Least Squares Estimation of Polynomial AR Models to Forecast Short Term Power Demand

The paper considers estimation problems of polynomial AR models based on least squares methods. The polynomial AR models are a kind of nonlinear AR models. System identification problems of the polynomial AR models are reduced to least squares problems using multiple data if the nonlinear functions are assumed to be polynomial. The estimated polynomial AR models are used as electrical power demand forecasting. A numerical example is shown to demonstrate the effectiveness of the proposed method about short term demand forecasting in Tokyo.