HESPER: An Expert System for Petrophysical Formation Evaluation

Abstract

A prototype expert system for formation evaluation is introduced and discussed, beginning with the rationale for its development and presenting briefly qualitative assessments and tests of its performance.

The background and definition of petrophysical formation evaluation is clarified, and the specific geological frame of the Permian Southern Basin of the North Sea, in which the system operates, is presented, with indications of future developments.

The different modules of the system are discussed in some detail, and the interaction between modelling, advice-giving, and the user interface is clearly shown to be a crucial factor. Of particular importance is the ability of the system of presenting the reasons for its choice of solution in a clear text and in a user-intelligible fashion.

Tests and performance of the system are discussed from a qualitative viewpoint, rather than after rigorous benchmark tests. This is considered appropriate at this stage, since the system is by no means in its final configuration.

The conclusion has been reached by Britoil that the system has demonstrated that the application of expert systems design and technology to petrophysical formation evaluation is both petrophysical formation evaluation is both feasible and desirable, and further development would therefore be beneficial.

2. DEFINITION OF THE PROBLEM

The HESPER system (or Heuristic Expert System for the Petrophysical Evaluation of Reservoirs) attempts to emulate the manual process of Petrophysical Interpretation. Petrophysical Interpretation. The system takes wireline log data, along with drilling coring and other data, and allows the petrophysicist to build and manipulate a petrophysicist to build and manipulate a geological model of the formation. Using petrophysical equations, synthetic logs can be petrophysical equations, synthetic logs can be generated from this model and compared, both visually and by the program, with the real logs recorded in the borehole. When a reasonable match is obtained with all pertinent logs, he model can be taken to represent adequately the formation under evaluation. This statement can be shown in diagrammatic form as a flow diagram (Fig. 1).

Note that we have separated the process into two distinct phases, a data capturing phase, and a data interpretation phase. As indicated in the diagram, HESPER concerns itself with the data interpretation phase. The reasons for this are both conceptual and practical. Data obtained at the wellsite (wireline logs, cores, cuttings, drilling logs, etc.) are in general supplied to the client by different service companies, at different times, using different systems. Often they are in a mutually incompatible format, and in general they all have to be reduced to a common, corrected version, eliminating the influences of the drilling environment, borehole, tool response, etc. In other words, the "raw data signals" have to be processed and deconvolved into "formation processed and deconvolved into "formation response signals."

P. 361