Reactive Power Optimization Of Distribution Network Based On Big Data Free Entropy And Scene Matching

Distribution network is at the end of the power system,and its operation state is greatly affected by the operation mode and load change.Distribution network is in a non-economic operation state in a certain region and time period,so it needs to coordinate and control the voltage,reactive power and other indicators.Reactive power optimization is an effective means to ensure the safe and economical operation of power system.With the continuous expansion of the power grid,the access of distributed power supply and new random load and the increase of control variables,the difficulty of reactive power optimization is increasing.In distribution network,fast growing multi-source heterogeneous big data with high dimensions,strong coupling and high randomness are generated in real time,which makes the application of big data technology in the field of reactive power optimization of distribution network possible.Based on ABB China research institute project "study on reactive power optimization and voltage management technology of distribution network based on big data random matrix and free entropy"(ABB20171128REU-CTR),this thesis proposes a reactive power optimization method of distribution network based on big data free entropy and scene matching.This method abandons the traditional idea of optimal dispatching of distribution network based on model and parameter,applies the big data modeling and analysis method to distribution network reactive power optimization field,from the point of view of the data driven to realize the reactive power optimization of distribution network.The main content of the thesis is as follows:(1)Combined with the operation status of distribution network,the IEEE-37-node distribution network model including distributed power supply and electric vehicle based on OpenDSS was established,and the traditional reactive power optimization method based on OpenDSS is proposed and its correctness and effectiveness are verified,which lays a foundation for the follow-up research.(2)The technical framework of big data for reactive power optimization of distribution network is proposed.Based on the relationship between reactive power optimization of distribution network and scene matching,this thesis demonstrates the feasibility of finding out the reactive power optimization scheme suitable for the moment to be optimized from the historical database,and establishes the big data technology architecture suitable for the reactive power optimization of distribution network,and preprocesses the data.(3)A cluster method of reactive power optimization scene in distribution network based on big data is proposed.In scene matching,clustering analysis is carried out first,and then the matching of similar scenes is targeted,which can speed up the matching speed and improve the accuracy.An improved K-means clustering algorithm suitable for the historical big data of distribution network is proposed,and the historical scenes of the IEEE-37-node distribution network model are clustered.The clustering results show that this method can effectively divide the scenes with different characteristics,which verifies the correctness and effectiveness of this clustering method.(4)A reactive power optimization method of distribution network based on big data free entropy and scene matching is proposed.The free entropy index which can reflect the scene feature of distribution network is defined,and the extraction method of scene feature of distribution network is proposed.Pearson correlation coefficient is used to weight the free entropy index and define the overall deviation degree,quantify the difference between scenes,and give the specific process of reactive power optimization of distribution network based on the matching of big data free entropy and scene matching.(5)Verification and comparative analysis of reactive power optimization effect of distribution network.The joint simulation platform of MATLAB/OpenDSS distribution network was established,and the daily simulation analysis and semi-annual simulation analysis were conducted on the reactive power optimization of IEEE-37-node distribution network.Compared with the traditional algorithm and particle swarm optimization,the proposed method is proved to be correct and effective.