Oilfield Production Forecasting Model Based on Bi-directional Long-Short Term Memory Network - NanLiang Intelligent Hydraulic Pumping Oil Field as a Case in Point

Abstract

With the widespread use of hydraulic pumping machines in oilfields, the productivity and energy efficiency of oilfields has been greatly enhanced. A large amount of data is generated during the production process in oilfields, which contains information that is closely related to oilfield production. Therefore, using this data for oilfield production prediction is currently one of the key elements in the research of big data and artificial intelligence for oil and gas field production. But traditional methods suffer from poor adaptability and high data requirements when solving this type of forecasting problem. Therefore, in this paper, according to the characteristics of the daily production of oilfield with time-series, the production prediction model is constructed by using Bi-directional long-short term memory network, which reduces the influence of production data fluctuation on the prediction model; and the structural parameters of the prediction model are optimized by using quantum genetic algorithm, which can quickly derive the optimal prediction model structure in space and avoid the model falling into the local optimal solution. The validity of the method is verified in a the Nanliang oilfield with an intelligent hydraulic press in Changqing, and the results show that the method has better prediction accuracy and generalization ability to adapt to complex production environments. The method provides a theoretical basis for the development of intelligent oilfields in the context of big data.