Machine Learning Prediction Versus Decline Curve Prediction: A Niger Delta Case Study

Several analytical techniques have been identified to obtain reliable estimates of production. Out of these numerous methods, decline curves are the most extensively used technique for the production forecast of Niger Delta Reservoirs. However, a major setback in applying the decline curve is its inability to adapt predictions to different past operational scenarios and uncertainties. With the emergence of big data and increasing computational power, machine learning techniques are increasingly being used to solve problems like this in the oil and gas industry. The objective of this paper is to present the application of a machine learning-based framework to predict the future performance of producing wells in some reservoirs in Niger Delta. In this paper, a machine learning model (Neural Networks model) was used to detect the non-linear relationship between the inputs in the production data and predict the future production rate of wells. The model is trained using available data from a Niger Delta Reservoir. Further data, excluded from the training data set, was used to assess the ability of the neural network to rapidly learn the basic shape of the time series data and model the non-linear relationship of the data for prediction. The different case studies are compared to forecasts from conventional decline curves to demonstrate the advantage of applying machine learning techniques to production forecasting. The proposed technique indicates high accurate prediction and learning performance for crude oil forecast of producing wells, especially for cases with changing operating conditions. The study also reflects that the performance of the model is largely influenced by the model-optimization technique. The research work provides empirical evidence that the proposed model can be applied to production forecasting, addressing complexities that other statistical forecast methods cannot implement. The proposed application of computational techniques in forecasting problems has proven to be a robust and reliable method of forecasting the future performance of producing wells. The procedures adopted in this work can also be extended to wells outside of the Niger Delta.