在管道油气泄漏检测中最主要的难题是管道距离过长的问题, 利用传统检测模式效果不好。提出一种基于光纤技术的管道油气泄漏共轭梯度回声状态网络(ESN)检测方法。首先, 利用光纤探测技术对长距离油气管道的实时数据进行监测和采集, 并利用数据串联方式, 结合粗糙集算法对采集的泄露数据采样无标记数据的维度压缩方式, 实现数据的降维和简化；其次, 选取ESN对降维后的油气管道泄露数据进行分析和处理, 并基于正则化方式对ESN的权重进行初步的优化, 在此基础上利用Hessian矩阵的半正定性, 对初步的权重结果进行深入的学习优化, 实现ESN对油气管道泄露数据的识别能力；最后, 通过实测的数据仿真验证, 提出的油气管道的泄露监测模型具有更高的识别精度和计算效率。

The main problem oil and gas leakage in pipeline detection is that the pipeline distance is too long, and the effect of traditional detection mode is not good. A detection method of oil and gas leakage in pipeline using conjugate gradient echo state network(ESN) based on optical fiber technology is proposed. Firstly, the real-time data of long-distance oil and gas pipeline is monitored and collected by optical fiber detection technology, and the dimension compression method of unmarked data is used to reduce the dimension and simplify the data by combining the data series method and rough set algorithm. Secondly, ESN is selected to analyze and process the dimension reduced oil and gas pipeline leakage data. The weight of ESN is optimized preliminarily, and then the preliminary weight results are further optimized by using the semi positive qualitative of Hessian matrix to realize the recognition ability of ESN to the leakage data of oil and gas pipelines which is based on regularization. Finally, through the simulation of the measured data, the proposed leakage monitoring model of oil and gas pipeline has higher recognition accuracy and calculation efficiency.