Short term load forecasting based on wavelet decomposition and random forest

Data mining algorithm 1 are main stream and intelligent algorithms for short term load forecasting. However, traditional data mining algorithms (neural network and support vector machine) are subject to some hardly conquerable drawbacks such as being easy to fall into local optimum, poor generalization capability, and difficult to determine the model. To overcome the above issues, a new short-term load forecasting method based on wavelet decomposition and random forest regression(RF) is proposed. On the one hand, wavelet decomposition algorithm is a valid method to extract the load of different components as the training set. On the other hand, RF algorithm suffers less from the problem of over fitting and determining difficultly the model parameters. Wavelet analysis and RF are introduced and applied for short term load forecasting, which is helpful to improve the accuracy of load forecasting. The historic load data is selected from a certain area of Anhui province in this paper. Compared with the traditional BP neural network, support vector machine, and not improved RF, the proposed method has higher forecasting accuracy.