Research On Knowledge Mining Method For Power Flow Section Power Adjustment Of Large Power Grid

With the development of my country’s economy,the scale and load level of the power grid continue to increase,and the scale of the power grid continues to expand.The traditional power grid simulation analysis mode has a large number of manual participation,showing obvious shortcomings.Therefore,it is necessary to use artificial intelligence methods for power grid simulation analysis to replace the current manualbased power grid simulation analysis work mode.In actual projects,an important part is to use the knowledge base to provide strong support for power grid simulation analysis.However,at present,there is less knowledge obtained from experts through questionnaires,which poses great challenges to the construction of the knowledge base and subsequent management Therefore,it is proposed to dig more knowledge from the behavior data of experts to supplement the knowledge base.At present,three application scenarios are mainly involved in the simulation analysis of large power grids,namely,power flow calculation adjustment,power flow section power adjustment,and temporary stability control measures designation.According to the current actual situation,it is decided to conduct related research on the actual scenario of power adjustment of tidal current section.The power adjustment of the power flow section is a sub-problem of the power flow state adjustment.Its goal is to adjust the power of a certain section of the power grid to a specified range.At first,it was planned to mine relevant knowledge from the manual adjustment sequence data of experts.However,due to the difficulty of obtaining manual adjustment data by experts,the current amount of data is very small and relevant research cannot be carried out.Therefore,the automatic adjustment model developed in this project is used to generate the Adjust the sequence data for research.The automatic adjustment model is based on reinforcement learning and can complete the cross-section power adjustment task with a maximum adjustment success rate of 95%,and the working mode in the adjustment process is similar to that of an expert,which is reasonable to replace an expert.This paper aims to study the knowledge mining method of the related automatic adjustment model based on reinforcement learning and the adjustment sequence data generated under the actual scenario of power adjustment of the tidal current section.It is divided into the three main parts,include the study of imitation learning method based on CEPRI36 example,the study of imitation learning method based on the example of Northeast Power Grid and of decision tree strategy extraction based on simulation experiment.In the process of research,the B&KD VIPER algorithm improved based on VIPER algorithm,feature selection method based on power grid structure information and Decision tree strategy extraction methods(knowledge extraction)based on simulation experiments,etc.,and their effectiveness is verified through experiments.