An evolutionary system for exploitation of fractured geothermal reservoirs

In this work, we present a novel methodology to integrate one of the most advanced technique for modeling fractured media for underground systems with a semantics-based genetic programming technique. The objective of the study is to develop a global framework to forecast the temperature of fractured reservoirs. The numerical method used to solve the physical equations is able to handle different fracture distributions without changing the background computational grid, i.e. the mesh of the rock matrix, as well as letting geometrically uncoupled the one co-dimensional fracture meshes. In the context of temperature forecasting, the use of a recently defined variant of genetic programming is taken into account for finding (quasi-)perfect solutions with high probability and for generating models able to produce near optimal predictions also on unseen data. The proposed computational intelligence technique integrates, in a recently developed version of genetic programming that uses semantic genetic operators, a “greedy” crossover and a self tuning algorithm. Experimental results confirm the suitability of the proposed method in predicting the correct temperature distribution in probes inside the domain.