Application of rough set theory and artificial neural network for load forecasting

Load forecasting plays a key role in power system operation and planning. However, accurate forecasting model requires an ability to select relevant factors so that the influences of irrelevant factors can be reduced substantially. Rough set theory in data mining, which provides a useful tool to analyze data can help solve the above problem. This paper proposes a novel hybrid method to integrate rough set theory, genetic algorithm and artificial neural network. Our method consists of two stages, in the first procedure, rough set theory and genetic algorithm are applied to find relevant factors to the load, the results are used as inputs of neural network, in the second procedure, active selection of training set is carried out by k-nearest neighbors, artificial neural network is used to predict load. The method is characterized not only by using attribute reduction as a preprocessing technique of artificial neural network but also by presenting an improved attribute reduction algorithm. The prediction accuracy is improved by applying the method on a real power system, which shows that the proposed method is promising for load forecasting in power system.