基于BP神经网络优化算法的供热管网泄漏诊断

发布时间：2018-02-27 11:05

本文关键词： 集中供热管网 BP神经网络粒子群算法泄漏诊断相似实验　出处：《大连海事大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着集中供热管网规模的持续扩大以及部分管线运行年数的增加,管网在供热运行中不可避免的会发生各类故障,其中管网泄漏问题更为常见。这不仅会给人们冬季的生产生活带来安全隐患,还会造成能源浪费带来巨大的经济损失。因此如何对管网运行数据进行科学的分析、预测制定良好的诊断方案,尽早的发现管网泄漏故障,具有切实的经济意义和社会意义。本文结合实际供热工程,首先,以供热信息管理平台为依托,从供热运行实测数据入手分析,通过利用人工神经网络良好的非线性逼近能力,选择BP人工神经网络建模,将平台采集的正常供热工况下的室外温度、二次网供水温度数据作为学习样本对二次网回水温度进行建模,较为准确的构建了回水温度的正常工况下的预测模型,误差范围与模型精度能够满足要求。其次,由于传统BP神经网络初始权值和阈值不确定,网络学习中易于出现收敛速度慢、运行结果不稳定的现象,为此,采用粒子群优化算法改进BP神经网络,经MATLAB仿真研究,新算法加快了BP神经网络的收敛速度,提高了网络预测精度。最后,设计了一个半实物仿真实验装置,通过相似实验验证:借助对二次网回水温度的实时预测进行管网泄漏的诊断,实验中模拟了一系列的泄漏故障,结果表明采用该方法进行快速检测和诊断具有一定的有效性。随着集中供热的不断推广,针对供热管网泄漏故障的关注也将不断提升,以实现不断改进安全性,提高经济性的目的。本方法在供热领域的工程化应用中有待持续完善,同时,本研究对燃油、燃气、自来水等管道传输工程的故障诊断,也有一定的借鉴意义。
[Abstract]:With the continuous expansion of the scale of the central heating network and the increase of the number of years of operation of some pipelines, all kinds of failures will inevitably occur in the operation of the pipe network. The leakage problem of pipe network is more common. This will not only bring safety hazard to people's production and life in winter, but also cause huge economic loss of energy waste. Therefore, how to scientifically analyze the operation data of pipe network, It is of practical economic and social significance to make a good diagnosis plan and find the leakage fault of the pipe network as soon as possible. This paper combines with the actual heating engineering, first of all, based on the heating information management platform, Based on the measured data of heating operation and the good nonlinear approximation ability of artificial neural network, BP artificial neural network is selected to model the outdoor temperature under normal heating condition collected by the platform. The secondary network water supply temperature data is used as a learning sample to model the backwater temperature of the secondary network. The prediction model of the backwater temperature under normal working conditions is constructed accurately. The error range and the model precision can meet the requirements. Because the initial weight and threshold of traditional BP neural network are uncertain, the convergence speed is slow and the running result is unstable in the learning of the network. Therefore, the particle swarm optimization algorithm is used to improve the BP neural network, and the simulation results of BP neural network are studied by MATLAB simulation. The new algorithm accelerates the convergence speed of BP neural network and improves the prediction accuracy of the network. Finally, a semi-physical simulation experimental device is designed. Through similar experiment verification: through real-time prediction of backwater temperature of secondary network to diagnose pipe network leakage, a series of leakage faults are simulated in the experiment. The results show that this method is effective for rapid detection and diagnosis. With the continuous popularization of central heating, the attention to leakage faults of heating network will be raised continuously, so as to realize the continuous improvement of safety. The purpose of improving economy is to improve the engineering application of this method in the field of heating. At the same time, this study has some reference significance for the fault diagnosis of pipeline transmission projects such as fuel, gas, tap water and so on.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TU995.3;TP183

【相似文献】