Many wells drilled and completed to produce oil and gas must be stimulated to increase flow rates and ultimate recovery. In low permeability reservoirs, long hydraulic fractures are required to optimize the depletion of the reservoir. In high permeability reservoirs, formation damage during drilling must be overcome using short, highly conductive hydraulic fractures. Most recently, in ultra-high permeability, poorly consolidated formations, a new stimulation method that combines both hydraulic fracturing and gravel packing principles (the FRACPAC) has been successfully used to stimulate oil and gas production.
To optimize production, one must optimize the fracture fluid properties, the treatment volume, and the fracture conductivity. Choosing the correct fluid and additives is extremely important to be sure the proppant is placed successfully. We also want to use a fluid that breaks and cleans up properly. Ideally, every fracture treatment should be designed by an expert with adequate data available. However, most fracture treatments are not designed by experts; instead, most stimulation treatments are designed by inexperienced engineers without all data they need.
We have developed a PC-based interactive computer model to help an engineer choose the best fluid, additives, and propping agent for a given set of reservoir properties. The computer model also optimizes the treatment volume based upon reservoir performance and economics. To select the fluids, additives, and propping agents, the expert system uses rules developed by surveying stimulation experts from different companies, reviewing the literature, then incorporating the knowledge into rules using an expert system shell. Our design expert queries the user for necessary reservoir data, then using the knowledge base, the expert system recommends the best materials for the treatment.
In this paper, we have explained the logic behind the rules we have developed, and we explain our optimization procedure. Although it is physically impossible to include every rule we have developed, we reference the source of many of our rules and present examples to illustrate our methodology. We expect the information contained in this paper will be useful to others working on a stimulation expert system.
The information contained in knowledge bases are the core of any expert system. The most difficult task when building an expert system is knowledge acquisition. As one improves the quantity and quality of the knowledge that is programmed, the expert system will better simulate the behavior of a human expert.