Using deep learning artificial intelligence and multiobjective optimization in obtaining the optimum ratio of a fuel cell to electrolyzer power in a hydrogen storage system

Abstract

Summary Due to the depletion of fossil fuel sources and their pollution, renewable energy sources have attracted researchers' attention for energy supply; power generation using solar energy is one of the most important examples of these sources. This energy needs to be saved because of climate changes, to be used with high confidence. In this investigation, the heating and cooling of 100 Conex in Riyadh, Saudi Arabia, was studied, which used hydrogen storage as the central storage system. The main purpose of this project is to determine the ratio of fuel cell power to the electrolyzer power in this storage system, to reach the minimum dependency of the system on the urban power grid and at the same time, the minimum cost of electrolyzer and fuel cell. Because the shelters were simulated by Energy Plus and Open Studio Software and the energy supply system by TRNSYS Software, the optimization was not possible. Therefore, this system was changed into a network in MATLAB Software after simulation with deep learning artificial intelligence; then, it was optimized by a genetic algorithm. In order to accurately form a network, different parameters of artificial intelligence were studied precisely. The best result was obtained from a neural network with two hidden layers, each one containing 10 neurons, sigmoid activation function, and without dropout, with a correlation coefficient of 0.9998, which shows an excellent accuracy. The results of optimization also indicated that at all points, the Pareto front for fuel cell power is less than or equal to the power of the electrolyzer, and this ratio is 0.47 at the optimum point in terms of cost and electric consumption.

Formulation and Data-Driven Optimization for Maximizing the Photovoltaic Power with Tilt Angle Adjustment

“…The meteorological data were extracted from TRNSYS [29], a transient systems simulation software. TRNSYS has already been proven to be capable of dynamic simulation in the literature [30]. Data extracted from the commercial simulation are reliable and do not require further processing to detect outliers and missing data.…”