Development of a Frequency-stabilizing Scheme for Integrating Wind Power Generation into a Small Island Grid

Transcription

1 Preprints of the 8th IFAC World Congress Milano (Italy) August 8 September, Development of a Frequencystabilizing Scheme for Integrating Wind Power Generation into a Small Island Grid K. Yamashita, O. Sakamoto, Y. Kitauchi, T. Nanahara, T. Inoue* H. Fukuda, T. Shiohama** *Central Research Institute of Electric Power Industry, Tokyo, 85 Japan ( yama@criepi.denken.or.jp, orie@criepi.denken.or.jp, kitauchi@criepi.denken.or.jp, nanahara@criepi.denken.or.jp, inotoshi@criepi.denken.or.jp). **Okinawa Electric Power Company, Okinawa, 96 Japan ( Hitoshi_Fukuda@okiden.co.jp, Tomohiro_Shiohama@okiden.co.jp) Abstract: Integrating wind power generation into small islands has been one of the demonstration projects in Okinawa Prefecture, Japan. Since such integration could deteriorate power quality, including frequency, in an island grid, a frequencystabilizing system using flywheels has been integrated into a small island. In order to establish a proper frequencystabilizing scheme for a small island, an accurate model of a diesel generator including a governor is vital. Therefore, a model was developed through generator dump tests. A new frequencystabilizing scheme was also developed through the timedomain simulation of the island grid model, which consists of the above mentioned diesel generator model and negative load change representing wind power variation. The developed stabilizing scheme was applied to the flywheels in the island grid and revealed great performance for mitigating frequency variation. Keywords: power system, wind power generation, small island power system, flywheel energy storage system, diesel generator. INTRODUCTION A project to increase wind penetration for small remote islands in Okinawa Prefecture, Japan was started through the local government of Okinawa Prefecture, including Okinawa Electric Power Company Incorporated, in 9. One of the purposes of the project is to introduce two wind power generators (5 kw in each) and eight flywheels (3 kw 3 s in total) onto a small island with a maximum demand of 63 kw (Refer to Fig. ). The flywheels (referred to FWs, hereafter) are expected to alleviate the impact of wind penetration. Because a power system on a small island is an isolated power system, the increasing wind penetration causes the increase of frequency variation. In order to mitigate the increasing variation, the following three items are required: ) development of accurate diesel generator models with a governor action; ) extraction of wind power variation characteristics; 3) establishment of an effective frequencystabilizing scheme using FWs. FWs are known as one type of energy storage system and are suited for alleviating the high frequency fluctuation of output of wind power generation. Although various frequency stabilization schemes using FWs have been proposed, only a few schemes can be applied for an isolated grid. Thus, a scheme for stabilizing system frequency using FWs in a small power system including a wind farm was proposed (Takahashi et al. [7]). In this paper, system frequency is used as the input of the proposed scheme. However, diesel generator models with a governor action were represented by a general IEEE standard model for hydraulic turbines, and the validation of the scheme was investigated using only simulation analysis. Another scheme for stabilizing system frequency using FWs in an isolated grid including wind generation was proposed (Hamsic et al. [7]). The actual power of the wind generation and system frequency was used as the input of the proposed scheme. The scheme was applied to real FWs installed in the Flores power system and verified its fundamental performance in the Flores isolated grid. However, no detailed control logic for FWs has been described in the paper. Diesel Generators Rated Capacity 5 kw 6G 7G 8G 9G 5 kw 3 kw 3 kw FW Rated Capacity 5 kw WT 5 kw WT Load Rated Capacity: kw (W7 cm H7.5 cm D8 cm 8 sets) Wind Generators Flywheel Energy Storage System Fig.. A small remote islanded power system. In order to establish a proper control scheme, accurate diesel generator models with a governor action were firstly examined based on the measured values of generator dump tests in this study. Then, a new frequencystabilizing scheme was developed using timedomain simulation, which consists of the developed diesel generator models and an equivalent load change representing wind power variation. Copyright by the International Federation of Automatic Control (IFAC) 873

