Experimental Research on Prediction of Remaining Using Life of Solar DC Centrifugal Pumps Based on ARIMA Model

In order to improve the stability and reliability of the solar DC centrifugal pump real-time operation and prevent the centrifugal pump failure caused by the unexpected shutdown of the system, a set of accurate and efficient centrifugal pump condition monitoring systems was built. A time series-based strategy for predicting the remaining using life (RUL) of centrifugal pumps was proposed. The time series of head and efficiency of centrifugal pumps at specific flow conditions were measured, the corresponding failure thresholds were set, and different differential autoregressive integrated moving average (ARIMA) models were developed to predict the remaining useful life of the pumps. The results show that the maximum prediction error of the head ARIMA model established under the design conditions of the pump was 0.040%, and the head time series reaches the failure threshold of 8 m at the 653rd data point; the maximum prediction error of the efficiency ARIMA model was 0.042%, and the efficiency time series reaches the failure threshold of 16% at the 672nd data point. According to the proposed prediction strategy, the RUL of the centrifugal pump under the design condition was 53 h. The head time series of the pump at high flow conditions reaches a failure threshold of 5 m at the 640th data point; the efficiency time series will reach a failure threshold of 12.5% at the 578th data point, and the RUL of the centrifugal pump at high flow conditions was 78 h. The established ARIMA model has a high prediction accuracy and can effectively predict the RUL of centrifugal pumps.