Repurposing of abandoned oil and gas wells as geothermal power plants: A comprehensive sensitivity analysis and AI based performance prediction

Geothermal energy is a renewable source of energy with immense untapped potentials. The high upfront cost, has made geothermal development projects, economically viable only for high temperature reservoirs. Repurposing of abandoned oil and gas wells as heat exchangers, relieves the drilling cost and presents a promising path for more affordable geothermal energy development. In this manuscript, the energy conservation in a double pipe heat exchanger, has been numerically solved using OpenGeosys, to predict the geothermal energy production from an oil well. The simulation results, under different parameter combinations have been used to rank the involved features, based on a novel perspective for two sets of scenarios with weak and efficient inner pipe insulation. At the end, different machine learning algorithms, have been trained to predict the output temperature of the well. The results showed that the features influencing the performance indicators, depend on the inner-pipe insulation and the output of interest. According to our analyses, in scenarios with weak insulation, the choice of the working fluid and the operational parameters are more important than the screening factors. Among the used ML algorithms, PCA-enhanced SVR demonstrates the best accuracy, with an R2 score of 0.96 and an RMSE of 1.88.