Partial Discharge (PD) Monitoring is the ultra-sensitive early warning system for high-voltage transformers. A partial discharge is a microscopic, localized spark inside the transformer's insulation that does not immediately bridge the gap between two conductors. Over months, this tiny spark slowly burns a conductive 'tree' through the insulation until it finally connects, triggering a massive, explosive short circuit. PD monitoring listens for these microscopic sparks so engineers can shut down the transformer weeks before it explodes.
At 400kV, even a microscopic void or air bubble inside the solid epoxy or paper insulation is subjected to immense electrical stress. Because the gas bubble has a lower dielectric strength than the surrounding oil/paper, the gas breaks down and sparks internally. This releases a tiny burst of electromagnetic energy (UHF waves), acoustic sound waves, and chemical breakdown gases. Advanced PD monitoring uses Ultra-High Frequency (UHF) antennas mounted inside the transformer tank to capture these nanosecond electromagnetic pulses, or piezoelectric acoustic sensors mounted on the outside of the steel tank to literally 'listen' to the sparking.
Sensor Deployment
UHF antennas are permanently installed into the drain valves of the transformer, converting the entire steel tank into a massive faraday cage that traps and amplifies the PD signals.
Phase-Resolved Pattern Analysis (PRPD)
The software plots millions of tiny sparks against the 60Hz AC sine wave. The exact shape of the pattern tells the engineer if the spark is a harmless floating particle or a alarming 'corona' eating through the primary winding.
Time-of-Flight Localization
By comparing the microsecond arrival time of the acoustic ping at 4 different sensors on the tank wall, software can triangulate the exact 3D location of the spark deep inside the core.
- Noise Filtering: A 400kV substation is incredibly noisy. Corona discharge from nearby overhead wires can easily trick the sensors into triggering a false alarm, requiring advanced AI algorithms to filter out the 'noise'.