Industrial control anomaly detection method and system based on graph nerve and gating circulation network

• Main Authors: TANG CHAOFAN, LI XIN, ZHAO DAWEI, TONG FENGHUA, CHEN CHUAN, XU LIJUAN
• Format: Patent
• Language: Chinese; English
• Published: 22.11.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; CONTROL OR REGULATING SYSTEMS IN GENERAL; CONTROLLING; COUNTING; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; HANDLING RECORD CARRIERS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS ORELEMENTS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS; REGULATING

The invention belongs to the technical field of industrial control system anomaly detection, and provides an industrial control anomaly detection method and system based on graph nerves and a gated loop network, and the method comprises the steps: firstly converting the time series data of a sensor into an embedded vector form, explicitly carrying out the modeling of the correlation of different features, a structure learning method is combined with a graph neural network, a potential relationship between multivariate time sequences is deeply mined by using a gated loop network, and meanwhile, an attention mechanism is combined to provide interpretability for detected anomalies; according to the invention, data anomaly detection of the industrial control system can be effectively realized, and excellent detection performance can be realized. 本发明属于工业控制系统异常检测技术领域，提供了一种基于图神经和门控循环网络的工控异常检测方法及系统，首先将传感器的时序数据转化为嵌入向量的形式，显式地对不同特征的相关性进行建模，将结构学习方法与图神经网络相结合，利用门控循环网络深入挖掘多元时间序列间的潜在关系，同时结合注意力机制为检测到的异常提供可解释性；本发明能够有效实现对工业控制系统