2 Preprints of the 8th IFAC World Congress Milano (Italy) August 8 September, The developed stabilizing scheme was applied to the actual frequencystabilizing system in the small island and verified its performance through connecting to the island power system.. DEVELOPMENT OF A DIESEL GENERATOR MODEL INCLUDING A GOVERNOR. Validation of the Governor Model of the Diesel Generators A governor model is a key generator controller for analyzing wind power impact on system frequency deviation in a small isolated power system. Although various types of governor models of steam turbines and hydro turbines have been used for power system planning and analysis, few governor models of diesel engines have been developed. Therefore, a generator dump test was performed for main diesel generators in the island grid in order to develop the governor model of a diesel engine. The following findings were obtained through the generator dump test. Static and dynamic response characteristics for frequency change vary depending on the manufacturers of engines and/or governors. Dynamic response characteristics of the same diesel generator vary depending on the magnitude of frequency deviation. Based on the above findings, the governor model was developed considering the following two conditions: The transient increase of generator power output and postfault steady state generator output obtained from timedomain simulation should be consistent with the same increase obtained from the result of the generator dump test respectively with regards to a few different magnitudes of frequency deviations. The time to reach maximal generator output obtained from timedomain simulation immediately after a generator tripping should be consistent with the time obtained from the result of the generator dump test. Fig. shows the block diagram of the developed governor model of the diesel engine. The model parameters for each diesel engine are provided in Table. It should be noted that the model was not developed using a componentbased approach, but it was developed using a measurementbased approach. The measured frequency is inserted into the input of the governor model as the rotor speed deviation in Fig.. The calculated active power output was compared with the measured active power output. Based on the comparison, the construction of the model and the identification of its parameters were performed through trial and error. For example, one leadlag element in Fig. was applied for the governor model of 6G and 7G, while two leadlag elements were applied for the model of 8G and 9G. The feedback control loop, including an integral element in Fig., denotes a firstorder lag element with a restricted rate of change and enables emulating a slow trend of the increasing generator output after a generator tripping. Two leadlag elements in Fig. enable emulating a transient response of the generator output immediately after a generator tripping. Although the washout element in Fig. originally has the function of a highpass filter, this element here has a role of eliminating highfrequency components from the original signal (output of the signal transducer in Fig. ) through the accumulator element. That helps preventing excessive control. Postcontingency analysis showed steadystate frequency deviation obtained by the generator dump test was proportion to the output of the tripped generator. However, the largest frequency deviation was not proportion to the output. Therefore, the nonlinear element in Fig. was introduced to match measured values with simulated ones for three different magnitudes of frequency deviations (about.3 Hz,. Hz, and.5 Hz). The example simulation results obtained by CRIEPI's Power Analysis Tools (CPAT) are shown with the testing result in Fig. 3. It should be noted that the responses of 6G, 8G, and 9G are the result in case of 7G tripping, and the response of 7G is the result in case of 6G tripping. As shown in Fig. 3, the developed governor model can express the response of the generator dump test precisely.. Validation of the small island power system model The small island power system model including not only diesel generators but transformers and loads should also be validated. Therefore, the small island power system model was developed and was validated through the dump test. The outline of the system model shows the following: () Composition of the island power system model: The island power system model consists of diesel generator models for G6, G7, G8, and G9, a stepup transformer, transmission lines, and loads as shown in Fig.. Because the generator is almost directly connected to the loads, the reactance of the transmission lines was set to extremely low values. Sg Rotor Speed Deviation Signal Transducer T S K Governor Gain Fig.. Governor model of a diesel engine. Load Reference Nonlinear T 3 S T S Washout T S U L First LeadLag T 6 S T 5 S Second LeadLag T 8 S T 7 S TQT Mechanical Output 87

3 Active Power Output[kW] G9 [kw] G8 [kw] G7 [kw] G6 [kw] Preprints of the 8th IFAC World Congress Milano (Italy) August 8 September, G Tripping 6G Tripping Table. Generator constants Specification 6G 7G 8G 9G Rated capacity [kva] Rated power [kw] Inertia constant [s]* D axis reactance [p.u.] D axis transient reactance [p.u.] D axis subtransient reactance [p.u.] Q axis reactance [p.u.] Q axis subtransient reactance [p.u.] D axis open circuit time constant [s] D axis transient time constant [s] D axis subtransient time constant[s].... Armature leakage reactance [p.u.] Armature time constant [s] Zero phase reactance [p.u.] Negative phase reactance [p.u.] *The inertia constant of the generator and its engine was assumed to be.5 times larger than that of only one generator. Table 3. Initial conditions for the generator dump test Case Tripping Generator Generator Tripping Amount Active Power Output [kw] Total Generation [kw] 6G 7G 8G 9G Active Reactive Power Power 6G G Measured Value Simulated Value Fig. 3. Example of the active power outputs of diesel generators No. 6 through No. 9 with frequency data. 5 Table. Frequency response obtained from simulation results and testing results Case Item Testing ResultSimulation