I. Early "Data Center" Power Supply Systems
Computers emerged in the 1950s, initially using vacuum tubes and exhibiting slow processing speeds. It wasn't until 1964, when my country produced the world's first semiconductor computer, that computers entered the semiconductor field. At that time, computing itself was a system—a very early, rudimentary system. Figure 1 shows the most advanced computing speed of the time: 100,000 operations per second.
The schematic diagram of the 108A computer shows that each part of the computer is very large. Its components, such as the arithmetic unit, memory, converter, and console, must have their own separate DC power supply system, and the source of these DC power supplies is AC mains power, as shown in the figure.
Its cooling system was a centralized air conditioning system; the concept of precision air conditioning did not exist at the time. The air conditioner's power supply also came from the mains electricity.
As mentioned above, early computers were powered by mains electricity, so the quality of the mains power supply determined the fate of the computer system. Because early computers had limited functionality, data could not be directly input using a keyboard or scanner as it is today.
Before inputting any data, all data must be converted into the basic Boolean algebra form of "0"s and "1s". Then, a "punch man" uses a punch machine to punch holes in an 8-digit paper tape; each hole represents a "1", and the unpunched area represents a "0". The calculator then inputs the punched paper tape into the computer via a photoelectric generator for calculation. If the mains power suddenly fails during the calculation, all data in the computer will be lost—a blank screen. The next calculation must then re-input the original data via the photoelectric generator, which causes significant problems for the calculation.
Therefore, people envisioned a scenario where a fault signal could be sent to the computer the instant the power was cut off, while the power supply continued for 5 seconds. This would allow the computer to store the current computational information in its memory, and once the mains power was restored, the computer could continue processing from where it left off. This led to the development of the first generation of Uninterruptible Power Supply (UPS), as shown in Figure 2, a 20kVA flywheel energy storage UPS. This is a 5-ton flywheel mounted on the same shaft as the generator. It's an online UPS; when the mains power is connected, the generator directly supplies power to the load. If the mains power fails, the 5-ton flywheel, due to inertia (energy storage), will continue to drive the generator to rotate for 5 seconds to complete the computer's data storage process. Therefore, UPS was born to protect data, and some call it the computer's twin brother.
Nevertheless, the efficiency is still too low, and the backup time is too short. This only solves the problem of preserving the current state; if the mains power outage is prolonged, the computer still cannot continue to operate and remains dependent on the mains power supply.
With the maturity of rectifier semiconductor devices, AC mains power was rectified to charge the battery pack, simultaneously driving a DC motor to rotate. This, in turn, caused a coaxial generator to rotate and generate electricity. If the mains power failed, the energy stored in the battery pack allowed the DC motor-AC generator system to continue supplying power, with the power supply time determined by the battery pack's energy storage, as shown in Figure 3. With the maturity of thyristor devices in the late 1960s, designers replaced the noisy and bulky DC motor-AC generator system with thyristor inverter circuits. However, since the circuits at that time only employed full-bridge inverter technology, outputting three live wires, while the load required a three-phase four-wire system (220V/380V in my country), a DY-type output transformer had to be added. Thus, the rotating generator-type UPS evolved into the static converter UPS, and the UPS system was divided into two types. Until 1980, when IPM (International Power Corporation, later acquired by Exide) upgraded the multi-step waveform circuit of the static converter UPS's SCR inverter to a pulse width modulation (PWM) circuit for transistor inverters, significantly reducing the size and weight of the machine. Until the advent of the all-IGBT high-frequency UPS in the mid-to-late 1990s, this development of UPS was entirely a process of energy saving. Throughout this period, the primary power supply of UPS systems has revolved around computer systems. Although computer systems have now evolved into data centers, UPS systems have remained indispensable. It can be said that UPS is an indispensable direct power supply system for modern data centers. With the expansion and increasing importance of data centers, UPS power supply systems must also adapt to this change.
II. Development and Classification of UPS Power Supply Systems for Data Centers
In today's digital age, UPS systems have undergone significant development in both performance and variety to adapt to these changes. For example, static converter UPS systems have entered the era of all IGBTs; the flywheel energy storage structure of rotating generator UPS systems has also undergone substantial changes, with the size and weight of the flywheel significantly reduced due to increased rotational speed; DC voltage UPS systems are also being explored, and so on. All of these developments revolve around the goal of energy conservation.
