Study on the oil particle contamination forecasting Using LSTM network

Abstract

As one of the main techniques of equipment condition monitoring, oil monitoring technology plays an extremely important role in evaluating the current state of equipment and predicting the development trend of equipment. In this paper, the LSTM neural networks was established by the historical data collected by a power plant. Using the cross validation method, and compared whit the popular time series prediction algorithm LSM, ARIMA, BPNN, SVR and RFR in the same test set, LSTM got the lowest RMSE value 42.26, which validates the applicability and accuracy of the LSTM neural network in the prediction of oil particle contamination.