A new procedure for well productivity and injectivity calibration to improve short-term production forecast

Abstract

During data assimilation (history matching), wells are normally controlled by observed rates over the history period, typically liquid or oil rate for producer and gas or water rate for injectors. The well rates are imposed as target to be honored by the simulator. To perform production forecasts (immediately after the history period) wells control (operational conditions) must be changed to the forecast mode where wells are typically controlled by pressure: bottom-hole pressure (BHP) or well-head pressure. Especially for off-shore fields, group controls (that represent platform limits) should also be defined. The target constraint imposed over the history period may hide productivity and injectivity issues that arise in the beginning of the forecast causing fluctuations in the transition from history to prediction. Even when the BHP curves are well matched (within a tolerance), this problem occurs for most cases. These fluctuations may impair the use of simulation models especially for short-term forecast. Reservoir properties, such as absolute permeability and relative permeability, and well parameters used to adjust the well index, act at the same time in the productivity index (PI), making it more difficult to match all data simultaneously (rates, pressure and well productivity) in a probabilistic history matching approach. This paper addresses this intrinsic complexity by separating the process into two stages: the first one consists of the data assimilation process focused on well rates and pressure matching. The second stage consists of the well productivity calibration considering each model (selected from the first stage) separately, which permits to isolate the other effects that change the well productivity over time, such as fluid mobility, and also the well-block permeabilities, that affect the well index (the static part of the well productivity). When the model is fixed, the effect of the well productivity variation on the flow rate variation is linear, making it much easier the PI calibration. The proposed approach was successfully applied to a real field from Campos basin (Brazil), a deep-water heterogeneous turbiditic reservoir containing heavy oil at the initial pressure. The main contribution of this paper is to increase the simulation models reliability for production forecast, especially for short-term forecasts.