Magnetic resonance logs provide the capability of in-situ measurement of reservoir characteristics such as effective porosity, fluid saturation, and rock permeability. This study presents a new and novel methodology to generate synthetic magnetic resonance logs using readily available conventional wireline logs such as spontaneous potential, gamma ray, density, and induction logs. The study also examines and provides alternatives for situations in which all required conventional logs are unavailable for a particular well. Synthetic magnetic resonance logs for wells with an incomplete suite of conventional logs are generated and compared with actual magnetic resonance logs for the same well.
In order to demonstrate the feasibility of the concept being introduced here, the methodology is applied to a highly heterogeneous reservoir in East Texas. The process was verified by applying it to a well away from the wells used during the development process. This technique is capable of providing a better image of the reservoir properties (effective porosity, fluid saturation, and permeability) and more realistic reserve estimation at a much lower cost.
In a recent SPE paper1 it was shown that it is possible to generate virtual magnetic resonance logs using conventional wireline logs. The concept was tested on several wells from different locations in the United States and the Gulf of Mexico. It was demonstrated that using virtual intelligence techniques (artificial neural network in this case) it is possible to generate accurate virtual magnetic resonance logs. It was further demonstrated that using the virtual magnetic resonance logs for reserve calculation provides very accurate estimations (within 3%) when compared to reserve estimation obtained by actual magnetic resonance logs.
The major shortcoming of that study was the fact that the development and the testing of the process were performed on the same well. In that study part of the pay zone was used for the model development and then the model was tested on the rest of the pay zone. The main reason for that shortcoming was lack of data. The study was conducted on several fields but from each field data was available from only one well. It was mentioned that the ultimate test for this methodology would be when data from several wells in a particular field would be available, so the methodology can be tested in a manner that would simulate its actual use. This methodology would work best when conventional logs are available from most of the wells in the field and magnetic resonance logs are performed only on a handful of wells (these wells should also have the conventional logs). The wells with magnetic resonance log will be used for model development and consequent testing and verification of the model. Then the developed (and verified) model will be applied to all the wells in the field. This would generate a much better and more realistic picture of the reservoir characteristics for the entire field. Having such an accurate picture of reservoir characteristics would be a valuable asset for reservoir simulation, modeling, and reservoir management.
This paper provides the ideal test bed for this methodology. Here the methodology is applied to a field in East Texas (Cotton Valley formation) that is known for its heterogeneity as well as for the fact that the well logs and reservoir characteristics are non-correlatable from well to well. A recently published paper2 demonstrated the non-correlatable nature of formation characteristics and well logs in this formation.