Reservoir production capacity prediction of Zananor field based on LSTM neural network

Abstract

Abstract This paper aims to explore the application of artificial intelligence in the petroleum industry, with a specific focus on the prediction of oil well production. Using the Zananor Field as a case study and leveraging several years' worth of monthly oil production data, experiments were conducted to establish Long Short-Term Memory (LSTM) neural network models to accurately forecast monthly oil production in the field. In this study, the raw data was meticulously organized, and distinctions were made between different wells and their respective production stages. Additionally, data normalization was performed. Initially, a univariate LSTM neural network model was constructed, utilizing monthly oil production data as the input to predict the monthly oil output in the experimental oil field. Furthermore, a multivariate LSTM neural network model was introduced, utilizing various production data sets as inputs to enhance the accuracy of monthly oil production forecasts. To further enhance predictive accuracy, two different feature selection methods were compared in the experiments: Grey Relational Analysis and Principal Component Analysis. The experimental results revealed that the multivariate model outperformed the univariate model in terms of prediction accuracy, making it more suitable for forecasting monthly oil production. Furthermore, the experiments demonstrated that Grey Relational Analysis exhibited higher accuracy in feature selection and greater applicability compared to Principal Component Analysis, rendering it a more viable option. These research findings provide valuable guidance for production forecasting and operational optimization within the petroleum industry.