Research And Application Of Data Mining Algorithm On Power Big Data

The big data industry in our country is in a period of rapid development,of which the development of power big data is also an important part of national data strategy.With the development of sensors and intelligent devices,the sources of power data are increasing and the data types are becoming more and more complicated.Therefore,it is urgent to excavate the information hidden in the massive data with the technology of big data,which can quicken the development of power big data and improve the operating efficiency of the power system.Electrical load prediction is the basis for ensuring the safe production and power supply quality of power system,and it is helpful to the efficient operation of power system.The accuracy of the electrical load prediction is related to the economic operation of the whole system.The high accuracy prediction can not only meet the demand of users,but also help the power supplier to arrange the scheduling reasonably and reduce the cost,so as to improve the social satisfaction and resource utilization.In this paper,we improve the relevant algorithm to improve the accuracy of electrical load prediction based on the data mining technology and the application background of electrical load prediction.This paper improves the stability and generalization performance of Random Forest algorithm and Extreme Learning Machine algorithm to improve the accuracy of electricity load prediction and solve the problem of simulating the actual distribution of electricity load.In this paper,we implement the classification model,regression model and time series model of combining the WEKA which are based on random forest to forecast the electricity load data of a certain province.After a large number of experiments and evaluation on different models,we found that the three models can reasonably predict the future of electricity load data.In addition,under the same evaluation index the model which combine the random forest algorithm and time series of WEKA can get better result when predicting the future moment of electricity load data.Electrical load prediction plays an important role in power system planning and operation of smart grid.Extreme learning machine has been widely used for many prediction problems from different areas due to its advantage on fast convergence and good generalization performance.However,how to handle the large scale electrical load data with various types and low density of value and improve the stability of the ELM for electrical load prediction is still a challenging task.To improve the stability of ELM,a novel Extreme Learning Machine based on Improved Particle Swarm Optimization named IPSO-ELM is proposed in this paper.By combining the improved particle swarm optimization with ELM,IPSO-ELM can find the optimal number of nodes in the hidden layer as well as the optimal input weights and hidden biases.Furthermore,a novel mutation operator is also introduced in IPSO-ELM to enhance the diversity of swarm and improve the convergence speed of the random search process.Then,to handle large-scale electrical load data,a parallel version of IPSO-ELM named PIPSO-ELM is implemented with the popular parallel computing framework Spark.Experiment results of real-life electrical load data show that PIPSO-ELM can obtain better stability and achieve high efficiency and scalability in large-scale electrical load prediction.