Research On Power Load Forecasting Based On K-means And FP-growth Algorithm

With the rapid development of big data technology,more and more applications are obtained in the traditional industry.Big data combined with traditional power systems to build a smart grid,which realizes the safe,stable and efficient operation of the grid.Power big data is obtained by data sensors and RTUs(Remote Terminal Unit)that are widely distributed in the operation of power systems.The data structure is complex and large,and the hidden data value is also huge.In this dissertation,aiming at the data defects existing in the data collection of the Recurrent Nerual Network(RNN)short-term load forecasting model,resulting in low accuracy of load prediction,this dissertation uses association rules and clustering algorithm in data mining to pre-process the collected data sets,and realizes the short-term load forecasting of RNN model and improves the forecasting accuracy.Many scholars have done a lot of research on power load forecasting,but the single predictive model is not efficient for curre nt increasingly redundant load data sets.Therefore,this dissertation uses improved association rules mining algorithms based on data mining algorithms to process the collected data sets.Because the collected data sets contain multiple perturbation factors,in order to mine its main factors,the dissertation uses the improved K-means algorithm to cluster it,and uses the processed clustering data set to realize the load forecasting based on RNN model.The main research contents include:1.The problem of load clustering and load forecasting under power big data is studied.In allusion to the characteristics of the power load data set,the various processing methods of large data sets are studied.Combined the traditional power big data processing method with the traditional load forecasting model,the data mining processing method for the association rule mining algorithm and clustering algorithm is improved.2.Based on the density canopy K-means algorithm,the power load data set clustering is divided to improve the prediction accuracy of the RNN model.The basic principles of clustering and association rules mining algorithms are studied,and the superiority is compared.In allusion to fact that the collected data sets contain various factors affecting the power load,the dissertation uses the improved P-growth and NP-growth algorithms to mine and determine the most influential factors between the weather conditions and temperature.Before the RNN model is used for load forecasting,the K-means algorithm based on density canopy is used for data clustering,which improves the accuracy of load forecasting.3.Based on the association rule mining and clustering processing of the collecting data sets,using RNN prediction model to select the experimental data set for experimental comparison.The simulation results show that the prediction is improved.