Using Hybrid Machine Learning Methods to Predict and Improve the Energy Consumption Efficiency in Oil and Gas Fields

Abstract

Oil and gas will remain essential to global economic development and prosperity for decades to come, and the oil and gas industry is an energy-intensive industry. Thus, enhancing energy efficiency for producing oil and gas in oil and gas companies is an important issue. The intelligent energy consumption prediction method with the ability to analyze energy consumption patterns and to identify targets for energy saving proved itself as an effective approach for energy efficiency in many industrial domains. Moreover, prediction of energy consumption enables managers to scientifically plan out the energy usage of energy production and to shift energy usage to off-peak periods. However, it still remains a challenging issue to some degree with the unpredictability and uncertainty caused by various energy consumption behaviors, and this phenomenon is becoming more obvious in the oil and gas company. To this end, in our work, we primarily discussed the forecasting of the energy consumption in the oil and gas company. Firstly, four different forecasting models, support vector machine, linear regression, extreme learning machine, and artificial neural network, were trained on the training dataset and then evaluated by the test dataset. Secondly, in order to enhance the energy consumption prediction accuracy, the combinations of all these four models were examined with the RMSE value by taking the average of two models’ outputs. The outcomes show that these four different models are able to predict energy consumption with good accuracy, but the hybrid model—artificial neural network and extreme learning machine—would present higher accuracy. In addition, the hybrid model is installed in the energy management system of the oil and gas industry to manage oil field energy consumption and improve the efficiency.