Comparison of Predictions from Reservoir Models to Actual Production Data: Field Examples

Abstract

Reservoir models were constructed on two gas condensate reservoirs (M-4/M4-A and M-6) in the Gulf of Mexico. The gas reservoirs are at a depth of 15000 feet subsea, over pressured and supported by active aquifers. Initial gas production was essentially water-free with steep pressure decline that was later followed by rapid rise in water production. Peak gas production from either reservoir reached about 166 MMCFPD. The reservoir models were fully compositional. The models were history-matched with production data and then used to predict future performance. This paper compares predictions of the models to actual reservoir production data. The model for the M-6 reservoir predicted gas production with an accuracy of -0.72%, condensate production at +2.80% and water production at +12.4%. The predicted accuracies for the M-4/4A reservoirs are as follows: gas (-0.33%), condensate (-4.99%) and water (+25.56%). The main objective of this paper is to provide data on the errors associated with the use of history matched models to forecast reservoir performance. Surveys of the petroleum literature indicate that they are few technical papers that compare model predictions to actual reservoir performance data over an extended period of time. Mass publications of these types of results are essential in establishing reservoir simulation as a valuable analytical tool for reservoir engineers. It will also improve its acceptance in the petroleum industry as a reliable tool for prediction of reservoir performance. Introduction The M-4/M-4A and M-6 sands are mid-size, over pressured and moderately rich gas condensate reservoirs located in the Gulf of Mexico at a water depth of about 240 feet. At their peak, gas production rate from each reservoir averaged 166 MMCFPD. Production data show that both reservoirs were supported by very strong aquifers. Two fully compositional models were constructed for the M-4/M-4A and M-6 reservoirs. The models were executed on VIP-COMP simulator from Landmark Graphics Corporation. This work is a direct application of reservoir simulation technology for prediction of reservoir performance especially in terms of fluid production rates and cumulative recoveries. The techniques or work processes used in the study are basic and routine. The relevance of the work lies in comparing model predictions to actual production data from the reservoirs. Generally, few technical papers have been published that compare model predictions to actual reservoir performance data over an extended period of time. As reservoir simulation develops into a tool used by most reservoir engineers, experienced practitioners of the art of reservoir modeling must strive to provide to the technical literature data that quantify errors associated with predictions from reservoir models. Concerted efforts in this area is needed at this stage of evolution of reservoir modeling It will accelerate acceptance of reservoir modeling as a reliable tool for prediction of reservoir performance.