Research On Solving Strategy Of Security-Constrained Unit Commitment Based On Heuristic Method

With the rapid development of electric power industry in China,the scale of the electric power system is increasing.During this process,the unbalanced distribution of power resources and load centers have appeared,which ask for higher requirement for economical and efficient power dispatch.The securityconstrained unit commitment(SCUC)is an important part of the preparation of the core power generation plan in the optimization scheduling of the power system.Improving its solution efficiency is a vital method to optimize the structure of power generation resources and improve the efficiency of the power system.Aiming to cope with the inefficiency of traditional mathematical methods for solving SCUC problems,comprehensively considering the economic optimality and solution speed of SCUC problems,a series of machine learning and mixed integer programming(MIP)approaches are designed for SCUC.The improvement of the efficiency of solving the SCUC problem is mainly achieved from two aspects of optimizing the algorithm parameters and obtaining the heuristic knowledge.The specific research content is as follows.In order to make the Branch-and-Bound algorithm(B&B)get a faster solution rate in solving specific SCUC problems,this paper proposes an asynchronous optimization algorithm based on Bayesian Optimization to realize the adjustment of B&B parameters.This method adopts the reduction of solution time as the objective evaluation function,draws up an asynchronous optimization strategy that comprehensively considers the solution effects of a series of similar SCUC problem,and obtains the optimal parameters suitable for the solution of specific SCUC problems.Compared with the default parameters of the B&B algorithm,the simulation results show that the tuned parameters can effectively improve the solution rate of specific SCUC problems.For obtaining higher efficiency,this paper proposes a solution method that combines heuristic knowledge and MIP(HKMIP)from the perspective of effective use of historical data.Firstly,the method adopts the historical load data and the solution as training data set,as well as using deep learning to construct a mapping model between the load and the unit status.And the heuristic knowledge of the on/off status of some units through is obtained through the dual-threshold decision logic Secondly,using statistical theories to analyze historical data of unit output and heuristic knowledge of the output of some units is obtained.Under the new load scenario,the heuristic knowledge of the current case for the SCUC mode is obtained by the above method,which assists MIP method to get the final unit commitment plan.Simulation results show that this method can improve the efficiency of problem solving while obtaining highquality solutions.