Result Bottom frequency [Hz] 59.5 [Hz] Time to reach the bottom frequency.5 [s]. [s] Postdisturbance steadystate frequency [Hz] [Hz] Bottom frequency 59.6 [Hz] 59. [Hz] Time to reach the bottom frequency.7 [s]. [s] Postdisturbance steadystate frequency [Hz] [Hz] Table. Parameters of the developed governor models Variable 6G 7G 8G 9G T.... T T K T.... U L.... T T T7..5 T G 8G 7G Measured Value Simulated Value () Load model: As the component of the loads in the island power system is obscure, a static exponential model with typical load voltage parameters was applied. The load voltage characteristics were given as constant current characteristics for active power and constant impedance characteristics for reactive power. Load frequency characteristics were not considered for this study because conservative simulation results were preferable. 5 Fig.. Example comparison between measured and simulated value of frequency response and active power output response (Case ). (3) Diesel generator model: The generator model with daxis and qaxis rotor circuits was used because the transient response of generators is closely related to the dynamic behavior of system voltage and, eventually, loads and system frequency. Each rotor circuit includes one damper winding respectively. The generator constants are shown in Table. Although an inertia constant of generators and its engine is critical for this study, only the generator inertia was able to be derived from the design data. Therefore, the inertia constant of the engine was assumed to be half of the generator inertia, the ratio of which is the smaller value selected from its appropriate range. () Generator control model: The standard AC exciter model, with its standard parameters, was applied. The developed governor model was also used. Example initial conditions for the generator dump test are shown in Table 3. The characteristic frequency response is shown in Table. The bottom frequency, the time it takes to reach the bottom frequency, and the postdisturbance steady state frequency obtained from timedomain simulation are 5 875

4 Coherence Wind Generation Output fs [kw ] WP Output [kw] Preprints of the 8th IFAC World Congress Milano (Italy) August 8 September, consistent with those three indices obtained from the testing result, respectively. As shown as Fig., the transient response and the steady state response of each generator output obtained from timedomain simulation are also consistent with the same responses obtained from the testing result. Therefore, it can be concluded that the small island power system model expresses the frequency response properly. 3. DEVELOPMENT OF A FREQUENCYSTABILIZING SCHEME FOR THE FREQUENCYSTABILIZING SYS TEM The dynamic behavior of wind power generation is important for the frequencystabilizing system. Wind power generation output was represented as an equivalent timevarying negative load for timedomain simulation. 3. Variation characteristics of wind power generation The installed wind turbines, produced by Vergnet in France, adopt a fixedspeed twobladed wind turbine with active stall control. The characteristics of the fixedspeed twobladed wind turbine are the appearance of remarkable power variations the frequency of which is proportion to the rated rotating speed of the wind turbine (.769 Hz). Fig. 5 shows an example power spectrum of the system frequency and the wind power generation under its power output limitation ( kw). The variation characteristics of wind power generation are summarized as follows: () Two frequencies,.769 Hz and.5 Hz, were obvious for both the wind power variation and frequency variation. Note that.5 Hz is double the rated rotating speed of the wind turbine. () The coherence between wind power generation and the system frequency at the above two frequency components was high. Based on the above findings, it can be observed that the two frequency components of the wind power variation have substantial impact on frequency fluctuation. Therefore, mitigation of the two frequency components with FWs is the most Table 5. Frequency change of wind power generation output Rated Rotating Speed Frequency of "N" Frequency of "N" 6 rpm.769 Hz.5 Hz Output of Wind Power Generation 3 6N N N N Time Period [/Hz] 5 Wind Generation Output System Frequency 8 Coherence between Wind Generation Output and Frequency Fig. 5. Example of the power spectrum of wind power generation output and coherence between wind power generation output and system frequency. effective way to alleviate the frequency variation caused by wind power generation. 3. Development of a frequencystabilizing scheme FWs are known as one of the effective ways to alleviate the frequency variation caused by wind power generation. The developed frequencystabilizing scheme using FWs employs both wind power variation and the deviation of the system frequency as an input. Fig. 6 shows a block diagram of the FW model equipped with the developed scheme. The proper parameters of this scheme were derived considering rolesharing between diesel generators and FWs. The considera Frequency fs [Hz ] DF Input Part DF Frequency Deviation [PU] DP Input Part.s DP.8.s Active Power of Wind Generation [PU] Signal Transducer Signal Transducer pukw/puhz [PU] PU kw s PU rpm s 6[rpm] w REF Gain Washout [PU] Scale Conversion kw PU Inverter w w.s w Ms [PU] Gain 5s 5s Washout.[PU].[PU] PU kw 3 Scale Conversion s Center Frequency Control Fig. 6. Block diagram of the FW model equipped with the developed frequencystabilizing scheme. 5[rpm] 373[rpm] PU rpm 876

