基于复合BP神经网络的天然气工况监测系统研究

Research on Working Condition Monitoring System of Natural Gas Based on Composite BP Neural Network

天然气站场无人化建设规划中,对于供天然气发电机组的电厂站,由于机组负荷大,工况变化复杂,建立有效的工况分析和预警机制一直是管网调度的难题。因此,应用Levenberg-Marquardt法和共轭双极法各自训练2个独立的BP神经网络,分别用于整体计算和小流量区修正,以此通过复合网络的形式建立天然气电厂站输配调节的流调模型,在此基础上通过比对相同工况下现场反馈流量和模型计算流量来实现工况监测和预警的目的。

In the planning of unattended operation of the natural gas station, it is a difficult problem to set up an effective working condition analysis and early warning mechanism for the power station of natural gas generating set due to the heavy load of the unit and the complex change of the working condition. In this paper, Levenberg-Marquardt method and conjugate bipolar method are used to train two independent BP neural networks, which are respectively used for global calculation and small flow area correction, and then a flow regulation model for natural gas power station transmission and distribution regulation is established by means of the composite network. On this basis, early warning and the monitoring of the working conditions can be realized by comparing the field feedback flow rate with the model calculation flow rate under the same working condition.