Power Management Approach to Minimize Battery Capacity in Wind ...

7 downloads 0 Views 2MB Size Report
Sep 18, 2017 - Abstract—In a wind-battery hybrid power system, minimal bat- tery capacity is a crucial requirement to achieve economic opera- tion.
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 5, SEPTEMBER/OCTOBER 2017

4843

Power Management Approach to Minimize Battery Capacity in Wind Energy Conversion Systems Cong-Long Nguyen, Member, IEEE, and Hong-Hee Lee, Senior Member, IEEE

Abstract—In a wind-battery hybrid power system, minimal battery capacity is a crucial requirement to achieve economic operation. In this paper, an optimal power control strategy based on a first-order low-pass filter is proposed to minimize the battery capacity by adjusting the filter smoothing time constant. We demonstrate the mathematical relationship between the filter smoothing time constant and the fluctuation mitigation requirement during one sampling time, so the optimal filter smoothing time constant can be easily computed to minimize the battery capacity. Moreover, an online short-term power control is also considered to maintain the battery state of charge within a safe range and to regulate the battery power below its rating. The proposed power management approach is simple and easy to implement. In order to verify the effective features of the proposed power management approach, a case study is carried out along with some experimental verification. Index Terms—Battery energy storage system (BESS), optimization methods, short-term power dispatch, state of charge (SOC) control, wind-battery hybrid power system (WBHPS).

I. INTRODUCTION N RENEWABLE energy resources, wind energy is the leading candidate for electricity production due to its lower investment cost and well-developed technology in manufacturing high-power wind turbines (WTs) [1]. However, similar to other renewable resources, wind power is unsteady and uncontrollable because wind speed depends on natural and meteorological conditions. The inevitable power fluctuation of wind farms (WFs) would introduce serious technical challenges in the electric power grid, such as power system quality and reliability, system protection, and power flow control [2]. As a result, the fluctuation issue is a primary barrier, preventing wind power from penetrating the electricity grid. Use of a battery energy storage system (BESS) provides a feasible solution to mitigate wind power fluctuation. For instance, Muyeen et al. [3] introduced a BESS incorporated with a static

I

Manuscript received October 24, 2016; revised April 5, 2017; accepted April 22, 2017. Date of publication May 10, 2017; date of current version September 18, 2017. Paper 2016-SECSC-1169.R1, presented at the 2014 IEEE Energy Conversion Congress and Exposition, Pittsburgh, PA, USA, Sep. 14–18, and approved for publication in the IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS by the Sustainable Energy Conversion Systems Committee of the IEEE Industry Applications Society. This work was supported by the National Research Foundation of Korea funded by the Korean Government under Grant NRF-2015R1D1A1A09058166. (Corresponding author: Hong-Hee Lee.) ´ C.-L. Nguyen is with the Ecole de technologie sup´erieure, University of Quebec, Montreal, QC H3C 1K3, Canada (e-mail: [email protected]). H.-H. Lee is with the University of Ulsan, Ulsan 680-749, South Korea (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TIA.2017.2703143

synchronous compensator to smooth out the wind power so as to improve the system stability as well as the power quality. Qin et al. [4] proposed a wind-battery hybrid power system (WBHPS) in which the storage device was connected directly to the dclink of the power converter to supply stable power to the grid. Thanks to the mature development of battery technologies and rapid battery cost reduction, there are several real projects using the BESS to stabilize the wind power. Typical examples include the 254-MWh NAS battery installed to stabilize the 51-MW WF presented by Kawakami et al. [5], and the 2-MWh lithium-ion (Li-ion) battery built in the 4.5-MW WF of the Jeju smart grid test-bed [6]. For planning and design of the WBHPS, the most challenging technical aspect is to minimize the system cost by mean of using the optimal BESS capacity [7]. Because the power dispatch decides the required BESS capacity, recent works have concentrated on the optimization of the power dispatch strategy. Wang et al. [8] determined the optimal power dispatch and BESS capacity based on the benefit function to maximize the system income. However, the determined values of the dispatch power and the BESS capacity are optimal only in the investigated day. Therefore, it is hard to ensure that the system optimal in all time. Yao et al. [9] developed a dual-BESS, where the BESS capacity was determined based on the wind speed statistic and the requirement of charge and discharge time of each BESS. Although the dispatch power is constant, the large BESS capacity and the complicated switches to exchange the BESS role make the scheme infeasible. Teleke et al. [10] planned the power dispatch by taking the average wind power prior to each dispatching time interval. Meanwhile, other researchers introduced min–max methods where the dispatch power depends on the maximum and minimum levels of the wind power [11]. In spite of remarkable contribution in the dispatching methods in [7]–[11], they require an advanced wind power forecast of several hours and involve a complicated process to manage the power dispatch in the short-term. In order to overcome such problems, a first-order low-pass filter (FLF) is used to find the dispatch power for real-time operations [12], [13]. In the FLF-based dispatching method, the filter smoothing time constant (FSTC) decides the fluctuation mitigation level and the BESS capacity. So far, most of the power dispatching methods using the FLF has been concerned with a fixed FSTC [14]. Thus, they require a rather large FSTC to ensure the dispatch power meets the fluctuation mitigation requirement (FMR), which requires the big BESS capacity. Even though the FSTC was optimized in [15] and [16] by using

0093-9994 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

4844

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 5, SEPTEMBER/OCTOBER 2017

Fig. 2.

Fig. 1.

