Research on Abnormal Diagnosis Method of Benzene-ethylene Ratio Double-layer Architecture Control System based on OHEncoding-XGBoost

Aiming at a kind of control system based on MPC-PID double-layer architecture in the process industry, due to the complex structure, existing the correlation between control layers and the coupling between control loops, it is difficult to accurately diagnose anomalies of the system control loop. In this paper, an anomaly diagnosis method based on the OHEncoding-XGBoost algorithm is proposed for the Benzene-ethylene ratio system of the Alkylation unit in the styrene production equipment. For the Benzene-ethylene ratio control system with 1 main-loop and 4 vice-loops, combined with the prior knowledge of the system, 5 working conditions between normal and common anomalies are analyzed. According to the evaluation index of the classification model, the One-Hot classification variable coding method is used to construct the feature matrix, and the hidden abnormal information is revealed. The characteristic matrix is used as the input of the diagnosis algorithm, and the abnormal state of the control loop is used as the output, which realizes the effective diagnosis of the abnormal condition of the double-layer Benzene-ethylene ratio control system. Simulation experiments show that the accuracy of the proposed method is 5.16 % higher than that of the traditional XGBoost abnormally diagnosis model.