Modeling and prediction of geothermal reservoir temperature behavior using evolutionary design of neural networks

Analytic modeling of geothermal reservoir temperature behavior is such a complicated process that aspects of that have been investigated experimentally and modeled using generalized GMDH-type (Group Method of Data Handling) neural networks. The experimental data used for training GMDH-type neural network are extracted from six exploration wells in Sabalan geothermal site in Iran. The input-output data used for modeling consists of five variables as input data namely, northing and easting of the top point of the well, major depth, angle and azimuth of each inside point of the well in relation to the top point of the well and one output which is the temperature of each inside point of the wells. Further, comparison of actual values with the proposed GMDH model corresponding depicts the very good behavior of proposed model of this work. It is also shown that the two factors namely, northing of the top point of the well and azimuth of each inside point in relation to the top point of the well, do not influence the temperature of each inside point of the wells. Moreover, genetic algorithm (GA) and singular value decomposition (SVD) are deployed simultaneously for optimal design of both connectivity configuration and the values of coefficients of quadratic sub-expressions embodied in such GMDH-type networks, respectively. (C) 2014 Elsevier Ltd. All rights reserved.