OPTIMAL CONTROL STRATEGY FOR ENERGY STORAGE SYSTEM PARTICIPATING IN AUTOMATIC GENERATION CONTROL

Transcription

1 OPTIMAL CONTROL STRATEGY FOR ENERGY STORAGE SYSTEM PARTICIPATING IN AUTOMATIC GENERATION CONTROL Presenting Author: Xu XIE, North China Branch of SGCC, , Xu XIE, North China Branch of SGCC, , Haocheng LUO, Tsinghua University, , Siqing NIU, North China Branch of SGCC, , Zhe ZHANG, North China Branch of SGCC, , Zechun HU, Tsinghua University, , ABSTRACT With the merits of high ramp rate and short response time, a large-scale energy storage system (ESS) can reduce the area control error (ACE) effectively while participating in automatic generation control (AGC). Within a control area that both traditional thermal generators (s) and ESS contribute to AGC, the control strategy should be designed elaborately to avoid the waste of regulation capability or low available time of ESS. This paper proposes an optimal control strategy for ESS participating in AGC based on the current AGC practice, aiming at improving the frequency regulation performance and reducing the regulation burden supported by ESS. Considering the capacity limit and delay time of ESS, the proposed strategy shows good regulation performance when tested in real power system. The effectiveness of the proposed strategy is illustrated by the comparisons with other strategies. KEYWORDS: AUTOMATIC GENERATION CONTROL, ENERGY STORAGE SYTEM, PUMPED STORAGE SYSTEM, OPTIMAL CONTROL STRATEGY 1 INTRODUCTION With the increasing public awareness of the need for sustainable energy solutions, there has been rapid growth of renewable energy generation during the last two decades. However, power system are facing more challenges in frequency control due to the fluctuating power and weak inertia of renewable energy generation [1], [2]. As part of frequency control of power system, automatic generation control (AGC), also referred to as secondary frequency control, is also affected. Since the fluctuation of generation increases with increasing penetration of renewable energy, the costs for AGC to maintain the balance between generation and load increases [3]. To cope with this problem, admission of energy storage system (ESS) into AGC is suggested [4], [5]. Comparing with a thermal generator (), an ESS has the merits of high ramp rate and short response time. While participating in AGC, a large-scale ESS can reduce the area control error (ACE) effectively by responding to the regulation command sending from the system operator promptly and accurately [3], [6]. However, within a control area that both traditional s and ESS contribute to AGC, traditional control strategy may result in waste of regulation capability or low available time of ESS [7]. Thus, a control strategy should be designed to allocate appropriate regulation requirement respectively to s and ESS, which takes advantage of them. Recent literature has investigated the control strategy for ESS participating in AGC. Reference [8] proposed a fast algorithm to allocate regulation requirement based on Linear Quadratic Regulator (LQR). In this algorithm, ESS was regarded as a fast source and got a separate signal from the control center to take advantage of its fast response. However, the high participation of ESS in regulation may degrade its available time, which is critical to good performance of frequency regulation. In [9], a fuzzy gain scheduled proportional-integral (PI) controller is presented for the AGC of a two-area multi-unit power system with ESS. With the presented mechanism, ESS exert its full capability to suppress ACE fluctuation, which may lead to fast battery degradation. ESS has been integrated into AGC in some real power systems with specially designed strategy. In PJM, an independent system operator dispatch the transmission power system of 13 states including Pennsylvania, Ohio, and Virginia, the frequency regulation signals include a low-frequency AGC signal and a high-frequency AGC signal [10]. ESS is responsible to track the high-frequency signal, while s is responsible to track the low-frequency signal. Although the strategy allocates regulation requirement based on the characteristics of ESS and s, low-frequency signal and high-frequency signal sometimes may conflict. In this case, some regulation actions may be unnecessary, or even harmful to frequency regulation performance. Aiming at improving the frequency regulation performance and reducing the regulation burden provided by ESS, this paper proposes an optimal AGC control strategy based on the current practice. The proposed strategy minimizes the active power mismatch which cannot be balanced by s and ESS while guarantees the available time of ESS. The feasibility and effectiveness of proposed strategy 1