Wind-battery hybrid power system configuration.

a two-time-scale coordination control method and a dual-layer control strategy, respectively, they demanded a complex process to search the optimal FSTC and required a wind power forecast. Another dispatching method introduced in [17] is based on a zero-phase low-pass filter, where the phase delay in the bandpass area of the filter is eliminated so that the BESS power can be reduced. However, the system performance is still dependent on the accuracy of wind power forecast. In addition, these methods are conceptually based on the time-coordination technique which is presented in different hybrid energy systems [18], [19]. Along with battery capacity optimization, another important challenge in the WBHPS is to control the BESS power and the dispatch power in the short-term. Luo et al. [20] introduced a coordinated operational dispatch scheme, where the dispatch power was set at the upper and lower levels of the wind power forecast depending on the state of charge (SOC) of the battery. In addition, Li et al. [21] introduced an SOC-based smoothing control strategy to satisfy the power fluctuation rate limit and to maintain the SOC within a desire range. Applying the modern controls such as the fuzzy-logic control [22] and the Wavelet transform combined with a model algorithm control [23] has been presented to control the WBHPS. Another advanced short-term power dispatch control was the online-SOC control proposed by Nguyen et al. [24]. Although these control algorithms are able to regulate the power dispatch to satisfy the constraints on the SOC and the BESS power, they require wind power forecast and need a large amount of computation time because of the complicated process. In this paper, an optimal power dispatch strategy based on FLF is presented to minimize the BESS capacity, and a shortterm power dispatch control strategy is developed for effective system operation. In order to minimize the battery capacity, we demonstrate a mathematical relationship between the FSTC and the FMR during each sampling time, so that the optimal solution can be directly computed. Furthermore, the short-term power dispatch control is also considered to regulate the SOC and the power of the BESS within safe operation ranges. The effectiveness of the proposed power management approach is verified through a case study using a 3-MW WT with real wind power data measured on Jeju Island. In addition, a wind-battery test bed is set up in the laboratory to experimentally validate the case study. II. WIND-BATTERY HYBRID POWER SYSTEM CONFIGURATION Fig. 1 illustrates the WBHPS, where the WT can be either fixed speed or variable speed types. The WT generator is

Comprehensive scheme of proposed power management system.

connected to the common dc link via the power converter system (PCS1) that regulates the WT generator to capture the available maximum power. The BESS is connected to the common dc link via the bidirectional power converter system (PCS2). When the storage is charged, its power is a positive value, i.e., Pb > 0 and vice versa. Under the assumption that power loss in the PCSs is negligible, the WBHPS output power Pd dispatching to the grid can be calculated from BESS output power Pb and wind power Pw as Pd = Pw − Pb . The WBHPS is managed by a power management system (PMS), which determines a suitable power command Pb∗ for the PCS2 based on the information of SOC of battery, wind power, and dispatch power to successfully attenuate the wind power fluctuation. In order to avoid the negative impacts of wind power fluctuation on the power system, the maximum fluctuation level of the dispatch power in a time window is limited by the grid code requirements [25]. The maximum fluctuation level of dispatch power Pd in a κ– min time window is calculated as follows: ΔFdκ (t) =

MAX {Pd (τ )} − MIN {Pd (τ )}

t−κ≤τ ≤t

t−κ≤τ ≤t

PW TR

(1)

where PW TR is the WT power rating. To satisfy the FMR, the maximum fluctuation level of dispatch power must be ΔFdκ (t) ≤ γκ – m in

(2)

where γκ – m in denotes the maximum power fluctuation limitation imposed by grid codes, which is set at a 10% fluctuation level in 10-min time window in this paper, i.e., ΔFd10 (t) ≤ 10%. III. PROPOSED APPROACH TO MINIMIZE THE BATTERY CAPACITY The definition of BESS capacity is related to how the wind power is dispatched to the grid. Therefore, to determine the minimal BESS capacity, the optimal power dispatch should be defined primarily. Fig. 2 shows the proposed PMS used in this paper. In order to satisfy the FMR, the dispatch power command Pd∗ is determined by passing the wind power through the FLF, and the difference between the wind power and the dispatch power command becomes the BESS power. Consequently, the smoothing time constant Tc significantly affects the fluctuation mitigation level of the dispatch power and the BESS capacity. Therefore, it is important to find an optimal FSTC not only to meet the FMR but also to minimize the BESS capacity. In Fig. 2, the power control signal Pc aims to regulate the SOC and BESS power within the desired operation ranges. A. Discrete Model of the FLF-Based Power Dispatch Strategy During determination of the optimal FSTC, we can neglect the power control variable Pc to simplify the system control

NGUYEN AND LEE: POWER MANAGEMENT APPROACH TO MINIMIZE BATTERY CAPACITY IN WIND ENERGY CONVERSION SYSTEMS

4845

Fig. 3. Relationship between the BESS capacity and the fluctuation level of dispatch power with respect to the filter smoothing time constant.

Fig. 5. Flowchart of searching the power control P c when the battery is getting full charge.

Fig. 4.

Principle of regulating γκ − m in to control the battery SOC.

model. In addition, if the PCSs well operate, their output powers meet their power command at steady state (i.e., Pb = Pb∗ and Pd = Pd∗ ). To implement the system digitally, the discrete model of the dispatch power, the BESS output power, and the BESS energy are derived by using the bilinear integration rule, and they are given as follows [26]: Pd [n] =

2Tc −Δt Δt Pd [n − 1]+ (Pw [n] + Pw [n − 1]) 2Tc +Δt 2Tc +Δt (3)

2Tc − Δt 2Tc Pb [n]= Pb [n − 1]+ (Pw [n]−Pw [n − 1]) 2Tc +Δt 2Tc + Δt (4) 2Tc − Δt η Δt Tc Eb [n − 1]+ (Pw [n] + Pw [n − 1]) 2Tc +Δt 2Tc +Δt (5) where Δt is the system sampling time and set at 5 s in this paper, η is the BESS charge and discharge efficiency, and n is the discrete index for the present sample. Based on the BESS power response expressed in (4), the BESS power fluctuation depends on two components: One is the FSTC, and another is the WT output power fluctuation. First, the range of the FSTC Tc is theoretically [0, +∞). When Tc = 0, the wind power is totally dispatched onto the grid, which means the BESS does not need to compensate any fluctuated wind power components. Meanwhile, when Tc is infinity, the dispatched power becomes a constant. In this case, the proposed method operates as the average method [10] or min–max method [11], and the BESS is required to compensate all fluctuated wind power components. Therefore, this is the worst case when the BESS power fluctuation is considered. So, it is reasonable to consider the system under the infinity Tc to examine the impact of the output wind power on the BESS power fluctuation. Based on the reports on the power spectrum density of wind power Eb [n] =

