Standby power systems for hospitals
Listen to the on-demand webcast, " Standby Power Systems for Hospitals: The Debate: Is Paralleling Generators a Good Idea?"
The primary goal of a hospital emergency power system is to have adequate and reliable capacity to serve priority 1, 2, and 3 loads when utility power is interrupted. These priorities generally, but not always, correspond to life safety, critical care, and equipment branches in the hospital, respectively. In the event of a generator failing, priority 1 and 2 loads should be powered at a minimum. There are two ways to achieve this capacity, each with pros and cons.
1. Parallel: Provide a paralleling, emergency system with N+1 generators such that the N generators will serve the entire load without exceeding 80% of their individual unit rating.
2. Nonparallel: Provide a system of generators and transfer switches so that each generator serves a dedicated load (that does not exceed 80% of any unit rating) and the load of any generator can be transferred to other generators to maintain power reliability.
For the paralleling system, the pros are traditional system configuration and true N+1 configuration; the cons are higher cost, large space requirements, and single point of failure (SPOF) mode in the paralleling cabinet. For the multiple generator system, the pros are a smaller footprint, reduced cost, and no SPOF modes; the cons are that the configuration requires two transfer switches be placed in series at some point in the system. Which is the better approach—if, indeed, there is one?
Paralleling generators for hospital systems
An N+1 parallel generating system is a group of generators whose output terminals are electrically connected to each other and to the devices that start and stop the units and control their outputs (see Fig. 1). The total capacity of all the generators exceeds the load by at least the capacity of the largest unit, and the system can tolerate the failure of any of the generators while continuing to serve the load.
The primary advantage of a parallel system over a group of unconnected generators is the system's ability to tolerate the failure of a single unit. In any emergency power system, the most likely failure is an engine generator that does not start. In an N+1 parallel system, the failure of a single generator doesn't lead to the loss of any emergency load, because the remaining generators are adequate to serve the entire load.
The parallel system does create the problem of SPOF mode. However, it's much more important to avert likely failures than unlikely ones. The intent of paralleling generators and of providing N+1 redundancy is to avoid load loss under the most likely negative scenario, even if this might create a SPOF.
Using dual-redundant programmable logic controllers (PLCs) as master controllers for paralleling gear may enhance reliability by ensuring normal operation of the gear if one of the PLCs fails. The extent of this enhancement will depend on the imagination and ability of the programmer to foresee possible failure modes, write reliable code to detect them, transfer control to the remaining unit when one fails, and arbitrate between the two controllers when they disagree.
Cost, size, and space
A parallel system requires the generators and distribution that are needed in any standby system, as well as paralleling gear. Paralleling gear is available in a variety of configurations and construction techniques, so there isn't a hard-and-fast rule for estimating its cost. Costs vary widely, and can run from $20,000 to $110,000 for each generator cabinet.
The cost for paralleling gear typically increases as unit size decreases, and the cost of each unit of capacity varies for different-sized generators. An indirect cost component is the total space required for generation, which typically decreases as unit size increases. For hospitals, the minimum generator size is determined by the National Electrical Code requirement that critical loads are restored within 10 sec of a power failure, generally requiring that the smallest unit can support those loads. Maximum unit size is determined by the generator cost and/or space constraints. Finally, generator size impacts maintenance costs as well. Larger units typically employ more exotic components, and require more sophisticated and expensive maintenance activities.
In theory, it's possible to parallel any generators that operate at the same voltage and frequency. In practice, the operational and economic issues involved sometimes preclude paralleling of dissimilar units, so most installations use units of the same capacity and manufacturer. The primary consideration is to verify that the existing bus has adequate ampacity to serve the increased load. The available breaker position must be able to accommodate an appropriately rated breaker for the new generator. Priority loads must still be served by the smallest generator. If the new generator is so large that the system can't support the increased load without it, the system will lose N+1 redundancy. In such a case, the designer should consider replacing multiple units.
Also, the new generators must be compatible with the existing units. If the alternators are wound with different pitch, their output voltages will have different harmonic content, causing harmonic current to flow among them. The magnitude of this current is limited only by the generator reactances. It can become quite large, and can cause alternator overheating, disruption of monitoring and control equipment, and system capacity reduction.
The engines' transient response, or the speed with which they can increase or decrease their energy output, must be compatible as well. If the engines differ too much in this regard, it can be difficult or impossible to persuade the generators to share load under changing conditions. When the load increases, the more nimble unit will accept most of the increase, and if the load step is large enough, it can trip on overcurrent. Alternatively, when load decreases, the slower unit may not back off quickly enough, and the faster units may trip on reverse power. Transient response data for engines aren't normally published, but they can be obtained from most manufacturers.
