Mathematical Model of Utilization Mapping for Geothermal Energy Using Machine Learning Algorithms

Abstract

Geothermal energy has an important function and cannot be neglected in human life. Moreover, nowadays almost human activities depend on energy. Many countries are engaged in initiative that lead to alternative sources of energy. Indonesia is successfully included in the top two countries in the world that produce geothermal energy. Geothermal energy has huge potential in Indonesia. Nonetheless, the utilization capacity of geothermal energy is lesser compared to its natural potential. Our study, therefore, was motivated to explore the potential of geothermal energy utilization and laying design strategies in developing a single flash geothermal power plant. Machine learning approach in geothermal energy applications currently is not popular means of predicting performance in the geothermal industry, yet machine learning has exhibited potential in many engineering and science problems. As of recent, there is no efficient means of predicting if the potential site identified can produce the expected amount of energy without boring wells in that site. In a situation where the wells are found to be insufficient, another well is bored creating potential environmental destruction. Moreover, the cost of boring is high covering 45% of the whole geothermal project. The most important of geothermal energy sources utilization parameter is the temperature of the brine. Geothermal for power production use the thermodynamic properties of the brine to determine how much power can be produced from the geothermal site. Using the enthalpy and entropy potential, the production capacity in a single flash can be estimated. In this study, we perform a literature survey to enable understanding of the geology, reservoir characteristic of selected potential sites and collect suitable data for the supervised machine learning technique. Further, a utilization strategy to classify all the possible sites of geothermal potential resources in Indonesia was done. In addition, machine learning was used to predict power requirements on the grid, depth and lifespan of the wells. It was observed that the maximum power output can be achieved on the grid if the well temperature is between 120 C and 150 C. With respect of the organization of the thesis, Chapter 1 consist of introduction that gives the background of the study. Chapter 2 consist of review of geothermal energy in Indonesia as current status and prospects. Chapter 3 consist of the utilization mapping for geothermal energy with respect to temperature. Chapter 4 consist of machine learning model for improving single flash geothermal energy production. Chapter 5 consist of the experimental setting, results, and discussion. Chapter 6 consist of the conclusion of this thesis by summarizing all results and future observation through the researches. xii