Figure 4 shows a general classification of UPS power supply systems. As can be seen from the figure, UPS systems today are divided into two main categories: rotating generator type and static converter type.
1. Rotary generators are structurally divided into vertical and horizontal types, as shown in two examples in Figure 5. Figure 5(a)
The device on the left is a magnetic levitation flywheel energy storage unit, which can provide 15 seconds of backup time under full load operation; the energy storage flywheel in Figure 5(b) also provides 15 seconds of backup time under full load operation. Its working process is shown in Figure 6. The flywheel energy storage UPS operates in online hot backup mode, that is, it operates in bypass mode when the mains power is normal. The difference between it and the bypass mode of the static converter UPS is that it has a certain adjustment and compensation capability for the bypass voltage. However, there are some differences in the working principle between vertical and horizontal UPS, as shown in Figure 7.
Operating Principle: When the mains power supply is normal, the UPS bypass input circuit breaker S2 is closed, the bypass static switch is open, and the maintenance bypass switch S3 is open; the input circuit breaker S1 is closed, and the main static switch and output circuit breaker S4 are closed. When the input mains voltage fluctuates or there is interference, the inverter circuit should compensate or adjust accordingly. When the mains power fails, the rectifier-inverter system converts the energy stored in the magnetic levitation energy storage flywheel system to the same voltage as the mains power to continue supplying power to the load. During this period, the fuel generator system starts, and after the voltage returns to normal, it takes over from the rectifier-inverter system to supply power to the load, while simultaneously driving the magnetic levitation energy storage flywheel system into energy storage mode. After the mains power is restored, the UPS returns to mains power mode. When the load is overloaded, short-circuited, or the UPS fails, the bypass static switch closes and conducts, switching the load to pure mains power supply mode. It can then be manually closed. The working principle of a horizontal flywheel energy storage UPS is basically the same, except that the energy storage component is not a magnetic levitation structure but a flywheel in an inductive coupler, and its backup time is also 15 seconds.
2. Static Converter UPS
As the name suggests, this type of UPS appeared in the 1960s. The advent of the silicon controlled rectifier (SCR) opened the door to static converter UPSs and greatly extended backup time. This laid a solid foundation for the rapid development of data centers. Currently, there are over a hundred brands of this type of UPS. Figure 8 shows the appearance of several commonly used static converter UPS models.
III. Differences between Linear Frequency UPS and High Frequency UPS
1. Differences in circuit structure
(1) Circuit structure of line frequency UPS
The circuit structure of this type of product generally involves an input rectifier operating at an industrial frequency of 50Hz. Low-power products are typically single-phase input and single-phase output (1/1), with the input rectifier usually being a diode rectifier. These products are generally below 10kVA and include an output transformer. Products with a three-phase input and single-phase output (1/3) structure are generally below 30kVA. These products were previously more common. In recent years, due to the rapid development of data centers of all sizes, three-phase input and three-phase output (3/3) UPS products have been widely used. Figure 9 shows the general principle circuit structure of a three-phase industrial frequency UPS. The circuit structure of an industrial frequency UPS is characterized by an input rectifier operating at an industrial frequency of 50Hz and a full-bridge inverter, therefore an output isolation transformer is necessary. Because regardless of whether it's a single-phase input and single-phase output, or a three-phase input and single-phase output, or a three-phase input and three-phase output UPS, since all output lines are live wires during full-bridge output conversion, grounding any one of these live wires will lead to serious consequences. Therefore, an output isolation transformer is necessary to create an isolation grounding point. Furthermore, because the input circuit of a line-frequency UPS operates in step-down mode, its adaptability to input voltage fluctuations is very weak. Generally, the input voltage fluctuation range is required to not exceed ±15%, so its output must also have a step-up transformer. Therefore, this output transformer is indispensable. It has no other function than this. The view that transformers provide interference immunity is a misunderstanding of the function of transformers.
