ÓÈÎïÊÓƵ

Integrated reservoir management through advanced waterflood pattern balancing techniques was delivered through a partnership between an NOC, as the field operator, and ÓÈÎïÊÓƵ, acting as the technical partner. This enabled more efficient, smarter decision making based on a data-driven machine learning process and greater focus on analysis rather than data input.

Theoretical approach

Around 40% of the operator’s daily oil production comes from waterflooding projects implemented in sandstone reservoirs. While waterflooding began in only one formation, with the expansion of secondary recovery, operational complexity has significantly increased, to injecting in two or more reservoirs, including cases with commingled production.

The operator wanted to enhance waterflooding performance principally by improving the operational, waterflooding decision-making process for injection pattern analysis, and by quickly generating actionable insights. One of the main challenges to address was inefficiencies, due to the use of spreadsheets as the primary tool to support the process. This necessitated a significant amount of manual data entry, allowed for only one pattern analysis at a time, and made each iteration analysis extremely time consuming, taking around 23 hours. It also meant there was no forecast capability so the team couldn't set up what-if/optimization scenarios and compare their estimated outcomes. Each engineer was using their own spreadsheet, with no shared environment or visibility of analysis. As a result, decision making in the field was extremely inefficient.

There is also complex geology, in a field developed with a combination of deviated injectors and deviated and/or horizontal producer wells, therefore, improved management was essential to prevent a steep decline in producer wells, due to early water breakthroughs..

The project was conducted using a tool that combines data-driven machine learning (ML) with physics-based capacitance-resistance models (CRMs), and an adapted pattern flood management (PFM) version to support the operational waterflooding decision-making process, assess short term scenarios, generate actionable insights, and enhance waterflooding performance.

The scope of the tool covers:

• Allocation factors, for quantitative assessment of inter-well connectivity.
• Forecasting capability, to provide a short-term fluids forecast (liquid, oil, and water) for up to 90 days, on a well basis, and with pattern level aggregation.
• Water-cut forecasting, with the model being trained to better understand the waterflood projects by analyzing the historical data (injection, production, and intake pressure) in the block.
• Aquifer influence, to generate insights into aquifer contribution of production or injection lost to the aquifer.
• What-if and optimization scenarios.

The tool generates insights on injector-producer relationships for pattern analysis, waterflooding setting understanding, and short (weekly) to mid-term (90-day) forecasts. The team can now evaluate the impact on production and waterflooding response as a function of operational parameter changes, with a shared view on the Delfiâ„¢ digital platform. This has resulted in short-term optimization opportunities for increased oil production, prolonged the production plateau, and optimized the overall waterflooding response.

After successful implementation of this tool in the waterflooding surveillance process, the reservoir surveillance and engineering teams experienced a step-change in productivity. The time taken for a complete iteration analysis has been reduced from 23 to 5 hours with excellent allocation and prediction matches to traditional reservoir-simulator-based forecasts. Using the tool, the operator improved analysis efficiency by 85%, delivered proactive responses to multiple operational upsets, produced reliable risk based 90-day forecasts, and gained a better understanding of the injection-production relationship. Ultimately, the combination of these improvements achieved 7300 barrels of incremental oil production in the first year. As of 2024, an additional 15,000 barrels had been produced, significantly surpassing the expected target.

The tool’s adoption is now being expanded to other assets by the operator. This broader deployment aims to replicate the significant gains achieved in operational efficiency, production optimization, and predictive accuracy across a wider portfolio of fields, further enhancing the value delivered through advanced reservoir surveillance and proactive decision-making.

Prognostic vs real data. Increasing injection in Well A to see the benefit in the water balance pattern.