Solving the Challenges of Short and Long-Term Production Forecasting and Uncertainty Using a Fully-Coupled Implicit Integrated Production Modelling System

Abstract

Abstract Integrated modelling is of paramount importance while determining optimum development plans for any asset, especially for high-stakes offshore plays. It can be used for generating reliable reserve estimates and short to long-term forecasts. However, performing such studies has been a difficult proposition for the petroleum industry. A number of logistical and technical challenges are present for any integrated modelling project. The logistical challenges appear because of multiple engineers working on multiple tools for different parts of the modelling work. The lack of timely feedback between teams often leads to the forecasts being unreliable. Moreover, several different tools and models need to be maintained for various disciplines and for different stages of asset development – presenting training and staffing issues. The technical challenges appear because of missing features in current tools. For example, it is common in the current workflow to see discontinuities across the use of various model fidelities. This deficiency can be seen in the application of reservoir models, where complex reservoir structures and processes are described in standalone reservoir forecasts by refined discrete models (particularly for long-term forecasts and reserve estimates) and in integrated forecasts by material balance tank models (typically used for short- to medium-term forecasting). The latter approach is often not suitable for making longer-term forecasts, yet the current tools fail to suggest when such simplifications lead to significant errors. Furthermore, in assets with multiple reservoirs and complicated fluid blending in the production facilities, the traditional modelling approach may lack the required physics to adequately model operational constraints or perform allocations. Lastly, an end-to-end uncertainty and optimization framework is generally missing in the conventional approach, particularly where decisions are being made on the integrated production system. As a result, while users can often explore uncertainties and design decisions within the scope of their particular discipline, a coherent, integrated picture is difficult and often impractical to pull together. In this paper, an innovative solution to the above mentioned problems is discussed. For demonstration purposes, a synthetic asset model has been developed, which represents a complicated deepwater asset. The asset has two heterogeneous reservoirs, with unique and complex fluids, such that modelling should account for fluid blending in the shared production network. A multi-user, multi-discipline, multi-fidelity tool is used that allows for collaboration and sharing of data between reservoir, production, and geomechanics engineers through a relational database. Multiple fidelities are used in the reservoir, wells, production network and geomechanics, with lower fidelity proxies typically being automatically generated from the more rigorous corresponding model – allowing for an easy-to-use, consistent and fit-for-purpose integration of model components. Even so, the adequacy of lower fidelity models is ensured by quick and easy comparisons with high fidelity models. All the relevant physics are modelled for a gas re-injection project, system bottlenecks are identified, and reliable short, medium and long-term forecasts are created using a consistent modelling approach. Forecasts are optimized for production, and valuation of forecast options including the timing and design of in-field drilling and compression upgrades are considered within the stochastic uncertainty of the dynamic reservoir models.