Research On Short-term Load Forecasting Of Power System Based On Random Forest Algorithm

With the development of economy and society and the improvement of people's living standards,electric energy has become an indispensable energy in social production activities.Every link of electric energy from production to use is inseparable from the role of power system scheduling,and the rapid development of power grid makes the complexity of power system increasing.Accurate short-term load forecasting of power system is very important to the safe and stable operation of power system and the economic dispatch of power grid.It not only affects the daily life of residents,but also reduces the waste of resources.It has always been an important research topic in the field of electrical engineering.The task of short-term load forecasting becomes complicated because of the influence of many factors.This paper introduces the research status of short-term load forecasting.Based on the analysis of the characteristics and influencing factors of load forecasting,it compares different decision tree algorithms,and uses cart decision tree to construct random forest.The random forest algorithm is applied to the field of short-term load forecasting,and the better effect is obtained through the analysis and improvement of the algorithm.In the traditional random forest model,the number of decision trees and the number of split features are selected according to experience.However,in the face of different research objects,the parameters that can make the random forest performance reach the optimum are different.To solve this problem,this paper uses particle swarm optimization algorithm to optimize the parameters,and the optimal parameters of random forest applied to short-term load forecasting are obtained.Experimental analysis shows that the prediction model optimized by particle swarm optimization for random forest has higher prediction accuracy than traditional prediction model.The traditional random forest model and the particle swarm algorithm optimized random forest model are both single-layer models.According to the statistical learning theory,one-time algorithm can only read part of the effective information of the corresponding space,which affects the prediction performance.In response to this problem,this paper proposes to use the double-layer random forest algorithm to establish the model for short-term load forecasting,and at the same time use the particle swarm optimization algorithm to optimize the double-layer random forest.Connect the two layers of random forest through training residuals,that is,put the training residuals of the first layer of random forest into the training set to participate in the training of the second layer of random forest,so that it can read more effective information in the space.The final result of the model can be obtained by adding the results of the two layers of random forest.In addition,this paper selects the similar day of the day to be predicted by the grey relational analysis method,and adds the similar day data to the training samples to increase the comprehensiveness of the training data.The simulation results show that the short-term load forecasting model based on the double-layer random forest algorithm has higher prediction accuracy and more stable prediction performance.