Physics-informed neural networks for modelling power transformer’s dynamic thermal behaviour

This paper focuses on the thermal modelling of power transformers using physics-informed neural networks (PINNs). PINNs are trained to consider physical laws provided by general nonlinear partial differential equations (PDEs). The PDE considered for study transformer’s behaviour is heat diffusion equation with boundary conditions given ambient temperature at bottom and top-oil top. model one dimensional along transformer height. distribution estimated field measurements temperature, load factor. from a real provide more realistic solution, but also an additional challenge. Finite Volume Method (FVM) used calculate solution further benchmark predictions obtained PINNs. results estimating show high accuracy almost exactly mimic FVM solution.