With the rapid development of the Internet, the demand for high-power UPS and generators in data centers has increased dramatically, which has also created some new problems. This article provides a theoretical analysis and practical case studies on the impact of UPS input power factor and input filter on generators, aiming to clarify the causes of these problems and find solutions.
1. Matching issues between generator sets and UPS
UPS power supply system manufacturers and users have long been aware of the matching issues between generator sets and UPS systems, particularly the adverse effects of current harmonics generated by rectifiers on power supply systems such as generator voltage regulators and UPS synchronization circuits. Therefore, engineers designed input filters and applied them to UPS systems, successfully controlling current harmonics in UPS applications. These filters play a crucial role in the compatibility of UPS systems with generator sets.
In fact, all input filters use capacitors and inductors to absorb destructive current harmonics at the UPS input. Input filter design considers the inherent percentage of all harmonic distortion inherent in the UPS circuitry and the maximum possible percentage under full load. Another benefit of most filters is improving the input power factor of the UPS under load. However, another consequence of using input filters is a reduction in overall UPS efficiency. The vast majority of filters consume approximately 1% of the UPS power. The design of input filters is always a process of seeking a balance between the advantages and disadvantages.
To maximize UPS system efficiency, UPS engineers have recently made improvements to the power consumption of input filters. This increased filter efficiency largely depends on the application of IGBT (Insulated Gate Transistor) technology to the UPS design. The high efficiency of IGBT inverters led to a redesign of the UPS. The input filter can absorb some current harmonics while also absorbing a small portion of active power. In short, the ratio of inductive to capacitive factors in the filter is reduced, resulting in a smaller UPS size and improved efficiency. However, this has raised the issue of UPS compatibility with generators.
2. Power factor problem
Typically, attention is focused on the UPS's operation under full or near-full load conditions. Most engineers understand the UPS's operating characteristics under full load, especially the input filter's characteristics; however, few are interested in the filter's performance under no-load or near-no-load conditions. After all, the current harmonics of the UPS and its electrical system have minimal impact under light load conditions. However, the UPS's no-load operating parameters, especially the input power factor, are quite important for the UPS's compatibility with generators.
Newly designed input filters have shown good performance in reducing current harmonics and improving the power factor under full load. However, they result in an extremely low power factor with capacitive lead under no-load or very low load conditions, especially those filters designed to meet the 5% maximum current distortion requirement. Generally, when the load is below 25%, the input filters of most UPS systems cause a significant drop in the power factor. Nevertheless, the input power factor rarely falls below 30%, and some newer systems have even achieved no-load power factors below 2%, approaching the ideal capacitive load.
This situation does not affect UPS output and critical loads, nor do it affect the mains transformer and power distribution system. However, generators are different. Experienced generator engineers know that generators will not function properly when carrying large capacitive loads. When connected to loads with low power factors, typically below 15% to 20% capacitive, the generator may shut down due to system imbalance. Such a shutdown after a mains power outage, where the emergency generator system drives the UPS system load, will cause a catastrophic accident. The shutdown poses a danger to critical loads for the following two reasons: (1) The generator needs to be manually restarted, and this must be done before the UPS battery finishes discharging; (2) Before the shutdown, the generator may cause "overvoltage" in the system, which may damage communication equipment, fire alarm systems, monitoring networks, and even UPS modules.
Worse still, after an incident, it's difficult to determine responsibility, pinpoint the problem, and correct it. The manufacturer claims the UPS system tested perfectly and points out that similar issues didn't occur with other identical equipment. The generator manufacturer says it's a load problem and cannot be fixed by adjusting the generator. Meanwhile, the user's engineers explain the specifications, hoping for compatibility. To understand why the incident happened and how to prevent it (or find solutions in critical applications), it's essential to first understand the working relationship between the generator and the load.
2.1 Generator and Load
The generator relies on a voltage regulator to control its output voltage. The voltage regulator detects the three-phase output voltage and compares its average value with the required voltage value. The regulator draws energy from an auxiliary power source within the generator, typically a small generator coaxial with the main generator, which transmits DC power to the magnetic field excitation coils of the generator rotor. The rise or fall of the coil current controls the magnitude of the rotating magnetic field, or electromotive force (EMF), in the generator stator coils. The magnetic flux of the stator coils determines the generator's output voltage.
