A Coordinated Control Method for Integrated System of Wind Farm and Hydrogen Production: Kinetic Energy and Virtual Discharge Controls

This paper proposes a novel coordinated control method of wind farms (WFs) and hydrogen production systems (HPSs). In a grid-connected system where the WF and the HPS are connected to the grid, the WF can supply the power to the grid, producing hydrogen in the HPS. Moreover, the output fluctuation of the WF can be mitigated when the HPS produces hydrogen from the output surplus. The purpose of the grid-connected system is to smooth the WF output fluctuation sufficiently, produce hydrogen constantly, and maintain a high capacity factor in the HPS. The proposed coordinated control achieves the mitigation of the WF output and the high capacity factor in the HPS. The key ideas are 1) utilizing the kinetic energy of wind generators and 2) virtual discharge of the HPS. The fluctuation components of the WF output are compensated by both the WF and the HPS. The proposed coordinated controller enables us to produce hydrogen constantly in the HPS with a low-rated power. The advantage of the proposed coordinated control is verified by a comparative analysis with conventional methods through simulations using real wind data.


I. INTRODUCTION
The penetration of renewable energy power generation (REPG) into power systems has rapidly grown due to the energy crisis, global warming, and other environmental problems. In particular, wind power generation has been widely used because it is clean, sustainable, and cost-effective [1], [2]. However, the power system, including large-scale wind farms (WFs), faces degradation of the power quality such as frequency and voltage fluctuations because the WF outputs constantly fluctuate due to wind speed variations [3]- [5]. In the worst scenario, the supply-demand imbalance caused by the power fluctuation may lead to a power outage. Therefore, it is essential to mitigate the output fluctuation of the WFs [5]- [8]. To smooth the WF output, energy storage systems (ESSs) such as batteries have been investigated [5], [9], [10]. The ESS can mitigate the output fluctuation of the WF, and its charge/discharge efficiency is high. Nevertheless, it is difficult for an ESS to store the excess power of the WF for a long time due to the limited energy capacity of the ESS. In addition, an overperformance ESS is usually used since the rated power and the energy capacity of the ESS cannot be designed independently [11]. Nowadays, hydrogen is also spotlighted as an alternative source to fossil fuels because of environmental problems [2], [12]. Household and industrial fuel cells (FCs), fuel cell vehicles (FCVs), hydrogen powered heavy-duty trucks, and hydrogen stations for the FCVs have been developed as applications of hydrogen energy [2], [13]- [16]. Moreover, the hydrogen can be converted to other useful chemicals such as methane by blending the hydrogen with CO 2 [2]. Although hydrogen can be produced by reforming fossil fuels, the actual cause of the environmental problem is not solved through the conventional hydrogen production method. To overcome this problem, hydrogen production systems (HPSs) composed of electrolyzers (ELZs) have attracted attention [17], [18]. An ELZ produces hydrogen by water electrolysis, which is free from carbon dioxide using the power obtained with the REPG [1], [2]. The HPS is suitable for the absorption of surplus WF output for a long time since the hydrogen in the tank can easily be taken out, utilized, or transported at anytime. This property has a significant meaning, which is not found in the energy stored in the ESS, such as chemical energy in a battery and kinetic energy in a fly-wheel [19]. In addition, the rated power and the energy capacity of the HPS can be designed independently. Therefore, the HPS enables us to implement large-scale WFs, producing an alternative source to fossil fuels.
There exist two types of systems composed of the HPS and the REPG, such as the WF, i.e., stand-alone systems [20]- [23] and grid-connected systems [11], [12], [16], [24]- [30]. In the stand-alone system where the HPS and the WF are disconnected from the grid, hydrogen can be produced without considering the power quality in the power system. However, the stand-alone system has the disadvantage that the WF cannot supply its power to the grid. Namely, the WF cannot contribute directly to the power system.
In the grid-connected system where the WF and the HPS are connected to the grid, the WF can supply power to the grid while producing hydrogen in the HPS. Moreover, the output fluctuation of the WF can be mitigated when the HPS produces hydrogen by using the output surplus of the WF. Therefore, the grid-connected system is a key technology to successfully install large-scale WF in the grid and produce hydrogen simultaneously.
Specifically, the purpose of the grid-connected system is to smooth the WF output sufficiently, produce hydrogen constantly, and maintain a high capacity factor in the HPS. In [24]- [28], systems composed of WFs, HPSs, FCs, and energy storage systems (ESSs), such as a battery, were proposed. Moreover, some operation methods for the system were also investigated in [24]- [28]. In these methods, the combination of HPS and FC behaves like an ESS because the HPS produces hydrogen by consuming power, and the FC generates power by consuming hydrogen. However, a part of electrical and hydrogen energies is wasted when the hydrogen produced by the HPS is reconverted to the power in the FC to smooth the WF output [16]. In addition, if hydrogen is used in the FC in order to smooth the WF output, it is challenging to store hydrogen for applications such as FCVs. In [29], a cooperative operation method for the WF and the HPS based on Nash bargaining theory was proposed for a system composed of WFs and HPSs. The method can reduce the operation cost of the WF and the HPS. Nevertheless, mitigation of the WF output was not investigated. Smoothing control methods using HPS for the REPG, such as the WF and photovoltaics, were proposed in [11] [12], and [16]. These methods mitigate the output fluctuation of the WF without using the FC and the ESS. However, it is difficult to smooth the output fluctuation of the WF sufficiently because the HPS cannot discharge, unlike the FC and ESS. Moreover, a high-rated power of the HPS is required to consume the large output fluctuation. Since these methods only utilize the output fluctuation of the WF to produce hydrogen, the HPS is not able to produce hydrogen constantly even when the wind speed is constant. As a result, it is inevitable to lower the capacity factor in the HPS.
To the best of our knowledge, there exists no method for grid-connected systems composed of the WF and the HPS to smooth the WF output sufficiently and to produce hydrogen constantly, while keeping a high capacity factor in the HPS. This paper proposes a novel coordinated control method for the WF and the HPS. The proposed coordinated control achieves mitigation of the WF output fluctuation and high capacity factor in the HPS. The key ideas in the proposed coordinated control are 1) utilizing the kinetic energy (KE) of wind generators and 2) virtual discharge of the HPS. The fluctuation components of the WF output are divided into positive and negative parts and they are compensated by the WF and the HPS, respectively. The proposed coordinated controller enables us to produce hydrogen constantly in the HPS with a low-rated power. The advantage of the proposed coordinated control is verified by a comparative analysis with the methods proposed in [11] [12], and [16] through simulations.
The rest of this paper is organized as follows: Section II This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
To WG FIGURE2: Block diagram of system.
describes the system configuration and reviews the conventional control methods. We propose a coordinated control for the WF and the HPS in Section III. Section IV illustrates the effectiveness of the proposed method through scenario simulations with real wind data. Finally, we conclude this paper in Section V.

