scheme. Table 1 lists the cost of the main system components. The performance data of the equipment and the related cost are obtained from [27]. The cash flow diagram obtained during simulation with Homer Pro is shown in Figure 4. It shows the superior performance of the considered HPS, that is, Scheme 2 over Scheme 1. The analysis of NPC and COE is listed in Table 2.

CONTROL STRATEGIES
Different control strategies to control the system components described in Section 2 are discussed in this section. These strategies are vital for operation and control.

Solar PV control
The solar PV controller includes the maximum power point tracking (MPPT) controller to extract maximum power from the solar photovoltaic [28][29][30][31] and another control method to maintain constant DC-link voltage (700 V). The error signal of the MPPT controller is passed through the proportional integral (PI)-1 controller to reduce the steady-state error, providing D 1 (the MPPT control duties ratio). The second control loop comprises a DC-link voltage regulator, in which the DC-link voltage is compared with the reference value (700 V), and the error signal is transmitted using the PI-2 controller for a steadystate error reduction. The PI-2 output supplies the duty ratio D 2 for DC-link voltage variations. Comparing the aforementioned duty ratios, the duty ratio or duty cycle for the DC-DC boost converter is produced ( Figure 5(a)).

VSI control
A voltage source inverter (VSI) converts V dc _ act to 3-ϕ AC voltage, using a dual-loop control strategy ( Figure 5(b)). In a dualloop control method, the first loop maintains V dc_act , and the second loop regulates I d and I q . Inputs of the controller include load voltage (V abc ), load current (I abc ), and regulated V dc _ act . To produce reference voltages (U abc_ref ), the current regulator transforms V abc into V d and V q . The d and q components enable independent control of real and reactive power.
The reference for I q is considered as zero to ignore the reactive power generated by the VSI. The reference generator uses V dc_act and phase-locked loop (PLL) to provide U abc_ref . To produce VSIPWM signals, U abc_ref is compared with V d and V q .

PEMFC and battery control
The PEMFC and battery operate as the two-part control strategy shown in Figure 5(c), that is, battery management system (BMS) and PEMFC management system (PEMFCMS). During BMS, the BDCC operates in the buck and boost modes to charge and discharge the battery, respectively. Moreover, PEM-FCMS performs two operations: electrolyzer control, that is, the controller activates hydrogen production; and hydrogen discharge, that is, the controller enables the hydrogen discharge by activating the PEMFC. The occurrences of circulating currents, caused by the parallel operation of power converters, are avoided by the EMS, and the sources are isolated using a relay.

EMS ALGORITHM
The proposed EMS shares power among loads, generation, and MESS using the following system parameters: Modes are decided based on the availability of P pv . Each mode may have sub-modes depending on the availability of ESS. The sub-modes or selection of ESS depends on SOC bat limits, as discussed in Section 2. The flowchart is shown in Figure 6(a).

Mode 1: Power surplus mode (P pv > P Load )
In Mode 1, surplus P pv is available. A part of P pv feeds the load, and the remaining part is stored in the ESS. The ESS is selected based on SOC bat, which is provided as two sub-modes: i. Battery charging sub-mode (BCSM) ii. Hydrogen storage sub-mode (HSSM) The algorithm helps select battery as energy storage if SOC bat ≤ 80% and enter into the BCSM. If SOC bat ≥ 80%, the algorithm selects HSS for energy storage, which is termed as HSSM.
In the BCSM, BDCC operates as a buck converter to charge the battery with i bat_ref , given in (9), up to SOC bat = 80%. The difference between i bat_ref and i bat is fed to the PI controller (PI-3) (Figure 5(c)). The PI controller output is compared with a 15-kHz triangular carrier wave to generate the BDCC switching signal. P Load in BCSM is provided in (8). In HSSM, the electrolyzer generates hydrogen using the excess P PV and stores it in HSS. Here, the electrolyzer control ( Figure 5(c)) becomes active. The equation of P Load is given in (10).

Mode 2: Power deficiency mode (P pv < P Load )
In Mode 2, P pv is less than the load demand. Therefore, the MESS provides deficit power. The BDCC operates in the boost mode to discharge the battery until SOC bat is 20%. Depending on SOC bat , this mode can be further divided into (i) Battery discharge sub-mode (BDSM) (ii) Hydrogen discharge sub-mode (HDSM) In BDSM, the battery discharges to supply the deficit power, as defined in (11); yet, the control in BDSM is similar to that in BCSM.
The PEMFC uses the hydrogen from the HSS to generate power, P fc , given as If both the battery and hydrogen storage have sufficient capacity, the battery will be given priority owing to the high ramping rate of the battery.

Mode 3: Power balanced mode (P pv = P Load )
In Mode 3, P pv can feed P Load , as given in (13).
The flow chart of the EMS algorithm and the flow diagram indicating power flow in the aforementioned modes are shown in Figure 6(a) and Figure 6(b)-(d), respectively.

