Identifying Bad Data And Power Load Forecasting Of Power System Based On Spark

The continuous improvement of the information and intelligence of the power grid makes the power data more and more large,which brings great difficulties to the processing and analysis of the data.In the processing of the application of smart grid,big data,data storage,efficient processing and real-time visualization of multi-source heterogeneous data integration and data aspect is facing serious challenges,need further research on these aspects,and develop large data in ensuring the safe and stable operation of power grid.Power big data mining is the basis of the electric power data platform,through the big data platform which can realize the smart grid data sharing,for all kinds of data storage,processing,analysis and application of excitation power market potential,and excavate the value of the electric power data contains.For abnormal data in power system will Reduce the accuracy of power system state estimation results and traditional process mass high-dimensional data clustering algorithm of computing resources shortage,the Map-Reduce framework cannot effectively deal with problems such as frequent iterative calculation.This article from the abnormal data detection and correction and two respects of short-term load forecasting power big data and the depth of the smart grid integration,based on the Spark big data computing platform,an improved ISODATA clustering algorithm to detect abnormal data and correction;Based on Spark big data computing platform,XGBoost algorithm is used to change the history of electricity data,date,weather data,and other features of different kinds of data,through the feature extraction and feature conversion,build the model for short-term power load forecasting.