When a Good History-Matched Model Results in a Bad Forecast; Question your Petrophysical Data

Abstract

When implementing any field development plan or an Enhanced Oil Recovery (EOR) project, it is critical to understand the key variables that influence the success of the plan such as; reservoir parameters and fluid properties. Mature Oil & Gas fields' increasing complexity of recovery mechanisms dictates an improved understanding of the fields' behaviour and the technologies that must be applied to maintain and prolong oil production plateau and achieve ultimate recovery potential. History matching forms an integral part of the reservoir modeling workflow process. It is used to examine the field performance under different production and injection scenarios in order to select the best scenario for hydrocarbon production. However, the history-matching process can be very frustrating and time-consuming, even for fields that appear relatively small and simple in nature, because of the reservoir processes involved and the non-unique nature of the solution. Traditionally, history matching is conducted as a deterministic process with a single realization considered representative at a single point in time. Although, the input data usually go through a data analysis process where the major uncertainties and scenarios are defined, and uncertainty ranges are created, time and budget constraints usually result in significant reductions in the number of sensitivity runs and analysis for the input data validation and quality control that results in an incomplete investigation of the uncertainty quantification. Therefore, uncertainties inherited in the Petrophysical data are carried from the static model construction throughout the entire dynamic modelling process, ultimately leading to less-than-optimal models to be used as a decision making tool. Consequently, due to the non-uniqueness of the numerical solution, a good history-matched model might have geological and petrophysical properties quite far from those of the "Field" and therefore could lead to a bad forecast. As in any numerical model, petrophysical data quality is fundamental for model precision.