• Reliability considerations in simple paralleling applications

Rear view of male supervisor examining control room in industry

Learning objectives

  1. 1. Understand advantages and disadvantages associated with using paralleled generator sets as opposed to a single, larger generator set.
  2. 2. Know the codes that govern generator on-site power generation systems.
  3. 3. Identify best practices to maximize system reliability in simple paralleling applications.

Reliability in power generation systems, defined as the probability that power will be available at any point in time, is the primary reason standby generator sets are purchased. Using paralleled redundant generator sets is one method commonly used to enhance system reliability. Redundancy traditionally has been a requirement only in critical applications such as data centers and hospitals where an extended loss of power could result in loss of life or a substantial financial loss, as these were the only scenarios where the cost of a redundant generator and the associated paralleling switchgear could be justified. In recent years, however, the availability of lower cost power transfer devices and paralleling control systems has made redundant paralleled generators an attractive option in less critical standby power applications.

The decision of whether to use a single generator set or multiple paralleled generator sets typically is based on reliability and cost. After a decision is made to use paralleled generator sets, many considerations must be addressed to ensure a reliable system.

Reliability and redundancy

The purpose of redundancy is to eliminate a single point of failure from a system. While it is well documented that having redundant systems makes the overall system more reliable, this is based on the assumptions that single points of failure are truly eliminated and not just moved to another part of the system and that the controls enabling redundancy don’t introduce new failure modes that compromise reliability. Paralleled generator sets that rely on a single master control for signals to start and to close to a paralleled bus actually replace one failure point with two, as the master control and the communication link between the master and the generator sets each represent single points of failure.

Investing in a reliable standby generator set and a robust maintenance program so the generator doesn’t fail is often a better investment than installing a more complex system to compensate for a failed generator set.

Total system cost

In some instances, the cost of two small generator sets is less than the cost of one larger generator set. The total installed cost of the system is often overlooked in basic standby applications. Beyond the cost of the generator sets, the following factors must be evaluated:

  • Foundation: A larger generator set may require additional structural support as its weight will be concentrated on one spot; however, smaller generator sets may require pouring multiple concrete slabs.
  • Space requirements: Multiple generator sets and their associated switchgear will take up more space than a single larger generator set, although the smaller generator sets offer greater flexibility as they can be maneuvered into smaller spaces.
  • Cabling: Smaller generator sets enable the use of smaller cables and easier termination. However, paralleled generator sets require additional cable runs, which is labor-intensive, particularly if cable is run underground.
  • Commissioning costs: Start-up and testing costs of paralleled generator sets are substantially higher than those for a single generator set.
  • Maintenance costs: Replacement parts for smaller generators are less expensive than replacement parts for a larger generator. However, that difference is more than offset by the labor costs of maintaining two generators and switchgear rather than one single generator set.
  • Capital investment: When a facility’s power demand is expected to increase in the future, in some cases initial capital investment can be minimized by installing a smaller generator with the intent of adding paralleled generators in the future as demand increases. This will need to be balanced against the future investment required to add generators and switchgear and other required facility modifications.

Control system

A robust control system is critical for a reliable paralleling system. A control system must minimize single points of failure and have built-in fault tolerance measures. Key factors in a paralleling control include the following.

Eliminate single points of failure. The most effective way to eliminate single points of failure in a control system is to use distributed logic and control rather than centralized control. Critical control functions such as generator starting, bus voltage sensing, synchronizing, and closing to the bus should be executed by individual generator set controls rather than a master control. This way, the system will have redundancy in critical control functions in addition to redundant generators, eliminating the single point of failure.

  • Generator starting: In a simple standby isolated bus paralleling application, the start signal is sent directly from the transfer switches that sense the utility failure to the generator sets. Sending the signal through a master control adds no value and introduces an unnecessary failure mode. Sending the start signal directly to the generator sets for the transfer switch contacts is the simplest, most reliable means of starting the generators.
  • Paralleling bus voltage sensing: For reliable paralleling, each generator must sense the bus voltage independently rather than rely on a signal from a separate control
  • Closing to a dead bus: When closing to a dead bus, the system must include an arbitration scheme to prevent multiple generators from closing to the bus at the same time. To provide the fastest and most reliable service to a dead bus, the arbitration and breaker control logic must be resident in the generator controls rather than in a master control. Waiting for a permissive signal from a master slows the system down and adds an unnecessary failure mode.
  • Synchronizing and closing to a live bus: Generator sets synchronize reliably and quickly when the bus sensing and synchronizing logic is part of the generator set control. External controls adjusting the generator set frequency and voltage in an attempt to synchronize with the bus introduce unnecessary complexity into the system.

Figure 1: Implementing a load add function: Connecting the normally closed aux contact of each genset’s paralleling breaker in parallel to the ATS transfer inhibit input will prevent the ATS from transferring to the emergency source until all of the gener

Load add and load shed. Load add and load shed schemes ensure that there is always sufficient capacity to serve the most critical loads; less critical loads are served as capacity becomes available. Two levels of load add (one level for emergency loads and one level for all other loads) and one level for load shed (emergency loads are never shed) are sufficient for most simple, isolated bus paralleling applications. This can be implemented without the use of a master control. However, a master control may be required for additional levels of load add/shed. Although a master control can present a single point of failure, the system can be designed so that failure of the master will not impact the most critical loads.