2 are verified by experiments in a real power system. Comparisons with other strategies show the superiority of proposed strategy. The rest of this paper is organized as follows. An optimal control strategy for ESS participating in AGC is proposed in Section 2. In Section 3, the experiments system is introduced. In Section 4, experiments and results is described to illustrate the validity of proposed strategy. 2 OPTIMAL CONTROL STRATEGY AGC is usually executed by the transmission system operator (TSO), with a basic objective to regulate area frequency. In addition, AGC should also maintain the tie-line interchange power with neighboring control areas at scheduled values in a multi-area power system. Hence, the linear combination of frequency deviation and tie-line power deviation, also known as ACE, is calculated by the TSO: ACE = BΔ f +Δ P tie (1) where B is the frequency bias factor, Δ f is the frequency deviation, and Δ Ptie is the total tie-line power deviation with other areas. Then, the area regulation requirement (ARR) is further calculated. For the existing AGC system, PI controller is often employed to estimate the ARR due to its high reliability and robustness. The ARR is usually allocated by the proportion of regulation capacity or ramp rate of regulating units, which is called proportional allocation strategy (PAS). However, this widely used AGC strategy does not consider of the fast response speed, relatively small power capacity, and limited stored energy of ESS. Hence, an optimal control strategy is proposed to improve the frequency regulation performance taking into account different characteristics of ESS and s. 2.1 Thermal generator first strategy For the AGC system of a real power system, it is proposed that the ARR π AGC is allocated to s and ESS respectively, denoted as π, π ESS. Considering the constraints of ramp rate and regulation capacity of s and ESS, the power output changes of s and ESS within a AGC control cycle can be estimated, which are denoted as Δ p and Δ pess, respectively. To improve the frequency regulation performance, the regulation mismatch which cannot be covered by s and ESS should be minimized: min π Δ p +Δ p (1) subject to where and p0 ESS 0 Δ p π, πess ( ) AGC ESS π = π + π (2a) AGC ESS { p } min p0 + π, p0 + rδ, p 0, π > 0 = max { p0 + π, p0 rδ, p} p0, π 0 { p } min pess 0 + πess, pess 0 + ress δ, ESS pess 0, πess > 0 Δ pess = (2c) max { pess 0 + πess, pess 0 ressδ, pess } pess 0, πess 0 p respectively denote the output power of s and ESS at the beginning of the AGC cycle, r and r ESS denote the ramp rate of s and ESS, respectively. Considering the high ramp rate of ESS, it is assumed that r ESS r.the parameters p ( p ESS ) and p ( p ESS ) are the upper and lower limits of regulation capacity of s (ESS), respectively. δ is the time interval of an AGC cycle. Despite the merits of high ramp rate and short response time, the regulation capacity of ESS is usually much smaller than s. Once the output power of ESS reaches upper or lower limit of its regulation capacity, ESS cannot participate in bidirectional frequency regulation. In addition, sustained charging/discharging of ESS may reduce its available time for regulation because of its limited energy storage capacity. All of these factors are adverse to the performance of AGC. Hence, another objective to reduce the regulation power supported by ESS is considered: min Δ (3) π, πess Mathematical derivation from (1)-(3) results in an optimal control strategy to allocate regulation requirement: p ESS (2b) 2

