基于BP网络的热力系统参数仿真

针对热力系统的传感器损坏率较高，导致有些分析软件失效的现状，提出了一种改进ＢＰ算法的人工神经网络，对一些重要的热力参数进行了仿真研究．在该方法中，网络学习过程通过同时调整学习率和动量修正因子两个参数，使得收敛速度沿着最佳方向进行．经实例计算结果表明，真实值和仿真值之间的误差在１％以内，可以满足工程应用的需要．文中最后给出的在调峰机组的寿命管理系统中的应用实例，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的应用软件的可靠性具有较大的参考价值

Parameter Simulation in the Thermal System Using a Kind of Improved Artificial Neural Network Based on the BP Algorithm

In order to solve the invalidation of fault analytic software. a kind of improved artificial neural network based on the BP algorithm was proposed and used to simulate some key thermal parameters at real time. This method, which can antomatically and simultaneously adjust learning and momentum factor in training process, results in convergence speed up. The analytic result shows that the error between simulative and real velue is less than 1%. Finally, a sample successfully applied in the life management system of steam units starting up and shutting down frequently days and nights was given. Experimental and calculative results are very useful for online simulation of the thermal parameters of the dynamical systems and reduce maintenance of the sensors as well as increase reliability of the applied software based on data acquisition.