The Annual Maximum Flood Peak Discharge Forecasting Using Hermite Projection Pursuit Regression with SSO and LS Method

Accurate prediction of extreme flood peak discharge is essential in developing the best management practices to avoid and reduce flood disaster. In recent years, many techniques have been pronounced as a branch of computer science to model wide range of hydrological process. Nevertheless, exploration of more efficient technique is necessary in terms of accuracy and applicability. In this study, a novel hermite-PPR model with SSO and LS algorithm is proposed for designing annual maximum flood peak discharge forecasting model at Yichang station on Yangtze River in China. The statistical properties of the data series are utilized for identifying an appropriate input vector to the model and then the performance of the proposed models were compared with adaptive neuro-fuzzy inference system (ANFIS), artificial neural network (ANN) and multiple linear regression (MLR) methods in terms of root mean squared error (RMSE), mean absolute relative error (MARE), coefficient of correlation (CC), Nash-Sutcliffe efficiency coefficient (NSEC) and qualified rate (QR). The results indicate that the presented methodology in this research can obtain significant improvement in forecasting accuracy in terms of different evaluation criteria during training and validation phases.