in [9] and [27], the dominant power spectra cluster over a lowfrequency band (lower than 5 cycles/h or 12-min period). That is reason why the sampling time Δt is usually set at several seconds to investigate the WBHPS. For example, Jiang and Wang [15] used 10 s to sample the wind power, or the sampling time of the system in [28] is one minute. In terms of battery power response, the battery can handle its power rating within a time interval of some seconds without impacting the battery lifetime [1], [29]. Therefore, the smoothing time constant is not limited during the system optimization. The dispatch power fluctuation level and the FMR in the continuous time region shown in (1) and (2) can be alternated, respectively, in the discrete model as ΔFdκ [n] =

MAX {Pd [i]} − MIN {Pd [i]}

n −L ≤i≤n

n −L ≤i≤n

(6)

PW TR

ΔFdκ [n] ≤ γκ−m in

(7)

where L is the total number of samples in one κ−min interval L = 60

κ . Δt

(8)

For the optimization of battery capacity, the historical wind speed data at the WF must be collected before making any investment or design decision in order to evaluate the wind power availability and its variation characteristics. In time duration T of the given wind data, the minimum requirement of the BESS capacity is defined with two terms [8]: 1) the power rating Pbrat ; and 2) the energy rating Ebrat as Pbrat = MAX |Pb [n]|

(9)

1≤n ≤N

Ebrat = MAX {Eb [n]} − MIN {Eb [n]} 1≤n ≤N

1≤n ≤N

(10)

where N is the total number of samples in T (hour) interval N = 3600

T . Δt

(11)

4846

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 5, SEPTEMBER/OCTOBER 2017

Fig. 6. Performance of the proposed and conventional power dispatch methods. (a) Wind power and dispatch power. (b) Fluctuation level of dispatch power. (c) The smoothing time constant of proposed method.

B. Optimization of the Filter Smoothing Time Constant Based on the discrete model of the FLF-based power dispatch strategy expressed in (3)–(5) and definitions of the BESS capacity rating in (9) and (10), a remark is noted here. Remark I: “The higher the FSTC (i.e., Tc ), the smoother the dispatch power obtained, but a higher BESS capacity is required.” Fig. 3 shows an example to confirm Remark I, where we investigate the fluctuation level of the dispatch power in 10-min time window and the BESS energy rating required in line with several FSTCs. It is noted that the fluctuation exponentially decreases, and the required BESS capacity linearly increases when the FSTC becomes higher. The investigation is based on a 3-MW WT and one month of real wind speed data. From Remark I, the optimal FSTC means the smallest FSTC, which leads to the power dispatch satisfying a predetermined level of the FMR. For easy determination of the optimal FSTC, we set 2Tc − Δt . (12) λ= 2Tc + Δt

Fig. 7. Wind power profile used to determine the required BESS capacity. (a) Wind power in one month of spring season. (b) Wind power in one month of summer season. (c) Wind power in one month of autumn season. (d) Wind power in one month of winter season.

Then, the control variable Pd defined in (3) becomes

Pd [n] =

1 − λ[n] (Pw [n] + Pw [n − 1]) + λ[n]Pd [n − 1]. 2 (13)

NGUYEN AND LEE: POWER MANAGEMENT APPROACH TO MINIMIZE BATTERY CAPACITY IN WIND ENERGY CONVERSION SYSTEMS

TABLE I BESS CAPACITY REQUIRED IN FOUR SEASONS Conventional

Spring Summer Autumn Winter

IV. DEVELOPMENT OF AN ONLINE SHORT-TERM POWER DISPATCH CONTROL

Proposed

TC (s)

Power Rating (MW)

Energy Rating (MWh)

Power Rating (MW)

Energy Rating (MWh)

3700 3400 2700 4100

2.015 1.888 1.630 2.116

2.716 2.623 2.174 3.153

1.321 1.212 1.098 1.669

1.795 1.447 0.888 2.233

From (13), the dispatch power fluctuation level during a sampling time can be derived as follows: ΔFdΔ t [n] =

|Pd [n] − Pd [n − 1]| PW TR

|Pw [n] + Pw [n − 1] − 2Pd [n − 1]| = (1 − λ[n]) . 2PW TR (14) It clearly demonstrates that the fluctuation of the dispatch power during a sampling time is attenuated with a high value of λ. If the maximum allowable limit of the dispatch power fluctuation during a sampling time is denoted as γΔ t , the maximum dispatch power fluctuation in one κ– min interval is MAX {ΔFdκ [n]} = LγΔ t .

(15)

In order to obtain the FMR (i.e., ΔFdκ [n] ≤ γκ− m in ), the maximum dispatch power fluctuation is given as following: LγΔ t = γκ− m in .

(16)

From (8) and (16), the maximum allowable limit of the dispatch power fluctuation during a sampling time becomes Δt . (17) 60κ In other words, the FMR defined in (7) by the grid code is obtained if the following constraint is satisfied: γΔ t = γκ− m in

ΔFdΔ t [n] ≤ γΔ t .

(18)

By substituting (14) and (17) into (18), the variable λ at the current sample must satisfy the following constraint: λ[n] ≥ 1 −

Δt γκ− m in PW TR . 30κ |Pw [n] + Pw [n − 1] − 2Pd [n − 1]|

(19)

From Remark I and the constraint (19), the FSTC at the current sample is optimally defined as λopt [n] = 1 −

4847

Δt γκ− m in PW TR . (20) 30κ |Pw [n] + Pw [n − 1] − 2Pd [n − 1]|

The optimal FSTC is derived by replacing the FMR in a κ− min time window to the FMR in a sampling time, so that the dispatch power meets the FMR in both time windows. Moreover, compared with the conventional optimization methods presented in [15] and [16], the proposed method requires no time consumption for the searching process because the optimal FSTC is directly pointed out as shown in (20).

