New sensor technologies have vastly increased the amount of data that operators have about downhole conditions and well productivity. But how do you turn all that data into better decision making? Battelle is developing analytical tools that use predictive modeling to help oil & gas developers maximize well productivity and reduce potential risks.
The first tool uses a predictive model that forecasts the downhole pressure as material is injected into a reservoir. The predicted pressure is modeled as a function of recent lagged injection rates and pressures, and an alert is signaled to the operator when that prediction is straying outside of acceptable limits. The model is integrated into a TIBCO Spotfire®application to visualize results. It is designed to help operators:
- Improve overall production efficiency
- Reduce the occurrence of costly well shut-in events
Most oil & gas analytical tools on the market today focus on summary results, which provide important insights but aren’t always useful for on-the-fly decision making. For example, operators managing wells in the field have to balance injection rates and downhole pressure in order to meet production targets. Operators tend to rely on experience and intuition to adjust settings based on aggregate data collected on an hourly basis. This results in sluggish response to any change in the downhole environment and the need for costly shut-ins to control production. Real-time sensor data analysis provides the capability to make more timely responses, leading to more efficient production. Data-driven modeling can help experienced and inexperienced operators better predict the outcomes of different alternatives and select the settings that will maximize well production. Eventually, similar tools could be used to automate some of these decisions, freeing operators to focus on higher-order decision making.
The Battelle model leverages sensor data to forecast production, identify potential issues, and recommend corrective action to avoid a shut-in. Real-time data is analyzed using sophisticated statistical modeling methods similar to those used by the financial, healthcare and intelligence/defense sectors. An anomaly detector algorithm monitors the data and predictions to detect divergences between the forecasted and observed data that fall outside of the historical range of variation. This allows the model to detect anomalies that may arise due to introductions of external sources of variation that are not accounted for in the model.
The predictive model was built using data collected during a CO2 injection experiment at the AEP Mountaineer power plant site from October 2009 through May 2011. The dataset contains pressure, injection rate, and temperature data sampled at one-minute intervals over the nearly two-year period.
The injection rate and pressure model is the first of several analytical tools Battelle plans to develop for the oil & gas industry. It builds on Battelle’s long history of work in enhanced oil recovery (EOR) and risk modeling for oil & gas development. Similar modeling tools could be developed to help companies make better decisions on well siting, EOR options or risk mitigation.
Battelle has long been considered a leader in statistical modeling and predictive analytics. These same methods are already in use in the healthcare field to analyze quality indicator data and make clinical predictions. Battelle has also applied its analytical expertise to national defense, cybersecurity and genomics applications. For the oil & gas industry, they are now combining this expertise with deep experience in geophysical modeling, oilfield production, carbon sequestration and subsurface resource management to develop new statistical and modeling tools for the industry.