The internal resistance of the generator stator windings is denoted by Z, including both inductive and resistive components; the electromotive force (EMF) of the generator, controlled by the rotor excitation windings, is represented by the AC voltage source E. Assuming a purely inductive load, in the vector diagram, the current I lags the voltage U by exactly 90° in phase. If the load is purely resistive, the vectors of U and I will coincide or be in phase. In reality, most loads fall between purely resistive and purely inductive. The voltage drop caused by the current flowing through the stator windings is represented by the voltage vector I×Z. It is actually the sum of two smaller voltage vectors: the resistive voltage drop in phase with I and the inductive voltage drop leading by 90°. In this example, it is exactly in phase with U. This is because the EMF must be equal to the sum of the voltage drop across the generator's internal resistance and the output voltage, i.e., vector E = the vector sum of U and I×Z. A voltage regulator can effectively control the voltage U by changing E.
Now consider what changes occur in the internal workings of the generator when a purely capacitive load replaces a purely inductive load. The current is now the opposite of that under an inductive load. The current I now leads the voltage vector U, and the voltage drop vector I×Z is also in opposite phase. Therefore, the vector sum of U and I×Z is less than U.
Because the same electromotive force E under inductive loads produces a higher generator output voltage U under capacitive loads, the voltage regulator must significantly reduce the rotating magnetic field. In practice, the voltage regulator may not have sufficient range to fully regulate the output voltage. All generator rotors are continuously energized in one direction, containing a permanent magnetic field. Even with the voltage regulator fully off, the rotor still has enough magnetic field to charge the capacitive load and generate voltage; this phenomenon is called "self-excitation." Self-excitation results in overvoltage or voltage regulator shutdown, which the generator monitoring system interprets as a voltage regulator failure (i.e., "loss of excitation"). Either of these situations will cause the generator to shut down. The load connected to the generator output may be independent or in parallel, depending on the timing and settings of the automatic switching cabinet. In some applications, the UPS system is the first load connected to the generator during a power outage. In other cases, the UPS and mechanical loads are connected simultaneously. Mechanical loads typically have starting contactors, which require a certain amount of time to reclose after a power outage. Inductive motor loads that compensate for the UPS input filter capacitors also have a delay. The UPS itself has a period called a "soft-start" cycle, which switches the load from the battery to the generator, improving its input power factor. However, the UPS input filters do not participate in the soft-start process. They are connected to the input of the UPS and are part of the UPS. Therefore, in some cases, the main load first connected to the generator output during a power outage is the UPS input filter, which is highly capacitive (sometimes purely capacitive).
The solution to this problem is clearly power factor correction. There are several methods to achieve this, roughly as follows:
● Install an automatic transfer switch to connect the motor load before the UPS. Some transfer switches may not be able to do this. Additionally, during maintenance, the factory engineer may need to individually commission the UPS and generator.
● To compensate for capacitive loads, a permanent reactive reactor is added, typically using a parallel wound reactor connected to the EG or generator output parallel board. This is easy to implement and relatively inexpensive. However, regardless of whether the load is high or low, the reactor always draws current and affects the load power factor. Furthermore, the number of reactors is always fixed, regardless of the number of UPS units.
●Installing inductive reactors in each UPS unit precisely compensates for the UPS's capacitive reactance. Under low load conditions, the reactor activation is controlled by contactors (optional). This method provides more precise reactor selection, but requires a larger number of reactors and incurs higher installation and control costs.
● Install a contactor before the filter capacitor to disconnect under low load. Because the contactor timing must be precise and control is complex, it can only be installed at the factory.
Which method is best depends on the site conditions and the performance of the equipment.
2.2 Resonance Problem
Capacitor self-oscillation problems can be exacerbated or masked by other electrical conditions, such as series resonance. Oscillations can occur when the ohmic values of the generator inductive reactance and the input filter capacitive reactance are close together, and the system resistance is low, potentially exceeding the power system's rated values. Newly designed UPS systems are essentially 100% capacitive input impedance. A 500kVA UPS may have 150kvar of capacitance and a power factor close to zero. Parallel inductors, series chokes, and input isolation transformers are standard components of UPS systems, and these components are inductive. In fact, together with the filter capacitance, they make the UPS behave capacitively overall, potentially introducing some oscillations within the UPS itself. Adding the capacitive characteristics of the power lines connected to the UPS significantly increases the complexity of the entire system, exceeding the analytical capabilities of most engineers.
