基于AutoEncoder的油气管道控制系统异常状态监测方法

Pipeline control system fault detection method based on auto-encoder

压缩机控制电路的健康状态管理在管道运输中至关重要。通常油气管道压缩机系统部署地点远离城市,环境恶劣,且负荷高、工作时间长,因此故障频发。构建可靠的健康状态检测模型通常需要大量的故障样本,然而在实际数据中,故障样本相对稀缺。采用一种基于自编码器(auto encoder,AE)的单分类方法对油气管道控制系统的异常状态进行辨识。该模型仅需对系统的正常工作状态进行学习,通过编码器可实现特征的自适应提取,从而对数据进行抽象表示,并获得较好的非线性映射能力;当数据分布异常时,系统可区分其与正常信号间的差异,并进行预警。实验部分采用西部输油管道控制系统中实地获取的通信解码信号以及电源信号进行验证,并以单分类支持向量机方法作对比实验,表明了所提出方法的有效性。

The healthy management of compressor control circuit is important in pipeline systems. Usually, the compressor system of oil and gas pipeline is deployed far away from the city, with poor environment, high load and long working time, so the fault happens frequently. Usually, a large number of fault samples is required to build a reliable health state detection model, but in reality, fault samples are relatively scarce. To solve this problem, in this paper, one-class learning method based on auto-encoder is used to identify the abnormal state of oil and gas pipeline control system. The model only needs to learn the normal working state of the system, the encoder can realize the adaptive extraction of features, so as to abstract the data and obtain better nonlinear mapping ability. When the data distribution is abnormal, the system can distinguish the difference between abnormal and the normal signals and provide an early warning. In the experiment part, the communication decoded signal and power supply signal provided by petro china west pipeline company are used to verify the validity of the proposed method, and the one-class support Vector Machine method is used as a comparative experiment.