Power generation prediction of a geothermal-thermoelectric hybrid system using intelligent models

Abstract

This study focuses on the most effective use of energy resources, which is the main goal of sustainable energy systems. The first contribution of the research is to ensure that the waste and geothermal fluids generated during the geothermal energy production process are evaluated in the most efficient way and reused for energy production. For this purpose, a geothermal-thermoelectric hybrid experimental system has been designed to produce electrical energy through heat exchange and how waste thermal energy could be reused in this process has been examined. Experimental tests have been performed under electrical load and in the presence of real-time uncertainties such as temperature changes, thermal conductivity, electrical noise, heat losses and random variations in thermal interaction. The second main contribution of the study is to predict and evaluate different power generation performances in the Thermoelectric Generator (TEG) system, which is affected by the mentioned experimental limitations based on artificial intelligence models. In addition, the optimal prediction performance of artificial intelligence models using hyper-parameter optimization has also been compared. Experimental results have shown that intelligent models are a successful tool for energy prediction in geothermal power generation. As a result of the hyper-parameter optimization, each model is improved by ∼5 % to ∼30 % in various error metrics. However, it was observed that the decision tree-based learning techniques resulted in ∼46 % to ∼96 % less inaccurate predictions on various error metrics compared to other methods. Power production predictions obtained using intelligent models have been obtained with 98.7 % accuracy with LightGBM learning algorithms. Most of the models used to classify different hot and cold water levels exhibit very high classification performance. The results show the successful integration of artificial intelligence models in effectively utilizing waste thermal energy and adopting a sustainable approach to energy production.