II. SYSTEM DESCRIPTION
This section reviews the system configuration composed of a WF and an HPS. For simplicity, a time function and its Laplace transform are denoted by the same notation, such as P (t) and P (s). The latter is also used for H. In Fig. 2, η is the overall efficiency of AC-DC and DC-DC converters, and G H (s) denotes the transfer function from P ref H to P H . In this study, we assume that G H (s) ≈ 1 because the response speed of AC-DC and DC-DC converters is much faster [16], [20], and [26].

B. WIND TURBINE MODEL
The output characteristics of the WG are as follows [31]: In this paper, the dynamics of the back-to-back converter in the WG is ignored to shorten the simulation time, because it does not affect steady-state responses except transient responses, such as voltage dips. For this reason, we assume P ref i ≈ P i . The VSWT can operate at a high efficiency because of the maximum power point tracking (MPPT) which keeps C p maximum. We denote the maximum power coefficient and the optimum tip speed ratio as C † p and λ † , respectively. From a C p -λ curve of the wind turbine in [31], we have C † p = 0.48 and λ † = 8.1. Therefore, the MPPT output reference and its per-unit value are given by where P n is calculated by i , the MPPT can be achieved without measuring the wind speed. A pitch angle control system [32] is used to prevent the wind turbine from rotating faster than the rated speed.

C. ELECTROLYZER
In this section, we first review the ELZs for the REPG. Then, we explain the ELZ model used in this paper.

