Reservoir properties prediction with least square support vector machine

摘要

Subsurface reservoir properties are predicted despite limited availability of well log and multiple seismic attribute data. The prediction is achieved by computer modeling with least square regression based on a support vector machine methodology. The computer modeling includes supervised computerized data training, cross-validation and kernel selection and parameter optimization of the support vector machine. An attributes selection technique based on cross-correlation is adopted to select most appropriate attributes used for the computerized training and prediction in the support vector machine.

主权项

1. A computer implemented method of modeling a reservoir property of subsurface reservoir structure by support vector machine processing in the computer of input data available from the reservoir to form measures of the reservoir property at regions of interest in the subsurface reservoir by regression analysis of the available input data, the method comprising the computer processing steps of:
(a) receiving training input data about subsurface attributes from seismic survey data obtained from seismic surveys of the reservoir; (b) receiving training target data about formation rock characteristics from data obtained from wells in the reservoir; (c) partitioning the subsurface attributes training data and the formation rock characteristics training target data into a plurality of subsets; (d) selecting formation attribute parameters for support vector machine modeling by performing the steps of:
(1) cross-validating the subsets of subsurface attributes training data each with the other subsets of the plurality of subsets for a radial based kernel function pair comprising a kernel parameter value and a penalty parameter pair value;(2) forming an error function for each of the cross-validated subsets;(3) repeating the steps of cross-validating the subsets of subsurface attributes training data and forming an error function for a plurality of different radial based kernel function pairs; (e) optimizing the selected formation attribute parameters by determining a minimum error function of the formed error functions for the plurality of different radial based kernel function pairs; (f) providing the training data, the selected formation attribute parameters, the cross-validated subsets of subsurface attributes training data, and the error functions for the plurality of radial based function kernel pairs as training inputs for support vector machine modeling; (g) performing support vector machine modeling by regression analysis to determine a minimum error function of the error functions of the provided training inputs; (h) predicting the reservoir property based on the support vector modeling of the training inputs; and (i) forming an output display of the predicted reservoir property.