5 Rotating Speed [RPM] WP FW Output [kw] Active Power Output [kw] WP Output [kw] Active Power Output [kw] Preprints of the 8th IFAC World Congress Milano (Italy) August 8 September, tion for deriving the parameters is described below. () Wind power variation was used as the DP input of the developed scheme. The parameters of the DP input part in Fig. 6 were selected in order to extract a specified frequency component range (. Hz or higher) of wind variations, including.769 Hz and.5 Hz. () Deviation of the system frequency was used as the DF input of the developed scheme. The parameters of the DF input part in Fig. 6 were selected in order to extract a specified frequency component range (. Hz or higher) of the load variations, which is difficult for diesel generators to suppress. (3) Center frequency control was originally equipped in the frequencystabilizing system. This control helps FWs to avoid the depletion of charge or discharge energy. If the rotating speed of the FWs reaches the lowest or the highest speed, its charge energy or discharge energy will be depleted and that depletion leads to the undesirable sudden change of the frequency. Therefore, the rotating speed of the FWs should be controlled near the center of the rotating speed (,373 rpm) using the center frequency control, unless a large sudden frequency change occurs. The parameters of the center frequency control were selected in order to position the rotating speed of FWs back to the center quickly, after a large sudden frequency change. The time constant of the washout in the DP input part was set to 5 s in order to suppress the obvious highfrequency components of wind power vibration directly, while the time constant of the washout in the DF input part was set to s. Not only the center frequency control but these two short time constants of the washouts also help the FWs to avoid the depletion of charge or discharge energy Hz 3 Frequency G6 G7 G9 WP Output 6 9 (a) Simulation result without FW Frequency G6 G7 G Hz.39Hz As shown in Fig. 3, the rate of output change and the output change amount of 9G after the sudden frequency change were much larger than those of other diesel generators. It is considered that this difference is related to: the rated rotor speed of 9G, which is much higher than others, and the response time of its fuel supply control, which is much faster than others. Therefore, it might be difficult to adjust the power output of 9G within its proper operating range in case of large sudden frequency changes caused by wind power generation. Those two shortduration constants of washouts were selected considering the prevention of the abnormal operation of 9G WP Output, FW Rotating Speed FW Output 3.3 Validation of the developed frequencystabilizing scheme A startup field test was assigned to one of the wind turbines in Jan.. The data measured at the test was used for the verification of the performance of the developed frequencystabilizing scheme. In order to evaluate the high penetration of wind power generation (the interim goal was 5 min average penetration up to 5%) through timedomain simulation, the measured values were extended.75 times larger than the original measured values. Operating diesel generators were assumed to be 6G, 7G, and 9G. Their initial outputs were set to kw, kw, 8 kw, respectively (b) Simulation result with FW Fig. 7. Time series data of wind power generation output (.75 times larger than the actual measured data). Figure 7(a) shows an example simulation result without FWs. The average system frequency was 6.36 Hz (see Table 6) during the operation of the wind turbine because the Load Frequency Controller (LFC) was not equipped. If the average frequency was shifted to Hz with the AFC, the frequency deviation was plus or minus.39 Hz, which exceeded the

6 FW Output [kw] FW Output fs [kw ] Active Power [kw] Frequency fs [Hz ] Preprints of the 8th IFAC World Congress Milano (Italy) August 8 September, desired frequency deviation (normally between. Hz and.3 Hz). Moreover, the generator output of 9G was sometimes below zero, which might cause tripping of 9G due to its reverse power relay. In other words, this simulation result revealed that the small island power system cannot maintain power quality and power system stability without a frequencystabilizing system if high wind penetration such as 5% is pursued. Figure 7(b) shows the same example simulation result with FWs. As shown in this figure, FW output increased its output while wind power generation decreased its output and vice versa, if a specified frequency component range (. Hz or higher) of the wind generation and system frequency was focused on. If the average frequency was shifted to Hz with the LFC, the frequency deviation was plus or minus. Hz, which is much smaller than the desired frequency deviation (see Table 6). Figure 8 shows the power spectrum of the frequency with and without FW along with the power spectrum of the FW output. FWs mitigated a frequency component range of the system frequency between.33 Hz and 3.33 Hz effectively.. VERIFICATION OF THE PERFORMANCE OF THE DEVELOPED FREQUENCYSTABILIZING SYSTEM IN THE SMALL ISLAND GRID The developed stabilizing scheme was