基于BP神经网络汽水分离器组合建模研究

Modeling Research of Steam-water Separator Combination Based on BP Neural Network

在热工系统建模中,若建立一个全工况模型,传统上会采用集总参数建模方法,但在建模过程中,模型中存在许多经验参数,这将导致机理模型精度较低,然而很难通过简单的试验建模在全工况范围内建立模型。因此提出了一种机理建模和神经网络建模相结合的组合建模方法,它不仅提高了模型的精度,而且可以应用到更大的范围。根据某电厂1 000 MW超临界机组的历史运行数据,依据典型工况(例如不同时期的负荷),在Matlab平台上对汽水分离器组合模型进行仿真和测试,仿真结果显示,神经网络模型在训练之后,该模型能够有效地预测机理模型中的未知参数,并且所构建的组合模型也能够有效地模拟该系统现场典型工况,通过机理建模和神经网络建模相结合的组合建模方法具有一定的通用性。

In the modeling of thermal engineering systems,if a full working condition model is established,the lumped parameter modeling method is traditionally used,but in the modeling process,there are many empirical parameters in the model,which will lead to a lower accuracy of the mechanism model. However,it is difficult to establish a model over the full range of conditions through simple experimental modeling. Therefore,a combination modeling method combining mechanism modeling and neural network modeling was proposed. It not only improves the accuracy of the model,but also can be applied to a larger range. Based on the historical operating data of a 1 000 MW supercritical unit in a power plant,and based on typical operating conditions(such as loads at different periods),the steam-water separator combination model was simulated and tested on the Matlab platform. The simulation results show that after the neural network model is trained,the model can effectively predict the unknown parameters in the mechanism model,and the combined model constructed can also effectively simulate the typical working conditions of the system site. The combined modeling method of combining mechanism modeling and neural network modules has certain versatility.