Fault prediction of track circuit compensation capacitor based on MFO-LSTM

Based on time series data collected by the ZPW-2000A track circuit detection vehicle, which detects the induced voltage of compensating capacitor, this paper presents a method for predicting compensating capacitor faults using the MFO-LSTM prediction model. Specifically, an induced voltage sequence is established based on the section and month, and an LSTM network, optimized by the Moth-flame optimization algorithm, is employed to identify the induced voltage variation over time and predict the fault of the compensated capacitor induction voltage over a given period. Experimental results demonstrate that the proposed model can accurately fit the induced voltage curve, with small error, providing valuable technical support for maintenance of railway compensation capacitors.