多时间尺度调度

Coordination Between Short-Term and Real-Time SchedulingIncorporating Wind PowerKui Wang, Buhan Zhang, Jiajun Zhai, Wen Shao,Xiaoshan Wu and Chengxiong MaoState Key Laboratory of Advanced Electromagnetic Engineering and TechnologyHuazhong University of Science and Technology, Wuhan, Chinawangkui_hust@Keywords: short-term scheduling, real-term scheduling, coordination scheduling, wind power.Abstract. Coordination strategies between short-term (e.g., weekly and daily scheduling ) and real-time scheduling in wind power integrated system are disscussed. To cope with the uncertainty of wind power and load demands, weekly and daily rolling schedulings are applied. According to the latest updated prediction results of wind power and load demands, weekly rolling scheduling is applied to revise unit commitment and fuel allocation in remaining hours in a week. Daily rolling scheduling is applied to revise generation scheduling in remaining time in a day. A modified IEEE 118-bus system is applied to test the proposed approach. IntroductionGeneration scheduling are generally divided into long-term [1], mid-term [2], daily[3,4] and real-time [5] schedulings in time scales. The coordination among different time scales schedulings is very important, and effective coordination can ensure that the whole scheduling can be implemented smoothly. The coordination between long-term and short-term scheduling are discussed in references [6], and a predetermined minimum reserve energy was used as the coordination index. If the reserve energy was less than the minimum value, the long-term scheduling would be adjusted. In reference [7], the daily unit commitment was implemented with daily energies constraints, obtained from the long-term scheduling.However, there're few articles concerning the coordination between short-term and real-term scheduling, especially incorporating wind power. In weekly scheduling, the horizon of the study is one week with increments of an hour, mainly focusing on unit commitment and fuel allocation. The study of the daily scheduling is one day with increments of 10min, mainly solving economic dispatch problem. Real-time scheduling is used to adjust unit power in the nest 10min.The characteristics (e.g., randomness, volati-lity and unpredictable) of wind increase difficulty and uncertainty and lead to great difference between higher and lower time scales schedulings. To solve the above problems, we adopt rolling schedulings, including weekly and daily rolling schedulings. According to the latest updated prediction results of wind power and load demands, rolling schedulings are carried out to revise the results of the original schedulings. Multi-time scales generation schedulingsA. Weekly schedulingWeekly scheduling mainly focuses on unit commitment and fuel allocation in hours. The objective function consists of fuel cost )(⋅Gi F and start-up cost Git S .∑∑==−−+=week GT t N i Git t i it Git Gi it week S u u P Fu F 11)1(])1()([min (1)where t is the index of time in weeks, and i is the index of thermal units. week T equals 168h in oneweek. it u and Git P represent the commitment state and the output power of unit i at time t respectively. 1=it u if unit i is running at time t else 0=it u .Constraints of weekly scheduling are listed as follows. a) System power balancet P PP u Dt N j WjtN i Git it WG ∀=+∑∑== 11(2)where j and W N is the index and the number of wind farms respectively. Wjt P is the predicted output power of the wind farm j at time t , and Dt P is the predicted system load demands at time t . b) Thermal unit generation limitst i P u P P u Gi it Git Gi it ∀∀≤≤ max min (3) where min Gi P and max Gi P are minimum and the maximum power of unit i . c) Thermal unit ramp rate limitsi u t Gi Git i d P P ξξ≤−≤−−)1( (4)where iu d /ξ represent ramp down/up limits of unit i . d) Thermal unit minimum up/down time limits0)()()1()1(≥−∗−−−oni on t i it t i T T u u and 0)()()1()1(≥−∗−−−off i off t i t i it T T u u (5) where off on it T / represent on/off time of unit i at time t , and off on i T / represent minimum up/down time of unit i .e) Weekly energy constraintsiE T Pu E wGi wT t Git it w Gi week∀≤∆≤∑= max 1min(6)where w Gi E min and wGi E max represent minimum and maximum weekly energy of unit i . w T ∆equals an hour. f) Up reserve constraintstDt N i Git Gi iti uit US P d P Pu u G+∗≥−∑=%)]( , min[1maxξ (7)g) Down reserve constraintst Dt N i Gi Gitit id it DS P d P P u u G+∗≥−∑=%)]( , min[1min ξ(8)where ττDS US /represent up/down spinning reserve requirements caused by wind power, and can be calculated by the second model [3]. %d is spinning reserve requirements ratio of load demands. B. Daily schedulingAccording to the unit commitment an fuel allocation results, obtained from weekly scheduling, daily scheduling is used to deal with economic dispatch in one day with increments of 10min. The objective is to minimize the total generation costs of thermal units.∑∑===day GT N i Gi Gi iday P F u F 11)(min ττττ (9)where it i u u =τ, if t ∈τ. day T equals 144 in one day.Daily scheduling should consider energy constraints obtained from weekly scheduling.i E T P u E dGi dT Gi i d Gi day∀≤∆≤∑= max 1minτττ (10)where d Gi E min and dGi E max represent minimum and maximum daily energy of unit i . d T ∆equals 10min. C. Real-time schedulingAccording to daily scheduling, real-time scheduling is used to adjust unit power in the nest 10min. The objective is to minimize the total adjusting costs of thermal units.