人工神经网络在发电厂培训系统中的应用

Artificial Neural Network Designed by Genetic Algorithm and its Application to Power Plant Simulation Training System

为了解决用人工神经网络映射发电机暂态过程中，各个变量的变化曲线存在的网络结构以及权值和阈值的初始值难以科学地确定的问题，将遗传算法和传统的误差反传算法结合起来，对人工神经网络进行设计和训练。该方法在发电厂培训系统中的应用表明：这种方法提高了人工神经网络中权值和阈值的初始值的确定以及网络拓扑结构的设计的科学性，从而使人工神经网络容易出现的发散以及陷入局部极小点的问题得以避兔。

The genetic algorithm was utilized in design of topology of artificial neural networks and determi nation of the initial values of the weights and thresholds. The application of the artificial neural network designed by the above method in fitting of the generator variables curves in transient process has illustrated that it can make the design of topology of artificial neural networks and determination of the weights and thresholds more scientific. The divergence and local infinitesimal problems can be avoided.