Senslytics Presents on ResVoirX Product Capabilities for Reservoir Fluid Evaluation at 2024 GeoGulf Conference

Senslytics presented at the 2024 GeoGulf conference in San Antonio,TX showcasing collaborative work with GeoMark Research and Senslytics' advisor Hani Elshahawi and emphasizing the innovative ResVoirX product capabilities for Reservoir Fluid Evaluation. 

GeoGulf Paper Abstract: 
The complexities of reservoir fluid properties arise from the geological history of the field as well as diverse biochemical processes of the hydrocarbons once trapped. A unified geochemical, petrophysical, and biological model can be used to explain fluid properties found along a specific geographic trend or along depth. Previously an integrated model has been very computationally intensive and requires extensive, and therefore expensive, direct data calibration. CausX AI is a promising tool to create this unified predictive model in a fraction of the time and cost than previous methods, yet still produce accurate fluid property predictions. CausX AI is an epistemic, causal artificial intelligence framework that can mine experts’ experience and create a multi-view hypothesis set, which forms elements of situation-specific guardrails by vetting initial training data and ground truths. As more situational datasets are processed or more experts’ hypotheses are added, this scientifically driven hybrid AI model is refined; the vetting process removes human opinion biases and fills gaps in the knowledge space by using translated or extended guardrails from one situation to its neighbor. CausX AI takes the multi-view convergence approach to be more definitive in inferences. This multi-view approach enables CausX AI to create a reliable model with a few dozen data sets rather than the thousands typically needed by traditional machine learning. This study showcases CausX AI as applied to reservoir fluid property estimation and its ability to unify the geology, chemistry, and biological controls of the subsurface. CausX AI uses C1–C5 gas views as reservoir fingerprints for training purposes. The trained AI uses mud gas logs from standard or advanced mud gas traps as its data source for estimates. An initial study designed to estimate gas-oil ratio (GOR) in the Gulf of Mexico shows great promise and scalability, providing a higher fidelity in fluid property estimation than attained with traditional machine learning. This measurable improvement is achieved with no additional investment in mudlogging operations or data collection.

Read the full paper