Since 1991, earthquakes have been recorded in the Groningen gas field and have been shown to be induced by gas production. Following the 2012 Huizinge earthquake (magnitude 3.6), increased concerns about seismic risk led to a ministerial requirement to analyze the relationship between gas production and seismicity, as well as the potential for mitigating seismic hazard through operational intervention.
Outcomes of the research were immediately used in the policy setting process. From 2014 onward, gas production was progressively reduced below field capacity, creating degrees of freedom in the spatial distribution of production across the field. This raised the question of whether spatial redistribution of gas off-take could be used to reduce induced seismic hazard while meeting prescribed production volumes.
To address this question, the authors developed and applied a model-driven optimization framework that couples a dynamic reservoir simulator with a probabilistic seismic hazard and risk assessment (PSHRA) model. The framework evaluates reservoir pressure evolution, compaction, earthquake event rates, and ground-motion metrics for alternative spatial distributions of production. Optimization targeted several hazard- and risk-relevant objective functions, including event count, maximum peak ground acceleration (maxPGA), maximum peak ground velocity (maxPGV), and population-weighted PGV (pwPGV) as a proxy for seismic risk.
Across the first 4 gas years of the prescribed production profile, the optimization identified feasible redistributions of off-take from the field that reduced hazard proxies by 10–14% relative to the 2017 production distribution, while optimization on the pwPGV risk proxy yielded a 5% reduction. Because actual field off-take must continuously adjust to weather-driven demand swings, a single optimized production scenario was insufficient. The analysis, therefore, was expanded using machine learning proxy models. Random forest surrogates combined with partial dependence analysis were used to derive robust, objective-specific production startup rankings that remain valid under operational variability.
The results demonstrate that coupling reservoir simulation with PSHRA, complemented by data-driven proxy models, provides a practical and scientifically grounded approach for translating optimization under seismic hazard constraints into operational production strategies. The resulting guidance was subsequently adopted in regulatory decision-making and implemented in actual field operations, demonstrating the practical relevance of the proposed methodology.
This abstract is taken from paper SPE 234686 by L. Geurtsen, P. Valvatne, and A. Mar-Or, Nederlandse Aardolie Maatschappij; G. Kaleta, J. Limbeck, and G. Joosten, Shell; and J. Van Elk, Nederlandse Aardolie Maatschappij. The paper has been peer reviewed and is available as Open Access in SPE Journal on OnePetro.