applied to the frequencystabilizing system using FWs in the isolated small island grid. The performance of the frequencystabilizing system was verified under a 5 min average penetration of 8%. As shown in Fig. 9, the system frequency deviation was within plus or minus. Hz. That value was almost the same as the simulation result (. Hz) described in Section 3. That means the developed frequencystabilizing system successfully mitigated the frequency variation caused by wind power generation Power Spectrum of Frequency without FW Power Spectrum of Frequency with FW Power Spectrum of FW Output. Time Period [/Hz] Fig. 8. Power spectrum of frequency with/without FW and the power spectrum of the FW output. Table 6. Simulation result of frequency change Stabilizing System F max[hz] F min[hz] DF[Hz] F Median F Without FWs.75. ± With FWs.8. ± Performance of the frequencystabilizing scheme with regard to various frequency components of wind power generation Figure shows the filtered total output of wind power generation and the filtered FW output. The filter was designed for extracting a specified frequency range (. Hz or higher). The filtered FW output increased its output, while the filtered total wind power generation decreased its output and vice versa. That means that the frequencystabilizing scheme, especially the DP input part in Fig. 6, effectively mitigated the obvious frequency components of the system frequency variation caused by wind power generation. Figure shows another filtered total output of wind power generation and another filtered FW output. The filter was designed for extracting a specified frequency range (.5 Hz or higher). The filtered FW output increased its output, while the filtered total wind power generation decreased its output and vice versa. That means that the frequencystabilizing scheme, especially the DF input part in Fig. 6, effectively mitigated the obvious frequency components of the system frequency variation caused by wind power generation.. Performance of the frequencystabilizing system using flywheels Figure shows the example response of FWs and diesel generators when one wind turbine (5 [kw]) was tripped due to its power output limitation control. The FWs promptly controlled its output, which led to the mitigation of not only frequency fluctuation but the power output variation of the diesel generators within their proper operating range. The frequency fluctuation after the tripping was also effectively mitigated with the introduction of FWs even without excessive compensation. Figure 3 shows another sample response of FWs and diesel generators when G7 (5 [kw]) was intentionally tripped. Although only the DF input part in Fig. 6 is active, the FWs Total Output of Diesel Power Station Total Output of Wind Power Generation 5 Fig. 9. Example of measured data under high wind penetration (5 min average penetration: 8 %)

7 FW Output [kw] Active Power [kw] FW, WP [kw] Load [kw] FW [kw] WP [kw] FW Output [kw] FW, WP [kw] Active Power [kw] FW [kw] WP [kw] Preprints of the 8th IFAC World Congress Milano (Italy) August 8 September, Nonfiltered Total Output of Wind Power Generation Nonfiltered FW Output Filtered Total Output of Wind Generation (over. Hz) Filtered FW Output (over. Hz) 5 Fig.. Filtered FW output and wind power generation output (cutoff frequency:. Hz). 5 5 Total Output of Wind Power Generation FW Output Filtered Total Output of Wind Generation (over.5 Hz) Filtered FW Output (over.5 Hz) 6 Fig.. Filtered FW output and wind power generation output (utoff frequency:.5 Hz). promptly controlled its output from kw to 5 kw, which contributed to the mitigation of not only frequency deviation but also the power output variation of the diesel generators. It should be noted that generator tripping could cause larger frequency deviation than the cutout of a wind turbine. 5. CONCLUSION The developed stabilizing system using FWs is currently operating in the small island power system and has shown great performance for alleviating frequency variation caused by wind power generation. The FWs have contributed to increase in wind power generation and decrease in diesel fuel usage. Future work would be concerned with the development of a voltagestabilizing scheme and with the integration of the voltage stabilizingscheme into the developed frequencystabilizing scheme. ACKNOWLEDGEMENTS This work was supported in part by the Ministry of Economy, Trade and Industry in Japan under Grant Special Coordina Total Output of Wind Power Generation 6G, 8G, 9G Fig.. Example of measured data when one wind turbine stopped due to its power output limitation control G, 7G, 8G, 9G Fig. 3. Example of measured data in case of a 7G tripping. tion Funds for Promotion of Okinawa. REFERENCES CRIEPI (99). Integrated Analysis Software for Bulk Power System Stability, CRIEPI report, ET9, July 99. Hamsic, N, Schmelter. A, Mohd, A, Ortjohann. E, Schultze, E, Tuckey, A, and Zimmermann, J (7). Increasing renewable energy penetration in isolated grids using a flywheel energy storage system, IEEE International Conference on Energy and Electrical Drives, 7 pp. 95. Kundur, P (99). Power System Stability and Control, McGrawHill, New York. Takahashi, R, and Tamura, J (7). Frequency stabilization of small power systems with wind farm by using flywheel energy storage systems, IEEE International Symposium on Electronics and Drives, 7 pp