Quebec Exploration

Oil & Gas

DIAGNOS adapted its CARDS system for the interpretation and evaluation of seismic data for oil and gas exploration. The system was designed as a way to quickly identify zones worth further exploration using the tremendous volume of data already available to most oil and gas companies. Similar work performed using traditional GIS/Seismic software and rule-based models could take years and have minimal success.

CARDS learns the signatures of the positive and negative wells primarily from seismic data but can also include other types of geophysical and geological data

By employing the same methodology based on the artificial intelligence, statistics and classification theory, the innovative technology of DIAGNOS increases significantly the probability of drilling wells where there is oil and gas.

A prediction model is created starting from this training data set and is validated using a test data set. Following various iterations and when the model is considered to be adequate, it is used to make predictions.

Finally, ground data is treated by the prediction model and a favorability map is produced displaying the probability of finding oil and gas at each point along the lines of seismic.

Why is CARDS different?

Our solution extracts knowledge from data without bias from any prior assumptions and is able to learn from all available data, which may include existing analysis completed using independent tools or approaches from a specific knowledge base.

In traditional geological/geophysical modeling, a model is developed to explain the underlying physics of a situation, for example, to determine the depth to bedrock using magnetic data. However, Inverse problems* are common and require the defining of a set of parameters to solve. Consequently, constraints are imposed on the data set to force results to converge. These constraints typically reflect a set of assumptions made from scientific observations and hypotheses proposed to solve those same inverse problems.

The set of hypotheses is most often limited by the experience of the interpreters and the quantity and quality of data gathered. If the outcome from drilling a target determined by the inversion method is negative, blame is often placed on the set of modeling or inversion hypotheses. In fact, traditional models are often designed to look for positive confirmation of the hypothesis and to a certain extent negative drill hole results are most often ignored since they don’t confirm the analysis and are treated as unfortunate anomalies.

These limitations are quite similar to how meteorological models function where the variables and physical principles controlling a system are numerous and sometimes not fully understood. In this situation, it still remains difficult to get good predictions strictly from current atmospheric models. The experienced meteorologist who has seen all kinds of weather systems must rely on his experience to make an assessment, learn from his mistakes and refine the model. Senior geologists are often called to do the same.

* An inverse problem is one where the outcome of a particular physical event or process is known as well as several parameters, but many parameters remain unknown. Normal problems are those where the parameters are known but the outcome is unknown.

This model building is not part of our approach

The way DIAGNOS’ CARDS system tackles the problem is a combined AI / stochastic approach using data pertinent to O&G exploration. The strength of the approach is the Artificial Intelligence and pattern recognition techniques that analyze the relationship between spatial observations. The variables: geophysical or any other types of data that might have some influence, even if small, on the final results (positive or negative drill holes) are fed to the system. One step of the processing is to incrementally analyze each data point with respect its neighbors and find the relationships between them. This approach does not require the articulation of a physical model to explain the results of the mathematical model. The inversion problem can be solved explicitly by withholding both positive and negative results and determining if they are found in the final predictions.

Additional information in this section:

Contact us

For additional information on how CARDS can work for you, contact Mr Michel Fontaine at 450 678-8882 or toll free at 1 877 678-8882, ext. 222 or by Email at mfontaine@diagnos.com.