基于神经网络的天然气流量计检定工艺智能控制系统

Neural network based intelligent control system of natural gas flowmeter verification process

随着天然气贸易交接计量站点和交接量的逐年增多,天然气流量计检定需求量不断增大,常规人工检定方法已经无法满足客户需求,亟需一套安全、高效、智能的检定方法。为了提升天然气流量计检定站场检定效率,以广州天然气大流量检定站为例,基于BP神经网络算法,结合现场生产数据构建站场水力仿真模型和流量调节控制器,实现了流量计检定工艺的智能控制。研究结果表明:①天然气流量计的实流检定工艺复杂,检定准确度要求高,调节频繁难度大,常规的机理建模与简单反馈控制方法无法实现检定工艺的自动控制;②基于BP神经网络构建的检定站水力仿真模型在保证计算速度的前提下较好地模拟了站场的实际流动状态;③结合现场专家经验与神经网络的泛化能力建立的天然气流量计检定智能控制器可以根据当前工况给出合理的开阀方案;④以检定站水力模型与智能控制器为核心所构建的智能检定软件经过实验检验,可以较好地服务于现场。结论认为,该研究成果可以为天然气流量计检定智能化和应用落地提供有益的参考。

As the number of natural gas custody transfer measurement stations and the amount of natural gas custody transfer increase continuously year after year,the demand for natural gas flowmeter verification increases continuously,but the conventional artificial verification method cannot meet customers'needs,so a set of safe,efficient and intelligent verification method is in an urgent need to improve the verification efficiency of natural gas flowmeter verification station.Taking Guangzhou high-rate natural gas verification station as an example,this paper constructs a hydraulic simulation model and a flow regulation controller for verification stations based on BP neural network algorithm,combined with field production data,so as to realize the intelligent control of flowmeter verification process.And the following research results are obtained.First,the real-flow verification process of natural gas flowmeter is complicated,the verification accuracy is highly required and the regulation is frequent and difficult,so conventional mechanism modeling and simple feedback control method cannot realize automatic control of verification process.Second,the hydraulic simulation model of verification station constructed based on BP neural network can better simulate the real flow state of the station while ensuring the calculation speed.Third,the intelligent controller of natural gas flowmeter verification which is constructed by combining the field expert experience with the generalization ability of neural network can provide an appropriate valve opening scheme according to current working conditions.Fourth,the intelligent verification software developed by taking the hydraulic model and the intelligent controller of verification station as the core is experimentally tested,which indicates it can serve well on site.In conclusion,the research results can provide beneficial references for the intelligentization and application practice of natural gas flowmeter verification.