1) Water electrolysis systems
In [33] and [34], the water electrolysis technologies were investigated in detail. This paragraph reviews the result briefly.
The hydrogen production systems are partitioned into three kinds of electrolysis systems: alkaline water electrolyzers (AWEs), solid oxide electrolyzer cells (SOECs), and proton exchange membrane electrolyzers (PEMEs) [33]. Table 1 summarizes their main advantages and disadvantages. The PEME has attracted much attention as one of the hydrogen production systems from the output of the REPG due to the advantages listed in Table 1 as well as a fast dynamic response when compared with the AWE and the SOEC [33], [35]. Due to these characteristics, we use an ELZ model based on the PEME in this paper. VOLUME 4, 2016

2) Electrolyzer model
In this subsection, we explain the ELZ model used in this paper. The notations for the ELZ are listed in Nomenclature. The ELZ is modeled as a diode, a resistor R 0 , and an internal voltage E 0 under constant temperature and pressure [12], [18], as shown in Fig. 3. From Fig. 3 and Kirchhoff's voltage law, we have In this paper, we adopt the electrical model of [36] as the ELZ model. Fig. 4 shows the characteristics of a single ELZ module [36]. The module is comprised of 45 electrolytic cells, and its specification is listed in Table 2. From Fig. 4, we obtain a simplified model 1 : From (8), (9), and ηP H = V dc I dc (=P H ), H can be rewritten as follows 2 : where Using (10) and (11), we can calculate H from P H (See Appendix for a detailed calculation of H † ). An megawatt-scale HPS can be constructed by connecting modules of the ELZ in series and parallel [17], [20]. For instance, in the HPS which has four parallelized ELZ groups, each containing 28 cascaded ELZ modules, the rated power is about 5 MW, the rated gas flow rate is 840 Nm 3 /h, R 0 is 0.21 Ω, and E 0 is 2665.6 V. In addition, a larger-scale HPS can be achieved by connecting the megawatt-scale HPS,   including AC-DC and DC-DC converters in parallel [11], [12]. For such an HPS, its total H † becomes where n s is the number of cascaded ELZ modules in a group, n p is the number of paralleled groups, and n h is the number of HPSs connected in parallel. Note that PH n h in (12) denotes the consumed power in one of the megawattscale HPSs connected in parallel.
To evaluate the performance of the HPS, we define the capacity factor as where τ is the simulation period.

D. CONVENTIONAL SMOOTHING CONTROL USING HPS
This subsection briefly reviews a standard first-order lowpass filter (FLF) method and conventional methods proposed in [11] [12], and [16]. Fig. 5 shows the FLF controller [3], [5], [16]. In Fig. 5, P FLF is the output of the FLF used as P ref g , and T is the time constant. The FLF controller has been widely used in ESS because of its simple structure. However, as shown in Fig. 6a, an HPS with an FLF controller cannot smooth P WF sufficiently when P FLF > P WF since it cannot discharge. This is different from the ESS which can discharge. In addition, the capacity factor of the HPS is low because the HPS hardly operates. As a result, hydrogen gas cannot be produced constantly.

1) FLF
2) Method Proposed in [11] and [12] The FLF controller for the HPS was improved in [11] and [12]. In the method proposed in [11] and [12], P ref g and P ref H are given by where P σ is the standard deviation of P WF and it is given by Figs. 6a and 6b show that the method proposed in [11] and [12] can smooth P WF much better than the FLF controller. Nevertheless, P WF cannot be mitigated sufficiently when P ref g > P WF . Moreover, the capacity factor of the HPS is still low because the HPS cannot produce hydrogen when P ref g > P WF or (P WF − P FLF ) ≈ 0. It implies that hydrogen is not produced regardless of whether wind condition is good or not. In addition, as shown in 6b, the HPS with the high-rated power is required to compensate all fluctuation components of P WF .  [11], [12].
Time Power (c)Method proposed in [16] FIGURE6: Concept of conventional methods.
Note that it is assumed that the ELZ blocks in the HPS operate only at either P ref H = P n H or P ref H = 0 in [11] and [12]. However, the HPS (ELZ) can operate at 0 ≤ P ref H ≤ P n H . Therefore, for a fair comparison among methods for the HPS, we assume that the HPS controlled by the method proposed in [11] and [12] operates within 0 ≤ P ref H ≤ P n H .
3) Method Proposed in [16] The FLF controller for the HPS was also improved in [16]. The basic concept of the method proposed in [16] is similar to the method proposed in [11] and [12]. In [16], the HPS operates based on where β † (P WF , P FLF , T w ) (0 ≤ β † ≤ 1) is a gain used to modify P FLF , and T w is a buffer time. The calculation method of β † is given in [16] and omitted in this paper. As shown in Fig. 6c, the method of [16] can smooth P WF much more than the method of [11] and [12]. However, P g obtained by this method includes fluctuation components, as compared with P FLF . In particular, when P WF suddenly VOLUME 4, 2016 decreases, P g also decreases. It leads to the degradation of power quality in the power system. Moreover, it is difficult to produce hydrogen constantly and to reduce the rated power of HPS in the method of [16] as well as the method of [11] and [12]. As a result, the decrease of the capacity factor in the HPS is inevitable. In summary, there exists no method for the HPS, which can simultaneously smooth the output fluctuation of the WF sufficiently, produce hydrogen stably, and keep the capacity factor high.