By using the proposed power dispatch strategy and the longterm wind power data, the optimal system can be specified. In this section, we present an online short-term power dispatch control to effectively manage the system operation. During system operation in the short term, the PMS must regulate the dispatch power to meet the FMR in a time window under two conditions: the BESS power must be kept below its power rating, and the SOC must be maintained within a safe operation range. As shown in Fig. 2, the proposed PMS regulates the SOC and the power of the BESS through the FSTC Tc and the power control signal Pc . To prevent the battery from being damaged due to deep discharge or overcharged states, two following constraints must be satisfied during the system operation: 0 < SOC[n] ≤ 1.0

(21)

|Pb [n]| ≤ Pbrat .

(22)

To guarantee such constraints, we propose an online shortterm power dispatch control method. From (5), the net energy of the battery is obtained by replacing the variable Tc with λ λ[n] + 1 Δt (Pw [n] + Pw [n − 1]) . 4 (23) According to the energy response in (23), the SOC is calculated by dividing the battery energy to its energy rating value Eb [n] = λ[n]Eb [n − 1] + η

1+λ[n] Δt (Pw [n]+Pw [n−1]) . 4Ebrat (24) From (24), the SOC is an increasing function of the variable λ. This means the SOC can be regulated by adjusting λ in a corresponding manner. From (20), the optimal value of λ is a decreasing function of γκ− m in . As a result, a critical remark is deduced to control the SOC of the BESS. Remark II: “The lower the γκ− m in , the more energy stored in the battery. Vice versa, the higher the γκ− m in , the less energy stored in the battery.” From Remark II, the system can be controlled to satisfy the constraints in (21) and (22) by suitably adjusting γκ− m in . In order to meet the FMR, we have to guarantee that γκ− m in ≤ β, where β is the maximum allowable level of the FMR defined by the grid code. SOC[n] = λ[n]SOC[n − 1]+η

A. When the Battery Energy is Low: SOC < 0.3 The battery is operating under a critical condition that the battery energy is low; the control algorithm should inject power into the BESS to prevent the battery from reaching a deep discharged state. According to Remark II, we reduce γκ− m in on the principle that γκ− m in = 0 if SOC = 0, and γκ− m in = β if SOC = 0.3. The simplest approach to adjusting γκ− m in can be β SOC[n − 1]. (25) 0.3 Regulating γκ− m in with (25) results in γκ− m in not being over the maximum allowable level of the FMR. In other words, the γκ− m in [n] =

4848

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 5, SEPTEMBER/OCTOBER 2017

control algorithm guarantees that the power dispatch satisfies the FMR. Therefore, the power control Pc in this case does not need to be added, or Pc = 0. Once the variable γκ− m in is determined, λ is also computed by using (20). Afterward, the BESS power and the dispatch power are easily assigned by (3) and (4), respectively. B. When the Battery Energy is Under a Normal Condition: 0.3 ≤ SOC ≤ 0.85 Because the battery energy is under a normal condition, there is no need to regulate γκ− m in and Pc , and we assign γκ− m in = β and Pc = 0. This ensures that the dispatch power satisfies the FMR, and the required BESS capacity is minimized. C. When the Battery Energy is Close to Full: SOC > 0.85 In this case, the BESS should release the power to lessen its energy. According to Remark II, we increase γκ− m in . However, this increment can cause the dispatch power to fluctuate over the FMR. Therefore, in this case, the control algorithm also sets the FMR at its maximum allowable level, or γκ− m in = β. The regulation of γκ− m in to control the system in all cases is depicted in Fig. 4. It is seen that the FMR is set at the maximum allowable level over most of the SOC range; the required BESS capacity is minimized by using the proposed control algorithm. To guarantee the battery does not reach an overcharged state in this case, we regulate the power control signal Pc . It is desirable to release the BESS power, so a positive value is added to the dispatch power command, or Pc > 0. Basically, the battery energy is released faster if Pc is set at a high value. However, this might cause dispatch power to exceed the fluctuation limit, and the battery power might be higher than its rating. The process to search for the suitable power control Pc is shown in Fig. 5. Initially, the dispatch power is computed by the discrete FLF model. After one search step, the power control Pc is increased by ΔPc . The search process is terminated whenever one of the following three conditions is not satisfied: The first is the FMR defined in (7) with γκ− m in = β, the second is the constraint on the BESS power rating (i.e., |Pb [n]| ≤ Pbrat ), and the last is to ensure that the BESS is being discharged. As we can see, the proposed short-term power dispatch control is simple and straightforward. And, it does not require a wind power forecast. The searching process of the control signal Pc is basically a linear search, so it can be easily implemented. V. CASE STUDY In order to verify the proposed power dispatch strategy and the online SOC control during the short-term operation, we provide several numerical examples constructed using MATLAB software. Wind speed profiles measured at a WF on Jeju Island, South Korea with 5 s sampling time, i.e., Δt = 5 s, are converted into power data using a 3-MW WT model [30]. We utilized a Li-ion battery to construct the BESS, and the PCS regulates the battery charge and discharge current rate within 1–1.5 C. In this case, the charge and discharge efficiencies become almost 95% (η = 0.95) [31]. Additionally, the maximum

Fig. 8. BESS capacity with respect to different FMR levels. (a) BESS power rating with respect to different FMR levels. (b) BESS energy rating with respect to different FMR levels.

