(28h) Towards an Integrated Wide Approach for Sustainable Upstream Field Recovery

Integrated Asset Modelling is the modelling of an entire production facility consisting of both subsurface and surface elements. The intent is to holistically capture the complex interactions between the individual components of the system. Historically, asset modelling has entailed discrete reservoir and facility models; adopting an integrated approach provides a holistic overview to modelling the production system which enables the user to determine the system constraints and evaluate optimisation opportunities for maximising production. Within the industry, the full benefits of integrated models are not realised for several reasons, including extensive simulation time of complex models, siloed disciplinary approaches resulting in separate disciplines owning separate elements of the framework and simplification challenges.

This research focuses on the development of surrogates of an integrated field set up for both a gas producing and oil producing supplemented by water injection field case. Surrogates were developed via numerous techniques (Response Surface Models, ANN, ALAMO and RS-HDMR) and tested to determine the accuracy versus the industry state of the art tool. Global sensitivity analyses were performed on a selected number of input field variables (including time dependency) of the postulated surrogate models prior to attempting optimisation of the developed surrogates. Following this, a constrained non-linear optimisation framework of the derived proxy models was proposed with favourable results; further gradient free meta heuristic optimisation of the surrogates was also reviewed including genetic algorithms applied to neural networks. The surrogates developed provided appropriately accurate predictions when compared a commonly employed fully developed integrated tool deployed within the upstream industry with significant gain on simulation time. Deviations in the performance and optimisation results of the surrogates were observed at later field life. Lastly, machine learning was used to adapt a reduced order integrated model to predict the performance of a full physics integrated asset model with similar outcomes.

Deviations from the predicted profiles in comparison to the integrated asset model outputs were noted at later field life or in circumstances when production is significantly choked back which would conclude that the surrogate models developed are more suited to be used in certain phases (early) of a field development or life cycle of a field hence within an opportunity realization framework, within the assess, select concept framing stages. Instabilities within the predictions of the surrogates were noted and attempt was made to smooth these out to improve the surrogate model definition.

The proposed surrogate models developed as part of this research have demonstrated that with sufficient definition, they can be deployed to generate a production forecast and field pressure profile with demonstrated improvement in the simulation time in comparison to a typically employed industry software tool which is typically restrictive in its ability to communicate with third party packages (and comes with associated costs). The developed models would most likely be deployed in an appraisal or early phase of a field development cycle given the responses noted at later field life conditions. The application of integrated asset modelling facilitates the maximisation of production from existing fields in addition to mitigating potential further installation and drilling of new facilities and wells respectively, hence is a tool, driver and ultimately, an enabler for sustainable and economical production from existing fields. This is particularly pertinent given the forecasted demand of hydrocarbons within the energy mix to meet the global energy demand in the short to medium term.