Data-driven system efficiency prediction and production parameter optimization for PW-LHM

Abstract

Heavy oil wells in Tahe oilfield are exploited by injecting light hydrocarbon mixing into annulus to reduce viscosity, and system efficiency prediction is of great significance for the evaluation of the energy consumption and management level of the production system. However, system efficiency of the pumping unit well with light hydrocarbon mixing (PW-LHM) is affected by many factors, and it is difficult to use traditional methods to make comprehensive evaluation and quantitative prediction. In order to accurately predict the system efficiency and evaluate the viscosity reduction effect, this paper used the Pearson correlation coefficient analysis method to analyze the correlation between production data, diluting attributes and system efficiency, and principal component analysis (PCA) was utilized to conduct data dimension reduction and controlling parameters determination. Considering the change trend and correlation of the artificial lifting equipment and mixing equipment working conditions, a time series prediction model for system efficiency was established based on the long short-term memory (LSTM) algorithm. Filed application results show that the prediction model based on LSTM can perform accurate prediction of the system efficiency, early forewarning of the production conditions and timely adjustment of the lifting and mixing parameters of the PW-LHM.