基于改进神经网络腐蚀管线剩余寿命预测

Prediction of Residual Life of Corroded Pipeline Based on Improved Neural Network

为准确预测埋地油气管道腐蚀剩余寿命,构建基于改进神经网络腐蚀管线剩余寿命预测模型。首先,利用共轭梯度法对神经网络进行改进,保证目标函数在迭代n次之后找到全局极小点;其次,以国内某油田在役埋地原油管线为背景,确定影响因素;然后分析数据,采用拉依达准则对不良数据样本进行剔除,依据样本数据建立BP神经网络模型及共轭梯度法改进的神经网络模型;最后,以国内某油田在役埋地原油管线为实例,验证模型的预测有效性。结果表明:通过改进神经网络预测的管道剩余寿命预测结果与观测值基本相同,改进神经网络模型隐含层共计1个,平均相对误差为9.4%,预测结果更为精准,改进神经网络预测的管道剩余寿命误差更小。

In order to predict the corrosion residual life of buried oil and gas pipeline accurately,the residual life prediction model based on improved neural network was established.Firstly,the conjugate gradient method was used to improve the neural network to ensure that the objective function can find the global minimum point after n iterations;Secondly,the influencing factors were determined based on the in-service buried crude oil pipeline of an oil field in China;Then,the data were analyzed,and the Pauta Criterion was used to eliminate the bad data samples.Based on the sample data,the BP neural network model and the improved neural network model by conjugate gradient method were established;Finally,the effectiveness of the models were verified by an example of an in-service buried crude oil pipeline in an oil field in China.The results showed that the prediction result of pipeline residual life predicted by improved neural network was basically the same as the observed value.There was one hidden layer in the improved neural network model with an average relative error of 9.4%,and the prediction result was more accurate,the structure of the model was simpler and the error was smaller.