A Hybrid Artificial Neural Network Framework for High-Fidelity Estimations of Well Performance

This study develops a hybrid ML model that integrates an MLP neural network with a GA for reliable well performance forecasting. The model utilizes four readily available wellhead parameters: choke diameter, wellhead pressure, gas-liquid ratio, and basic sediment and water content, with the liquid production rate as the target output. A rigorous data preprocessing step, including outlier removal, was applied to enhance dataset quality. The optimized MLP-GA model demonstrated high predictive accuracy, achieving R_x۰۰۰۰__x۰۰۰۰_ values of ۰.۹۹۸۳, ۰.۹۹۸۹, and ۰.۹۹۱۴ on the training, validation, and test sets, respectively, while maintaining a very low MSE.

کلیدواژه ها:

MLP ، GA ، Wellhead Choke ، Production Optimization ، ML