【系统仿真学报】煤气管道服役期预估神经网络模型的建立与应用.pdf

第22卷第2期 系 统 仿 真 学 报© Vol. 22 No. 2 2010年2月 Journal of System Simulation Feb., 2010 煤气管道服役期预估神经网络模型的建立与应用 张乐乐，郭 佳，宋守信 （北京交通大学机械电子与控制工程学院，北京 100044 ） 摘 要:神经网络模型具备的针对大量无规则数据自学习的能力，适于解决城镇煤气管道布局复杂、 影响因素众多、服役期限难以计算的问题。基于BP神经网络算法建立的煤气管道年限模型，经过 训练对比，选择了最优的BP 网络模型，结合共轭梯度法，改进了该算法收敛慢、易出现震荡的问 题，实现了对城镇管道漏损的定量分析，结合某市煤气管道漏损的实际案例验证其可行性，满足 预估管道服役期限的工程要求。 关键词：煤气管道；漏损；BP神经网络; 服役寿命 中图分类号：TP182; TP183 文献标识码：A 文章编号：1004-731X (2010) 02-0407-04 Establishment and Application of Model for Lifetime Forecasting of Gas Pipeline Based on BP ANN ZHANG Le-le, GUO Jia, SONG Shou-xin (School of Mechanical, Electronic and Control Engineering, Beijing Jiaotong University, Beijing 100044, China) Abstract: For the complicated urban pipeline net, a comprehensive assessment model for pipeline leakage has been established which can use to predict the urban pipeline lifetime based on BP ANN by using the different between BP ANN which is good at dealing with many a data and studying by itself. The quantitative analysis has been done based on the assessment model. Combining with advantages of BP ANN, the best BP ANN model has been obtained after training network in software of Matlab. Parallely, the BP arithmetic was improved by the conjugate gradient method, which solve the problem of BP ANN is slowly to constringe and easily to concuss. The function of this model has been checked by fitting the output data of the network and compared with the pipeline leakage facts. Key words: gas pipeline; leakage; BP ANN; lifetime 引 言1