A load add scheme is required in a paralleling system when a single generator is not large enough to carry all of the loads in the system. A simple load add scheme with two levels can be implemented using the inhibit function of the non-emergency load transfer switches and the aux contacts of the genset paralleling breakers (see Figure 1). Emergency load transfer switches should not be inhibited and should close to the bus as soon as it is live. Non-emergency transfer switches can be inhibited until all of the generator sets come on line.

A load shed scheme is required so that when generators are overloaded, the noncritical loads can be taken off line so that there will be sufficient capacity to serve the critical loads. Most paralleling generator set controls have a load shed or load dump output that can be connected to the load shed input of the transfer switches that serve the non-emergency loads (see Figure 2). This will take the non-emergency loads off line in the event that the generator sets are overloaded. Note that to properly shed load, transfer switches must be three position switches with center-off positions. Using two position switches to shed load is not recommended as this often results in connecting the load to a failed utility.

Figure 2. Implementing a load shed function: Connecting the load dump contacts of each generator set in parallel to the ATS load shed input will cause the ATS to shed load in the event that any of the generator sets activate their load dump output. Most gIsochronous vs. droop load sharing. Most load sharing systems today are isochronous, meaning that the voltage and frequency are held constant. However, there are still controls being produced that use droop load sharing, which allows voltage and frequency to vary with load. Droop load sharing controls were once popular because they allowed generator sets to parallel with each other without communicating with each other. Due to the variation of frequency and voltage with a droop paralleling system, the quality of power provided to the load typically is not very good and may not be suitable for some electrical equipment. Isochronous load sharing is the appropriate technology to use.

Random access paralleling vs. exciter paralleling. Random access paralleling refers to a system in which the first generator at rated speed and voltage closes to the dead bus and then all the other generators actively synchronize and close to the bus. This type of paralleling is the most reliable method and is most commonly used in critical applications. A less expensive paralleling method known as exciter paralleling is used in some paralleling applications. In an exciter paralleling system, all of the generators start with their paralleling breakers closed and their excitation circuits de-energized. When the generators start, they are connected to the bus but produce no voltage. When all generators reach starter disconnect speed, their excitation circuits are energized and the generator bus voltage ramps up with the generators forcing each other into sync. Because exciter paralleling systems will not work until all generators either come up to speed or are locked out, they are not used in critical applications. Another disadvantage of an exciter paralleling system is that if one generator set shuts down it cannot be brought back on line unless the entire system is shut down and restarted. Random access paralleling with active synchronization should always be used when paralleling gensets in critical applications.

Fault tolerance—manual mode. A key consideration in assessing the reliability of a system is its ability to operate when certain components have failed. Having a predefined sequence of operations for a user to follow in a worst-case scenario to manually provide power to loads is often a requirement in critical applications. A user should be able to manually start generators, initiate synchronization, and close paralleling breakers. Manual operation does not mean that the generator control is not operating; it means that functions are user initiated rather than system initiated. All system protection functions will still be active. The control will not allow a paralleling breaker to close if the generator and the paralleling bus are not in phase with each other.

Installation considerations

Installing and commissioning paralleled generator sets is not a simple process. A qualified manufacturer will have experience with protective relaying, system grounding, and other paralleling issues beyond the generator set functionality. Working with a manufacturer that has substantial paralleling experience over a wide range of applications and will assume responsibility for a correct installation is a key to a successful project even in the most basic paralleling application. There are several considerations that an experienced installer will address.

Selective coordination. The National Electrical Code (NFPA 70) requires selective coordination for emergency and legally required loads (Sections 700.27 and 701.27). Any downstream breakers must coordinate with upstream overcurrent protection such as paralleling breakers or a genset mounted breaker. Coordinating with a generator set mounted molded case circuit breaker (MCCB) with an instantaneous trip will be very difficult and will require in most cases that the downstream breakers are supplied from the same breaker manufacturer as the genset mounted MCCB. It is much easier to coordinate with power breakers as are most often used in paralleling switchgear as they are typically equipped with a programmable trip unit specifically for the purpose of coordination. When the generator control includes integral, UL listed overcurrent protection, coordination between the genset and the paralleling breaker is simplified because the overcurrent trip curve is optimized to allow the maximum permissible time delay while still protecting the alternator.

Separation of circuits. The National Electrical Code (700.10(B)(5)) requires that emergency loads are separated from legally required and optional loads. NEC 700.10(B)(5)(d) allows the use of “single or multiple feeders to supply distribution equipment between the emergency source and the point where the combination of emergency, legally required or optional loads are separated.” In this context the “emergency source” can refer to either a single generator set or the bus in the case of paralleled generator sets. Emergency circuits must be separated from legally required and optional standby circuits with separate circuit breakers installed in separate switchboard sections or enclosures. Emergency loads must be switched using a dedicated transfer switch, separate from legally required and optional standby circuits.