3 π { p } min p0 + πagc, p 0 + r δ, p 0, πagc > 0 = (4a) max { p0 + πagc, p0 rδ, p} pt G0, πagc 0 π = π π (4b) ESS AGC Equation (4) can be interpreted intuitively as: allocate regulation requirement to s first considering their ramp rates and regulation capacities, then allocate the regulation requirement which cannot be provided by s to ESS. This strategy is so-called thermal generator first strategy (FS). 2.2 Complementary logic In the AGC system of a real power system, a dead band of ACE is usually set to avoid overregulation of regulation sources. When the ACE lies in the dead band, the ARR is set to zero, which means the regulation sources should stay at the current operating points. As to ESS, staying at the current operating point deviated from the base point persistently may lead to lower regulation capacity and lower available time. Hence, a complementary logic is added to FS. With this complementary logic, the available regulation capacity and available time of ESS can be improved. Complementary logic-1: Once the ACE lies in the dead band, the regulation command for ESS should be set at its base point, instead of the current target point. Although the response time of ESS is short, the delay time in communication and control loop can dramatically affect the regulation performance of ESS. For example, if the ACE enters the dead band from a negative value, the output power of ESS may still climb (when the last AGC command to ESS is power increase) due to the delay. In this case, ESS with high ramp rate can aggravate ACE fluctuation. Moreover, the complementary logic for ESS to return to base point may sometimes be adverse to the regulation of ACE. For example, when the ACE enters the dead band from a negative value, the complementary logic would be implemented. Once the output power of ESS is below the base point, ESS would increase its output power to return to base point. If the power output change of ESS is large enough, ACE may increase rapidly to a positive value out of the dead band. Thus the control of ESS results in unnecessary fluctuation of ACE. Due to the factors discussed above, the following complementary logic is further proposed to improve FS. Complementary logic-2: Once the ACE lies in the dead band, estimate if the regulation command to return to base point contributes to suppress ACE fluctuation. If so, the regulation command for ESS should be set at its base point; if not, the regulation command for ESS should be set at its current operating point. With complementary logic-2, the return-to-base-point command which can aggravate ACE fluctuation will not be implemented. Furthermore, the overshoot due to delay of ESS response can be restrained as far as possible. 3 EXPERIMENTAL SYSTEM To verify the feasibility and effectiveness of proposed strategy, experiments of AGC control strategies are conducted on a real power system. 3.1 System description The test system is a real power system in north China, with more than 120 generating units controlled by the control center. To maintain the balance between generation and load, the control center chooses some generating units to take part in AGC (according to their regulation performance), while the other generating units are dispatched to follow their scheduled outputs. Most of the generating units are s, while a pumped storage system (PSS) is also included in this system. The basic parameters of the power system are listed in Table 1. Table 1 Parameters of the power system Parameters Values Parameters Values Inertia constant(mw s) Frequency bias factor(mw/hz) 3400 Load damping coefficient(mw/hz) 2200 AGC cycle(s) 4 Regulation capacity of s(mw) 1000 Proportional gain -1 Regulation capacity of ESS(MW) 100 Integral gain Regulation performance of PSS Among the various storage technologies, battery storage system (BSS) and pumped storage system 3

4 (PSS) widely participate in power system dispatch now. However, the high investment cost of BSS restricts its application in the AGC of bulk power systems. The widely used and reliable large-scale PSS, whose ramp rate is lower than BSS, but still much higher than s, can participate in AGC as ESS. To confirm if PSS could be regarded as a fast frequency regulation resource, we conducted an experiment to identify its regulation performance. In this experiment, the output power of PSS with a rated power of 400 MW is set to climb from 300 MW to 400 MW. The target output and actual output are illustrated in Figure 1. It can be seen that the total time for the PSS to climb from 300 MW to 400 MW is about 45 seconds. During this process, the PSS receives its new target point at about 12 35, but the actual output stays at the original point due to the delay time in control loop; PSS starts to climb at about and ends to climb at about Hence, the ramp rate of ESS can be calculated as 33%/min even with the delay. On the other hand, the rated ramp rate of s is only stipulated as 1%/min ~ 2%/min. Thus, it is reasonable to regard ESS as a fast frequency regulation resource. Actual Output Target Output Power[MW] Time Figure 1. Target curve and actual output curve of a 400MW PSS 4 EXPERIMENTS AND RESULTS 4.1 Verification of complementary logic Since the complementary logic is critical to avoid the adverse actions, its implementation should be verified first. Figure 2 is given to illustrate effectiveness of the complementary logic-2. In the red circles of Figure 2, ACE enters the dead band from a positive value, while the output power of ESS is above the base point. In this case, the return-to-base-point command for ESS can aggravate ACE fluctuation. Hence, complementary logic-2 is implemented to avoid this adverse action, as the regulation command for ESS is to maintain its current operating point. In the blue circles, ACE turn to negative from a positive value, while the output power of ESS is still above the base point. In this case, the return-to-base-point command for ESS can suppress ACE fluctuation. Hence, complementary logic-2 is implemented to guarantee the available regulation capacity and available time of ESS, as the regulation command for ESS is set as base point. From the above analyses, it can be seen that complementary logic-2 can be implemented in the AGC system to improve the availability of ESS and avoid the adverse regulation actions. 4

