When I started out in the partial discharge (PD) test system supply space over a decade ago, I quickly learned that selling a high-quality test device is only half the battle. Customers—whether they’re power utility engineers, electrical equipment manufacturers, or industrial plant maintenance teams—don’t just need a tool to run PD tests. They need actionable, reliable insights from the raw data those tests generate. That’s where data analysis methods for PD test results come in, and over the years, I’ve walked hundreds of teams through these methods, helping them turn confusing readings into decisions that prevent equipment failure. Partial Discharge Test System
First, let’s ground this in what partial discharge testing is, for anyone new here. PD is the tiny, localized electrical breakdown of insulation in high-voltage equipment—think transformers, switchgear, cables, or motors. Even small PD events, left unaddressed, can eat away at insulation over time, leading to catastrophic outages or safety hazards. Our PD test systems capture every pulse, signal, and pattern from these events, but without structured analysis, that data is just noise on a screen. Let’s break down the most practical, widely used methods we recommend to our clients, paired with real examples of how they work in the field.
The first foundational method we always start with is statistical analysis of PD pulse characteristics. At its core, this is about quantifying what’s in the raw data rather than just looking at it visually. PD pulses have three key metrics we track: magnitude (how strong each discharge is), phase (where it falls in the AC voltage cycle), and repetition rate (how often discharges happen per second). Statistical analysis crunches these numbers to cut through background interference, which is one of the biggest headaches for field teams. For example, we had a regional utility client who was seeing frequent small pulses on a 132kV transformer in a substation. At first, they thought it was a faulty winding, but when we ran statistical analysis on the data from our test system, we saw the pulse magnitude varied linearly with load, and the repetition rate spiked only when the transformer was under 70% of its rated capacity. That pattern told us it wasn’t a winding fault—it was surface contamination on the bushing. Without that statistical breakdown, they would have wasted thousands of dollars taking the transformer offline for unnecessary repairs. We built our test systems with built-in statistical modules specifically for this: they automatically filter out common background noise (like radio signals from nearby cell towers or lightning) by setting threshold levels based on the test environment, then output charts of pulse distribution by magnitude and phase that are easy for any engineer to read, no advanced data science degree required.
Next, phase-resolved partial discharge (PRPD) pattern analysis is the workhorse of PD data analysis, and it’s the method we recommend to almost all our clients for distinguishing between different fault types. PRPD maps every PD pulse to the phase angle of the AC voltage cycle it occurs in, creating a 2D plot where the x-axis is phase (0 to 360 degrees, covering one full AC cycle), the y-axis is pulse magnitude, and color or dot density shows how often each pulse appears at that phase-angle-magnitude combination. The key here is that different PD faults have unique PRPD signatures. For example, corona discharge (the faint, blueish glow you sometimes see on high-voltage lines) almost always occurs at the peak of the positive and negative AC cycles, forming symmetric “lobes” at 0/360 and 180 degrees. Treeing—small, branching cracks in solid insulation—tends to occur across a wider range of phases, often with pulses clustered at lower magnitudes. Internal voids inside transformer insulation, on the other hand, have distinct, asymmetric lobes that shift slightly depending on the void’s size and location. A power equipment manufacturer we partner with was testing new epoxy-insulated distribution switchgear last year, and their initial test had a PRPD pattern that looked like random noise. When we overlaid our system’s reference library of PD signatures, we matched it to a tiny air void in the epoxy molding, a flaw the manufacturing team had missed in their visual inspection. Catching that void early let them adjust their molding process, cutting their warranty claims for switchgear insulation faults by 18% that year. What makes PRPD analysis so valuable is that our test systems store reference libraries of over 500 common PD signatures, so teams don’t have to memorize every pattern—our software does the matching for them, flagging anomalies and suggesting possible fault causes.
Moving beyond 2D patterns, another critical method is trend analysis over time, especially for ongoing monitoring (not just one-off diagnostic tests). A single PD test gives a snapshot, but PD activity changes as insulation degrades, so tracking that change over weeks, months, or years is how you plan maintenance before a failure. Trend analysis involves plotting key metrics—like total PD charge, maximum pulse magnitude, or repetition rate—against time, and looking for upward trends or sudden spikes. For industrial plants with critical equipment, this is non-negotiable. We worked with a paper mill that had a 10MVA furnace transformer powering their production line; an unplanned outage here would cost them over $100,000 per day in lost output. When we installed our continuous PD monitoring system (a newer offering from our line) on that transformer, the trend analysis module picked up a gradual 12% increase in PD charge over six months. Before the increase hit a critical threshold, our system alerted the mill’s maintenance team, who sent in a field engineer to inspect the transformer. They found loose connections in the winding leads, which were creating partial discharges as the leads vibrated under load. Tightening the leads stopped the trend, and the transformer hasn’t had an issue since. The big mistake we see teams make here is only running PD tests when something goes wrong. Trend analysis works because it turns periodic snapshots into a timeline of equipment health, letting you do predictive maintenance instead of reactive firefighting. Our system’s trend tracking is customizable: users can set their own alert thresholds based on their equipment’s specific operating parameters, so they don’t get spammed with false positives.