There is some ambiguity in the code regarding whether legally required and optional standby circuits need to be separated from each other. The language in the code doesn’t explicitly require separation of legally required loads from optional standby loads, but all of the exhibits in the NEC handbook show them as separate. Some authorities having jurisdiction (AHJs) have enforced separation of legally required loads from emergency standby loads.

Isolation of generators from the paralleling bus. To enable maximum reliability and safety, there must be means to individually disconnect each generator from the paralleling bus located at the paralleling switchboard. Without this disconnecting means, typically an incoming breaker, a fault on one generator can make all generators inoperable and all generators will have to be locked out to do maintenance work on any generator in compliance with NFPA 70E lockout-tagout requirements (see Figure 3). Without this disconnecting means, much of the value of having a redundant generator will be lost.

Figure 3: Distribution panel with no incoming breaker: Without the incoming breaker, a fault in one generator feed or maintenance on one generator will make the whole system inoperable. Courtesy: Cummins Power Generation

10-second start requirement. The National Electrical Code requires that emergency loads are served within 10 seconds of a utility failure. As 10 seconds typically isn’t enough time to start, synchronize, and close multiple generator sets, this requirement means that each of the generator sets in the system must be large enough to carry all of the emergency loads on its own. For example, a system of three paralleled 600 kW generator sets will not meet the NEC requirement if the emergency load exceeds 600 kW. To meet this requirement, the sequence of operations should not require any interaction with a master controller. All control functions should be carried out independently by the generator and the transfer switch controls with no communication required other than a genset start command.

Supplier of distribution switchboard. Sourcing the generator and paralleling controls separately from the switchboard complicates a project. The contractor and consulting engineer must have a clear definition of what is included in the scope of supply for each part of the system. Dividing scope of supply won’t be limited to equipment but will also include assignment of responsibility for meeting code requirements like selective coordination and separation of circuits, system testing and start-up, and ongoing maintenance and service. Having the same entity responsible for supply, installation, and maintenance helps promote an efficient maintenance program.

Service and support

One of the first questions that should be considered when choosing a supplier of a paralleling system is how the system will be supported in the future. Properly supporting all the system’s equipment requires a diverse skill set. Working with a supplier that has a proven track record of designing, installing, and maintaining complete paralleling systems is the best way to ensure reliable operation over the life of the system. Questions to ask include:

  • Have the service technicians been trained and certified by the manufacturer on all components of the paralleling system? Claims that an engine dealer technician can service a paralleling system should be viewed with skepticism.
  • Have the service technicians been certified with the make and model of engines being used in the application? Suppliers that use engines from different manufacturers may require different service organizations to support systems within the same geographical region.
  • Does the service organization offer comprehensive maintenance programs for the entire system?
  • Does the service organization have a demonstrated history of supporting paralleling systems in many different types of applications?
  • What is the availability of replacement parts? Power transfer devices will often need to be replaced after faults. Replacing a proprietary contactor rather than a paralleling circuit breaker can result in a system being down for a significant period of time. Replacing a proprietary component with a standard one may not be acceptable if the new device isn’t listed for use with the existing overcurrent protection.
  • If a control needs to be replaced, is custom programming required for the replacement control and who is authorized to do the programming? What is the time frame for replacing the control?


Paralleling systems are frequently expanded after they are put into service to accommodate increasing power demands. The ability to add a generator set and the associated switchgear in the future should always be considered. The system should have the flexibility to allow generator sets from a different manufacturer to be added in the future. Being locked in to a certain manufacturer limits the flexibility for future expansions. Additional question to ask concerning expansion include:

  • Are the generators properly isolated so that new generators can be added without taking the facility off line?
  • What is involved in modifying the control for expansion? If a different manufacturer’s generators are used for the expansion, what will be required for the generators to parallel properly?
  • Does the manufacturer have experience in implementing field expansions, including expansions that include generators from other manufacturers?
  • How can the system be modified to support utility paralleling if that is required in the future?


The decision on whether to provide backup power with a single generator set or with redundant paralleled generator sets will be based on reliability and cost. The key question is whether the redundancy coupled with the added complexity of a paralleling system increase the system reliability enough to justify the additional cost. When a decision has been made to parallel generator sets, the following should be addressed to maximize the reliability of the system:

  • The control system should be designed with critical functions distributed to the individual generator controls to minimize single points of failure.
  • The controls should have fault tolerance provisions such as load shed and manual operating modes.
  • The installation must meet code requirements for coordination, separation of circuits, and 10-second service to emergency loads and must allow proper isolation of the generator sets.
  • The system must be supported by an organization with a proven track record for servicing complete paralleling systems.

Rich Scroggins is a technical specialist at Cummins Power Generation. Scroggins has been with Cummins for 20 years in a variety of engineering and product management roles and is active in IEEE 1547 and NEMA SC 16. He received his bachelor’s degree in electrical engineering from the University of Minnesota and an MBA from the University of St. Thomas.

Powered by ContentStream®CSE_Logo_Color_ID