Physics-Informed AI-Assisted Well Portfolio Management: Integrated Diagnostics to Field-Proven Production Gains

Abstract

Abstract Managing mature oil and gas assets requires continuous production optimization across large well inventories characterized by multiple reservoirs, diverse completion intervals, varied artificial lift configurations, and dynamic operating conditions. Identifying underperforming wells and prioritizing high- impact interventions remains challenging in such environments, as engineers allocate significant effort to manual data consolidation and surveillance, which limits scalability and delays operational decisionmaking. This paper presents a field deployed Well Portfolio Management (WPM) framework implemented across the mature Greater Hapjan Field of Oil India Limited for 150 wells of which approximately 115 are producing and distributed across 40 reservoirs. The framework integrates structured and unstructured production, operational, and surveillance data within a unified analytics environment and combines conventional reservoir engineering diagnostics with physics-informed artificial intelligence (PI-AI) methods to enable scalable, engineering consistent production optimization. The workflow applies portfolio scale screening followed by diagnostic modules including water breakthrough analysis, decline curve analysis, productivity index evaluation, type-well comparison, gas-lift performance assessment, and production engineering diagnostics. Outputs are consolidated into a ranked, physics-constrained decision-support framework, with recommendations validated through engineering review and reintegrated through a closed-loop mechanism. Across the producing inventory the framework generated over 50 intervention recommendations. Four interventions were implemented during the study period two bean-size optimizations and two isolation repair with selective perforation workovers delivering a combined production increase of approximately 100 m3/d, with analysis-to-decision cycle times reduced from days to hours. In each case the productionlimiting mechanism identified by the framework was confirmed by independent engineering evaluation. The results demonstrate that combining engineering driven surveillance with physics-informed analytics can improve diagnostic consistency, accelerate intervention decision making, and deliver measurable production uplift across large mature well portfolios.