BP网络在电站故障诊断系统中的应用研究

针对目前大型电站测点传感装置损坏率高，导致有些分析软件失效的现状，应用人工神经网络原理，设计了一个自适应BP网络模型，进行实时判别及仿真。以某一实际机组为例，对其参数进行了仿真计算，绝大多数数据的相对误差在1．5％以内，可以满足工程实际的需要。同时，对输入／输出参数之间的关联程度，对影响输出结果的精度、收敛速度等因素进行了分析比较，这对今后的仿真结果有很好的借鉴意义。因此，该模型对动力系统的热力参数在线仿真，减少传感器的维护量，尤其是提高基于参数采集应用软件的可靠性具有较大的实用价值。

Application research of BP algorithm in fault diagnosis system of power plant

The failure of sensor equipment applied in large power plant causes the failure of analysis software.A self adaptable BP network model based on the theory of artificial neural network in proposed,which carries out the real time check and simulation.As a practice example,the simulative calculation for some key thermal parameters of the real units is presented.The results show that the relative errors of most data are below 1.5% and this model can meet the engineering requirement.Comparison and analysis are made for the relationship between input and output parameters,its effect on the accuracy of output results and the factors in convergence speed,which are referential for later simulation results.Therefore this model is very useful in the online thermal parameter simulation of dynamic systems to reduce sensor maintenance and increase software reliabillity applied in data acquisition.