A Short-Term Load Forecasting Model Based on Improved Random Forest Algorithm
Yiling Huang, Huang Shaofeng
Abstract
Aiming at the current situation of local optimization and over-fitting in current load forecasting algorithms, this paper proposes an improved power system load forecasting model based on random forest algorithm. Considering that the power system load has the characteristics of schedule periodicity and temperature correlation, this paper uses the same cluster of data obtained by fuzzy clustering as the input set to construct the decision tree when establishing the random forest. Considering that the random forest algorithm has the disadvantage of over-fitting, this paper uses rough set theory to generate compensation rules to modify the prediction results of random forest. Finally, the load forecasting model described in this article was used to predict the electricity load of a certain area on January 1st, 2020, and compared with the existing forecasting model. The results of calculation examples show that the method described in this paper has better performance.