汽轮机在线性能分析BP神经网络模型

为了提高汽轮机在线性能分析系统的健壮性、精确性和抗干扰能力，对BP网络算法进行拓展，增强其网络功能。使之能够分析网络输入参数偏差对输出参数的影响，建立了一种新的汽机在线性能分析模型．与常规在线分析模型相比，该模型只需常规模型输入参数个数的1/5，大大减小了测量设备损坏对在线系统的影响；而且该模型分析结果精确、可靠，测定的热耗结果与常规分析结果误差不超过1％．另外，该模型容错性能良好，参数在2％内波动对分析结果无太大影响．

Model of Turbine's Performance Analysis Based on Error Back Propagation Artificial Neural Network

In order to make the turbine's online performance analysis system stronger and more accurate, this article enhances the ability of error back propagation network to make it able to analyze the influence of inputs on outputs and builds up a new model to analyze the performance of the turbine. In contrast with the normal model it needs only 1/5 inputs that the normal model needs, and its results are more accurate than that of the normal model. In addition, the model is still strong with 2% input undulation. The model improves the quality of the turbine's online performance analysis system, and is of benefit to the economical operation in power plant.