∑=∆+∆+=GN i Gi i Gi i Gi i i real P a P b Pa u F 12])2[()(min τττττ (11)where i a and i b are consumption characteristic coefficients of unit i , and τGi P ∆ is the adjustingpower of unit i at time τ.Similarly to weekly scheduling, daily and real-time schedulings must also satisfy a series of constraints, such as system power balance, unit generation limits, unit ramp rate limits and up and down reserve constraints.Coordination strategiessBased on the unit commitment states and fuel allocation results of weekly scheduling, daily scheduling mainly focuses on economic dispatch in smaller increments (e.g., 10min). Real-time scheduling is applied to adjust unit power in next 10min based on the results of daily scheduling. The daily energy minimum and maximum limits can be calculated as follows [7].∑==weekTtGitwGidGidGiEEEE1minminand∑==weekTtGitwGidGidGiEEEE1maxmaxwheredTGiidGiTPuEday∆=∑=1τττ(12)The wind power and load prediction is more precisely in shorter term than longer term, so it's necessary to revise the results of longer term scheduling. According to the latest updated prediction results of wind power and load demands, weekly rolling scheduling is applied to revise unit commitment and fuel allocation in remaining hours in a week. Similarly, daily rolling scheduling is applied to revise generation scheduling of remaining time in a day.Solution methodsThe weekly scheduling is divided into external subproblem deciding unit commitment states, and inner subproblem solving economic dispatch. External subproblem is solved by discrete particle swarm optimization algorithm, and inner subproblem is solved by Lagrangian multiplier method based on the equal consumption increment principle. Daily and real-time scheduling can be solved by improved continuous particle swarm optimization effectively.Case studyA modified IEEE 118-bus system [1] is applied to test the proposed model. Thermal units on 36, 69 and 77 buses are replaced by three wind farms with the capacity of 250MW, 180MW and 100MW repectively. Hourly wind power of the first week in winter is depicted as Fig. 1, and 10min wind power in the first day of the above mentioned week is depicted as Fig. 2.one week farms in one dayThe planned output energy of thermal units in the first week in the winter is shown in Fig. 3. It can be seen that the lower unit generation cost is, the bigger unit planned energy will be. The planned output energy of unit 27 and 28, whose generation costs are lower, are 38831.19MWh and 37286.47MWh. Based on the results of weekly scheduling, the planned output energy of thermal unit in the first day is shown in Fig. 4.Short-term load and wind power prediction results at some 10min, used in daily scheduling are 3222.6732MW and 205.349MW respectively in a 10min. And ultra short-term prediction results of them, used in real-time scheduling, are 3500MW and 180MW respectively in the above mentioned interval. The adjusting power of thermal units is shown in Fig. 5. It can be seen that the lower unit adjusting cost is, the bigger unit adjusting power will be. Due to ramp limits, units 5, 10, 11, 20, 21, week. Fig. 4. Units output energy in a day. Fig. 5. Units adjusting power inreal-time. ConclusionsWeekly scheduling mainly focuses on unit commitment and fuel allocation, and daily scheduling canbe treated as economic dispatch problem based on the results of weekly scheduling. At last, real-time scheduling is applied to adjust output power of thermal units. In order to handle the future uncertainty, weekly and daily schedulings are implemented at regular intervals based on the lastest updated load demands and wind power. The case study shows the effectiveness of the coordination strategies of short-term and real-time scheduling. AcknowledgementsThis work was supported by National High Technology Research and Development Program of China(863 Program) (2011AA05A101), National Basic Research Program of China (2009CB219702) and National Natural Science Foundation of China (50837003). References[1] Yong Fu, Mohammad Shahidehpour and Zuyi Li, ''Long-term security-constrained unitcommitment: hybrid Dantzig-Wolfe decomposition and subgradient approach'', IEEE Trans on Power Systems , vol. 20, no. 4, pp. 2093-2106, Nov. 2005．[2] C. G, Baslis and A. G. Bakirtzis, ''Mid-term stochastic scheduling of a price-maker hydroproducer with pumped storage'', IEEE Trans on Power Systems , vol. 26, no. 4, pp. 1856-1865, Nov. 2011．[3] C. L. Chen, “Optimal wind-thermal generating unit commitment, IEEE Trans on EnergyConversion, vol. 23, no. 1, pp. 273-280, Mar. 2008．[4] J. Hetzer, D. C. Yu, and K. Bhattarai, “An economic dispatch model incorporating wind power,IEEE Trans on Energy Conversion , vol. 23, no. 2, pp. 603-611, Jun. 2008．[5] Y. Li, L. Ge, and Y. Lin, “Real-time power dispatch strategy considering energy saving andenvironmental protection, Electric Power Automation Equipment , vol. 29, no. 3, pp. 42-45, Mar. 2009(in Chinese)．[6] M. K. C. Marwali, ''Coordination between long-term and short-term generation scheduling withnetwork constraints'', IEEE Trans on Power Systems , vol. 15, no. 3, pp. 1161-1167, Aug. 2000． [7] E. Handschin, and H. Slomski, ''Unit commitment in thermal power systems with long-termenergy constraints'', IEEE Trans on Power Systems , vol. 5, no. 4, pp. 1470-1477, Nov. 1990．Renewable and Sustainable Energy II10.4028//AMR.512-515Coordination between Short-Term and Real-Time Scheduling Incorporating Wind Power10.4028//AMR.512-515.700。