Incipient Fault Detection of Combined Heat and Power Networks: A Mechanism-Data Co-Driven Approach

The deep coupling of combined heat and power (CHP) networks makes it difficult to detect faults at an early stage. To address this issue, we develop an approach that combines a mechanism model with a data-driven model, which is promising in accurately and promptly detecting incipient faults of the complex system. First, the mechanism model for CHP networks is constructed and used to generate a large amount of data under different system operation states. The data lays the foundation for the subsequent data-driven system analysis tasks. Then, by combining random matrix theory (RMT) and support vector machines (SVMs), a data-driven approach is proposed to study the data correlation and judge the network’s operation state. In our proposed mechanism-data co-driven approach, the mechanism model solves the problem of data shortage for the data-driven model, and the trained RMT-SVM-based data-driven model can accurately detect the faults of the complex system due to its sensitivity to the variation of data correlations. Case studies on the Barry Island CHP networks validate the effectiveness and advantages of our approach for incipient fault detection.