Jung, Hye Young (2015-12). Application of History Matching Quality Index with Moving Linear Regression Analysis. Master's Thesis. Thesis uri icon

abstract

  • History Matching is the process of calibrating uncertain parameters of a reservoir model in order to reach the best plausible match with the observed data. By integrating the dynamic data of a reservoir, the reservoir properties can be estimated so that it is a key step in developing reservoir performance, which is normally time consuming and computationally infeasible. With the rise of global energy demand, the reliance on enhanced oil recovery (EOR) has increased and the model calibration for chemical flooding also becomes significant. However, due to large amounts of uncertain parameters and complicated relationships among them, it is hard to apply a traditional manual history matching with a single deterministic model to chemical flooding. Instead, a stochastic method using the genetic algorithm (GA) can be efficient in that it can consider several parameters simultaneously. However, this probabilistic-assisted history matching generates several updated models, all of which have a potential to be good matches. Therefore, there is a need to evaluate history matching results consistently without any subjectivity. In addition, the assessment of results from model calibration is also difficult when it comes to large field cases, which are involved with a number of wells and different types of objectives. Since each well and objective presents contrasting results, a comprehensive decision making for selecting a better history matching model is necessarily complicated. However, current approaches mostly rely on reviewers' experience, which is too subjective or uses a misfit function without any consideration for the data. We first introduce a History Match Quality Index (HMQI) in assessing the quality of history matching and ranking among those results. This method assigns index a value of either 0 or 1 based on the quality of the match. Moreover, combining the HMQI with a Moving Linear Regression Analysis (MLRA) provides the more robust assessment by removing outliers which come from a variety of sources of errors. Secondly, we apply the HMQI to the synthetic case of alkaline-surfactant-polymer (ASP) flooding as well as that of polymer flooding. Moreover, we compare the results with other method for evaluating the quality of history match to prove the feasibility of our approach. Lastly, field-scale simulations are conducted to demonstrate the reliability and robustness of our methodologies. The HMQI with the MLRA has proven its ability to identify outliers of data using the case study from synthetic to field. In addition, in comparison with the misfit calculation, it has been shown to eliminate subjectivity, using normalized values without the bias toward outliers.
  • History Matching is the process of calibrating uncertain parameters of a reservoir model in order to reach the best plausible match with the observed data. By integrating the dynamic data of a reservoir, the reservoir properties can be estimated so that it is a key step in developing reservoir performance, which is normally time consuming and computationally infeasible.
    With the rise of global energy demand, the reliance on enhanced oil recovery (EOR) has increased and the model calibration for chemical flooding also becomes significant. However, due to large amounts of uncertain parameters and complicated relationships among them, it is hard to apply a traditional manual history matching with a single deterministic model to chemical flooding. Instead, a stochastic method using the genetic algorithm (GA) can be efficient in that it can consider several parameters simultaneously. However, this probabilistic-assisted history matching generates several updated models, all of which have a potential to be good matches. Therefore, there is a need to evaluate history matching results consistently without any subjectivity.
    In addition, the assessment of results from model calibration is also difficult when it comes to large field cases, which are involved with a number of wells and different types of objectives. Since each well and objective presents contrasting results, a comprehensive decision making for selecting a better history matching model is necessarily complicated. However, current approaches mostly rely on reviewers' experience, which is too subjective or uses a misfit function without any consideration for the data.
    We first introduce a History Match Quality Index (HMQI) in assessing the quality of history matching and ranking among those results. This method assigns index a value of either 0 or 1 based on the quality of the match. Moreover, combining the HMQI with a Moving Linear Regression Analysis (MLRA) provides the more robust assessment by removing outliers which come from a variety of sources of errors.
    Secondly, we apply the HMQI to the synthetic case of alkaline-surfactant-polymer (ASP) flooding as well as that of polymer flooding. Moreover, we compare the results with other method for evaluating the quality of history match to prove the feasibility of our approach. Lastly, field-scale simulations are conducted to demonstrate the reliability and robustness of our methodologies.
    The HMQI with the MLRA has proven its ability to identify outliers of data using the case study from synthetic to field. In addition, in comparison with the misfit calculation, it has been shown to eliminate subjectivity, using normalized values without the bias toward outliers.

publication date

  • December 2015