Abstract
With the digital transformation of the oil and gas industry, new processes like the proposed alternative history matching methodology are required to leverage the growing variety of production data. This enhances the strategy of building trustworthy reservoir simulation models that more realistically represent reservoir dynamics. Conventional history matching typically uses an objective function with limited observed data like pressures and rates. This overlooks other factors controlling fluid flow and forecast reliability. This PhD research proposes and investigates an enhanced technique incorporating non- conventional data such as logs, 4D seismic, and tracers into the validation process. The methodology integrates three components: 1) a flexible classification method to categorise diverse data types, 2) tailored match metrics for each data class, and 3) a robust multivariate objective function integrating the parameters based on a multi-criteria framework. By expanding beyond traditional production data mismatches, key findings demonstrate superior reservoir characterisation and better representation of the flow dynamic in highly heterogenous water flooding reservoirs. Integrating multiple static and dynamic data sources provides comprehensive insights into displacement patterns, heterogeneity, and forecast accuracy. Other practical implications include better development planning, drilling optimisation, and production forecast reliability. The novelty behind the proposed methodology lies in the flexibility to incorporate any surveillance data, complementing rather than compromising conventional methods through optimised leveraging of expanded information. This provides a capable framework to create more reliable reservoir models by improving the understanding of fluid dynamics derived from all available data, thus unlocking the full potential of models to define new development opportunities. The approach was evaluated on a semisynthetic reservoir model using a state-of- the-art benchmarking workflow to generate and select matched models across an uncertainty envelope. Comparative analysis of representative models from this process proves the methodology's capabilities for creating more reliable reservoir realizations by extracting maximum value from the incorporation of diverse dynamic data.Keywords: history matching, simulation models, multivariate data, reservoir characterisation, binary metrics
| Date of Award | 11 Jun 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Hom Dhakal (Supervisor), Mohamed Galal Hassan Sayed (Supervisor) & Reza Malakooti (Supervisor) |