Frontier Issues in Managing Oil and Gas Production

Wednesday, February 9, 2000 - 11:00am - 12:00pm
Keller 3-180
Steve Bryant (The University of Texas at Austin)
Despite enormous advances in reservoir simulation technology, in some respects hydrocarbon production remains a 'seat of the pants' operation. Consider the role of simulation in process- control based industries. In a petrochemical plant, for example, the behavior of individual reactors, distillation columns, and pumps is monitored continuously, and process parameters are changed automatically in real time to maintain desired setpoints. The behavior of the entire plant is also monitored and setpoints can be altered in order to maximize profit on a daily or even hourly basis in response to variations in feedstock and product prices and utility costs. In comparison, the oil industry frequently uses a 'black box' approach in the critical area of forecasting and maintaining production from a field. Production data is rarely compared with expectations from simulations done in the early stages of field development, and 'problem wells' are often treated on an ad hoc, individual basis without considering the influence of the other wells.

One hindrance to routine application of simulation to reservoir management is the huge practical difficulty of determining reliable values for the large number of parameters that any engineering analysis requires (permeability, porosity, relative permeability, capillary pressure, and their variations through the reservoir; fluid properties; reservoir geometry.) Moreover, it often takes too long to collect the necessary data, to set up simulations, to run simulations and interpret the results.

During the next decade or so, technological advances will make possible a new paradigm for reservoir management. For example, permanently installed fiber optic instruments will be available to transmit pressure, temperature and flow rate information continuously from every well in a field. Seismic profiling and fluid tracer technology will increase the resolution with which fluid movements can be tracked. With these sources of increasingly detailed information, it may become possible to operate a reservoir in the same closely controlled, continuously optimized fashion as a petrochemical plant. The heart of the real-time production controller will be the forward simulator of fluid flow in the reservoir. Like the process model in a petrochemical controller, the simulator must be fast and must capture the essential physics of the processes in the reservoir at all the relevant scales, which range from kilometers to centimeters and below. The challenge of speed is huge, and this talk will discuss some of the options for meeting this challenge. We will also consider the related challenge of interpreting the huge volume of simulation results, which imposes special requirements on the interface between scientific computing and visualization.

Beyond the fast forward simulator is the challenge of opening the black box of the reservoir rock. One vision is to use the real-time stream of data from the wells in the field to solve continuously an inversion problem. In essence the simulator would continuously condition the values assigned to rock properties throughout the geological model so that the simulated production and injection rates best match the actual rates. This continuously updated model would thus be the best tool possible for forecasting field performance. Such a model would establish a much more reliable basis for planning recovery strategies based on all the economic factors that influence field development.