TY - JOUR
T1 - An improved capacitance-resistance model for analysing hydrogen production with geothermal energy utilisation
AU - Liu, Zhengguang
AU - Shi, Minghui
AU - Mohammadi, Mohammad Hadi
AU - Luo, Haizhi
AU - Yang, Xiaohu
AU - Babaei, Masoud
PY - 2024/8/12
Y1 - 2024/8/12
N2 - The integration of geothermal and hydrogen energy through advanced predictive modeling not only enhances energy production efficiency but also contributes significantly to the global transition towards renewable energy sources. This study commences with the establishment of a geothermal hydrogen production model using Aspen Plus, incorporating physical constraints, derived from the Capacitance Resistance Model (CRM) model. Refinements to Long Short-Term Memory (LSTM) neural networks based on historical data and CRM constraints tailor them for geothermal reforming-based hydrogen production. Production analysis indicates that the hybrid CRM-LSTM model adeptly predicts heat and hydrogen production, improving system performance. The approach demonstrates superior accuracy, with hydrogen production predictions deviating by less than 2%. The comparison highlights the hybrid model’s advantage in handling nonlinear characteristics but it also shows the hybrid model requires longer training times. Sensitivity analysis reveals significant implications for investment decisions, with CRM predicting a 26-year cost recovery period under standard conditions, potentially underestimating actual outcomes by over eight months. Such discrepancies underscore the importance of accurate predictive models in guiding investment decisions for sustainable energy projects. This model contributes to achieving sustainable development goals by integrating geothermal and hydrogen energy, advancing the transition towards renewable and environmentally friendly energy sources.
AB - The integration of geothermal and hydrogen energy through advanced predictive modeling not only enhances energy production efficiency but also contributes significantly to the global transition towards renewable energy sources. This study commences with the establishment of a geothermal hydrogen production model using Aspen Plus, incorporating physical constraints, derived from the Capacitance Resistance Model (CRM) model. Refinements to Long Short-Term Memory (LSTM) neural networks based on historical data and CRM constraints tailor them for geothermal reforming-based hydrogen production. Production analysis indicates that the hybrid CRM-LSTM model adeptly predicts heat and hydrogen production, improving system performance. The approach demonstrates superior accuracy, with hydrogen production predictions deviating by less than 2%. The comparison highlights the hybrid model’s advantage in handling nonlinear characteristics but it also shows the hybrid model requires longer training times. Sensitivity analysis reveals significant implications for investment decisions, with CRM predicting a 26-year cost recovery period under standard conditions, potentially underestimating actual outcomes by over eight months. Such discrepancies underscore the importance of accurate predictive models in guiding investment decisions for sustainable energy projects. This model contributes to achieving sustainable development goals by integrating geothermal and hydrogen energy, advancing the transition towards renewable and environmentally friendly energy sources.
KW - Geothermal energy
KW - Hydrogen
KW - Capacitance Resistance Model
KW - Machine learning optimisation
KW - Long Short-Term Memory
M3 - Article
SN - 0360-3199
JO - International Journal of Hydrogen Energy
JF - International Journal of Hydrogen Energy
ER -