allowable power fluctuation level was set to 10% of the WT power rating in 10-min time window, i.e., κ = 10 min and β = 10 %. A. Evaluation of the Proposed Power Dispatch Strategy Performance of the proposed power dispatch strategy is shown in Fig. 6, in which the conventional method presented in [12] and [13] is also evaluated for comparison. In the conventional method, the FSTC is a fixed value during the system operation. The FMR in 10-min time window during the longterm dispatch was set at the maximum allowable level, i.e., γ10− m in = β = 10%. We investigate the WBHPS with wind speed data in one day, and the FSTC of the conventional dispatching method should be set at 2000 s. In Fig. 6(a), the wind power profile and the power dispatched by the proposed method and the conventional method are shown. The fluctuation of the dispatch power in 10-min time window is shown in Fig. 6(b). In addition, the optimal FSTC of the proposed method is depicted in Fig. 6(c). We can see that the optimal FSTC is varied so as to satisfy the FMR during system operation, and it is kept less than 2000 s applied to the conventional method. This results in a smaller BESS capacity with the proposed method to dispatch the wind power. To numerically demonstrate the effectiveness of the proposed methods, we investigated the WBHPS with the wind power data in four different seasons of a year: March for spring shown in Fig. 7(a), June for summer shown in Fig. 7(b), September for

NGUYEN AND LEE: POWER MANAGEMENT APPROACH TO MINIMIZE BATTERY CAPACITY IN WIND ENERGY CONVERSION SYSTEMS

4849

autumn shown in Fig. 7(c), and December for winter shown in Fig. 7(d). We can see that the wind power is essentially dependent on the season and highly fluctuated, and the fluctuation level of wind power in 10-min time window is up to 83.4% in March, 79.2% in June, 68.1% in September, and 87.9% in December. Using these data, the dispatch power is calculated from (3), where the smoothing time constant is a fixed value in the conventional dispatching method. Meanwhile, the smoothing time constant in the proposed dispatching method is optimized at each control sampling time as shown in (20). Next, the BESS power and energy response are calculated based on (4) and (5). Eventually, the BESS capacity can be defined from (9) and (10). Table I presents the results including the BESS capacity required by the conventional and proposed dispatching algorithms with respect to the season. In order to satisfy the FMR, the smoothing time constant Tc of the conventional dispatching method with respect to each season is defined based on the relationship between the smoothing time constant and the fluctuation level of the dispatch power as shown in Fig. 3. We can see that the winter season requires the largest Tc due to the highest fluctuation in wind power. Therefore, the largest BESS capacity is required in winter. Based on the results, the proposed method significantly reduces the BESS capacity compared with the conventional one in all seasons. In order to investigate the effect of the FMR levels on BESS capacity, we study several cases of maximum power fluctuation limitation γ10 – m in with the wind power data in winter. The results are shown in Fig. 8; both BESS power and energy ratings are firmly dependent on the FMR level. In detail, the wind system demands a larger BESS capacity when γ10− m in is smaller. These results are reasonable because the lower value of γ10− m in means the fluctuation of the dispatch power needs to be strictly mitigated. Compared with the conventional dispatching method, the proposed method significantly reduces the BESS power and energy ratings at any FMR level, which verifies the advantage of the proposed power dispatch strategy. B. Short-Term Power Dispatch Control To evaluate the system in terms of short-term power dispatch, we utilized three days of wind speed data. Based on Table I, a BESS where capacity is 2.0 MW and 2.3 MWh was installed to ensure that the WBHPS can dispatch stable power satisfying the FMR under all wind conditions. The maximum allowable fluctuation level of the dispatch power during short-term power control was set at 10% of the WT rating, i.e., β = 10%. Fig. 9 shows the system performance: the wind power and the dispatch power in Fig. 9(a), the battery power in Fig. 9(b), the SOC of battery in Fig. 9(c), and the power control Pc in Fig. 9(d). When the BESS is in the critical state (SOC > 0.85), the positive value of the power control Pc is added into the power dispatch to discharge the BESS, which prevents the BESS from being in overcharge state as shown in Fig. 9(c). The control algorithm determines the optimal value of the smoothing time constant which is shown in Fig. 9(e). We can see that the smoothing time constant is set at a high value when the wind power is highly fluctuated. Meanwhile, the smoothing time constant is

Fig. 9. Simulated results of proposed system during short-term power dispatch. (a) Wind power and dispatch power. (b) BESS power. (c) SOC of BESS. (d) Power control P c . (e) Updated smoothing time constant. (f) CDF of maximum fluctuation of wind power and dispatch power.

4850

Fig. 10.

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 5, SEPTEMBER/OCTOBER 2017

Configuration of the test bed.

set at a low value when the wind power fluctuation is low. To investigate the fluctuation level of the dispatch power and the wind power, the cumulative distribution function (CDF) of the maximum fluctuation in 10-min time window of the wind power and dispatch power is shown in Fig. 9(f). Based on the CDF of the maximum fluctuation in 10-min time window of the dispatch power, we can see that the dispatch power fluctuation level is kept less than 10% to meet the FMR, despite the fact that the wind power fluctuation level is up to 64%. In addition, the proposed control algorithm is able to maintain the BESS power below the rating while the SOC is kept in a safe range. These performances verify the effectiveness of the proposed control algorithm.

Fig. 11.

Control scheme for the WECS and BESS in the test bed.

Fig. 12.

Control scheme for the inverter in the test bed.

VI. EXPERIMENTAL VALIDATION In order to evaluate the proposed management system, a 1 kW wind-battery hybrid power test bed is set up in the laboratory [32]. Due to the limited conditions, the BESS power rating in the test bed is scaled down compared with that in the case study in Section V-B; in the test bed, a 72 V Li-ion battery with 0.5 kW power rating and 0.75 kWh energy rating is utilized. In spite of the downscaled test bed, it may be sufficient to evaluate the performance of the proposed control system because its control algorithms are perfectly the same as those used in the case study. As shown in Fig. 10, the DSP TMS320F28335 is utilized to realize the control algorithms, and its serial communication interface module is used to share the required data between the DSP and a laptop where the short-term control algorithm is executed. A. Wind Power and BESS Control Sides In the test bed, a battery emulator with the model pCUBE MWBFP3-1250 from Myway Company is used to emulate the BESS operation, and a programmable dc power source with the model NF-DH40012 is utilized as a wind energy conversion system (WECS). Fig. 11 shows the WECS and the BESS control scheme. The wind power reference Pw∗ as the desired WECS output power in the test bed is defined by the maximum power point tracking algorithm from the wind speed information. From the dc link voltage and the wind power references, the WECS

