Hydrogen production via renewable-based energy system: Thermoeconomic assessment and Long Short-Term Memory (LSTM) optimization approach

Abstract

A clean and carbon-free energy market can be more sustainable and safer that can be approaches by investing on hydrogen energy. Further, a renewable energy-driven energy production cycle coupled with an energy storage option is of major significance. In the present article, a hybrid process based on geothermal and solar energies is conceptually designed and simulated. The process is comprised of a solar-organic Rankine cycle (ORC) unit, a geothermal cycle, an electrolysis unit (based on alkaline electrolyzer stack, AES), and a nanomaterial-based unit (i.e., alkaline fuel cell stack, AFC) to store hydrogen. Combining nickel nanoparticles thermally enclosed in multiwalled carbon nanotubes (MWCNTs) with a composite thermoset anion-exchange membrane enables alkaline water splitting. When the grid demands are less than generated electricity rate, additional electricity is fed into an AES to generate hydrogen fuel. Further, when the grid demands are more than generated electricity rate, stored hydrogen and oxygen are fed into an AFC to back convert the chemical energy in the fuel into electrical energy. The thermodynamic and thermo-economic evaluations of the considered hybrid process are comprehensively presented. Furthermore, multi-objective optimization of the introduced process has been developed in accordance with the genetic optimization technique to maximize the rate of electricity and minimize the cost rate. Additionally, the Long Short-Term Memory (LSTM) -based optimization technique is established and discussed. The outcomes indicated that under the LSTM-based genetic optimization technique the output electricity and electricity and hydrogen costs of the considered hybrid process can be improved by 34.55% and reduced by 40.3% and 26.8%, respectively. It was also found that the estimations obtained from the LSTM technique are very reliable, and can be relied on with acceptable accuracy. The estimation based on LSTM has an accuracy level of more than 98.1%.