For the past 10+ years, the oil and gas industry has been developing shale reservoirs using long horizontal well bores and multi-stage hydraulic fracturing. These reservoirs are complex, layered systems of mixed lithology and grain size, with various amounts of total organic carbon. The commonality is that these reservoirs are laminated, with low porosity and permeability. Rock properties can also vary laterally along the length of a horizontal wellbore or between wellbores. Industry has done a reasonable job of determining how to 'optimally develop'each reservoir in specific parts of a specific basin essentially by trial and error. Using these multi-million dollar wells to learn what works and what does not work represents an approach to technology development that is costly and has little or no predictive capability away from the geographically limited area where the technology approach is developed.
Even though the industry has done a reasonable job of 'optimizing' in a single formation in a single location in many cases, every horizontal well is a research project. The industry needs models to help predict how to drill, complete and produce shale wells to optimize development in new formations and new geographic areas.
We need predictive geologic models as input to decide where to drill wells, how long the horizontal wells should be drilled, and where to land the laterals. The geologic models should be used as input into both the hydraulic fracturing and the reservoir simulation models.
Once the geologic predictive models have been used to generate a detailed 3D geologic data set, then the hydraulic fracturing modelers need to use the information to determine how the hydraulic fractures will propagate from the horizontal wellbore and how natural fractures might contribute to flow. The industry still has a long way to go before rigorous models are developed to simulate fracture propagation from multiple injection clusters in thick, layered reservoirs.
The ultimate goal is to develop reservoir models that can take the input from the geologic models, the hydraulic fracture propagation models, and additional data such as PVT properties, pressure data, well test data, and production data to match both the reservoir behavior and be capable of predicting production from the reservoir as functions of well bore landing zone, well bore length, and hydraulic fracture properties. Such matches and predictions can only come if we can model the reservoir around the well bore in detail and then learn how to translate that into practical models that can be used to simulate the entire reservoir.