Research On Exception Analysis And Forecasting Of Line Loss Based On Big Data Theory

Line loss management plays an important role in the development of power grids in a region.The abnormal analysis and accurate prediction of line loss can play a guiding role in power grid development planning and loss reduction measures.With the popularization of intelligent electric meter and the continuous improvement of information collection system of electricity,the era of large power data has come.The existing method of identifying line loss anomaly is based on manpower and has limitations in the improvement of timeliness and accuracy.In addition,the traditional method of line loss prediction can not accurately reflect the data sequence characteristics in the case of large datasets,and the prediction precision is limited.So how to combine the data mining and depth learning theory to solve the problem of traditional methods in the abnormal analysis and prediction of line loss is of great significance.This paper mainly studies the power line loss analysis method based on large data technology,which is mainly as follows(1)This paper investigates the current research situation of line loss calculation,abnormal line loss analysis and line loss prediction,and analyzes the current situation and future development trend of data mining,deep learning and big data technology.(2)The principle and results of several kinds of theoretical line loss calculation methods(RMS current method,equivalent resistance method and line loss calculation method based on power flow calculation)are compared,and the advantages and disadvantages of different methods and the applicable scenes are obtained.(3)Based on the combination model of three different data mining technologies,the paper analyzes the abnormal line loss,establishes the abnormal sample database of the network operation data,puts forward the distinguishing of the abnormal occurrence of the line loss,the time of abnormal occurrence and the location of the anomaly location,and finally verifies the accuracy and feasibility of the method.(4)Based on the theory of depth learning,this paper analyzes the defects of the standard recurrent neural network to the long sequence modeling,puts forward the advantage of using the Long Short-term Memory Network for the learning long-term dependence problem,establishes the line loss prediction model based on the long short-term memory network under the complete dataset.By utilizing the memory unit's powerful memory,the characteristics of long time series of line loss are extracted,stored,used in proportion,feedback to network and then extracted in order to obtain the change rule of line loss curve effectively.The prediction accuracy of the model is improved by comparing with the predicted results of the models and standard RNN and traditional neural network.(5)In the case of lack of data,based on the theory of large data,a method of data restoration for sequential patterns is proposed,and a numerical example is given to show that the method can effectively solve the problem of low accuracy of data incomplete using traditional methods under missing data.With the development of the related technology of smart grid,the high quality power data can be gained and the data diversity is increasing.The analysis methods of large data analysis,data mining and depth learning theory not only have low data integrality requirement,but also have strong adaptability and precision,and will increase the function of the future line loss analysis,and lay the foundation for the future fine management of power system.