Best Practices of Assisted History Matching Using Design of Experiments

Abstract

Summary Assisted history matching (AHM) using design of experiments (DOE) is one of the most commonly applied history-matching techniques in the oil and gas industry. When applied properly, this stochastic method finds a representative ensemble of history-matched reservoir models for probabilistic uncertainty analysis of production forecasts. Although DOE-based AHM is straightforward in concept, it can be misused in practice because the work flow involves many statistical and modeling principles that should be followed rigorously. In this paper, the entire DOE-based AHM work flow is demonstrated in a coherent and comprehensive case study that is divided into seven key stages: problem framing, sensitivity analysis, proxy building, Monte Carlo simulation, history-match filtering, production forecasting, and representative model selection. The best practices of each stage are summarized to help reservoir-management engineers understand and apply this powerful work flow for reliable history matching and probabilistic production forecasting. One major difficulty in any history-matching method is to define the history-match tolerance, which reflects the engineer's comfort level of calling a reservoir model “history matched” even though the difference between simulated and observed production data is not zero. It is a compromise to the intrinsic and unavoidable imperfectness of reservoir-model construction, data measurement, and proxy creation. A practical procedure is provided to help engineers define the history-match tolerance considering the model, data-measurement, and proxy errors.

Clay–rock fracturing risk assessment under high gas pressures in repository systems

“…As the phase-field approach is costly regarding computation time, we studied the risk using a proxy model. For this purpose, we employed a workflow (Figure 1) similar to an approach used elsewhere to study uncertainties in the context of reservoir engineering [4] or waste repositories [7] based on a history matching and Design-of-Experiments Approach and also very similar to a proxy-based approach for studying sensitivity and uncertainty of fracture properties of nanocomposites with the phase field method [18]. As no detailed data on parameter distributions for our set of parameters was available, we used a uniform probability density function for all parameters.…”

“…However, these fracturing simulations are computationally costly. Furthermore, computational costs become infeasible if we are to estimate the sensitivity of the parameters and assess risks under various scenarios, For this reason, we have embedded the fracturing simulation in a risk analysis tool, the Design of Experiments (DoE) [4,5,6,7], which develops a computationally inexpensive proxy model based on a reduced set of significant parameters known as heavy hitters.…”

Reservoir automatic history matching: Methods, challenges, and future directions

“…Offline methods build a sufficiently accurate surrogate model that can completely replace the role of numerical simulation and approximate the entire search space. Most studies (Dachanuwattana et al, 2018;Li et al, 2019) use sensitivity analysis methods to select some key parameters and then use traditional machine learning methods to construct alternative models. The emergence of deep learning makes it possible to directly establish the mapping from high-dimensional spatial parameters to reservoir dynamics without sensitivity analysis.…”

Deep Conditional Generative Adversarial Network Combined With Data‐Space Inversion for Estimation of High‐Dimensional Uncertain Geological Parameters

“…In recent years, the deep‐learning‐based surrogate model has been widely researched in inverse modeling, which can completely replace the role of numerical simulation and approximate the entire search space. Different from traditional studies that try to select the key parameters through sensitivity analysis and then use the machine learning method to establish the surrogate model (Bhark & Dehghani, 2014; Dachanuwattana et al., 2018; Li et al., 2019). The proposal of deep learning makes it possible to directly input high‐dimensional uncertain parameters into the surrogate model without sensitivity analysis to select key parameters.…”