论文部分内容阅读
结合模糊逻辑与神经网络优点,提出T-S模糊神经网络(TS-FNN)多元变量年径流预测模型,以新疆伊犁河雅马渡站年径流预测为例进行实例分析。利用实例前16年实测资料对TS-FNN模型进行训练,后7年资料进行预测,预测结果与文献IEA-BP模型、GA-Elman模型和多隐含层BP神经网络模型(MHL-BP)的预测结果进行对比。结果表明:TS-FNN模型对实例后7年年径流量预测的平均相对误差绝对值和最大相对误差绝对值分别为3.51%、6.87%,预测精度及泛化能力均优于相关文献的预测结果。TS-FNN模型兼顾了神经网络与模糊逻辑的优点,具有较好的预测精度和泛化能力。
Combined with the advantages of fuzzy logic and neural network, a multi-variable annual runoff forecasting model based on T-S fuzzy neural network (TS-FNN) was proposed. The annual runoff forecast of Yalongdu Station in Yili River of Xinjiang was taken as an example to conduct an example analysis. The TS-FNN model was trained using the measured data of the first 16 years and the data of the latter seven years were predicted. The prediction results are in good agreement with those of the literature IEA-BP model, GA-Elman model and multiple implicit BP neural network model (MHL-BP) The forecast results are compared. The results show that the absolute value of the relative error and the maximum relative error of TS-FNN model are 3.51% and 6.87%, respectively. The prediction accuracy and generalization ability of TS-FNN model are better than those of the related literature . TS-FNN model takes into account the advantages of neural network and fuzzy logic, and has better prediction accuracy and generalization ability.