Research On Data-driven Method For Completing Missing Load Dat

Power system analysis and management rely heavily on load data,and load analysis,load management,user classification,demand-side response,and load forecasting all greatly benefit from accurate measurement data.However,the issue of missing data cannot be avoided in any part of data processing,which will have a negative effect on further applications in the future.A load missing data completion framework based on modal decomposition and rearrangement components is suggested as a solution to this issue in order to mine more load data information and create a load data completion model.By resolving the objective function,the missing value recovery is achieved.This essay primarily does research on the following three aspects:(1)This study summarizes the missing mode of the load data according to the actual operation of the power system and analyzes the causes of the missing data based on the features of the load data.At the same time,the corresponding completion performance evaluation index is given for the subsequent missing value completion research.In order to mine more load data features,the modal decomposition of load time series is studied,and the random components,oscillation components,and trend components of load data are extracted.(2)In order to solve the problem of missing load data at a single station,a matrix completion framework based on modal decomposition and reconstruction was proposed,which decomposes the load time series in a station region to address the issue of missing load data at a single station.A low-rank matrix completion model is created at the same time as the component matrix is reorganized in line with the sample entropy feature and additional load data information is collected by regularizing the truncation function.The low-rank matrix completion model is built as part of the optimization process,and the alternating direction method of multipliers(ADMM)is used to solve the Lagrangian function.The ability of this approach to complete the missing load data for a single station region and the effectiveness of this framework to improve recovery are confirmed by simulation research.(3)In order to solve the problem of missing load data in multi-station areas,a tensor completion framework based on modal decomposition and reconstruction was proposed in this paper.Because the changes of load data with strong correlation show similar characteristics,the same component matrices of different stations are combined into tensor structures with natural high sparsity.Making use of the smoothness of load data in various dimensions,total variational regularization constraints are added to the decomposition of tensor expansion matrix,a low-rank tensor completion model is constructed,and the block coordinate descent method(BCD)is used to solve the problem,and achieved the completion of missing data.The simulation results show that the method makes use of more internal characteristics of load data and can complete the missing data of multiple stations at the same time.