Research On Optimal Day-Ahead Dispatch Of Large Scale Power Systems Considering Wind Generatin Integration

With high penetration wind power being increasingly integrated into the modern power grid and the regional power grids more closely interconnecting with each other, the day-ahead generation dispatch has revealed two major challenges: 1) how to efficiently solve much more complex day-ahead generation dispatch with consideration of stochastic wind power, 2) how to ensure the secure and economic operation of the whole multi-area power grids while maintaining the independence of each regional power gird. Based on the newest stochastic optimization methods and decentralized optimization methods in the optimization theory, this paper aims at studying the day-ahead generation dispatch of large-scale interconnected power system with wind power integration. The main research contents can be summarized as follows:Based on scenario technology, multi-cut method with dynamic reduction is proposed. The stochastic unit commitment problem with wind farm integration is decomposed into the master problem and the sub-problem. The master problem determines on/off status of generation units and output scheduling corresponding to the forecasted wind power scenario, while the sub-problem determines generation scheduling corresponding to the sampling scenarios. The master problem and the sub-problems are connected by optimality cuts and are solved alternately. The redundant optimality cuts are dynamically reduced during iterations and the number of optimality cuts can remain constant. Through this method, the size of the master problem can be reduced and the disadvantage of multi-cut method that the size of the master problem keeps increasing as the number of iterations grows can be solved. Computational results on a real provincial power system including a large scale wind farm and 180 thermal units demonstrate that this multi-cut method with dynamic reduction is superior in terms of convergence, computational efficiency and computer memory.Furthermore, in order to study the influence of wind power spillage and security constraints on the economic dispatch, a bi-level dynamic economic dispatch method for power plant/network considering stochastic wind power and network security constraints is proposed. The upper-level aims at the network’s security and economic operation and determines the output scheduling of power plants and wind power curtailment, the master problem and sampling scenario sub-problems are solved alternately to minimize operation cost of power network and accommodate the volatility of wind power. The lower level minimizes the generation cost of each power plant and determines the output scheduling of generation units by solving the power plant sub-problems. The multi-cut algorithm is used to solve this bi-level model,and two kinds of optimality cuts are generated to approximate the sampling scenario sub-problems and the power plant sub-problems. The computational results on a real provincial power grid with three wind farms show the effectiveness of the proposed method.A decentralized dynamic economic dispatch of the multi-area system with stochastic wind power integration is proposed. Firstly, the uncertainty of wind power is not considered, the multi-area system is decomposed into independent areas using the splitting variables method. Next the analytical target cascading method is used to solve the multi-area dynamic economic dispatch of the forcasting scenario in a decentralized way. If the deterministic model is solved, the day-ahead generation outputs and the tie-line power flows will be fixed by the subproblems of error scenarios. After each area solves their own stochastic economic problems, the analytical target cascading method is adopted again to ensure the satisfication of coupling constraints. By solving the decentralized problems and the stochastic problems iteratively, the dispatch independence and data confidentiality of different areas can be maintained, the regional power grid can efficiently cope with the stochastic wind power and the whole mulit-area power system can operate securely and economically. The 3-area IEEE RTS system is tested to demonstrate the effectiveness of the proposed method.A cutting plane consensus algorithm is implemented in this paper to solve the multi-area dynamic economic dispatch of large-scale power systems in a fully decentralized way. A local master problem is constructed in each area to approximate the original problem, cutting planes are transmitted among different areas, so an upper coordinator is unnecessary. Two modifications have been developed: firstly, only the newest cutting planes are transmitted among different areas to reduce amount of information transferred; secondly, the cutting plane will be deleted if it could be inactive during consecutive iterations to reduce the iterations. The convergence and correctness can be guaranteed for power system with a wide range of scales, and the convergence speed is independent from parameter tuning. Hence, the proposed method can be directly used in different power systems without adjusting parameters. The 3-area IEEE RTS system, a real 4-area 2298-bus provincial power system and an 8-area test system are tested to demonstrate the effectiveness of the proposed method.