Prediction of geothermal wells location and hot water temperature using machine learning

Abstract

With the world facing difficulties like global warming there is need to curb down the use of renewable sources to a certain extent and employ the use of systems which use non-renewable energy sources more. In this paper we try to predict the surface temperature because the geothermal wells are highly dependent on surface temperature. Moreover, we conducted another experiment where we predicted hot water temperature coming out of the heat pumps. These two experiments were conducted keeping in mind the need of renewable energy and how geothermal wells as well as heat pumps can be beneficial to the human generation. The algorithms implemented for both the experiments were Support Vector Machines and Random Forest algorithm due to their higher efficiency and better accuracy as compared to other regression models. The output will be in the form of Root Mean Square error (RMSE) and accuracy for both of these algorithms in order to predict surface temperature. RMSE tells actual difference between actual and predicted values that can be really helpful for prediction. These models can then be used on a larger scale if the output comes better for larger datasets provided pre-processing steps are being properly carried out.