Artificial Neural Network–Particle Swarm Optimization of a Photovoltaic Multi-Channel Latent Heat Thermal Energy Storage System: Toward a Fast-Charging Mode

Document Type : Research Paper

Authors
1 Department of Computer Engineering and Information, College of Engineering in Wadi Alddawasir, Prince Sattam Bin Abdulaziz University, Al-Kharj, Saudi Arabia
2 College of Applied Medical Sciences in Wadi Alddawasir, Prince Sattam bin Abdulaziz University, Saudi Arabia
3 Department of Mechanical Engineering, GLA University, Mathura, UP, India
4 Department of Mathematical Sciences, Saveetha School of Engineering, SIMATS, Chennai, India
5 Physics, Faculty of Science, Carthage University, Tunisia
10.22055/jacm.2026.49772.6233
Abstract
The use of latent heat thermal energy storage (LHTES) systems integrated with photovoltaic (PV) systems has attracted increasing attention; however, achieving optimal performance remains challenging. This study introduces a photovoltaic multi-channel LHTES system filled with pure paraffin wax and evaluates its charging and discharging processes. Porous copper heating coils supply heat during charging, whereas water flows through adjacent channels during discharging. The reference model was simulated using the governing partial differential equations (PDEs) and the enthalpy–porosity method. Following a mesh-independence analysis and validation against published data, the effects of total photovoltaic power (PPV), channel length (l), wall thickness (t), and channel acute angle (θ) were investigated. Relative to the reference design, the selected geometry and higher photovoltaic input reduced the charging time by approximately 34% while satisfying the maximum PCM-temperature constraint (Tmax,PCM ≤ 100 °C). A dataset of 1,768 time-dependent records was then used to train a multilayer perceptron artificial neural network (MLP–ANN) for performance prediction. The trained network was coupled with particle swarm optimization (PSO) to identify a candidate optimum within the investigated design space. The simulation screening and ANN–PSO search were used to examine the charging-time and temperature tradeoff. The optimum design, with PPV = 1.04 kW, l = 50 mm, t = 3.6 mm, and θ = 60°, reached a melt volume fraction (MVF) of 0.95 in 41 min; the corresponding simulation maximum temperatures were 95.2 °C in the heating-coil channels and 86.5 °C in the clear channels.
Keywords
Subjects


Articles in Press, Accepted Manuscript
Available Online from 30 October 2026