Mitigating Temperature Effects in Rogowski Coils’ Accuracy Through Physics-Informed Model

Low-Power Instrument Transformers (LPITs) are increasingly being deployed across power networks, either to replace or operate alongside traditional inductive instrument transformers. While LPITs offer several advantages, they are also known for their sensitivity to various influencing factors. This research focuses on one specific type of LPIT-the Rogowski coil-and examines the influence of temperature on its performance, considering both the ratio error and phase displacement components. Leveraging standardized testing procedures as defined by IEC standards, the study captures the coil's behavior at three controlled temperature points. These empirical results are then used to train a physics-informed deep learning (DL) model capable of generalizing the coil's accuracy indices across a continuous range of temperatures. The proposed approach enables the estimation and compensation of temperature-induced errors without requiring real-Time thermal calibration or additional sensing hardware, thereby reducing complexity and cost. By embedding physical knowledge into the data-driven model, the methodology ensures that predictions remain robust and interpretable, even outside the originally tested temperature points. The results demonstrate that the integration of domain knowledge with modern DL tools can significantly mitigate one of the most critical error sources in Rogowski coils, enhancing their reliability for grid monitoring, protection, and control applications. This work paves the way for more resilient LPIT deployments in smart grid environments.