Lastly, the control systems must be compatible. In general, the older the existing system, the less compatible it will be with modern controls. There are many control schemes in service in existing paralleling gear. Some older installations might require replacement of some or all of the controls in order to accommodate a new unit. Under these conditions, a relatively lengthy outage of the existing system may be required, and designers should factor the costs of accommodating that outage into design decisions.
Nonparalleled generators for hospital systems
While traditional electrical design uses paralleling equipment for multiple generators to serve hospital emergency power and lighting systems, there is a viable alternative that may provide equivalent reliability at significant cost savings. In this nonparalleling system (NPS), paralleling equipment is eliminated and each generator provides power to one or more transfer switches, whether priority 1, 2, or 3 (see Fig. 2).
To accomplish this goal, an additional transfer switch must be inserted in the line side of the existing transfer switch so that there is an alternate generator that can feed the priority 1 and 2 loads if the lead generator fails to start. When additional transfer switches are included, the priority 1 and 2 loads should always have three sources: utility, generator 1, and generator 2. To facilitate the ease of shedding priority 3 loads when one generator out of service, these loads would have only two sources using one transfer switch.
Elimination of the paralleling equipment reduces the level of sophistication of the emergency system. For facilities with maintenance staff experienced in the operation and maintenance of paralleling systems, this is less of a problem. However, smaller hospitals have smaller maintenance staffs that often do not have the experience to deal with the complexity of a paralleling system. Without some very complex options, there are several SPOF modes for the paralleling system.
Some paralleling systems attempt to mitigate some of the SPOF opportunities by having multiple, redundant PLCs controlling the system. However, there is always at least one point in any paralleling system where a SPOF will create a total blackout of the emergency system. For even the most reliable paralleling systems, a short circuit on the main, paralleling bus would be the SPOF mode. These failure modes do not exist in NPS when the series transfer switches are used.
Cost, size, and space
While there are many advantages to using transfer switches to provide redundant emergency power to critical healthcare facility electrical loads, the construction cost savings are the greatest immediate benefit. The NPS requires no paralleling equipment, only the generators, switches, and other equipment needed for the standby system.
Additionally, using multiple transfer switches requires less floor area where space for the emergency electrical system is at a premium. This may allow a redundant power system to be installed without expanding the building, while having paralleling equipment might force the construction of an addition to the building. The capital funds made available by not installing the paralleling equipment may then be used to help with other portions of the project.
When transfer switches are used in lieu of the paralleling system, adding a load, another generator, or a series of generators and switches is just as simple as the initial design, and virtually all of the existing systems may be conserved. In the NPS, existing generators may be used without modification and new generators may be added to accommodate new loads and provide redundancy. There is no need to consider the ultimate load on the emergency system as in the case of the paralleled system, since additional generation may be added at any time.
Generators in the paralleled system may be of differing sizes; equally sized sets can easily carry equal parts of the load, but distributing load on unequally sized generators can create some problems. With the NPS, the loads are distributed during the design process so you can have generators of different sizes and still match the load to the generator capacity.
With the NPS, load shedding may be designed into the system and will not require a load-shedding control system. Since any load shedding involves removing the priority 3 loads from the system, designing a single transfer switch to feed these loads will automatically not power the loads when one of the generators does not start. Using less transfer switches for the priority 3 loads, as well as not buying the load-shedding system or controls, saves money.
Using multiple transfer switches
Using multiple transfer switches has many advantages for systems up to loads of 3,000 kVA. However, for larger systems, the implementation of multiple transfer switches becomes unwieldy. Accommodation of the larger loads is considerably easier with the use of a paralleling system. The second disadvantage of multiple transfer switch schemes is that they require placing two transfer switches in series. Although there is no electrical or code problem, one of the basic design precepts of a good electrical design is to avoid serried transfer switches.
To each facility its own
With all of pros and cons of the paralleled and NPS emergency power systems, the bottom line is that both systems have a place in critical power applications. For smaller power systems (nominally 3,000 kVA), using multiple transfer switches in series provides an inexpensive alternative to the significant investment in a paralleling system. When the calculated emergency load exceeds 4,000 kVA, the use of multiple transfer switches becomes unwieldy and a paralleling system becomes attractive.
It should be noted that the 3,000 kVA breakpoint is not hard and fast, but just a guideline. Complex systems with many smaller transfer switches would benefit from having a paralleling system even with emergency load totals on the order of 2,000 kVA. On the other hand, systems with fewer large transfer switches might still be feasible for emergency loads on the order of 4,000 kVA.
