A Data-Driven Approach to Modelling and Optimization for a North Sea Asset Using Real-Time Data

Abstract

Increased numbers of sensors on wells do not result in increased production. New methods of data analysis are needed to capitalize on the new data. A novel approach to modelling gas coning has been developed for one of BP's assets. It integrates several data-driven models representing different aspects of a well's performance.Prior art, as exemplified by other approaches in BP and other case studies in the literature, is discussed. A description of the field and its operational issues around gas production is given. The details of the methodology that has been used are provided. The novelty relative to the prior methods described in literature is discussed. Advantages of the data-driven approach are explored, incorporation of the models into real-time operational support systems is considered, and other opportunities for leveraging available data to maximize the value from our operations are mentioned.The approach has been tested on a North Sea reservoir. An interconnected suite of models for reservoir pressure, indication of the presence of gas coning, virtual flow measurement for fluids and gas production, and dynamic prediction for future fluids and gas rates, was developed. The suite of models predicts fluid and gas rates for use in short-time-loop optimizations. Models use the available real-time data, which includes bottom-hole pressure, wellhead pressure, choke changes, and gas lift rate. The data is captured for production, injection and observation wells.The methodology assists in the prediction of the dynamic well behavioral changes which result from the onset or occurrence of gas coning. It enables the asset to leverage its real-time sensor data to manage and optimize its wells while respecting the facility's constraints. It works for similar assets where dynamic well behavior must be managed.