Optimized LSTM Model Using Artificial Bee Colony Algorithm for Crude Oil Production Forecasting in Nigeria

In the oil and gas sector, economic planning and decision-making depend heavily on forecasting crude oil production. Crude oil production has been predicted using a variety of methods. Techniques based on deep learning show promise because they have been successfully implemented in many industries and can be used at many phases of the oil exploration and production process. Still, the oil industry needs more work in this regard. This study proposes an optimized Long Short-Term Memory model by Artificial Bee Colony ABC for oil production prediction. The methodology employed is the CRISP-DM methodology (Data Mining) for a structured approach. The proposed model was applied using real data from 323 oils that were developed and 265 oil wells from flow stations. The model was developed in a Google Colab environment, trained, tested, and evaluated well. The result of the model prediction is 1,428,751 barrels of crude oil production per month, the LSTM model was optimized using the Artificial Bee Colony Algorithm successfully and all three regression models namely the GA-GB model, Bagging Regressor, and KNN had lower scores when compared with the optimized model. Further studies suggest exploring novel techniques and methodologies to enhance model interpretability and scalability using robust real-life data