论文部分内容阅读
本模型基于BP神经网络仿真学原理,结合水文学理论研制。模型输入为:何巷闸下水位、区间流入量(包括何巷闸流入量和四方湖闸流入量)、新胡洼闸过去时段的出流量和闸上水位,输出即为预见期的新胡洼闸上水位。采用Sigmoid激活函数输出限幅改进算法,分别建立了胡洼闸上8h、16h、24h三种滞时BP神经网络洪水预报模型。经检验分析,三种滞时BP神经网络洪水预报模型预报精度均达到洪水预报规范中甲级精度要求。
The model is based on BP neural network simulation theory, combined with hydrological theory development. The model inputs are: the water level of the alluvial gates, the inflow of sections (including the flow of the alluvial gates and the flow of the Si Fanghu gates), the outflow of the Xinhu gate in the past periods and the water level on the gates, and the output is the Xinhu Wa brake on the water level. Sigmoid activation function output limiter improvement algorithm was used to establish Huwa brake on the hysteresis of 8h, 16h, 24h three delay BP neural network flood forecasting model. After the test and analysis, the forecasting accuracy of the three kinds of delayed model of BP neural network flood forecasting model meets the Grade A precision requirements of the flood forecasting standard.