output current reference Iw∗ is obtained, and, finally, the boost converter duty cycle dw is determined to control the WECS through the PI controller. In order to control the switches in the PCS2 in the BESS control side, two gating signals are generated from the duty cycle db . To define db , the BESS current reference is determined from the power reference Pb∗ and the dc link voltage reference. B. Inverter Control Side The dispatch power is controlled by a the-phase voltage source inverter including an LCL filter in the PCS3. The test bed inverter scheme is shown in Fig. 12, where the dc link voltage is kept constant to guarantee that the overall system operates under the balanced power condition among the dispatch power, the wind power supply, and the BESS power. To improve the dynamic performance, the desired q-component of the load current is obtained from the dc link voltage by adding a feed-forward current component IqFF , which is defined by using the BESS power and the wind power references  Pw∗ − Pb∗ FF (26) Iq = 1.5Ra

NGUYEN AND LEE: POWER MANAGEMENT APPROACH TO MINIMIZE BATTERY CAPACITY IN WIND ENERGY CONVERSION SYSTEMS

Fig. 13. Dynamic response of the wind-battery test bed. (a) Output current of WECS and BESS, dc link voltage, and load current. (b) DC link voltage and the load current at the moment t1 . (c) DC link voltage and load current at the moment t2 .

where Ra means the phase load. From the d- and q-components of the voltage vector, the six gating signals for the inverter are generated with the aid of the space vector pulse width modulation method. C. Experimental Results In Fig. 13, the dc link voltage is controlled to be a constant value with 145 V. Fig. 13(a) shows the dynamic response, where the BESS output current (Ib ), the WECS output current (Iw ), the dc link voltage (Vdc ), and the load current in phase A (ia )

4851

are plotted from top to bottom. In Fig. 13(a), the wind power and the load power reference have been 300 W and 400 W, respectively, before the time. Therefore, the BESS should supply 100 W, which means –100 W, to satisfy the desired load power before t0 . We can see that the output currents of the BESS and the WECS are successfully controlled to satisfy the load power commands. In order to find the dynamic response, the load power decreases from 400 to 200 W at t0 , while the wind power is still kept at 300 W. In this case, the BESS must be charged to absorb the surplus power 100 W. Subsequently, at t1 , the wind power increases from 300 to 600 W, and also, the load power reference is increased from 200 to 400 W; 200 W must be charged into the BESS, that makes the output current of the BESS change to 1.4 from 0.7 A. At t2 , the wind power decreases from 600 to 400 W, and the load power reference is decreased from 400 to 200 W; the BESS power is kept at 200 W. From Fig. 13(a), we can say that the test bed has a good dynamic performance. In addition, Fig. 13(b) and (c) show the dc link voltage and the load current when the wind power and load power references are changed at the moments t1 and t2 . It can be seen that the dc link voltage is kept constant regardless of the wind power and load power variation. Fig. 14 shows the experimental performances of the proposed management control strategy in one day. The reference of the wind power is randomly generated within the test bed power rating. The actual wind power response is shown in Fig. 14(a), and the CDF of the maximum power fluctuation in 10-min time window of wind power is shown in Fig. 14(b). We can see that the fluctuation level of the wind power is up to 40%, which prevents the wind power from dispatching into the grid. With the aid of the BESS, the dispatch power is stabilized to meet the FMR as shown in Fig. 14(a) and (b). The smoothing time constant is regulated and its updated value is shown in Fig. 14(c). Thanks to the proposed power management control strategy, the BESS power is maintained below its rating while the SOC is kept in a safe range as demonstrated in Fig. 14(d) and (e). To verify the control schemes for the PCSs, the dc link voltage and the BESS voltage are shown in Fig. 14(f). Both voltages are well stable at 145 and 73.5 V, which prove the good performance of the PCSs control schemes. As a result, these experimental performances verify the effectiveness of the proposed control algorithm. It is noted that the initial SOC of BESS in Fig. 14 is set at 50%, which may be a preferable condition for the system at the initial moment. To verify the performance of the proposed control method under any SOC conditions, we carry out the experiments with two critical initial SOCs, i.e., almost full charge and empty conditions. Fig. 15 shows the system performance with the full charge state where the initial SOC is set at 90% (SOC0 = 0.9). The wind power is generated randomly with a maximum fluctuation in 10-min time window up to 61% as shown in Fig. 15(a) and (b). Although, the wind power is highly fluctuated, the dispatch power is stabilized with a maximum fluctuation lower than 10% as shown in Fig. 15(b). To obtain the desired FMR, the smoothing time constant is regulated, and its updated value is shown in Fig. 15(c). Due to the high SOC in beginning, the power control Pc in Fig. 15(d) must be added to the dispatch power to reduce the BESS energy. Therefore, the SOC is kept

4852

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 5, SEPTEMBER/OCTOBER 2017

Fig. 14. Experimental performance of the proposed control strategy. (a) Wind power and dispatch power. (b) CDF of maximum fluctuation of wind power and dispatch power. (c) Updated smoothing time constant T c . (d) Power of BESS. (e) SOC of BESS. (f) DC link voltage and battery voltage.

Fig. 15. Experimental performance of the proposed control strategy when the BESS is near full charge state at initial moment. (a) Wind power and dispatch power. (b) CDF of maximum fluctuation of wind power and dispatch power. (c) Updated smoothing time constant T c . (d) Power control Pc . (e) Power of BESS. (f) SOC of BESS. (g) DC link voltage and battery voltage.