RESULTS AND DISCUSSION
A robust EMS should meet the load demand, under all circumstances, while managing energy sources and energy storage. The performance of the proposed EMS is analysed considering the following four scenarios in real-time using OPAL-RT OP5700 ® with RT-LAB ® [32] and Simulink ® interface: (i) Scenario 1: Variable P pv and constant P Load (ii) Scenario 2: Variable P pv and variable P Load (iii) Scenario 3: Effect of fault on the EMS (iv) Scenario 4: HPS analysis using daily average solar insolation profiles.
For each scenario, the values of different parameters of each component discussed in Section 2 are listed in Table 3. During the study, the parameters monitored are MESS response; MESS voltage and current response; load voltage and current; and VSI output. In each scenario, three different cases, namely, Case-1: SOC bat ≤ 20%, Case-2: SOC bat = 50%, and Case-3: SOC bat ≥ 80%, are considered. The SOC bat is selected based on the upper and lower limits of the battery SOC, as discussed in Section 2. In the first three scenarios, the same changed pattern of irradiance is used, whereas in Scenario 4, simulations are performed for a whole day with varying irradiance. The study is evaluated based on the tracking of the load profile using the HPS.

Scenario 1
In Scenario 1, P Load is constant, and P pv varies (Table 4) owing to a change in solar irradiance.

Case 1 (SOC bat < 20%)
In this case, SOC bat = 10%, and the solar irradiance varies, as listed in Table 4. The system response, as shown in Fig-  P pv > P Load and SOC bat < 20%. Therefore, the previous mode and sub-modes remain active. The voltage and current waveforms of the system (Figure 7(b)) shows that irrespective of the operating condition, V dc _ act = 700 V. Under varying P pv , V pv is maintained. However, I pv changes with P pv , which reflects the resilient nature of the EMS.

Case 2 (SOC bat = 50%)
The response of Case 2 is similar to Case 1 but with battery only. Using the irradiance levels listed in Table 6, the response is plotted (Figure 8(a)) at 0-0.1 s and P pv = 59 kW (<P Load ), at which

Case 3 (SOC bat > 80%)
In Case 3, the battery is charged fully. Therefore, the electrolyzer uses excess P pv . Figure 9(a) shows the system response with the same irradiance levels and P pv as in Case 1. At 0-0.1 s, P pv < P Load ; therefore, BDSM (in Mode 2) gets activated. The battery complements the solar PV to furnish the remaining power of 16 kW. At 0.1-0.2 s, P pv > P Load . During this period, the HSSM (in Mode 1) gets activated to feed the extra power to the electrolyzer. A similar operation has been observed at 0.2-0.3 s. The voltage and current of various sources are shown in Figure 9(b). v e and i e are shown in Figure 9(c). At 0.1-0.2 s, i e is 8 A, which increases to 50 A at 0.2-0.3 s. At 0.3 s, P pv becomes low. Thus, i e becomes zero, and the battery provides 32.6 A of current. The VSI outputs, V abc and V dc-act , are shown in Figure 9(d). V dc-act and V abc are maintained irrespective of the varying irradiance. The in-phase operation of the system is shown in Figure 9(e). The lifetime of VSI depends on the operating environment and load. P pv at input terminals reduces the life of VSI; however, the ESS, acting as a buffer, absorbs sudden changes in the power to increase the life of VSI [15,16]. Figure 10 shows the VSI loading with and without MESS.

Scenario 2
In Scenario 2, a random load variation marked as P load_ref in Figure 11(a), is used for analysis. The change in load at each time interval is listed in Table 5. Based on SOC bat , the following cases have been considered:

Case 1 (SOC bat < 20%)
For SOC bat = 10%, the system operates the BCSM in Mode 1 and the HDSM in Mode 2. The power flow among the components is shown in Figure 11 The current and voltage generated by each source are shown in Figure 11(b) and Figure 11(c). It shows that the changes in P pv and P Load follow the current. The voltage and current at VSI terminals are shown in Figure 11(d), where the variations in P Load are reflected in the load current.

6.2.2
Case 2 (SOC bat = 50%)  Figure 12(b) and Figure 12(c), respectively. P pv and P Load track the changes in the current source. The voltage and current at VSI terminals are shown in Figure 12(d), in which the variations in P Load are reflected in the load current. Moreover, the load current follows the load changes instantaneously.

