Machine learning-based optimal design of a phase change material integrated renewable system with on-site PV, radiative cooling and hybrid ventilations—study of modelling and application in five climatic regions

The widespread application of advanced renewable systems with optimal design can promote the cleaner production, reduce the carbon dioxide emission and realise the renewable and sustainable development. In this study, a phase change material integrated hybrid system was demonstrated, involving with advanced energy conversions and multi-diversified energy forms, including solar-to-electricity conversion, active water-based and air-based cooling, and distributed storages. A generic optimization methodology was developed by integrating supervised machine learning and heuristic optimization algorithms. Multivariable optimizations were systematically conducted for widespread application purpose in five climatic regions in China. Results showed that, the energy performance is highly dependent on mass flow rate and inlet cooling water temperature with contribution ratios at around 90% and 7%. Furthermore, compared to Taguchi standard orthogonal array, the machine-learning based optimization can improve the annual equivalent overall output energy from 86934.36 to 90597.32 kWh (by 4.2%) in ShangHai, from 86335.35 to 92719.07 (by 7.4%) in KunMing, from 87445.1 to 91218.3 (by 4.3%) in GuangZhou, from 87278.24 to 88212.83 (by 1.1%) in HongKong, and from 87611.95 to 92376.46 (by 5.4%) in HaiKou. This study presents optimal design and operation of a renewable system in different climatic regions, which are important to realise renewable and sustainable buildings.