Figure 9 Circuit structure of two types of UPS
Figure 9(b) shows the general circuit structure of a high-frequency UPS. It differs from a line-frequency UPS in that its rectifier and inverter are both IGBTs or other high-frequency devices, operating at industrial frequencies several tens of times higher than 50Hz. Because it uses half-bridge inverter technology, the output isolation transformer is eliminated. Furthermore, since its input rectifier operates in boost mode, it has a strong adaptability to input voltage variations; generally, input voltage variations exceeding ±30% are permissible. For example, a single-phase UPS may allow an input voltage range of 80~280V. Because the transformer is eliminated, the system efficiency becomes very high, typically reaching 95% at half load, which is unattainable for a line-frequency UPS.
2. Differences in input power factor
A typical line frequency UPS has an input power factor of about 0.65 during single-phase input and about 0.8 during three-phase 6-pulse rectification. Therefore, in applications with higher requirements, an active filter must be added or the rectification level increased to 12 pulses. Some manufacturers also offer passive filters that can achieve an input power factor of 0.9.
High-frequency UPS systems, whether single-phase or three-phase, can achieve an input power factor of over 0.98 without any external interference.
3. Requirements for load balance
Early line-frequency UPS systems had stringent requirements for load balance due to the mutual influence between the three-phase voltages. Figure 10(a) shows the flow path of the three-phase voltage and current in this type of UPS. As shown in the figure, the positive half-wave current of UAB flows through Q5-Q4 back to the negative terminal of the power supply; the negative half-wave flows through Q3-Q6 back to the negative terminal. The positive half-wave current of UAC flows through Q5-Q2 back to the negative terminal; the negative half-wave flows through Q1-Q6 back to the negative terminal. The positive half-wave current of UBC flows through Q1-Q4 back to the negative terminal; the negative half-wave flows through Q3-Q2 back to the negative terminal. It can be seen that each of the three bridge arms in the circuit is used twice, meaning it is used by two phases of the power supply. Therefore, each pair of power supplies influences the other; if the three-phase currents are different, it will lead to differences in the output voltage. Although some adjustment and control components were added later to largely solve the problem, the root cause of mutual influence still exists, increasing the complexity of the circuit.
Figure 10(b) shows the flow path of the three-phase voltage and current in a high-frequency UPS. As shown in the figure, the positive half-wave of UAN flows only through Q5, while the negative half-wave flows only through Q6; the same applies to the other two phases, UBN and UCN. That is, in a high-frequency UPS, one bridge arm can output one phase voltage, so the three-phase voltages do not interfere with each other, and therefore there is no requirement for the three-phase current to be balanced.
Figure 10 Current flow paths for the two types of UPS
4. Differences in the concept of circulation
All line-frequency UPS units have output isolation transformers at their output terminals, so parallel connection of UPS units is essentially parallel connection of the secondary windings of these output transformers. This leads to circulating current within the transformers because, generally, no two transformers are identical; there is usually a voltage difference between the windings of two transformers. Since this voltage difference is an electromotive force, it will inevitably generate circulating current, as shown in Figure 11(a), and the path of this circulating current is completely unobstructed. This circulating current is at its maximum under no-load conditions, resulting in additional power consumption. However, the voltage difference in parallel connection is generally less than 1V, so when a load is applied, the voltage drop due to the line's own resistance automatically adjusts the balance, generally preventing adverse consequences and only increasing some additional power consumption.
Figure 11(b) shows the circulating current situation when a high-frequency UPS is connected in parallel. Since a high-frequency UPS does not have an output isolation transformer, the circulating current is much less, or even non-existent. As can be seen from the figure, if there is a voltage difference between the two parallel circuits, for example, the voltage of the one on the left is slightly higher, the current path is as shown in the figure. It has to pass through two diodes and the internal resistance of two sets of batteries. The path is quite bumpy. Even a voltage difference of a few tenths of a volt is consumed along the way, and there is no power to form a circulating current.
IV. Current UPS Power Supply Technologies
The emergence of high-frequency UPS technology has also introduced many new technologies, such as wireless parallel connection in multi-stage parallel operation; the ability to perform full-load and overload tests on the machine without an external load; upgrading the unreliable operation of ECO (Economic Operating Mode) to BSS (Reactive Power Compensation) technology, achieving a power supply system efficiency of up to 98%; ensuring the UPS always maintains a high-efficiency energy-saving operating mode; and a half-bridge inverter circuit with one battery, etc. These advancements have laid a solid foundation for optimizing PUE in cloud computing and big data power supply systems.
