High-precision energy consumption prediction method based on numerical interpolation and multilayer long short-term memory for offshore oil and gas platforms

Abstract

In the global context of advancing the “Dual carbon goals”, accurate prediction and evaluation of offshore oil and gas field carbon emissions are crucial for the energy industry’s green low-carbon transformation. To address the bottlenecks (harsh offshore monitoring environment, data coordination difficulties, core algorithm deviations, and immature intelligent prediction models) that cause the prevalent “prediction-decision disconnection” in existing carbon management systems, this study proposes a deep learning-based energy consumption prediction and evaluation framework for offshore platforms. Innovatively adopting “numerical interpolation + multi-layer stacked LSTM network”, the framework achieves high-precision prediction of key energy consumption indicators (e.g., fuel gas, vent gas) via systematic data processing and multi-dimensional model optimization. Based on historical data from an offshore oil and gas platform in the Bohai Sea, the data were proportionally divided into training, validation, and test sets, with optimal model configuration determined by parameter tuning. It results that the method effectively captures energy consumption’s dynamic and nonlinear characteristics provides reliable technical support for energy scheduling optimization, refined carbon management, and safe production, This study has important theoretical and practical significance for the oil and gas industry to achieve the “Dual carbon goals”.