III. COORDINATED CONTROL OF WF AND HPS
This section presents a novel coordinated controller for the WF and the HPS. The proposed coordinated controller is based on KE of the VSWT and virtual discharge of the HPS.
Kinetic energy control (KEC) for the VSWT has been widely investigated [37]- [39]. In the KEC, the KE is stored/released in the WG via rotor speed deceleration/acceleration so as to smooth the WG output. In other words, the WG does not operate based on MPPT. Nevertheless, the KEC cannot smooth the output fluctuation of the WF sufficiently because the available KE is low unlike the energy of ESS. In particular, excessive release of the KE may put a stop to the WG. Namely, it is not suitable to operate the WG in P i > P † i . In contrast, the operation in P i ≤ P † i does not destabilize the WG because the pitch controller protects the WG even if ω i is over 1 p.u.
As mentioned in Section II, it is difficult for the HPS to smooth the output fluctuation of the WF because the HPS cannot discharge, unlike ESS. In addition, hydrogen production in the HPS using fluctuation components of P WF lowers the capacity factor of the HPS and requires a highrated power of the HPS.
The proposed coordinated controller can overcome the problems in the KEC and the control of the HPS. In the proposed coordinated controller, the WF does not release the KE excessively. Specifically, the WF operates in P i ≤ P † i . Although the proposed KEC cannot remove all fluctuation components of the WF output, the HPS with the proposed controller compensates the residual fluctuation components. Therefore, the WF operates stably, and its output fluctuation can be mitigated sufficiently. In addition, since the HPS need not consume all fluctuation components of WF output, the HPS with a large-rated power is not required. This leads to an increase in the capacity factor in the HPS. Fig. 7 shows the proposed coordinated controller. As mentioned before, the fluctuation of P WF is smoothed by P i and P H whose references are generated by the proposed coordinated controller as P ref  [12], [26]. The proposed coordinated controller consists of an FLF to extract the fluctuation component in P † WF = n i=1 P † i , a KEC for the WF, and a virtual discharge control (VDC) for the HPS. In Fig. 7 positive and negative parts, 3 and P b ≥ 0 is an offset value for the control of the HPS. As shown in Fig. 7, ∆P WF and ∆P WF are given by the limiters and they are compensated by the KEC and the VDC, respectively.
First, we present the KEC in Fig. 7. In the proposed KEC, the WGs operate based on In (19), ∆P WF is allocated to each WG by the second term of the right hand [39]. From (19), It is clear from (20) and Fig. 8a that ∆P WF can be compensated by the proposed KEC. Since the proposed KEC compensates only ∆P WF , the WF does not release the KE excessively. In other words, the WGs operate based on MPPT when P † WF ≤ P FLF . Although it is inevitable to degrade the efficiency of the WG with the KEC, the operation based on the MPPT during P † WF ≤ P FLF leads to efficiency improvement compared to an operation in which the KE is constantly used.
Next, we describe the proposed VDC for HPS. Although the HPS cannot discharge, the proposed VDC enables the HPS to discharge virtually. Therefore, the HPS with the proposed VDC can mitigate ∆P WF (≤ 0). As shown in Fig. 7, P ref H is given by Note that P ref (20) and (21), P g is given by where P FLF = P † WF − ∆P WF as shown in Fig. 7. A constant power, P b , is taken out from P FLF and used for hydrogen production. From (22) and Fig. 8b, the HPS discharges ∆P WF from the constant P b supplied by the WF. As a result, smoothed power can be supplied to the grid.  The proposed coordinated controller cannot only smooth the WF output but also produce hydrogen constantly because the HPS operates around P b . In particular, when P b is set to P n H , a high capacity factor of the HPS can be achieved. Although the HPS with the conventional controller cannot produce hydrogen when ∆P WF ≈ 0, the HPS with the proposed coordinated controller generates hydrogen in such a condition. Moreover, the rated power of the HPS can be designed to a small value since the HPS only compensates ∆P WF . It also leads to an increase in the capacity factor.

