The present disclosure relates to power electronics, and more specifically, but not exclusively, to a system and method for generation and distribution of high voltage direct current (HVDC) within a contained power domain.
A data center is a contained and controlled environment where server, storage, and networking gears run grouped in racks (e.g., commonly 19″ Electronic Industries Association (EIA) racks also known as “standard racks” following EIA-310 specifications) lined up in data center rows. These rows are normally powered by Alternative Current (AC) (from power grid) distributed throughout the data center at various voltage levels depending on locations. This internal data center power distribution is supported by a centralized (or in-row) Alternative Current Uninterruptible Power Supply (AC UPS), to guarantee the supply of AC power to the server racks during external forces including AC power grid outages, dips, sags, and power line disturbances. Data centers deploy and operate racks to match business needs, such as for an Information technology (IT) capacity. IT racks need to be continuously powered and service must not be interrupted by those external forces.
Conventionally, following an AC power grid outage, data center emergency generators (e.g., GenSet) startup to provide AC power again until the grid becomes available again; however, there is some latency before the generators get online. Accordingly, in the meantime, IT racks run using energy from batteries. Normally, the batteries do not last long (typically less than a few minutes)—only enough for the generators to start and get online. Once online, the generators might stay ON for hours and/or days, depending on the wait for the normal AC power grid to get back online.
Similarly, conventional IT racks may temporarily use high levels of power (higher than rack power rating), for example, during random repetitive ‘peaks’ or ‘surges’ resulting from microprocessors engaging ‘TURBO’ functions and/or similar temporary higher power functions. This may not be easy to handle since power availability to each rack is limited by their own source rating and breaker. Normally the power provisioning for a data center row is approximately the sum of nominal power rating of the racks in the row. Therefore, the row normally cannot handle load peaks above that sum.
Any power solution making use of local batteries for backup can include extraneous functionalities, provided that topology allows to share the energy from the batteries with the energy from the AC grid. For ‘Surge Mode’ (and consequently ‘Peak Shaving’ of the AC powering the rack) to work properly, large (or very large) battery banks may be needed; otherwise, these functions may be compromised by the quick discharge rate of the batteries.
Conventional approaches avoid using high voltage direct current due to safety aspects and limited availability of standard compatible components on the load side: for example, commercially available IT racks are powered from the AC grid with an AC plug, and are not available powered from high voltage direct current (HVDC) with a DC plug.
In view of the foregoing, a need exists for an improved system for electrical power design and distribution in an effort to overcome the aforementioned obstacles and deficiencies of conventional power systems.
It should be noted that the figures are not drawn to scale and that elements of similar structures or functions are generally represented by like reference numerals for illustrative purposes throughout the figures. It also should be noted that the figures are only intended to facilitate the description of the preferred embodiments. The figures do not illustrate every aspect of the described embodiments and do not limit the scope of the present disclosure.
Since currently-available power management systems are deficient because of the need to use high levels of power and latency delays from alternating current (AC) power grid outages, dips, sags, and power line disturbances, a system for improved electrical power design and distribution can prove desirable and provide a basis for a wide range of power management applications, such as for efficiently powering information technology (IT) racks deployed to data center environment, data center infrastructures, hyperscale scale data centers, IT environments, and providing features and functions highly desirable for the specific application. Additionally, the systems disclosed herein provide the advantage to be a very effective solution for capital expenditures, operating expenses, and for cost amortization over time. This result can be achieved, according to one embodiment disclosed herein, by a power management system 100 as illustrated in
Turning to
Turning to
The batteries 202 can keep service during AC input power grid outages (for minutes) and seamlessly compensate during AC grid sags/hiccups/disturbances. In some embodiments, the batteries 202 can include one or more Lithium Ion (Li-Ion) battery backup units (BBUs) that are coupled to the HVDC busway via intermediate converters 203 used to regulate voltage when BBUs discharge to the HVDC busway, share BBU output currents, enable “surge mode” functions, and charge BBUs. In other words, the batteries 202 can keep service during AC grid outages and seamlessly compensate for grid disturbances, sags, and so on. In some embodiments, the intermediate converters 203 can include one or more chargers and/or one or more optional dischargers. Although shown and described as separate functional units, those of ordinary skill in the art would understand that the intermediate converters 203 can reside on the same platform and/or circuitry as the batteries 202. By way of example, a charger circuit can reside directly inside each of the batteries 202.
