Realization of an electrical aircraft demands a low-weight electric distribution and propulsion system. The use of high voltage and power electronics operated at high switching frequencies is essential to achieve this objective. However, printed circuit boards (PCBs) in electrified aircraft are in a harsh working environment, which can make PCBs more susceptible to generate partial discharges (PDs). The current PD detection technology has poor immunity to the electromagnetic interference and acoustic interference in the operating environment of aircraft. Therefore, this article proposes an optical-based PD detection method for PCBs, which is effectively immune to electromagnetic and acoustic interference. This method uses fluorescent fiber as a PD optical signal sensor and then collects the optical signal by the avalanche photodiode (APD). Experiments have verified that the detection sensitivity, sensing range, and anti-interference performance of this method are well satisfied with the PD detection. In addition, single PD pulse, optical phase resolved PD (PRPD) patterns, and PD inception voltage (PDIV) under different air pressure and voltage conditions are investigated. Finally, the relationship between the optical signal and PD amplitude is found to be proportional, which proves that the severity of the PD on PCBs can be effectively detected and evaluated by this method.

Optical partial discharge (PD) detection is an efficient means of diagnosing the insulation status of power equipment. C4F7N/CO2 gas mixture is a very potential environmentally-friendly SF6 substitute gas, and its PD optical characteristics need to be studied to guide the PD diagnosis of novel C4F7N/CO2 equipment. Therefore, this article proposes a multispectral microarray detection technology, which can achieve high-sensitivity detection and PD diagnosis by simultaneously collecting the spectral characteristics of multiple bands. By setting up an experimental platform, the PD experiments of four typical defects in the C4F7N/CO2 gas mixture with five different proportions and pure SF6 are carried out. Based on the analysis of PD multispectral features, the correlation between different gases and the difference between different defects are obtained. Finally, by combining multispectral detection with a t-distributed stochastic neighbor embedding (T-SNE) feature extraction algorithm, a PD diagnosis method that can adapt to both C4F7N/CO2 gas mixture and SF6 is proposed, which provides a reference for the PD detection of novel C4F7N/CO2 equipment application.