针对大型火电机组广泛采用的对分、表面式凝汽器，结合其结构、性能特点和现场实际情况，确立 了便于工程应用的对分式凝汽器典型故障知识库。在此基础上采用ＢＰ神经网络方法实现对分式凝汽器故障诊断。提出一种恒误差修正率控制的网络学习率自适应调整方法，大大缩短了网络训练的收敛时间。 Aiming at the widely used sub-surface and surface condensers of large-scale thermal power units, a typical fault knowledge base of the fractional condenser for engineering application is established based on its structure, performance characteristics and field conditions. On this basis, BP neural network method is used to realize fault diagnosis of fractional condenser. This paper proposes a method of adaptive adjustment of network learning rate controlled by constant error correction rate, which greatly shortens the convergence time of network training.