Case 3 (SOC bat > 80%)
In this case, SOC bat = 90%. The system operates in HSSM (in Mode 1) and BDSM (in Mode 2). P pv , P fc , P electr, and P bat are shown in Figure 13(a), where at 0-0.1 s, P pv = 59 kW and P Load = 75 kW. As P Load > P pv , the remaining power of 16 kW is supplied by the battery. At 0.1-0.12 s, P pv = 79 kW and P Load = 75 kW; the excess power of 4 kW is used by the electrolyzer. At 0.12-0.18 s, P Load = 100 kW and P pv = 79 kW. The battery supplies the remaining power of 21 kW. At 0.18 s, P Load is reduced to 50 kW and P pv is 79 kW. As P pv > P Load , the excess power is used by the electrolyzer. At 0.2 s, P pv = 100 kW and P Load = 50 kW, the excess P pv is stored in the battery. At 0.25 s, P Load increases to 125 kW and P pv remains at 100 kW; the battery supplies the remaining power. At 0.34 s, P pv is 79 kW and P Load is 125 kW. The battery supplies the remaining power of 46 kW. At 0.34-0.4 s, P Load is 60 kW and P pv is 79 kW. The electrolyzer uses excess power of 19 kW. The current and voltage of each source  Figure 13(b) and Figure 13(c). In this case, the load current reflects the variation in P Load , as shown in Figure 13(d).

Scenario 3
In this scenario, a line to ground fault is created at load terminals. Case 1 (SOC bat < 20%) has been considered for analysis. Similar responses are obtained for Cases 2 and 3; hence, their results have been omitted owing to space constraints. The fault is applied during 0.1-0.15 s. Figure 14(a) shows that during the fault, the phase voltage reduces to zero, and the current increases. The voltage and current regain their previous states immediately after the fault clearance. Figure 14(b) shows that the sources kept generating during the fault. The EMS maintains its performance during the fault conditions, and P Load tracks P Load_ref . P Load gets disturbed during the fault, but as soon as the fault is cleared, the EMS starts tracking P Load_ref . The  comparison of the proposed EMS with the existing EMS [33][34][35] is listed in Table 6. It shows the superior performance of the proposed EMS over other algorithms.

FIGURE 15
Monthly average solar insolation level [27] FIGURE 16 Typical load profile

Scenario 4
For long-term performance evaluation, simulations are conducted over 24 h using average solar insolation data [27]. The monthly average data for solar irradiance, with a resolution of 1 min, from 9:00 AM to 5:00 PM is shown in Figure 15. For the remaining period, the MESS should support the load. The average load profile of the day is shown in Figure 16. The load is of an academic building [27], varying from 50 to 125 kW. The load schedule is listed in Table 7. Figure 17(a) shows 24-h energy management for Case 1 (SOC bat < 20%). In this case, the EMS activates Mode 2 (HDSM). As a result, the PEMFC supports the loads whenever P pv is unavailable. The battery charges itself, using the excess P pv , whenever available. Figure 17(b) shows energy management for Case 2 (SOC bat = 50%). In this case, the battery supports the loads in the absence of P pv . The excess P pv is diverted to charge the battery. The energy management for Case 3 (SOC bat > 80%) is shown in Figure 17(c). As shown in Figure 17(c), in this case, the battery would support loads whenever P pv is not available. The excess P pv is diverted to HSS. In Figure 18, the voltage and current of each source are shown.
The three-phase current, voltage, and DC voltage are shown in Figures 19 and 20 for all cases. The load change does not affect the voltage and current of the VSI or the DC-link voltage. The variation of current and the tracking performance of the power are shown in Figure 19(b) and Figure 20(c), respectively. The response times are tabulated in Tables 8 and 9. As the response time is similar in all cases, the results are shown    After each set of simulations, that is, of all the three cases for each scenario, P loss equal to 0.3-0.5% of the total capacity (100 kW) has been obtained. Considering the aforementioned value, along with converter and FC heat losses, efficiency of ≈93.5% is achieved. In all the cases, the efficiency of the system is higher compared with that of [11] because the system uses an electrolyzer instead of dump loads.

CONCLUSION
This study proposes a dynamic robust EMS for solar PV/PEMFC/BATTERY/HSS. The proposed design helps replace dump loads with adequate storage to improve system performance and reliability. Solar PV and PEMFC are the primary sources of energy, while the battery and HSS form the MESS. The proposed EMS is tested under different operating conditions: (i) variation in loads, (ii) variation in sources, (iii) fault at load terminals, and (iii) 24-h operations. Based on the SOC of the battery and the Solar PV output, the EMS switches between the MESS and Solar PV. Under all circumstances, the EMS is found to do the following functions: (i) Respond faster to the dynamic changes in solar PV/ load power (ii) Maintain DC-link voltages The system reliability improves with the simultaneous operation of the battery and electrolyzer, which maintains the power flow. The reliability and longevity of the inverter increase by supplying constant voltage to its input terminals. The proposed EMS monitors and allocates power sources as soon as the load or power supply variation occurs. The results show the superior performance of the proposed EMS.
To prove the practicality of the proposed standalone HPS, the system has been tested for 24 h, and the responses are found to be satisfactory. Finally, the economic analysis of the system has been performed using Homer pro. HPS is an economical alternative for the existing power system. The COE of the proposed HPS is 0.270 $/kWh, which is much lower compared with the grid-connected operation owing to the inclusion of MESS.