Assessing the Accuracy of History-Match Predictions and the Impact of Time-Lapse Seismic Data: A Case Study for the Harding Reservoir

Abstract

Abstract The oil industry benefits from reservoir models that are accurate predictors of future performance, but a "good" match to historical well data doesn't guarantee a "good" prediction. This paper presents a case study from the Harding field in the North Sea, illustrating that we can greatly improve our chances of getting "good" predictions from our reservoir models by including time-lapse (4D) seismic data as part of the history matching process. The value of the surveillance is also discussed in the light of its use in improving reservoir performance prediction. The first part of the case study addresses the accuracy of prediction. The Harding field (Figure 1) started production in 1996, with the first time-lapse seismic survey being acquired in 2000. Using computer-assisted history matching techniques,1 multiple alternative reservoir models were generated that matched the well surveillance to 2000. In a blind test of predictive capability, forward predictions from these alternative models were compared with the subsequent production data from 2000 to 2003, allowing us to compare quantitatively the prediction quality against the history match quality for the selected cases. The time-lapse seismic data from 2000 was then included in the history matching process and the impact on the resulting prediction quality through 2003 was evaluated. Addition of the 4D seismic data enabled the selection of a subset of reservoir models which better predicted the (known) future of the field. The best hypotheses are those that make predictions that can be tested. If 4D seismic information is effective in reducing uncertainty, then there should be a way of making a quantitative estimate of the value of a future seismic survey. The impact and value of acquiring a future time lapse seismic survey, and a discussion around the metric for value of surveillance, are discussed in the second part of the case study.