Research On Short-term Power Load Forecasting Based On Spark Platform

Short-term power load forecasting has attracted more and more attention with the growth of power energy consumption in recent years.Accurate short-term power load forecasting is not only the decisive factor to ensure the safe and stable operation of power system,but also an important support for the government’s economic planning.This paper proposes a shortterm power load forecasting model based on time series decomposition to improve the accuracy of short-term load forecasting,designed the parallel computing of the predictive model on the Spark platform,and built a power load data analysis platform that integrates data visualization,data analysis,and data management.The main work and results of this paper are as follows:(1)In order to solve the problem that a single model is difficult to effectively fit the trend of load data,this paper adopts the time series decomposition method EEMD and combines meteorological factor data to put forward the EEMD-SVR-ARIMA load forecasting model.The model decomposed the load data sequence into several sub-sequences with different frequencies,and decomposed the meteorological factor data sequence to obtain the meteorological data components.Pearson correlation analysis method was adopted to select the appropriate components as the characteristic input of the model,and then combined with ARIMA and SVR algorithm to forecast the load data sub-series.Finally,the prediction model results of each sub-sequence are superimposed to get the final prediction results.The experiment proves that this method can effectively improve the prediction accuracy of load data.(2)The EEMD-SVR-ARIMA model improves the prediction accuracy of the model at the expense of time and space complexity,and the model needs to be trained for a long time in a stand-alone environment.Therefore,the parallel calculation of the prediction model is designed and realized in combination with Spark distributed computing platform.The experimental results show that the prediction effect of the model is very close to that of the single machine environment,and the running time of the model is greatly shortened.(3)In order to realize the load data of visual management,this article with the aid of the Spring Boot development framework,design and implements the power load data analysis platform,the platform consists of user management,data analysis and data management of the three major modules,data management,integrated with load load data outlier detection,load forecasting,load data visualization,and other functions,Through the test to the function of the platform,it shows that the platform has good interaction and relatively complete function.