Artificial neural network based multivariable optimization of a hybrid system integrated with phase change materials, active cooling and hybrid ventilations

Abstract

Utilising diversified forms of energy in combination with advanced energy conversions and thermal energy storages is an effective way of developing high energy-efficient renewable systems for green buildings. In this study, a novel hybrid system for the energy cascade utilisation has been proposed, integrating the hybrid ventilations, the active photovoltaic cooling, the radiative cooling and the phase change materials’ storages. An enthalpy-based numerical modelling using the finite-difference method, which has been developed earlier, was used to characterize the sophisticated heat transfer process. A generic optimization methodology with competitive computational efficiency was applied by implementing the supervised machine learning and the advanced optimization algorithm. Multivariable optimizations for geometrical and operating parameters have been conducted and contrasted between the teaching-learning-based optimization and the particle swarm optimization. The results illustrate that the developed artificial neural network-based data-driven learning algorithm is more accurate and more computational-efficient than the traditional ‘lsqcurvefit’ fitting methodology for the characterization of the optimization function. In addition, the optimal case through the teaching-learning-based optimization is more robust than the optimal case through the particle swarm optimization in terms of the equivalent overall energy generation. This study presents a novel hybrid system for the energy cascade utilisation and a new generic optimization methodology, which are important for the promotion of green buildings with high efficiency of renewable energy utilisation.