Imagine that you are running a data centre responsible for a large internet service provider's (ISP) server network. A storm cripples the power grid and the centre suffers a power outage. If there is no ups infrastructure in place and the servers' power supply is severed, the loss of revenue to both the ISP and data centre could be catastrophic.
Having planned for such an occurrence and by operating an ups system, this scenario can be averted. The ups provides protection against, and assurance that, any loss of power will not detrimentally affect power-reliant operations.
Now, imagine the above happening and the ups failing or it being the subject of a regular maintenance visit. This does not bear thinking about. However, despite the ups offering 99·999% availability, such vital applications should not be entrusted to a solitary ups.
To protect against the second scenario, organisations are demanding parallel redundant ups systems, where multiple supplies are configured in parallel incorporating an external static switch (bypass). This provides a multi-module support system with increased fault-clearing capacity (fcc) compared to the internal static switch of the individual ups.
The premise behind this model is that a parallel system offers increased reliability; should one ups fail, additional units are available to support the load, thereby offering 99·999% availability. The rule for the configuration of such systems is n+1, where a 1200 kVA load would be supported by three 400 kVA (n) units plus an extra 400 kVA ups (+1), thus increasing the fcc and mean time between failure (mtbf).
The implementation of a parallel redundant system is widely acknowledged as a way of offering increased mtbf, high power availability and integrity. However, the means by which the fault or short circuit sharing capacity is managed is one that creates controversy among manufacturers.
External versus internal
UPS vendors implementing such parallel redundant systems use either an internal, integral ups module static switch from each of the units to operate in parallel (figure 1) or an external load-rated static switch with a very high kA rating. However, with the likes of data centres and air traffic control installations requiring secured 99·999% power availability, the overall reliability and functionality of the internal static switch means that it cannot consistently satisfy such high demands during short circuits on the load side.
The premise behind the implementation of parallel redundant systems is that they will offer maximum power protection against virtually all eventualities. However, due to the technical nature of the internal integral static switch, such protection cannot be provided. If we take the earlier example and relate this to the arguments that follow, it will become apparent that the two methodologies differ greatly.
To protect against an outage, the data centre has in place a parallel redundant system (n+1) operating with integral internal static switch methodology – four units each at 400 kVA (n+1) protecting a 1200 kVA load. Let us assume that one of the units fails, resulting in reduced fcc values. At this point, the 1200 kVA load is only supported by a parallel implementation of 3 x 400 kVA. The fcc is now subject to the individual capacity of the internal static switches (400 kVA each) and the cable impedances.
In order to clear the short circuit, the fault must be routed to each of the supporting ups systems via their internal static switches. However, as all static switches cannot operate simultaneously (nor can the electromechanical breakers for each supply), the load is therefore exposed and without either ups or mains support.
It is clear that such systems require an extremely detailed study with regard to the selection of input side circuit- breaker sizes, thus ensuring that the network discrimination is not left to chance. Albeit for a millisecond, such a moment of exposure could bring down the data centre, potentially costing the organisation millions.
Should the data centre have installed the same parallel system using the external static switch methodology (figure 2), there would have been no exposure. External static switches can be rated to any load and are not reliant on individual ups static switch limitations. The application could have been designed with four 400 kVA units (n+1) supporting a 1200 kVA load with a 1200 kVA external static switch. At the point of solitary ups failure, the system's fcc would not have been compromised as the external static switch instantaneously transfers the short circuit to the bypass, avoiding any risk to the load or exposing the centre.
Ultimate protection
From this example, we can see that an external static switch can provide an organisation with a far greater degree of assurance. Due to its independence, the external static switch functions as a separate fail-safe in the event of an ups malfunction or downstream short circuit.
Due to its ability to expand, the external static switch offers a far greater degree of load protection. However, internal static switch technologies can be adapted to provide similar levels of protection, but with higher component counts.
Consider a system operating five 200 kVA units (n+1) supporting a load of 800 kVA. The external static switch in place would be 800 kVA, thus supporting the total load in the unlikely event of ups failure. If it were a super-critical application, a second 800 kVA static switch could be installed to offer additional protection.
To provide the same level of assurance with an internal static switch this model would have to implement four static switches, each being a 800 kVA-rated ups, resulting in a total rating of 3200 kVA – far greater than necessary.
Such practice would be nonsensical, as this renders the system impractical economically.
Nevertheless, vendors providing parallel redundant systems to ventures such as data centres enjoy two advantages: size of footprint and cost.
Despite the inherent technical and functional disadvantages, the internal static switch system does offer a slightly smaller footprint as all components are housed within each ups. In the eyes of the data centre, this means that the ups infrastructure needs less space, and as such allows more server racks to be rented, thus bringing in more revenue. This, coupled with the lower initial purchase price, has allowed the internal static switch model to retain a prevalent position in the market, despite its flaws.
To combat such thinking, providers of ups systems with external static switch technology need not issue special discounts to entice customers, or instil the virtues of the technical advantages of its offering, but ask a simple question: how much indemnity cover does your consultant possess?
Not a strange question in reality, if we take an air traffic control centre and imagine that a design consultant has specified a parallel ups system using internal static switch methodology. Naturally, to facilitate purchase, a professional recommendation has been made that the system suggested will meet the protection needs of the installation in question (99·999% availability). Now imagine that a short circuit on the load side takes place and the millisecond taken to send the fault upstream causes data exposure, radar screens going blank and planes being placed on hold position. Does the consultant have the indemnity cover to protect against this eventuality? They had better hope so. Despite the minimal cost saving, the risk to passengers, loss of revenue and faith could be millions of pounds.
In such a scenario, users face a situation where they cannot realistically argue in favour of lower capital outlay and increased risk. Instead, common sense must prevail and the external static switch design should be chosen.
Source
Electrical and Mechanical Contractor
Postscript
Shri Karve is business development manager at MGE UPS Systems.
No comments yet