The paralleling system has two flaws (the SPOF of the paralleling cabinet and the paralleling bus) that also have solutions, but the cost of the improved system rises even higher than the basic paralleling system. The NPS offers significant cost savings but may not satisfy the needs of larger facilities. Selection of the appropriate emergency power system, whether a 1,200 kW with two 800 kW generators and four transfer switches, or a paralleled system with six 1,500 kW generators, requires a complete engineering analysis. The emergency load, available budget for the equipment, user-required reliability, space requirements, and numerous other factors must be considered to select the appropriate system.
Divine is project manager and electrical engineer at Smith Seckman Reid Inc. Lovorn is president and chief engineer at Lovorn Engineering Assocs.
About the webcast
On May 14, 2009, Consulting-Specifying Engineer hosted the first webcast in its Critical Power University series, entitled “Standby Power Systems for Hospitals: The Debate: Is Paralleling Generators a Good Idea?” To expand on the points for each side of the debate and to address the many outstanding questions posed by the more than 600 attendees, this article revisits the issue of paralleling generators in hospital emergency power systems. To view the original webcast, visit www.csemag.com/webcast/ondemand .
Selected questions from the webcast are answered in the sidebar beginning on page 14
Webcast Q & A: Here are selected questions/answers from the webcast.
Q: How can a paralleling cabinet be considered a single point of failure if it has three inherent levels of redundancy? And isn't a single ATS a single point of failure just the same as a paralleling bus?
A: Even for systems that have multiple redundant controls and other components, a major short-circuit on the paralleling bus will take down the entire system. In an NPS, while each transfer switch is a SPOF for the load that it feeds, that failure is not a single point that takes down the entire emergency system; it only takes down the specific load to which it is connected.
Q: For both approaches, I see shutdowns required for maintenance. Has anyone provided parallel transfer switches so that the transfer switches can be maintained without a shutdown?
A: T he standard transfer switch configuration to allow for maintenance on switches feeding highly critical loads is the isolation/bypass (I/B) switch. However, there is a significant cost differential between the standard ATS and the I/B-ATS. Unless your client has an unlimited budget, care should be taken on which circuits I/B switches are required and on which circuits they would be helpful, but not essential.
Q: How is sequencing set between the two ATS layers?
A: There is no sequencing equipment used between the two ATS layers. The first layer of transfer switches sense voltage on the utility side. The transfer sequence is initiated when the utility fails and a preset voltage and frequency is sensed by the transfer switch, signaling it to transfer the load to the emergency side. Likewise, when the transfer switch between the second generator and the utility/first generator senses that there is no power available from feeder for the first transfer switch, it then awaits the transfer level of voltage and frequency so the load may be picked up by the second generator.
Q: If the ATS is between the paralleling gear and the distribution equipment, why would you have to shut down the normal distribution to service the paralleling gear?
A: If you do not shut down the normal distribution in a scheduled manner and the paralleling gear is out of service, what would you do if there is an unscheduled utility power outage? We always recommend that both the utility and the generators be isolated from the paralleling gear during any maintenance operation. However, as you point out, maintenance might be completed without interrupting power to the load.
Q: Are the transfer switches in the transfer system closed transition? How do you achieve the 10-sec requirements for priority 1 and 2 loads?
A: The transfer switches were not closed transition. We have considered using them in the past but have not done so to date. As for timing, the generators should start simultaneously, so the delay between the downstream transfer switch realizing that both the utility and the upstream generator are not going to provide power should be around 7 sec. Since the downstream generator has already started, there is only a delay of ATS transfer time, which is on the order of three cycles, so the 10-sec requirement is easily achieved.
Q: How difficult was it to coordinate trip units between normal, 1,500 kW, and 800 kW and have enough energy to trip under fault while on 800 kW?
A: For the main breakers for each generator, the typical thermal-magnetic breaker will never trip due to a fault prior to the field collapsing to the forced-field condition. However, most generators today are internally protected so that they take themselves offline when the downstream fault does not clear. There is no issue with the smaller branch circuit breakers since they will have adequate tripping current available.
Q: Will a larger unit take longer to come up to a load acceptance speed and voltage? Does this support or argue against either side when speed to serve loads is considered?
A: We were assisting one of our clients on a study which included an older, 1,200 rpm, 1 mW generator. They could not get the genset to start and assume load within 10 sec and the transfer switches were set to transfer load at 90% voltage and 95% frequency levels. At these setpoints, the generator would not assume the step load (of about 700 kW), so we recommended that the transfer switch timers be reset to 20, 40, and 60 sec, respectively. At 20 sec, the generator would accept 30% of the load and the remainder was split between the other two delayed transfer switches, solving the problem.