Two additional factors have recently made these problems more prevalent in critical applications. First, driven by users' demands for high-reliability data processing, computer equipment manufacturers are increasingly incorporating redundant power inputs into their equipment. Typical computer racks now come with two or more power cords. Second, facility managers are requiring systems to support online maintenance; they want critical loads protected even when UPS systems are shut down for maintenance. These two factors have led to an increase in the number of UPS units installed in typical data centers, while reducing the load capacity per UPS. However, the increase in generators has not kept pace with the increase in UPS installations. In the eyes of facility managers, generators are typically backups, easy to schedule maintenance for. Furthermore, in some large projects, financial constraints limit the number of expensive, high-power generator sets. The result is that each generator powers more UPS units—a trend that pleases UPS manufacturers but troubles generator manufacturers.
The best defense against self-excitation and oscillation is based on fundamental physics principles. Engineers should carefully determine the power factor characteristics of the UPS system under all load conditions. After UPS equipment installation, the owner should insist on comprehensive testing, carefully measuring the operating parameters of the entire system during commissioning and acceptance. When problems are discovered, the best approach is to establish a project team composed of the manufacturer, engineers, contractors, and the owner to fully test the system and find solutions.
3 Typical Cases
The following is a case study of a UPS and generator compatibility issue that occurred during commissioning of a newly built data center by an online service provider. It illustrates how manufacturers, engineers, and users can identify and resolve such problems.
The site is equipped with three MGE UPS 3000kVA systems, each consisting of four 75kVA IGBT modulated pulse width modulation (PWM) and frequency modulation (FM) modules, expandable to six. The modules are designed for a 65% load rate and are equipped with input isolation transformers and a maximum 5% input current harmonic filter. All modules are connected to two parallel generator buses, each with three 1600kW generators, expandable to six. Each generator is equipped with an electronic voltage regulator. The power conversion plan for each parallel bus is to wait for the two generators to connect before the first batch of loads is connected. The first batch of loads includes one UPS unit from each system and some air conditioning loads.
As subsequent generators were added, the same load as the first batch was subsequently added. During fault mode testing, the operator observed that if one of the two generators carrying the first batch load failed, the other would trigger an overvoltage alarm and shut down after 2 seconds. However, the first batch load was far below the capacity of a single generator, as the UPS load was very light at this time. Further testing was then arranged to determine the impact of the UPS on a single generator. Since the initial suspicion was a problem with the UPS input circuitry affecting the voltage regulator, the UPS was tested without a load, or with the UPS inverter shut down. The testing setup included DC voltage and ammeters, directly monitoring the field excitation coils, as these parameters are controlled by the voltage regulator and would immediately reflect the regulator's action. Simultaneously, the generator's own meters were used to monitor the load's power (W), current (VA), and voltage (var).
First, a baseline was established using a purely resistive load. This showed that the excitation current and voltage increased as expected with increasing load. A larger load current generates a larger voltage drop I×Z across the generator's internal resistance Z, which must be overcome to maintain a stable output voltage U. Next, the effect of the UPS on the generator was tested, adding one UPS at a time. The UPS was left unloaded, and the soft-start process of the UPS rectifier was observed. The test results clearly showed that the voltage regulator's operation was the opposite of that under a purely resistive load. After connecting two UPSs, the voltage regulator was already close to the edge of its allowable range; adding another UPS caused the generator to enter an overload state after 2 seconds.
At this point, note the load value corresponding to a single 750kVA UPS. It causes the generator to shut down but there is no real load. The capacitive reactance of each UPS, which is close to 230kvar, makes the power factor 0.
The project team, comprised of engineers, the owner, contractors, suppliers, and manufacturers, selected a solution involving installing reactive reactors on each capacitive load after considering all possibilities. Based on previous test data, the manufacturer designed a 200kvar parallel reactor for each UPS, controlled by a contactor. The contractor installed it in parallel with the UPS's input filter on-site. The engineers designed an external control circuit that measured the generator load, allowing the reactor to be connected only when the UPS was powered by the generator and the generator's total load was below a (adjustable) setpoint. The project team retested the system with a modified UPS connected to a generator.
At this point, the effect of the capacitors still exists; the reactor can only balance part, not all, of the capacitors. Therefore, as the number of UPS units increases, the excitation current gradually decreases, but this does not cause a problem. This is because six UPS units exceed the capacity of one generator, while the voltage regulator remains normal and controls the output voltage.
