PIML-DCNN on Photovoltaic system cooling.

Dapeng Wang, Zhaojian Liang, Ziqi Zhang, Mengying Li

  • Propose a combination of Physics-Informed Machine Learning and Deep Convolution Neural Network (PIML-DCNN) to quantify cooling performance of PV systems under different configurations

  • Propose a new loss function - Pocket Loss to increase PIML-DCNN's interpretability by guiding model's outputs enter numerical zones with physcial interpretability through gradient and magnitude control.

  • The model is much faster than Computation Fluid Dynamics (CFD) and more accurate than empirical methods.

  • Supervised by Dr. Mengying Li..

  • Sponsored by PolyU Undergraduate Research & Innovation Scheme (URIS).

  • Paper accepted: "Efficient estimation of convective cooling of photovoltaic arrays: A physics-informed machine learning approach".Energy and AI (IF 9.8)

  • My Contribution:
    • Design & develop PIML-DCNN
    • Design & develop Pocket Loss
    • Papaer writing