Electrical Output Stabilization of Nanofluid Spectral Splitting Photovoltaic/Thermal System by Regulating Parameters Based on Machine Learning Algorithms

Abstract

The nanofluid spectral splitting photovoltaic/thermal (NSS-PV/T) system offers flexible regulation compared to traditional PV/T systems. However, there is currently a lack of research on regulating the NSS-PV/T system to achieve stable electrical output. Therefore, this paper provided two methods of parameter stepping regulation and parameter optimization that were applicable in both single parameter and multi-parameter application conditions. Firstly, the energy equilibrium model of the NSS-PV/T system was used to simulate actual operating conditions and serving as training data to train the subsequent prediction models. Two machine learning algorithms, including support vector machine and particle swarm optimization-back propagation neural network, were employed to design parameter stepping regulation models. Subsequently, parameter optimization was performed using particle swarm optimization and pattern search algorithms. The results illustrate that the relative errors of single parameter optimization do not exceed 0.070%, 0.099%, and 0.120%, respectively. The relative errors of multi-parameter optimization are not more than 0.369%. The parameter stepping regulation models still face challenges in operating conditions that deviate significantly from the electrical output standard. The method of parameter optimization can achieve lower deviations, and the variations of the parameters are smaller using multi-parameter optimization, which can further reduce fluctuations and improve regulation speed.

Keywords: Nanofluid spectral splitting PV/T system, stabilized electrical output, stepping regulation, Parameter optimization, Machine learning