Optimization of fin arrangement in solar thermal storage devices and convolutional neural network modeling

Abstract

Solar energy, a renewable and eco-friendly source, faces challenges in aligning energy production with storage durations. Phase change materials (PCMs) effectively tackle this issue, yet their poor thermal conductivity limits solar thermal storage device performance. Conventional solutions involve adding fins, but limited volume compromises overall heat storage capacity. Our study enhances device performance by rearranging fins while maintaining the same area. The H/R = 0.5 fin arrangement with equal spacing significantly improves heat storage, reducing total melting time by 54.31% with a higher entropy value. Further, arranging five fins, predominantly in the lower part, shortens PCM melting time without compromising upper material melting, enhancing device temperature uniformity. Utilizing computed data, we trained a Convolutional Neural Network (CNN) model, maintaining a maximum error within 5%, with an actual maximum error of only 1.56%. The efficacy of CNN in addressing prediction challenges is noteworthy, showcasing improved solar thermal storage device performance.