Load Forecasting Based On The Analysis Of User Electricity Behavior

With the advancement of smart grid construction,how to mine key information from the massive user data provided by smart meters,grasp the difference in power usage rules between different users,accurately predict the user's power load,and make scientific and reasonable for users.The green energy-saving electricity plan is an important criterion for measuring whether each power company has core competitiveness.The load forecasting in an open sales environment is different from the traditional extensive forecasting for the entire power supply area,which is a refined prediction based on the user.The research program is roughly divided into three parts,and the specific work of each part is as follows:1.The overall power usage behavior is preferred.In order to reduce the time complexity and computational cost of cluster analysis without affecting the clustering accuracy,the overall electricity behavior evaluation system is constructed,and the overall electricity consumption behavior of the user is screened.The screening results show that the system shows the system.Scientific and practical.2.Cluster analysis.In order to overcome the K-means algorithm,it is necessary to determine the K value in advance,and the clustering result is greatly affected by the initial clustering center.The Canopy clustering algorithm is combined with the K-means algorithm.The hybrid clustering algorithm can avoid the trial work of K value,reduce the randomness when the initial cluster center is selected,and significantly improve the convergence speed of the error square sum.3.Load forecasting.In view of the lack of online learning ability of I-ELM neural network and the need of OS-ELM to artificially determine the number of hidden layer nodes,I-ELM and OS-ELM two extreme learning machine algorithms are combined to propose I-OS-ELM.model.The model has both I-ELM's ability to find the optimal number of hidden layer nodes and the online learning function of OS-ELM.At the end of the paper,the I-OS-ELM model was used to train the Canadian electricity consumption data.The results show the validity and feasibility of the model.