IV. SIMULATION VALIDATION
In this section, a comparative analysis is performed with the methods proposed in [11], [12], and [16] to validate the proposed coordinated controller through the simulations. We demonstrate the comparative analysis for a WF (500 MW) composed of five WG groups, and each group contains 20 WGs (5 MW/WG), assuming that WGs involved in each group operate under the same wind conditions [11], [40]. The evaluation is conducted based on a power ramp requirement in Japan [41]. The grid code is as follows: • The maximum power change per five minutes is within 10% of the WF power rating. The fluctuation ratio ∆F (x(t)) in a five-minute window is defined as where x(t) is a signal such as P g (t) and P WF . From (23), the grid code is defined by The actual wind speed data 4 sampled every 3 s at a WF in Hokkaido, Japan, are used in the simulations. Fig. 9 shows one scenario and Table 3 lists the data of WF output obtained with all scenarios. In Table 3, SD stands for the standard deviation. In the simulation, η = 0.85 is assumed. The simulation is performed on MATLAB/Simulink 2019b and the simulation period is 3600 s.

A. ANALYSIS OF PARAMETERS IN CONTROLLERS
In this subsection, we determine T in the FLF and P n H for the proposed and conventional controllers through scenario simulations. Note that the rated power of the HPSP n H is estimated by ηP n H . Similarly, T w in the method proposed in [16] is also determined by scenario simulations. Note that P b in the proposed coordinated controller is set to P n H . First, we investigate T in the FLF and P n H for the proposed coordinated controller. Figs. 10a and 10b show the maximum ∆F (P FLF ) vs. T and the maximum |∆P WF | vs. T , respectively. 5 It is clear from Figs. 10a and 10b that the HPS with the proposed coordinated controller can sufficiently compensate ∆|P WF | when T > 314 s and P n H > 0.33 p.u. (the base value is P n WF = 500 MW in p.u.). It implies that the grid code can be guaranteed when T > 314 s and P n H > 0.33 p.u. Then, we also investigate T and P n H for the method proposed in [11] and [12]. Figs. 11a and 11b show the maximum ∆F (P g ) vs. T and the maximum P H vs. T , respectively. As shown in Fig. 11a, the grid code cannot be satisfied in all scenarios. From Fig. 11b, P n H > 0.37 p.u. is required. The method proposed in [16] has three design parameters: T , T w and P n H . We set T = 350 s based on the above analysis. Fig. 12a shows the maximum ∆F (P g ) vs. T w . The method proposed in [16] cannot satisfy the grid code for two scenarios even though T w is large. Although we investigated the results for numerous combinations of T and T w other than Fig. 12a through simulations, the grid code cannot be FIGURE11: Method proposed in [11] and [12].
satisfied. Fig. 12b shows the maximum P H vs. T w . From Fig. 12b, P n H > 0.46 p.u. is required. Finally, we list the specification and the parameters of the HPS in Table 4 based on the above analysis and Table 2. The HPS with the conventional controllers cannot satisfy the grid code. Furthermore, the conventional controllers require larger HPS than the proposed coordinated controller.
Note that the parameters of the HPS can be determined in a similar way as above in the real system through measured wind speed data. Moreover, the stability of the system is not affected by the proposed coordinated controller, due to its feed-forward structure as shown in Fig. 7. FIGURE12: Method proposed in [16].

