基于BP神经网络的尼尔基水库短期洪水预报研究

Research on Short Term Flood Forecast for Nierji Reservoir Based on BP Neural Network

人工神经网络(ANN)诞生于20世纪40年代,兴起于80年代,以其非线性特性、大量的并行分布结构以及出色的学习和归纳能力,广泛应用于水文各要素的预测分析。本次研究利用ANN中最常见的BP(Back Propagation)神经网络,由上游3个控制站(即柳家屯、库漠屯和科后)的流量,模拟出未来24 h的尼尔基入库流量,分析二者的非线性关系以用于水库短期洪水预报。预报方案中各预报要素精度均达甲级水平,可用于实际作业预报。

Artificial neural network was emerged in the 1940s,rose in the 1980s.And it is widely used in forecast and analysis of the hydrological elements due to its characteristics of the non-liner,a large number of parallel distribution structure and excellent study and conclusion capacity.In the study,the BP（Back Propagation） neural network,the most common in ANN,was used to simulate the inflow to the Nierji Reservoir in future 24 hours based on discharges from three control stations（Liujia tun,Kumo tun and Kehou tun） on the upstream to analyze the nonlinear relationship for the short period forecast of reservoir flood.The accuracy of each forecasting element in the forecast scheme reaches the level of Grade A and may be used in practical forecast work.