Induced Draft Fan Early Anomaly Identification Based On Sis Data Using Normal Behavior Model In Thermal Power Plant

In this work, a new idea was proposed that establishes normal behavior model (NBM) with multiple inputs and multiple outputs for each specific equipment based on Principle components analysis- Nonlinear autoregressive exogenous model (PCA-NARX) a kind of ANN. The operating parameters interested in condition monitoring are selected from SIS as an aggregation for a certain equipment, and the corresponding NBM is constructed based on the co-relation among parameters and the autocorrelation in each parameter. Each operating parameter can determine a reasonable range in real time by NBM, so it can detect abnormal operation parameters more quickly than the traditional fixed threshold method. Combining the historical operational data of the No. I induced draft fan of No. 3 generating unit in Shajiao C Power Plant in China, and the aggregation for induced draft fan covers 12 operating parameters interested in condition monitoring. This work used MATLAB to verify and analyze the proposed method. It is found that the NBM for induced draft fan early anomaly identification established in this work can achieve rapid response to the fault and give an alarm in the early stage of the fault. Moreover, the method can be easily applied to other mechanical equipment in thermal power plant and has good engineering application value.