Pump-off identification with time series analysis of transient water discharge rate

Abstract

To advance the prognosis of progressing cavity Pumps (PCPs) used for artificial lifting, the pump-off need to be identified to forestall failure. This study developed a new technique for determining the Pump-off events Activation Times (PATs) of the PCPs using the transient Water Discharge Rates (WDRs) from coal seam gas producing wells. The Gaussian distribution function parameters of the rolling standard deviations of the water discharge rate (RSWR) and the transition probability of the rolling standard deviations of the water discharge rate (TP_RSWR) were used to build the model. By determining the anomalies in the RSWR signals with the bottom-up segmentation technique and computing the statistical characteristics at the changepoint locations, the steady-state of the WDR signals was established. This steady-state signal, which represents the Operation Transition Level (OTL) between the Normal Operation (NOP) and the Pump-off Event (POE) was used for monitoring the transition of the PCPs' operating status. An algorithm was developed in Python and tested it on field data from 36 coal seam gas wells. The performance of my technique was determined with precision, recall and F1 score, which gave an average value of 94.94%, 92.63%, and 93.56% respectively. It is expected that the implementation of this technique in the real-time estimation of PATs will be vital for reducing PCPs faults seeing that poor PATs detection results in PCPs running dry and consequently failures due to the extreme temperatures and abrasions.