Genetic Algorithm in Subsidence Modeling in the Cerro Prieto Geothermal Field, Baja California, Mexico.

Abstract

Comparison between observed subsidence rate and 30 years of fluid extraction in the Cerro Prieto Geothermal Field (CPGF) suggests that the observed subsidence is mainly of anthropogenic origin (Glowacka et al., 1999, 2000). Additionally, 8 years of continuous observations of extension at the Imperial fault and field observation of the Cerro Prieto fault indicate that most of the subsidence is bounded by these faults. In this work we use the precision leveling data obtained by CFE (ComisiÛn Federal de Electricidad, Mexico) during 1994-1997. We choose the mathematical model of tensional cracks of Yang and Davis (1986) as the one most appropriate to represent the deformation of sediment layers produced by fluid extraction in a reservoir bounded by faults. A genetic algorithm was used to fit the crack parameters: x, y, z (center of crack), p (crack closure), c and c1 (crack dimensions), and azimuth and dip (crack orientation) by minimizing the RMS error between observed and modeled subsidence rate. The algorithm works in a stripping mode: after fitting a crack to the data, its effects are removed and new cracks are fit to the residuum. After calculating a few tens of models, we analyze the physical interpretation of the calculated parameter values and compare it with a crack model based on the known hydrological model (Sarytchikhina et al, this volume). Except for the depth of the main reservoir, the genetic solutions agree, in general, with the known structure of the reservoirs. The analysis shows that vertical data alone cannot resolve the dependency between depth and closure of the cracks, and points out the necessity to have both vertical and horizontal measurements in order to obtain unique solutions.