In this paper, we first provide guidelines for selecting the most appropriate permanent downhole sensor or combination of sensors for reservoir monitoring, given fluid and rock characteristics. This selection is applied to pressure, electrical, and seismic sensors, based on their respective response equations to formation and rock properties.
We then present two synthetic cases illustrating the interpretation of permanent monitoring data, with emphasis on the benefits of data fusion and data assimilation, referring respectively to the use of multisensor and time-lapse data; more specifically, we show how those approaches enhance the convergence of the inversion process towards the solution.
Introduction
One important challenge for reservoir management in the coming decade is to monitor fluid movements in hydrocarbon reservoirs, with the goal of optimizing their drainage. Three main physical principles can be envisaged: resistivity, pressure response to well tests, and acoustics (or seismics), since changes in resistivity, mobility, and elastic properties are expected at a front location in a hydrocarbon reservoir.
To this end, permanent downhole sensors such as pressure gauges, electrode or geophone arrays have therefore been1 or are currently being developed. Arrays of such sensors would allow repeatedly conducting reservoir surveys around the instrumented wells.
Because different tools are sensitive to different reservoir and fluid properties, sensor-screening criteria have to be established to take the full benefit of investing in such a technology. This is dealt with, in this paper, by proposing a simple graphical method that will help in choosing the optimal single sensor or the combination of sensors best suited to a given problem and environment.
Once the selected sensors have been deployed, interpreting the data recorded by the sensors to obtain information on the fluid movement within the reservoir requires an inversion process. This inversion process can be improved by taking advantage of time-lapse acquisition (i.e., data assimilation2) and/or by the concurrent use of different types of sensor (i.e., data fusion3).
In this paper, two synthetic examples are used to illustrate the use of permanently acquired data in a waterflood characterization context. The first example deals with the determination of a water-injection front movement from a simultaneous inversion of pressure, electrical potential, and seismic data, showing how their joint use can alleviate the indetermination on the front geometry parameters to be inverted.
In the second example, we show how information about relative permeabilities can be derived from interpreting a continuous stream of flow rate and electrode-array potential data. In particular, we show how the time-lapse nature of the acquisition can be used to obtain better parameter estimates.
Sensor Response Equations
Selection of a wellbore sensor, such as a rate measurementdevice, is relatively straightforward, given the anticipated multiphasic well production and the device's nominal characteristics. By contrast, selecting one (or several) reservoir sensor(s) may be a much more difficult task, given the large number of reservoir and fluid properties that are influencing the measurement(s). One should therefore return to the response equation of each sensor to evaluate the sensitivity of a particular measurement to a given set of reservoir parameters.
The following equations correspond to the response of resistivity, seismic, and pressure sensors for the situation depicted in Fig. 1; i.e., the simple case of a vertical water-front movement across a homogeneous oil-bearing reservoir. The sensors are located in the producing well, and the distance between the sensor and the front at time? is L(t).