In some embodiments, additional value is provided by this design such as ‘peak sharing’, ‘surge mode’, and ‘peak shaving’.
Peak sharing is the capability of the data center rack 300 to operate at power levels above its maximum source rating, as long as the system power domain POD (where the rack belongs to) stays below its maximum power rating. During peak sharing, the batteries 202 are not required.
Surge mode is the capability of the system power domain 200 to operate at power levels above its maximum rating and source rating. During ‘surge mode power peaks,’ the extra energy is supplied by the batteries 202. Individual rack power peaks may be as high as +50% nominal. In some embodiments, an in-rack power solution can be used to handle the additional power peaks as desired.
Peak shaving is an indirect effect of the surge mode. Specifically, during surge mode, the input energy from the AC grid can be capped (load power peaks are not transferred to the AC grid).
Electrical efficiency of the overall power distribution and conversion is comparable (or exceeds) best AC approaches. By way of example, best AC approaches are defined in the standardized server system specifications for scale computing of the Open Compute Project (OCP) Server Project.
Turning to
By way of example, the batteries 202 of
An AC power grid (e.g., a three-phase 380 VAC or 480 VAC) powers a first POD (e.g., the system power domain 200A) through a redundant high-power conversion stage, consolidated and installed to a separate cabinet within the cabinet 500 together with its own set of batteries 202. As shown in closer detail in
The power management system 100 can include one or more switches 450. With reference to
As also shown in
With this configuration, for example, one HVDC cabinet (e.g., the cabinet 500) powers two rack PODs (e.g., the system power domain 200A and 200B), each POD composed of six racks: e.g., (6×15 KW)=90 KW, with POD peak power capability as high as (90 KW+50%)=135 KW during surge mode. A fully deployed assembly of the cabinet 500 and the data center racks 300 are shown in
In some embodiments, handling batteries 202 so they can be kept charged, releasing power during an AC power grid outage, and/or adding extra energy during POD surge mode events (peak power demands above POD rating) includes the following process:
The cabinet 500 includes converter modules to produce HVDC from an AC input power grid, the batteries 202, and also further power converter modules called ‘DSCHG’ (the optional discharger module, the charger module, and/or the intermediate converter 203). These DSCHG modules are optional but can be used for best system performances as desired. The convertor modules can be embedded into the cabinet 500. The DSCHG can advantageously be used to keep the individual battery backup BBU modules charged (charging them properly and individually) and to keep the voltage across the HVDC busway 400 regulated during the discharge of the batteries 202 following an AC outage (backup sequence), or during surge mode. In fact, the DC-DC power converters installed in the IT racks (necessary to produce low voltage, e.g., 12 VDC for servers and IT gears) can be designed very efficiently and inexpensively provided that HVDC input voltage variation is limited: this issue is solved by using the DSCHG modules that keep HVDC bus voltage regulated during batteries discharge.
As another advantage of a DSCHG, during backup sequences or surge modes, the HVDC current in the busway never exceeds the max rating; being the load at constant power, lower bus voltages cause higher currents. This can be important when data center bus-bar power distribution and breakers are sized to avoid unwanted breakers tripping, for correct provisioning, to limit power and voltage loss during backup sequence or surge mode, and to avoid distribution bus overheat. In fact, without the DSCHG modules, the HVDC voltage during backup sequence or surge mode (events supported by the batteries) would lower substantially with the electrical current increasing in the opposite direction (batteries slowly discharge to lower voltages while load power remains constant). Furthermore, the DSCHG allows control of the battery modules to discharge current individually (optimum BBUs discharge current share), and make the application safer because no HVDC voltage from BBUs can reach cross slots in the rack (blocked by DSCHG).
The DSCHG can engage surge mode functions by sharing output current with the output of the AC-DC HVDC converter during surge mode.
This DSCHG ‘discharge converter and battery charger’ is normally OFF (or standby), thereby maximizing electrical efficiency during online operations. An exemplary current-voltage (IV) characteristic of the system power domain 200 during surge mode is shown in
An exemplary current-voltage (IV) characteristic of the system power domain 200 during backup mode is shown in
The power management system 100 can also be implemented by paralleling the batteries directly to the HVDC BUS. In this case the battery chargers would sit directly inside the battery modules and DSCHG modules are not required. This embodiment can be simpler and cheaper, but can sacrifice desirable system performances and safety guards, as previously explained.
