Strongly reduced computation time of thermal modelling of primary components using a convolution-based approach

Accurate thermal models for current-carrying components in power systems are essential for increasing ampacity by exploiting cyclic load patterns, thereby mitigating grid congestion. Traditional models, such as those based on the finite element or differences methods offer high accuracy but are computationally intensive, while others are fast but conservative and therefore less ideal for dynamic ampacity calculations. This paper proposes a fast and generic calculation method for primary components that balances accuracy and speed by using heat decay responses to the load. By mapping physical quantities to a limited set ofeffectivemodel parameters, derived from either an accurate reference model or sufficient measurement data, the method enables rapid temperature calculations from load profiles using convolution of such profiles with an asset-specific kernel. This approach can compute a full cable sheath temperature profile from a year-long load profile within milliseconds, with minimal error compared to physics-based models. It is also applicable to other components such as transformers.