B. ANALYSIS OF TIME RESPONSES
In this subsection, we demonstrate the performance of the proposed coordinated controller, comparing it with the conventional methods. Note that the time responses from scenarios 2 to 5 are not included in this paper because of page limitations. Fig. 13 shows the WF output and the power supplied to the power system. Fig. 14 shows ∆F (P g ). It can be seen from Fig. 13a that the proposed KEC can compensate ∆P WF . In addition, P g obtained with the proposed coordinated controller is similar to (P FLF − P b ). Indeed, the proposed coordinated controller can mitigate the WF output as shown in Fig. 14. In contrast, as shown in Figs. 13b and 13c, P g obtained with the conventional methods include the fluctuation components, compared with P FLF . In particular, as shown in Fig. 13c, the method proposed in [16] causes significant fluctuation in P g when P WF suddenly decreases. As a result, it is difficult for the conventional methods to satisfy the grid code as shown in Fig. 14. Fig. 15 shows the consumed power in the HPS. It is observed from Fig. 15  (b)Method proposed in [12] and [11]. (c)Method proposed in [16].
FIGURE13: WF output and power supplied to power system. the proposed VDC operates at around the rated power. As a result, the capacity factor is increased. In contrast, the conventional methods require the HPS with high rated power to compensate the large-output fluctuation of WF instantaneously and operate at low power when (P WF − P ref g ) is small. Therefore, it is inevitable to decrease the capacity factor. Fig. 16 shows the hydrogen gas produced in the HPS. It is clear that the HPS with the proposed VDC can produce more hydrogen gas than that with the conventional methods.
Efficiency decrease of the WF is inevitable with a KEC in comparison with the MPPT. As can be seen from Fig. 17, the WF with the proposed KEC operates at a different rotor speed from the WF controlled by the MPPT. It implies that the efficiency in the WF with the proposed KEC also decrease. We investigate the efficiency of the WF with the proposed KEC based on whereP WF is the WF output obtained by the MPPT. Fig. 18 shows the efficiency vs. T . It can be observed from Fig. 18 that the deterioration of efficiency is about 5% even if T is large. Table 5 summarizes the simulation results. It contains the average values of all scenarios. Table 6 shows the performance indices of the system composed of the WF and the HPS through simulation results. Although the average value of P g obtained with the proposed coordinated controller is 47.7% lower than that obtained with the method proposed in [11] and [12] and 34.0% lower than that obtained with the method proposed in [16], the hydrogen gas obtained with the proposed coordinated controller is 72% higher than that obtained with the method proposed in [11] and [12] and 39% higher than that obtained with the method proposed in [16]. Nevertheless, the rated power of HPS is the lowest among the control methods. In particular, the rated power of the HPS with the proposed controller is 29% lower than that with the the method proposed in [16]. As a result, the proposed coordinated controller achieves a capacity factor as high as VOLUME  (b)Method proposed in [11] and [12]. (c)Method proposed in [16]. 87.6% while those of conventional methods are less than 40%. It is evident from the simulation results that the proposed coordinated controller can mitigate the WF output sufficiently and achieve a high capacity factor in the HPS. Avg. value of Pg obtained with the pro. controller . b Power ratio in HPS = P n H of the HPS with the conv. controller P n H of the HPS with the pro. controller . c Hydrogen gas ratio = Hydrogen gas obtained with the conv. controller Hydrogen gas obtained with the pro. controller . d Capacity factor ratio = Capacity factor in the HPS with the conv. controller Capacity factor in the HPS with the pro. controller .
coordinated controller is smaller than that obtained by the conventional methods. Nevertheless, the proposed coordinated controller allows us to introduce a large scale WF to power systems because the grid code can be satisfied. This implies that the average value of P g can be increased in the proposed coordinated controller.

V. CONCLUSION
This paper proposed a novel coordinated control method for the system composed of a WF and an HPS. The proposed KEC and VDC could smooth the fluctuation component of the WF sufficiently without increasing the rated power of the HPS. Moreover, the HPS with the proposed controller produced more hydrogen gas while keeping the capacity factor high. Comparative scenario simulations between the proposed coordinated controller and the conventional controllers demonstrated that the HPS with the proposed coordinated controller could mitigate the WF output fluctuation with a lower HPS rated power and produce hydrogen gas constantly. The simulation results validated that the proposed coordinated controller was more effective than the existing control methods for systems composed of WF and HPS.
In the coordinated control, it is inevitable that the efficiency of the WG degrades. In addition, time-consuming simulations are often needed to design T in the FLF and P b . Future work includes the optimization of WG's efficiency and the parameter design of the coordinated controller.