Offshore Production Surveillance and Intervention Using Multi Agent AI

Abstract This paper introduces a pioneering Agentic Artificial Intelligence (AI) framework designed for offshore production surveillance and intervention. Agentic AI is a novel framework comprising a collection of AI-models, operating autonomously yet collaboratively to achieve a common goal. Each model specializes in performing a certain task to streamline production monitoring, root cause analysis, predictive maintenance, and optimization workflows. It utilizes comprehensive datasets, including production history, well-coordinates, intervention history, petrophysical, and completion information, to support dynamic decision-making across the asset. The system utilizes independent AI agents for specific tasks, interacting conversationally with users: Data-QC Agent: Detects and corrects anomalies in production data, improving allocation and workflows. Well-Surveillance Agent: Monitors production trends and reservoir performance, identifying issues such as decline, water breakthrough, and liquid loading. Asset-Surveillance Agent: Analyzes network, facility, and equipment performance, identifying bottlenecks, flow assurance risks, and optimization opportunities. Well-Screening Agent: Performs analyses (e.g., decline curve, Chan plot) to identify well candidates and intervention types. Model-Management Agent: Updates simulation models and runs sensitivity analyses. Log-Interpreter: Interprets log data. Additional agents support domain knowledge, email alarms, and ad-hoc plot generation. Preliminary results demonstrate significant operational improvements. Key use cases include enhanced production surveillance, root-cause analysis, and predictive maintenance. In one example, North Sea-operated data from subsea pipeline inspections and surface facilities were analyzed using an Object Detection Vision Agent, identifying integrity and corrosion issues, saving 30% of manual effort. The chat-driven interface automated data quality control, simulation updates, and alarm management, resulting in time savings 1. Predictive maintenance agents flagged early-stage failures, reducing downtime. Moreover, the system saved 80% of manual time in identifying well intervention candidates, optimizing asset management. The chat-driven model management also halved simulation run times, while visualization and notification agents streamlined data interpretation. Continuous anomaly detection minimized downtime, enabling early intervention. The unified platform empowers operators to make faster, data-driven decisions. This approach transforms offshore production management by integrating production-reservoir data with AI analytics, and intuitive user interaction. The multi-agent AI system combines petroleum engineering expertise with state-of-the-art large language models and advanced machine learning techniques, driving faster decision-making, optimized workflows, and enhanced asset performance.