A Non-Contact Transient Voltage Inference Method Based on Multi-Fidelity Data Fusion

The development of power systems requires advanced technologies for transient voltage measurement. In non-contact measurement, accurately determining the coupling parameters between the sensor and the measured conductor is critical. However, these parameters can be influenced by factors such as relative position and the electromagnetic environment, which complicates efforts to maintain measurement stability. To this end, this paper proposes a transient voltage inference method based on multi-fidelity data fusion. In the frequency domain, the proposed method effectively integrates the stable low-frequency data from voltage transformers with the high-frequency trend information from non-contact sensors, allowing for accurate wideband measurements. Notably, this method adaptively determines coupling parameters in a data-driven manner, eliminating the need for the cumbersome manual calculations and ongoing calibrations required by existing methods. An experimental platform was established to conduct lightning pulse tests across multiple measurement points. The results indicate that this method can adaptively adjust coupling parameters according to different measurement points, achieving high measurement accuracy with an average relative error of 2.37% in amplitude and 3.00% in full width at half maximum.