After 10 years of successful uninterrupted operation, the UPS systems in our data center failed. Although this incident stranded us without power for only minutes, the recovery process took months. In the post-mortem search for a smoking gun, we removed buckets of failed components, including diodes, capacitors, printed circuit boards, SCRs, fuses, and even control logic power supplies. While there was plenty of smoke, we never found the gun. In fact, our opinion changed from dismay at a failure to amazement that the box was able to operate and produce power at all. We came to understand that simply because a UPS is up and running, it doesn't mean it's fully operational.
Consequently, we revamped our testing procedures to begin functionally testing our UPS equipment along with its associated batteries. We decided on a course of action primarily based on standard factory acceptance testing — broken down into three phases: physical preventive maintenance (PM), protection settings and calibration, and functional load testing (including battery discharge).
Let's take a look at each of these phases in more detail.
This includes infrared scanning prior to shutting down the unit. A poor power connection will generate a higher temperature condition than the surrounding temperature. This condition shows up on an infrared scan. By purchasing a relatively inexpensive camera, you can perform an independent scan whenever the box is open or specific inspection is warranted.
You must check and tighten all connections. A minor ground may induce transients on the logic power supply that will affect the system control. Using the infrared scanner, check for loose connections on power supplies, diodes, capacitors, inverter gates, and drive boards (see the Photo, on page 59). Inspect capacitor cans for swelling, leaking, or a pressure pop out.
Protection settings and calibration
Check protection settings and calibrate the UPS unit. By simulating an alarm condition, the technician can verify that all alarms and protective settings are operational. If the UPS system metering is out of calibration, it may affect its operation. For example, a meter that monitors DC bus voltage affects the DC overvoltage or undervoltage alarm and protection. Should the UPS transfer to battery operation, the ensuing battery discharge might continue too long. As the battery voltage drops, the current must increase to continue supporting the same power. When current becomes too high, the temperature in the conduction path increases to the point of destruction.
Functional load testing
You can break functional load testing down into six distinct operations, including steady-state load test, harmonic analysis, filter integrity, transient response test, module fault test, and battery rundown. For systems with multiple modules, you should also perform a module fault test and restoration.
Steady-State Load Test
Under a steady-state test, you should check all input and output conditions at 0%, 50%, and 100% load. Make sure to test input voltage, output voltage, input current, output current, output frequency, input current balance, and output voltage regulation. The analysis will reveal if input currents match across all phases of a module as well as determine if all modules equally share the load. Because power is equal to voltage times current, a degraded voltage from a single module or phase means other modules or phases must produce more current to accommodate the voltage drop.
You should monitor the input and output of the UPS for harmonic content during the steady-state load test. Observing the harmonic content at 0%, 50%, and 100% load allows you to determine the effectiveness of the input and output filters. Most manufacturers design UPS filtering systems for the greatest efficiency at full load. Consequently, the harmonic distortion is greatest when the system is least loaded and smooths out as load increases. As long as the distortion is consistent across phases and modules, there is no reason for alarm.
Total harmonic distortion (THD) is essentially a measure of deviation from a perfect sinusoidal wave. A load with high THD requires more energy to sustain than a unit with low THD. The wasted energy dissipates as heat. Although inefficient, a UPS operating with high THD at low load is in no danger of damaging conduction components. A UPS operating with high THD at high load is extremely taxed because the unit produces energy to sustain the load along with additional heating. You should test harmonics at the following points: input voltage, output voltage, input current, and output current.
Most UPS systems have three basic filter elements: input, rectifier, and output. The filters commonly use an arrangement of resistors, inductors, and capacitors to remove unwanted waveform components. Thermal scans typically pick up indications of an inductor failure. Capacitors are more prone to failure under stress. Depending on the capacitor, they're subject to rapid expansion and leakage of acidic electrolyte when overstressed or drying out over time.
In large three-phase UPS systems, virtually hundreds of capacitors are wired in series and arranged in banks. Identifying a single failed capacitor can be difficult and time-consuming. Performing a relative phase current balance is a simple means of checking filter integrity.
By checking phase current through each leg of the filter, you can make a quantitative evaluation. A marked difference in phase currents drawn through a filter assembly indicates that one or more capacitors have degraded.
Transient Response Load Test
The transient response test simulates the performance of a UPS with large instantaneous swings in load. The UPS accommodates full-load swings without a distortion in output voltage or frequency. Periodically testing the response to load swings and comparing them with the original specifications will help keep your UPS healthier.
To observe the response to rapid load swings, connect a recording oscillograph and load banks to the UPS output to monitor the three phases of voltage along with a single phase of current (the current trace acts as a telltale sign when you apply or remove load). As you apply and remove load, you can see the effect on the voltage waveform (see Figure, at right). The voltage waveform should not sag, swell, or deform more than 8% under any given transient. You can perform similar tests on loss of AC input power and then measure the AC input restoration. Conduct load testing using the following power increments:
- 0% — 50% — 0%;
- 25% — 75% — 25%; and
- 50% — 100% — 50%.
Module Fault Test
You should also functionally test multimodule systems to verify that the system continues to maintain the critical load in the event of a module failure. By applying full system-rated load (via load banks), a single module should be simulated to fail. The system should continue to maintain the load without significant deviation of voltage or frequency as verified by a recording oscillograph. Test each module similarly. Failing multiple modules will test the system transfer to bypass.
Your final test should be a battery rundown. Often a battery system may be subject to failures that go undetected. The most meaningful test of a battery is to observe temperature, voltage, and current under load conditions. Dissimilar voltages from cell-to-cell or string-to-string are a clear indication of battery degradation. The load test will depend on the type of batteries connected to the UPS. Wet-cell batteries are more robust and are capable of going a year between full battery-rundown tests. Valve-regulated batteries are easily overstressed by a full battery rundown, so you should test them more often.
While conducting a discharge test, you should continuously check the battery cell voltages manually or with a cell monitor. When the battery string is no longer able to provide designed discharge time (e.g., 15 min), or 80% of the rated capacity, then it's time to consider full battery replacement.
Effective UPS testing is a complex and potentially dangerous operation if you don't do it correctly. Hiring a test-engineering group with strong experience in UPS maintenance and testing is usually best. Although thorough testing of your UPS unit can be costly and time-consuming, the costs associated with an unanticipated loss of business and production far exceeds those incurred from testing.
John E. Diamond is director of engineering and property manager for Cushman & Wakefield, New York.