Low oil recovery reached after more than 60 years of production in a rather deep heavy oil (9.5-11 °API) field in South America encourages looking for a thermal method (steamflood) which potentially promises a tenfold increase in the recovery factor.
Steamflood forecasting for such a large field is highly challenging. Direct thermal reservoir simulation seems impractical because of the size of the model (about 120 million grid-blocks). A recently developed correlation-based probabilistic technique has met the forecasting challenge. First, a relatively small set of thermal reservoir simulation runs with sector models provided a basis for production/injection correlations between key rates as well as cumulative quantities and the models' geologic and engineering input. All the runs included pressure depletion and steamflood stages. Next, about 350 forecasting elements (1 sq. km each) covered the entire Northern part of the field, which was subject of the forecasting; every element received probabilistic distributions of the key geologic parameters involved in correlations. Then, the Monte-Carlo simulation united the correlations with the distributions and allowed selecting combinations of the parameters corresponding to P10-P50-P90 probabilities of the dynamic performance. Reconstruction of oil and water production as well as steam injection P10-P50-P90 time-profiles completed preparation of the forecasting elements.
Reservoir depth and quality cut-offs restricted the area of the potential future steamflood application. For several sets of the cut-offs, wells were placed within the areas in accordance with inverted 7-spot patterns at either 8 or 24 acres/pattern. Every well adopted production or injection profiles from its forecasting element-host. Drilling timetable controlled the wells coming on primary production. Steamflood started in a 100-pattern cluster after 4-5 years of primary production. Heat management governed time-profile of steam injection into the cluster. On reaching a steam-oil-ratio cut-off, steam injection shifted to the next 100-pattern cluster. Summarizing individual well production or injection time-profiles, arranged in accordance with the drilling schedule, generated field P10-P50-P90 forecasts for several cases of interest.