An Artificial Neural Network Model for Na/K Geothermometer

Abstract

In this study, a brief explanation is first given on solute Na/K geothermometers developed until now, and a new Na/K geothermometer model is derived after presenting a world geothermal database (n=212) to a neural network as a training set and another database (n=112) as a validation set. In this model Na and K values are treated as input values and geothermometer temperatures as output values. A multilayer feedforward neural network is trained using a genetic algorithm for optimizing hidden layer neuron weights and linear regression for optimizing output neuron weights. The model is successfully evaluated and compared with actual deep temperature measurements to avoid training bias.