Session: 01-03: AI for Energy Sustainability III

Paper Number: 140345

140345 - Energy Consumption and Greenhouse Gas Emissions Prediction for Gas-Oil Separation Plants Through Application of Machine Learning 

Abstract: 

A Machine Learning (ML) based method is presented to accurately predict energy demand and associated greenhouse emissions for Gas-Oil Separation Plants (GOSPs).  The oil reservoirs maturity and increase in water-cut are key contributing factors in determining the energy required to process the multiphase crude extracted from the wells. ML prediction models are then utilized to establish challenging Energy Intensity (EI) targets, considering the planned production from the GOSPs, to drive energy efficient operation.

To address the challenge of predicting energy demand with changing composition of crude and production target, Saudi Aramco team developed an AI-powered predictive solution, based on machine learning approach. The solution predicts the energy required for crude separation while considering the injected and disposal water associated with Gas and Oil Separation Plants (GOSPs). The solution uses state of the art data analytics algorithms to predict water-cut for the target oil production and the energy required to separate oil, gas, and water. The solution covers individual facility as well as overall network of multiple fields feeding the GOSPs.

Deploying the aforementioned solution has resulted in significant decrease in energy consumption for the hydrocarbon production processes. The early assessment indicates a reduction of 2-4% energy consumption per annum for the selected GOSPs, with associated CO₂ emission reduction. The predicted energy requirements for several GOSPs enables the change in operational philosophy by prioritizing the production from less energy intensive GOSPs.  The solution also identified high impact facilities for electrification to reduce further energy consumption and associated GHG emissions.

Utilization of ML allows retraining of energy prediction models, using the newer data, if the forecasting accuracy drops. The analysis of historical data reflected a continued increase in water injection to maintain the reservoirs’ pressure and sustain oil recovery. Higher water injection and water-cut cause a range of operational and environmental challenges, including increased energy consumption, reduced oil and gas production rates, and increased greenhouse gas emissions. This imposes a critical sustainability challenge, and impacts environmental aspect of ESG framework. 

While many oil and gas companies are pursuing long-term net-zero carbon emission goals, the challenge of predicting the future energy demand is adding more complexity to the equation. The method described in this paper streamlines the energy prediction process.

Energy prediction using AI and machine learning based models is novel to oil and gas industry. The development of site structure and identification of the suitable variable for modeling requires understanding of the plant operation. Furthermore, the application of a robust energy modeling software reduces data processing time and prediction accuracy, which are the main pillar for energy consumption forecasting work.

 

Presenting Author: Muhammad Abbas Saudi Aramco

Presenting Author Biography: Muhammad Abbas has more than 28 years of experience in operations, energy efficiency and project engineering. For the last 21 years, he has been solely involved in the areas of energy efficiency and environmental sustainability. He has worked in different roles to drive the energy efficeincy agenda for IBM Canada, SNC-Lavalin, City of Calgary, and Suncor Energy. Currently, he is working with Saudi Aramco as Engineering Specialist in Energy Systems Division. He is a registered professional engineer in Canada, Certified Energy Manager, ISO 50001 Lead Auditor, LEED AP BD+C, and ASHRAE member.

Authors:

Muhammad Abbas Saudi Aramco
Mussa Alamri Saudi Aramco
Mubarak Alotaibi Saudi Aramco