BP神经网络在渠系自动控制中的应用

Application of BP Neural Network to Canal System Automation

就渠系自动控制的问题，比较分析了常用的两种渠系建模方案，提出利用改进BP网络进行渠系辨识，然后在线自动控制。通过改进BP网络算法及模糊技术，克服BP网络所存在的容易陷入局部极小点、对初值要求高、训练时间长等缺点。仿真研究表明，在渠系自动控制中，利用神经网络技术是一种比较理想的方法。

This paper proposes a scheme using an improved BP netWork for canal system recognition andthen automatic controlling on-line, comparing and analysing the two commonly used modeling methods ofcanal system. Drawbacks of the conventional BP network, such as easilly falling into the local minimum, requiring strict initial conditions, long training period and so on, can be overcome by means of the modifiednetwork algorithms and fuzzy techniques. By computer simulation,it is shown that the nearal network techniquecan be used for the canal system automation, effectively and accurately.