Perpustakaan Digital ITB

This study presents an innovative approach to optimising infill well placement by integrating machine learning techniques, Voronoi diagrams, and GridSearchCV optimisation. The main objective is to enhance oil recovery by strategically determining the optimal locations for infill wells within a multi-layered reservoir. The methodology involves a comprehensive analysis of reservoir characterization data, including porosity, water saturation, and hydrocarbon pore volume (HCPV). Machine learning models predict the current oil in place (COIP). Voronoi diagrams partition the reservoir into manageable regions, each representing a potential infill well location, acting as proxies for drainage areas, thus allowing for precise targeting of high-potential zones. GridSearchCV is utilised to fine-tune the hyperparameters of the machine learning models, ensuring the best possible predictive performance. The approach is validated through a case study of Field X, which encompasses 24 wells across three formation layers: UBR, UBR-B, and UBR-C. The study finds that this integrated method significantly improves the accuracy of infill well placement, resulting in enhanced oil recovery and more efficient reservoir management. The results indicate a substantial increase in the predicted oil after adding new infill wells, demonstrating the effectiveness of combining machine learning with Voronoi diagrams and GridSearchCV optimisation in optimising well placement strategies. This approach offers a valuable tool for reservoir engineers to maximise hydrocarbon recovery and improve field development planning.