Long-term runoff forecast method based on dynamic adjustment particle swarm optimizer algorithm and Holt-Winters linear seasonal model

To improve the reservoir long-term runoff forecast accuracy and speed, dynamic adjustment particle swarm optimizer algorithm (DAPSO) was proposed. Prediction function of Holt-Winters (H-W) linear seasonal model was used to deal with runoff forecast, the combination parameters of Holt-Winters linear seasonal model was solved and optimized by using DAPSO algorithm. The hybrid algorithm of the dynamic adjustment particle swarm optimizer algorithm and Holt-Winters linear seasonal model was developed. It can automatically determine the parameters of the Holt-Winters linear seasonal model. The long-term runoff forecast model was formed based on the hybrid algorithm. The reservoir long-term runoff forecast was carried out by using the method and history runoff data. The result shows the convergence of method was faster and forecast accuracy was more accurate than that of the particle swarm optimizer algorithm-H-W model and Holt-Winters linear seasonal model. The method improves forecast accuracy and forecast capacity of the H-W linear seasonal model. It had a high computational precision, and in 95% of confidence level the average percentage error was not more than 6%. The hybrid algorithm model can successfully improve the reservoir long-term runoff forecast accuracy and speed problem in Shiquan Reservoir. The model has better prediction accuracy and may be used for long-term runoff prediction of the reservoirs and rivers.