Random forest based ensemble system for short term load forecasting

The short term load forecasting plays an essential role in the operation of electric power systems. Plenty of features involved in the forecasting cause a complex system and the long training time. The curse of dimensionality also downgrades the generalization capability of the predictor. This paper applies the random forest based ensemble system to load forecasting application. Rather than selecting a subset of features, which may cause the information lost, all features are considered in the proposed method. Different feature sets are used to construct regression systems and the average method is used as a fusion. The performance of the proposed model is compared with another existing method based on mutual information feature selection using real load datasets in New York and PJM. Experimental results show our method achieves a better result in term of higher accuracy.