Combination model for short-term load forecasting

Abstract

Gas demand possesses dual property of growing and seasonal fluctuation simultaneously, it makes gas demand variation possess complex nonlinear character. From previous studies know single model for nonlinear problem can't get good results but accurately gas forecast were essential part of an efficient gas system planning and operation. In recent years, lots of scholar put forward combination model to solve complex regression problem. In this paper, a new forecasting model which named regression combined neural network is presented. In this approach we used regression to model the trend and used neural network for calculating predicted values and errors. And to prove the effectiveness of the model, support vector machines(SVM) algorithm was used to compare with the result of combination model. The results show that the combination model is effective and highly accurate in the forecasting of short-term gas load and has advantage than other models. © Chen et al. Licensee Bentham Open.