Evaluation and Optimization of Heat Extraction Strategies Based on Deep Neural Network in the Enhanced Geothermal System

Abstract

Production strategies and parameters control the efficiency of geothermal energy extraction related to the thermal stability and economic benefits of a geothermal system. The optimization strategies of geothermal energy extraction play a critical role in engineering and are generally determined through a numerical simulation approach. Considering the correlation among production parameters, numerical simulation requires numerous runs and manual adjustments, resulting in lower calculation efficiency and limited or local optimizations. This study proposes a high-efficiency network based on a three-dimensional heterogeneity model in the Gonghe Basin in China to achieve a high-efficiency and high-precision production strategy. The neural network was successfully established as a surrogate of the numerical model for the repetitive forward simulation. Meanwhile, the neural network is integrated with the Harris Hawks algorithm to optimize extraction strategies for sustainable heat extraction. This paper focuses on the effects of human-controlled operational parameters on geothermal systems. Results indicated that the maximum electrical power can be guaranteed 5.2 MW during a 50-year production period at an injection temperature of 60°C, an injection rate of 39 kg/s, and a well spacing of 380 m. The study provides important operational guidance for sustainable utilization in the Gonghe Basin. This simulation-optimization approach can be applied to other geothermal sites for sustainable energy production.