CONSTRUCTION METHOD OF VIRTUAL POWER PLANT INTERACTION MODEL BASED ON WKNN AND KELM-GPR
Currently,the participation of the virtual power plant in distribution network dispatching mostly depends on physical model.However,due to the diversity,time-varying,and temporal coupling of the aggregated members of virtual power plant,its analytical modeling becomes more difficult,and it is difficult to meet the timeliness of the intra-day scheduling requirements of distribution network,and there are privacy security issues.Therefore,a virtual power plant interaction model construction method based on weighted K-nearest neighbor(WKNN)and kernel extreme learning machine-Gaussian process regression(KELM-GPR)is proposed.Firstly,to improve the prediction accuracy of the interactive model,a method of uniformly generating training sets is proposed;Secondly,a scheduling instruction feasibility model is established through the WKNN algorithm to measure the dispatchable boundary of the virtual power plant;Next,GPR is introduced as an error compensation model,and combined with KELM to construct a virtual power plant interaction cost model based on KELM-GRP to participate in the economic dispatch of the distribution network;Finally,to verify the feasibility of the proposed method,based on the model of scheduling instruction feasibility and interactive cost of virtual power plant,a virtual power plant participation in the intra-day optimization scheduling of the distribution network model is constructed.Simulation results show that the proposed method can significantly reduce the model optimization solution time and protect the internal information security of the virtual power plant.
virtual power plantmachine learningoptimal schedulingerror compensationmodeling method
NETL
NSTL
万方数据
