Application of moving windows autoregressive quadratic model in runoff forecast

This paper describes a novel method to mid-long term runoff prediction using moving windows autoregressive quadratic model which combines autoregressive quadratic model and moving windows method to improve prediction capability of natural runoff. The parameters of the model are determined in light of the joints of half-sine function, self-adaptive optimization, smoothly moving windows and generalized likelihood uncertainty estimation. The application shows that the model can not only improve prediction capability but keep robust, and shows that the model has simpler structure and less parameter than artificial neural networks model, and avoids locally minimal point and excess study, etc. Therefore, the moving windows autoregressive quadratic model is a promising tool for mid-long term runoff forecast.