A Novel Modular Data Acquisition System for Distributed Monitoring of Partial Discharge in Industrial Drives

Insulation fault detection in rotating machinery using Partial Discharge (PD) has been a hot research topic over the last decades. In recent years, attention has also been paid to low voltage machines, following the advent of wide bandgap semiconductor devices. However, most of the solutions currently available are based on analog technology, which entails high costs, bulky wiring, and long installation times, making them unattractive for the market. This paper presents a low-cost, modular, and full-digital PD monitoring device, which implements a Capacitive Coupling (CC) front end for PD conditioning and an innovative, digital monitoring network, based on the Automotive Audio Bus (A2B), particularly suitable for deployment in large industrial plants, ensuring complete coverage and continuous monitoring of critical machinery without suffering of electromagnetic disturbances. The proposed system has been simulated in PSpice, then built and tested against a reference, consisting of a laboratory grade oscilloscope, using two different PD interfaces: CC and High Frequency Current transformer (HF-C). The comparison was carried out by subjecting a low voltage induction motor to a standard PD test to evaluate PD inception voltage detection with the different methods. The digital oscilloscope demonstrated a slightly better sensitivity concerning the onset PD voltage threshold, 1.3 - 1.35 kV. The proposed monitoring device on the other hand exhibits a greater robustness at transition voltages, being less prone to false positives. The two systems are equivalent at 1.4 kV while the proposed solution performs significantly better at higher voltages, with 6 PD events per cycle detected at 1.5 kV in place of 3. The HF-C transducer demonstrated even greater sensitivity at these voltage levels, with 7.6 PD detections per cycle at 1.5 kV. The results show that the proposed PD CC sensor has satisfactory sensitivity towards detecting onset of PD activity and appears less susceptible to noise than the digital oscilloscope CC sensor.