论文部分内容阅读
根据超超临界锅炉汽水分离器的结构特点建立了三维有限元模型,以某1 000MW机组为例模拟了其启动过程的总应力场.将有限元法和神经网络法相结合,以有限元计算结果作为训练样本,以介质压力和筒体壁温序列为辅助变量,建立了基于Elman神经网络的分离器应力动态软测量模型,通过模型的训练,确定了准确的应力预测模型结构.应用电厂实际运行监测数据对所建立的Elman网络软测量模型进行验证,结果表明:模型计算结果可很好地逼近有限元结果,预测精度高,实时性好,可为锅炉寿命的在线监测提供数据支持.
According to the structural characteristics of the ultra-supercritical boiler water separator, a three-dimensional finite element model was established, and the total stress field during the startup process was simulated with a 1000MW unit. The finite element method and the neural network method were combined, As a training sample, based on the pressure of medium and the cylinder wall temperature sequence as auxiliary variables, the stress dynamic soft-sensing model of splitter based on Elman neural network was established. Through the training of the model, the accurate stress prediction model structure was determined. The monitoring data validates the established Elman network soft-sensing model. The results show that the calculated results of the model can approximate the finite element results well, and the prediction accuracy is high and the real-time performance is good. It can provide the data support for the on-line monitoring of boiler life.