🌍 Pleased to share our progress in history matching with seismic data assimilation 🚀
We’ve been investigating how AI can support the Ensemble Smoother with Multiple Data Assimilation (ES-MDA) to improve history matching and reservoir model updates. By using AI to reduce seismic data volume, we’ve been able to save time while achieving results comparable to the standard ES-MDA approach.
📊 Key Achievements:
✅ Reduced seismic data using pre-trained AI for 2D and 3D models 🤖
✅ Estimated time reduction between 50% and 90% ⏲️
✅ ES-MDA + AI (2D and 3D) performs similarly to standard ES-MDA
✅ Both methods include localization techniques for reliable integration
The attached images compare the reference Porosity and Seismic observable data (delta P-impedance) from the PUNQ Benchmark reservoir with results from standard ES-MDA, ES-MDA + AI (2D), and ES-MDA + AI (3D).
This work suggest a practical way to speed up history matching and 4D seismic applications while maintaining reliability.
📖 Read more in our recent publications:
📌 Feature Extraction in Time-Lapse Seismic Using Deep Learning for Data Assimilation
🔗 https://lnkd.in/dwPJyCTY
📌 Deep Feature Extraction for Data Assimilation with Ensemble Smoother
🔗 https://lnkd.in/duxCr8Qg
Thanks to Petrobras CENPES and LTrace coworkers for the collaboration on this project!
#Geosciences #SeismicData #AI #ESMDA #HistoryMatching #4DSeismic #Petrobras #LTraceGeosciences #ReservoirSimulation #ReservoirManagement