Physical information entropy-driven performance deterioration diagnosis for control systems of coal power plant

Abstract This article primarily focuses on addressing the problem of performance deterioration diagnosis for coal power plant control systems. First, a digital twin dynamics model of real industrial control systems driven by physical state variables is constructed based on Newton’s second law. This digital twin is designed to achieve an accurate and lightweight digital description of the control system using two-dimensional orientations. Then, to enable physical information entropy-driven performance deterioration diagnosis (PIE-PDD) of the digital twin dynamic model, an adaptive overcomplete dictionary is established using the K -singular value decomposition algorithm in a lower triangular sparse structure. Furthermore, low-rank representation learning based on digital twin (DTLRR) is used to develop the PIE-PDD for control systems. This process aims to diagnose, identify, and perform root cause analysis of the performance degradation of the physical control system at the dynamic model level. Finally, recent results are compared with other deterioration diagnosis methods presented in this paper, and an experiment involving a coal mill primary air damper temperature control system is conducted to verify the effectiveness of the proposed method.