Research Of Micro-grid Load Forecasting Based On Cloud Computing And Machine Learning Algorithms

With the energy and environmental pollution problems become more and more serious,comprehensive development and rational utilization of new energy sources is imperative,and the construction of the micro grid can make new energy consumptive and optimize energy structure.Accurate load forecasting can not only provide an important basis for optimal operation and energy management,but also can ensure the efficient operation of micro grid.Therefore,this paper focuses on the short-term load forecasting problem of micro grid,which has important theoretical significance and practical value for the optimal operation of micro grid system.Based on the analysis of the micro grid load forecasting,using the improved shuffled frog leaping algorithm to optimize the kernel extreme learning machine parameters(ISFLA_KELM),with the Spark on YARN platform,the algorithm is improved by parallel computing,while ensuring that the prediction accuracy of the parallel computing to deal with the challenges of big data.And the use of a micro grid real load data validation accuracy and efficiency.This article mainly carries on the following several aspects of the work.(1)The problems of micro grid load forecasting are analyzed,and the advantages and disadvantages of different forecasting methods are studied.According to the characteristics and influencing factors of micro grid load forecasting,an intelligent optimization algorithm named shuffled frog leaping algorithm is suitable,aiming at the disadvantage of it,this article puts forward an improved shuffled frog leaping algorithm(2)Analysis of the principle of the SFLA optimization algorithm with its characteristics,determine the relative advantage of other optimization algorithms.Then the improved SFLA optimization algorithm and kernel extreme learning machine combination,this paper presents a novel micro grid load forecasting algorithm(ISFLA_KELM).The combination parameters of the kernel function extreme learning machine as the frog swarm optimization algorithm is optimized.And the parallel design of the improved algorithm is proposed.(3)Put forward the Spark ISFLA_KELM micro grid load forecasting algorithm based on the power of big data under the single computing resources,respectively in KELM and ISFLA time-consuming calculation algorithm for parallel design,combined with the Spark machine learning repository and distributed file system,improve the efficiency of the algorithm.(4)Experimental test and example analysis.The UCI standard data sets to provide real data,based on the proposed algorithm were tested in 8 node laboratory built Spark on yarn memory computing,and the existing load forecasting methods are compared.The experimental results show that the proposed algorithm is better than the existing algorithm,which can provide the effective basis for the load forecasting,and has better parallel performance.