基于混沌理论的自适应模糊推理网络系统在径流时间序列预测中的应用

Application of ANFIS in Runoff Time Series Forecast Based on Chaos Theory

水文期预报对水资源管理、调度及社会的生产、生活具有十分重要的意义。针对常规混沌预测方法的局限性,提出基于混沌理论的自适应模糊推理网络系统的径流时间序列预报方法。该方法径流时间序列被分解为趋势项、周期项和随机项,对随机项进行混沌辨识,然后建立有自适应能力的神经网络模糊推理模型对随机项进行预测,最后将各项线性叠加进行径流预报。实例表明,该方法预测精度较高,具有良好的泛化推广能力。

Hydrological forecasting is of great significance to the successful water resources management and scheduling,social production and life.According to the limitation of the past chaos forecasting methods,the combined method of runoff time series forecast was developed,based on ANFIS and chaos theory.The runoff time series were decomposed as trend component,periodical component and random component,and the existence of chaos was determined in the random component.Then the model of ANFIS was built to forecast the random component.Finally,all the components were superposed in a linear way to forecast the runoff.Examples showed that the method had good prediction accuracy and generalization ability.