5 Actual Output Target Output Base Point ACE Positive Dead Band of ACE Negative Dead Band of ACE Power[MW] Stay at the current operating point Return to base point ACE '00'' 33'05'' 33'10'' 33'15'' 33'20'' 33'25'' 33'30'' 33'35'' 33'40'' 33'45'' 33'50'' 33'55'' 34'00'' 34'05'' 34'10'' 34'15'' 34'20'' 34'25'' 34'30'' 34'35'' 34'40'' 34'45'' 34'50'' 34'55'' 35'00'' 35'05'' 35'10'' 35'15'' 35'20'' 35'25'' 35'30'' 35'35'' 35'40'' 35'45'' 35'50'' 35'55'' Time Figure 2. Implementation of Complementary Logic Experiments and results The experiments duration is from 14:00 p.m. to 18:00 p.m. on 24 th, May, Besides FS with complementary logic-2, two more strategies are implemented during the experiment time for the purpose of comparisons: PAS and FS with complementary logic-1. For PAS, the regulation requirement is allocated by the proportion of regulation capacity of s and ESS. The root mean square values (RMS) of ACE with different strategies are listed in Table 2. It can be seen that the participation of ESS as a fast resource improves the performance of AGC and the FS with complementary logic-2 is more beneficial for suppressing ACE fluctuation, compared with the other strategies. The reason for the superiority of FS with complementary logic-2 can be summarized as follows. 1) With FS, ESS provides the regulation power which cannot be provided by s to minimize the regulation mismatch when regulation requirement is large. Whereas, PAS never considers the ramp rate of s and ESS, which results in the waste of regulation capacity. 2) With complementary logic-2, FS can both guarantee the availability of ESS and avoid the adverse actions, while FS with complementary logic-1 neglects the adverse actions which may sometimes aggravate ACE fluctuation. Table 2 Regulation performance of AGC with different strategies Experiment time 14:10-15:10 15:20-16:20 16:30-17:10 FS with FS with Control strategy PAS complementary logic-1 complementary logic-2 RMS of ACE Conclusions In this paper, we propose an optimal control strategy for ESS participating in AGC based on the current practice. With the reasonable consideration of capacity limit and delay time of ESS, the proposed strategy aims at improving the frequency regulation performance and reducing the regulation power supported by ESS. Experiments are conducted to show the feasibility and effectiveness of proposed strategy. Experiments results show that a large-scale PSS with the merits of high ramp rate and short response time can reduce ACE effectively while participating in AGC as a fast resource. Moreover, the proposed strategy shows best regulation performance among the strategies implemented in the AGC system of a real power system. For future work, the differences of characteristics among s should be also taken into consideration to make better use of them. In addition, another interesting topic is to consider the costs for ESS and s participating in AGC. Acknowledgment This work was supported by State Grid of China (No. SGHB0000DKJS ). Reference [1] R. Wiser and M. Bolinger, 2012 wind technologies market report, Lawrence Berkeley National Laboratory, Berkeley, CA, Rep. LBNL-6356E, July

6 [2] H. Chavez, R. Baldick and S. Sharma, Regulation Adequacy Analysis Under High Wind Penetration Scenarios in ERCOT Nodal, IEEE Transactions on Sustainable Energy, vol. 3, no. 4, pp , Oct [3] Y. V. Makarov, J. Ma, S. Lu, T. B. Nguyen, (2008, June). Assessing the Value of Regulation Resources Based on Their Time Response Characteristics, Richland, WA. [Online]. Available: 20resources-Report%20final-1.pdf [4] K. Vu, R. Masiello, R. Fioravanti, Benefits of Fast-Response Storage Devices for System Regulation in ISO Markets, in Proc. IEEE PES General Meeting, Calgary, AB, 2009, pp [5] C. Lu, C. Liu and C. Wu, Effect of battery energy storage system on load frequency control considering governor deadband and generation rate constraint, IEEE Transactions on Energy Conversion, vol. 10, no. 3, pp , Sep [6] KEMA, (2011, Dec.). KERMIT Study Report. KEMA, Chalfont, PA. [Online]. Available: m-kema-final-study-report.ashx [7] Z. Hu, X. Xu, F. Zhang, J. Zhang, Y. Song, Research on Automatic Generation Control Strategy Incorporating Energy Storage Resources, Proceedings of the CSEE, col. 34, no. 29, pp , Oct [8] R. D. Masiello and W. P. Katzenstein, Adapting AGC to manage high renewable resource penetrations, in Proc. IEEE PES General Meeting, San Diego, CA, 2012, pp [9] Sathans and A. Swarup, Intelligent Load Frequency Control of Two-Area Multi Unit Power System with SMES, in Systems Engineering (ICSEng), st International Conference on, Las Vegas, NV, 2011, pp [10] C. Pilong, (2016, June). PJM Manual 12: Balancing Operations. PJM, Audubon, PA. [Online]. Available: 6