The power management system 100 advantageously provides seamless continuity of power to the IT racks after any input AC power grid loss or sudden sags by using an alternative HVDC (High Voltage Direct Current) approach with local batteries, and fractioning the power domains to multiple independent cells called ‘PODs’. One POD of IT racks can be for example 6 racks, with each rack rated for example 15 KW (max).
Another advantage of the power management system 100 is a much higher electrical efficiency (cheaper power bill), a much lower up-front capital expenditure cost for implementation and deployment, and lower maintenance cost vs. the classical AC Universal Power Supply (UPS). Finally, grouping battery banks together for all racks in the PODs allows to efficiently use the energy stored in the batteries; in fact, power consumption of individual racks in the POD varies and cannot be controlled.
The power management system 100 also allows the racks to temporarily use up to +50% rack power on top of their normal max load rating (power converters installed in the IT racks are sized accordingly). This can happen because the source of power is consolidated and shared among all the racks belonging to the same POD (e.g., same row), so racks with lower consumption can give excess power to other racks in same POD that momentarily need more power (Peak Sharing); or even using battery banks charged at the HVDC voltage to compensate for power peaks exceeding the POD power rating itself (Surge Mode) while the energy from AC power grid would stay capped to the max provisioned value (Peak Shaving).
Moving out from IT racks into data center infrastructure part of the power conversion circuitry normally included in the racks, reduces cost overtime because racks are swapped in average every three years due to new generation of IT gears getting available (when this happens, rack frame and in-rack (or in-chassis) power converters also get disposed). With this approach, part of the power converters are not disposed with the racks (specifically the PFC circuitry (Power Factor Correction) and all of the related front-end AC circuitry of the server power supply), rather that stays in the data center infrastructure indefinitely in a consolidated fashion, until failure (note: the presently disclosed system can include redundancy). This is a winning cost amortization model, a substantial reduction of IT rack power cost over time.
Depending on how much battery capacity is installed, the power management system 100 facilitates the data center power provisioning in the sense that ‘Peak Shaving’ functionality pushes the <average used power> value towards the <peak provisioned power> value, because of the dumping factor provided by the battery banks that are connected to the HVDC bus. The AC input power from the grid can be capped (Peak Shaving) and so AC power cannot surpass the provisioned value. Stated in another way, the AC grid power provisioning, which usually equals the peak load and not to the average load, can be sized with precision and utilized in its fullness, with substantial cost savings. In fact, the power management system 100 compresses the ‘average power’ to ‘peak power’, so that the difference of the ‘AC peak power provisioned’ and the ‘AC average power used’ is minimized.
The disclosed embodiments are susceptible to various modifications and alternative forms, and specific examples thereof have been shown by way of example in the drawings and are herein described in detail. It should be understood, however, that the disclosed embodiments are not to be limited to the particular forms or methods disclosed, but to the contrary, the disclosed embodiments are to cover all modifications, equivalents, and alternatives.
Accordingly, persons of ordinary skill in the relevant art will understand that, although particular embodiments have been described, the principles described herein can be applied to different types of environments and solutions. Certain embodiments have been described for the purpose of simplifying the description, and it will be understood to persons skilled in the art that this is illustrative only. It will also be understood that reference to particular hardware or software terms herein can refer to any other type of suitable device, component, software, and so on. Moreover, the principles discussed herein can be generalized to any number and configuration of devices and protocols, and can be implemented using any suitable type of digital electronic circuitry, or in computer software, firmware, or hardware. Accordingly, while this specification highlights particular implementation details, these should not be construed as limitations on the scope of any invention or of what may be claimed, but rather as descriptions of features that may be specific to particular embodiments of particular systems.
This application claims the benefit of, and priority to, U.S. Provisional Application Ser. No. 62/636,424, filed on Feb. 28, 2018, the disclosure of the provisional application is hereby incorporated by reference in its entirety and for all purposes.
Number | Date | Country | |
---|---|---|---|
62636424 | Feb 2018 | US |