PROBLEM STATEMENT

Building a reservoir model, using the traditional simulation and modeling techniques, is impractical in many cases. Lack of expert manpower, capital or most importantly data can contribute to this impracticality. The objective is to build a reservoir model that can be a) validated for accuracy and b) be used to identify the most appropriate locations in the field for infill drilling. Traditionally reservoir models are built using a bottom-up approach. Meaning that geology is always the starting point. In the Top-Down, Intelligent Reservoir Modeling the reservoir model is built and then validated using the measured data such as production data and well logs as the starting point. In this projecy a top-down model was developed and validated for the Wattenberg field producing from Codell and Niobrara formations in the D.J. Basin.

METHODOLOGY

The methodology for Top-Down, Intelligent Reservoir Modeling is multi-folds. It integrates traditional reservoir engineering techniques such as Decline Curve Analysis, Type Curve Matching , single-well History Matching, Volumetric Reserve Estimation and calculation of Recoverable Reserves with state-of-the-art in Artificial intelligence and Data Mining (AI&DM). The AI&DM techniques that are used to build the final cohesive full field model incorporates artificial neural networks, genetic optimization and fuzzy set theory. Implementation of Decline Curve Analysis in IPDA™. Implementation of Type Curve Matching in IPDA™. Implementation of Single-Well History Matching in IPDA™. Implementation of Volumetric Reserve Estimation in IPDA™. Implementation of Recoverable Reserves in IPDA™. Upon completion of all the above traditional reservoir engineering analyses, a wealth of data about the field has been generated. The nex step is to build a series of individual, data driven, predictive models for all the wells in the field. Building individual, data driven predictive models in IPDA™. Once predictive models for all the wells in the field have been prepared, ISI's Fuzzy Pattern Recognition technology is used to fuse all the individual models into a cohesive and comprehensive full field model. Building a comprehensive and cohesive predictive models for the full filed in IPDA™.

RESULTS

Above methodology was applied to the Wattenberg field producing from Codell and Niobrara formations in the D.J. Basin. Following are samples of reservoir engineering analysis on individual wells. DCA and TCM for a well in the D.J. Basin. Once all the traditional analyses was completed on individual wells (a section of the field with 137 wells were used in this analysis) a full filed model was developed for the field. Following figures show the Fuzzy Pattern Recognition and deliniation of the field into different Relative Reservoir Qualities as a function of time. Fuzzy Pattern Recognition model for the full filed three months into field production, Wattenberg field producing from Codell and Niobrara formations in the D.J. Basin. Relative Reservoir Quality deliniations are identified. Fuzzy Pattern Recognition model for the full filed three years into field production, Wattenberg field producing from Codell and Niobrara formations in the D.J. Basin. Relative Reservoir Quality deliniations are identified. Upon development of the full field model 3 dimensional map of permeability, drainage area and fracture half length in the filed can be developed.. Fuzzy Pattern Recognition model use to develop the 3D maps of permeability, drainage area and fracture half length for the Wattenberg field producing from Codell and Niobrara formations in the D.J. Basin. In this study the model was validated by removing several (most recently completed) wells from the dataset, building the model and using the model to predict the performance of the removed (blind) wells. The model was able to successfully predic the removed wells' production characteristics.

REFERENCES
Intelligent Production Data Analysis; IPDA™
Paper in SPE Journal of Reservoir Engineering & Evaluation