NGUYEN AND LEE: POWER MANAGEMENT APPROACH TO MINIMIZE BATTERY CAPACITY IN WIND ENERGY CONVERSION SYSTEMS

4853

in a safe range while the BESS power is maintained below its rating as demonstrated in Fig. 15(e) and (f). To verify the control schemes for the PCSs, the dc link voltage and the BESS voltage are shown in Fig. 15(g). The dc link voltage is well stable at 145 V even though the BESS voltage is significantly changed due to the large variation of the SOC. Another critical condition with a deep discharge state is tested with the 10% initial SOC (SOC0 = 0.1), and the system performance is shown in Fig. 16. The wind power is generated randomly with a maximum fluctuation in 10-min time window up to 35% as shown in Fig. 16(a) and (b). Although, the wind power is highly fluctuated, the dispatch power is stabilized with a maximum fluctuation lower than 10% as shown in Fig. 16(b). To obtain the desired FMR, the smoothing time constant is regulated as shown in Fig. 16(c). Due to the low SOC at beginning, the smoothing time constant becomes a big value up to 5000 s at the beginning time to keep the dispatch power almost zero. This leads the BESS to be charged with all wind power. Therefore, the SOC reaches the normal range as shown in Fig. 16(e). In addition, the BESS power is maintained below its rating as demonstrated in Fig. 16(d). Based on the experimental results, we can say that the proposed system works very well to satisfy all requirements regardless of the wind power fluctuation level as well as the initial SOC condition. VII. CONCLUSION In this paper, we propose an optimal power control strategy to minimize the BESS capacity in WECSs. The dispatch power is determined based on an optimal FLF where the smoothing time constant is optimized in each sampling time. The proposed method significantly reduces the BESS capacity compared with the conventional dispatching methods. Additionally, the shortterm power dispatch control is also considered to regulate the SOC and the battery power within desirable operation ranges. The proposed power dispatching method defines the optimal FLF by directly computing the smoothing time constant, so that it is simple and easy to implement; the proposed method is a feasible solution to manage a real WBHPS. In the shortterm power dispatch, the control algorithm can keep the battery power below its rating and can maintain the SOC in a safe range to prevent the battery from being overcharged or too deeply discharged. To evaluate the performance of the proposed power control approach, we numerically analyzed a 3-MW WT generator using real wind power data, and a wind-battery test bed is set up in the laboratory. The simulated and experimental results verify that the proposed approach can be applied to real systems. REFERENCES Fig. 16. Experimental performance of the proposed control strategy when the BESS is closed to deep discharge state at initial moment. (a) Wind power and dispatch power. (b) CDF of maximum fluctuation of wind power and dispatch power. (c) Updated smoothing time constant T c . (d) Power of BESS. (e) SOC of BESS. (f) DC link voltage and battery voltage.

[1] H. Zhao, Q. Wu, S. Hu, H. Xu, and C. N. Rasmussen, “Review of energy storage system for wind power integration support,” Appl. Energy, vol. 137, pp. 545–553, Jan. 2015. [2] C. Luo and B. Ooi, “Frequency deviation of thermal power plants due to wind farms,” IEEE Trans. Energy Convers., vol. 21, no. 3, pp. 708–716, Sep. 2006.

4854

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 5, SEPTEMBER/OCTOBER 2017

[3] S. M. Muyeen, R. Takahashi, T. Murata, J. Tamura, and M. H. Ali, “Application of STATCOM/BESS for wind power smoothening and hydrogen generation,” Elect. Power Syst. Res., vol. 79, no. 2, pp. 365– 373, Feb. 2009. [4] Z. Qin, M. Liserre, F. Blaabjerg, and H. Wang, “Energy storage system by means of improved thermal performance of a 3 MW grid side wind power converter,” in Proc. 39th Annu. Conf. IEEE Ind. Electron. Soc., Nov. 2013, pp. 736–742. [5] N. Kawakami et al., “Development and field experiences of stabilization system using 34 MW NaS batteries for a 51 MW wind farm,” in Proc. 2010 IEEE Int. Symp. Ind. Electron., Jul. 2010, pp. 2371–2376. [6] Experience of Jeju smart grid test-bed & national strategy. [Online]. Available: www.ifema.es/ferias/genera/2013/jornadas/ree28/man_geun.pdf [7] X. Y. Wang, D. V. Mahinda, and S. S. Choi, “Determination of battery storage capacity in energy buffer for wind farm,” IEEE Trans. Energy Convers., vol. 23, no. 3, pp. 868–878, Sep. 2008. [8] X. Wang, M. Yue, E. Muljadi, and W. Gao, “Probabilistic approach for power capacity specification of wind energy storage systems,” IEEE Trans. Ind. Appl., vol. 50, no. 2, pp. 1215–1224, Mar./Apr. 2014. [9] D. L. Yao, S. S. Choi, K. J. Tseng, and T. T. Lie, “Determination of short-term power dispatch schedule for a wind farm incorporated with dual-battery energy storage scheme,” IEEE Trans. Sustain. Energy, vol. 3, no. 1, pp. 74–84, Jan. 2012. [10] S. Teleke, M. E. Baran, S. Bhattacharya, and A. Q. Huang, “Optimal control of battery energy storage for wind farm dispatching,” IEEE Trans. Energy Convers., vol. 25, no. 3, pp. 787–794, Sep. 2010. [11] C. L. Nguyen, T. W. Chun, and H. H. Lee, “Determination of the optimal battery capacity based on a life time cost function in wind farm,” in Proc. 2013 IEEE Energy Convers. Congr. Expo., Sep. 2013, pp. 51–58. [12] K. Yoshimoto, T. Nanahara, and G. Koshimizu, “New control method for regulating state-of-charge of a battery in hybrid wind power/battery energy storage system,” in Proc. IEEE Power Syst. Conf. Expo., Nov. 2006, pp. 1244–1251. [13] Y. Zhou, Z. Yan, and N. Li, “A novel state of charge feedback strategy in wind power smoothing based on short-term forecast and scenario analysis,” IEEE Trans. Sustain. Energy, vol. 8, no. 2, pp. 870–879, Apr. 2017. [14] L. Wei, G. Joos, and J. Belanger, “Real-time simulation of a wind turbine generator coupled with a battery supercapacitor energy storage system,” IEEE Trans. Ind. Electron., vol. 57, no. 4, pp. 1137–1145, Apr. 2010. [15] Q. Jiang and H. Wang, “Two-time-scale coordination control for a battery energy storage system to mitigate wind power fluctuations,” IEEE Trans. Energy Convers., vol. 28, no. 1, pp. 52–61, Mar. 2013. [16] Q. Jiang, Y. Gong, and H. Wang, “A battery energy storage system duallayer control strategy for mitigating wind farm fluctuations,” IEEE Trans. Power Syst., vol. 28, no. 3, pp. 3263–3273, Aug. 2013. [17] C. L. Nguyen and H. H. Lee, “Optimization of wind power dispatch to minimize energy storage system capacity,” J. Elect. Eng. Technol., vol. 9, no. 3, pp. 1080–1088, May 2014. [18] A. K. Kaviani, B. Hadley, and B. Mirafzal, “A time-coordination approach for regenerative energy saving in multiaxis motor-drive systems,” IEEE Trans. Power Electron., vol. 27, no. 2, pp. 931–941, Feb. 2012. [19] W.-S. Im, C. Wang, L. Tan, W. Liu, and L. Liu, “Cooperative controls for pulsed power load accommodation in a shipboard power system,” IEEE Trans. Power Syst., vol. 31, no. 6, pp. 518–5189, Nov. 2016. [20] F. Luo, K. Meng, Z. Y. Dong, Y. Zheng, Y. Chen, and K. P. Wong, “Coordinated operational planning for wind farm with battery energy storage system,” IEEE Trans. Sustain. Energy, vol. 6, no. 1, pp. 253–262, Jan. 2015. [21] X. Li, D. Hui, and X. Lai, “Battery energy storage station (BESS)-based smoothing control of photovoltaic (PV) and wind power generation fluctuations,” IEEE Trans. Sustain. Energy, vol. 4, no. 2, pp. 464–473, Apr. 2013. [22] X. Li, “Fuzzy adaptive kalman filter for wind power output smoothing with battery energy storage system,” IET Renew. Power Gen., vol. 6, no. 5, pp. 340–347, Sep. 2012. [23] X. Han, F. Chen, X. Cui, Y. Li, and X. Li, “A power smoothing control strategy and optimized allocation of battery capacity based on hybrid storage energy technology,” Energies, vol. 5, no. 5, pp. 1593–1612, May 2012. [24] C. L. Nguyen, H. H. Lee, and T. W. Chun, “Cost optimized battery capacity and short-term power dispatch control for wind farm,” IEEE Trans. Ind. Appl., vol. 51, no. 1, pp. 595–606, Jan. 2015.