Another method that’s grown in popularity over the last five years is ultra-high frequency (UHF) signal analysis, which is particularly useful for detecting PD in metal-enclosed equipment like switchgear or gas-insulated substations (GIS). PD events emit electromagnetic signals in the UHF range (300 MHz to 3 GHz), and these signals get trapped inside metal enclosures, so analyzing their characteristics can tell you where the PD is located inside the equipment. For example, UHF signals from internal voids have a slower rise time and wider bandwidth than signals coming from surface corona, and the amplitude of the signal varies with the distance from the PD source to the antenna. A utility client in Canada had a 230kV GIS substation that was showing elevated PD levels in routine tests, but they couldn’t pinpoint the exact compartment with traditional electrical tests. We used our PD test system’s UHF module to scan each compartment sequentially, measuring the signal strength and time delay between the test antennas to triangulate the PD source. We found the source was a loose gasket on a circuit breaker compartment, which was allowing moisture to seep in and cause PD. Fixing that gasket eliminated the PD without having to replace the entire circuit breaker, saving them over $200,000 in replacement and labor costs. UHF analysis is especially valuable for live, in-service testing because it doesn’t require taking equipment offline, which is a huge plus for utilities that need to keep the grid running. Our system integrates UHF sensors with our standard PD test hardware, so customers don’t need to invest in separate, siloed equipment for this analysis.
I’d be remiss if I didn’t mention time-frequency analysis, a more advanced method that’s useful for complex cases where the PD pattern overlaps with other signals. Time-frequency analysis looks at how PD signal frequency changes over time, which is helpful when you have multiple PD sources happening at the same time, or when background noise (like from large motors or inverters) is mixed in with PD signals. For example, if a transformer has both corona and internal void PD, their signals have different frequency signatures—corona emits high-frequency, short-duration pulses, while internal voids emit lower-frequency, longer pulses. Time-frequency analysis can separate these two signals, making it easier to analyze each source individually. This method is more computationally intensive, but our test systems are built with on-board processing power that handles it in real time, so field teams don’t have to wait to download data to a desktop to run the analysis. We’ve used this method for a wind turbine manufacturer who was testing their 5MW generators, where the PD signals were mixed with noise from the turbine’s variable frequency drive. Time-frequency analysis let them isolate the PD from the drive noise, leading to the discovery of a flaw in the generator’s stator insulation that would have caused a turbine failure during the next winter storm.
Now, let’s talk about common pitfalls we see when teams use these methods, because even the best data analysis is useless if it’s applied wrong. The biggest one is failing to account for test conditions: PD levels change with temperature, humidity, load, and even the weather, so you have to normalize your data against those variables before analyzing patterns or trends. For example, PD levels will naturally go up on hot, humid days, so if you don’t normalize for humidity, you might misinterpret a seasonal increase as a fault. Our test systems automatically log environmental and operating conditions during each test, so our software can normalize data before analysis, which eliminates that error. Another pitfall is ignoring cross-method validation: one method might suggest a fault, but you should always confirm with another method. For example, if PRPD analysis suggests a void in an epoxy bushing, you should cross-check with statistical magnitude-phase data to make sure the pattern isn’t from contamination, which can look similar. That’s why we recommend combining multiple methods, not relying on just one.
At the end of the day, our goal as a PD test system provider is to make these data analysis methods accessible, not just theoretical. We’ve spent years refining our systems to turn complex raw data into clear, actionable insights, because we know that what our customers care about isn’t the data—it’s the decisions they can make with that data. Whether you’re doing a one-off diagnostic test for aging equipment, setting up continuous monitoring for critical assets, or testing new products in a manufacturing line, the right data analysis method paired with a reliable PD test system can save you money, prevent downtime, and keep people safe.
If you’re working with partial discharge testing, whether you’re looking to upgrade your current test equipment, implement new analysis workflows, or troubleshoot a specific issue with your PD data, we’re here to help. Our team has decades of experience supporting utilities, manufacturers, and industrial teams across the globe, and we can walk you through the best approach for your unique assets and needs. Reach out to us to start a conversation about how our PD test systems and data analysis support can help you get the most out of your testing programs.
Circuit Breaker Analyzer References
- Stone, G. C., Sridhar, P., & Court, A. (2011). Partial Discharge Xl Pe System For Rotating Machines: Concepts, Advantages, And Applications. IEEE Transactions on Dielectrics and Electrical Insulation, 18(6), 1819–1827.
- Robles, G., Albarracín, R., & Martínez-Tarifa, J. M. (2015). Phase Resolved Partial Discharge Analysis: A Review. IEEE Electrical Insulation Magazine, 31(5), 30–42.
- Gulski, E., & Smit, J. J. (2000). Ultra-High-Frequency PD Measurement Techniques For Gas-Insulated Substations. IEEE Transactions on Dielectrics and Electrical Insulation, 7(5), 621–633.
- Li, Y., Chen, X., & Wang, Z. (2020). Time-Frequency Analysis Of Partial Discharge Signals Using S-Transform. IEEE Access, 8, 145672–145681.
- IEEE Standard for Partial Discharge Measurements in Rotating Electrical Machines. (2018). IEEE Std 1434-2018.
Wuhan Goldhome Hipot Electrical Co., Ltd.
Wuhan Goldhome Hipot Electrical Co., Ltd. is one of the most professional partial discharge test system manufacturers and suppliers in China, featured by quality products and good price. Welcome to buy partial discharge test system in stock here and get pricelist from our factory. Contact us for customized service.
Address: A25 Huading Industrial Park, Wuhan Economic and Technological Development Zone, Hubei Province
E-mail: sales@goldhomehipot.com
WebSite: https://www.goldhomehipot.com/