[25] M. Tsili and S. Papathanassiou, “A review of grid code technical requirements for wind farms,” IET Renew. Power Gen., vol. 3, no. 3, pp. 308–332, Sep. 2009. [26] C. L. Nguyen and H. H. Lee, “Optimization of power dispatch to minimize battery storage capacity in wind farm,” in Proc. 2013 IEEE Energy Convers. Congr. Expo., Sep. 2014, pp. 420–427. [27] S. Kennedy and P. Rogers, “A probabilistic model for simulating long-term wind-power output,” Wind Eng. J., vol. 27, no. 3, pp. 167–181, 2003. [28] Q. Li, S. S. Choi, Y. Yuan, and D. L. Yao, “On the determination of battery energy storage capacity and short-term power dispatch of a wind farm,” IEEE Trans. Sustain. Energy, vol. 2, no. 2, pp. 148–158, Apr. 2011. [29] F. D. Gonz´alez, A. Sumper, O. G. Bellmunt, and V. Robles, “A review of energy storage technologies for wind power applications,” Renew. Sustain. Energy Rev., vol. 16, no. 4, pp. 2154–2171, May 2012. [30] V112-3.0 MW. [Online]. Available: http://www.vestas.com/Files/Filer/ EN/Brochures/Productbrochure_V112_Offshore_UK.pdf [31] J. Kang, F. Yan, P. Zhang, and C. Du, “Comparison of comprehensive properties of Ni-MH (nickel-metal hydride) and Li-ion (lithium-ion) batteries in terms of energy efficiency,” Energy, vol. 70, no. 6, pp. 618–625, Jun. 2014. [32] C. L. Nguyen and H. H. Lee, “A novel dual-battery storage system for wind power applications,” IEEE Trans. Ind. Electron., vol. 63, no. 10, pp. 6136–6147, Oct. 2016.

Cong-Long Nguyen (M’16) received the B.S. degree from the Ho Chi Minh City University of Technology, Ho Chi Minh City, Vietnam, in 2010, and the Ph.D. degree from the University of Ulsan, Ulsan, South Korea, in 2016, both in electrical engineering. He was at Intel Corporation, Vietnam, as a Production Engineer to set up the CPU and chipset test systems. He is currently a Postdoctoral Fellow with ´ the Department of Electrical Engineering, Ecole de technologie sup´erieure, University of Quebec, Montreal, QC, Canada. His research interests include applications of energy storage devices in renewable energy conversion systems, power electronics in electric vehicles, power quality control, dc and ac microgrids, and microbial fuel cells.

Hong-Hee Lee (SM’11) received the B.S., M.S., and Ph.D. degrees in electrical engineering from Seoul National University, Seoul, South Korea, in 1980, 1982, and 1990, respectively. From 1994 to 1995, he was a Visiting Professor at Texas A&M University, College Station, TX, USA. Since 1985, he has been with the Department of Electrical Engineering, University of Ulsan, Ulsan, South Korea, where he is currently a Professor in the School of Electrical Engineering. He is also the Director of the Network-Based Automation Research Center, which is sponsored by the Ministry of Trade, Industry, and Energy. His research interests include power electronics, network-based motor control, and renewable energy. Dr. Lee is a member of the Korean Institute of Power Electronics (KIPE), the Korean Institute of Electrical Engineers, and the Institute of Control, Robotics, and Systems. He was the President of KIPE in 2014.