24 Hour: 0800 978 8988 Int: +44 (0)1844 340122

Parallel UPS – a new generation

Published by Jason Koffler on 27 February 2012

Parallel UPS (uninterruptible power supply) configurations have been in existence since the 1970s, but due to their cost and complexity have been deployed primarily in large industrial installations. Smaller operations have been unable to justify the cost of purchasing two or more UPS. However, paralleling UPS technologies have moved on in the past 40 years and the way today’s offerings, from manufacturers such as EATON, for example, manages the four complex challenges of paralleling: control, synchronisation, load balancing and selective tripping, is revolutionary and opens the market to a whole new span of UPS applications.

UPS (uninterruptible power supplies) can be paralleled to offer greater redundancy and/or extended capacity. Two or more UPS are electrically and mechanically connected to form a unified system with one output. An N+1 redundant configuration, for example, means that there is at least one more UPS than is required to support connected loads. In normal operation, each UPS shares the load equally but is also ready and able (in terms of having enough capacity) to take over either the total load or load sharing with the rest of the UPS in the system should one module fail.

If there is only one UPS in the system, if it goes offline through developing a fault condition or for any other reason, it automatically switches to by-pass supply, normally direct mains utility power. Today’s sensitive electronic equipment, particularly computers and networking technology, require a continuous source of clean electricity. Raw, unconditioned mains power is ‘unclean’ and prone to power problems which can damage or disrupt connected loads.

However, there are several challenges associated with traditional UPS paralleling techniques:

• Inefficiency – load sharing amongst paralleled UPS means that each unit is operating for most of the time at a fraction of its optimum capacity, which is highly inefficient.

• Complexity – controlling how each separate UPS operate as part of a unified system is complicated, as is synchronising the output of each and balancing the load equally amongst all UPS.

• Difficulty – If trouble occurs, and one or more UPS in the system fail, locating the source of the problem is difficult due to the complexity (particularly in the wiring) of the system.

• Challenging – how each of the four key challenges: control, synchronisation, load balancing and selective tripping, of parallel UPS are managed is also challenging.

Problems associated with some traditional parallel UPS architectures

In a traditional parallel UPS system, control is usually managed in a master/slave arrangement that relies on a highly complex, ‘spaghettified’ mass of wiring between UPS modules and other parts of the system, such as the bypass cabinet. This, in essence, is the brains that determines how UPS synchronise their outputs, share loads and decides what information they should respond to. Having such complexity exist in a control system is not the most reliable way forward.

Failure of the control system to automatically switch to a redundant control path is the leading cause of failures in mission critical power systems.

In a conventional UPS, synchronisation depends upon a master controller. If that controller goes offline, the whole system is in trouble. UPS modules would get out of synch with each other causing overload conditions, which could subsequently lead to a total breakdown.

Load sharing is typically dependent on a load-share loop, whereby UPS modules continually communicate their status to each other through the control wiring. But if any part of that spaghetti should get into a twist, the whole system fails. Electrical noise is also a common problem in this type of load sharing system and with this type of ‘close wiring’.

In a conventional UPS paralleling system, identifying the cause of a problem is hugely difficult. A faulty module may be signalled by the whole system going into bypass – which UPS is it?

New parallel UPS peer-to-peer control

Based on the successful example set by peer-to-peer computing, newer parallel UPS control systems such as EATON’s adopt a peer-to-peer control system rather than master/slave. Each UPS assesses its own operating parameters and determines how to interface with the others, thus negating the need for all the spaghetti wiring. No matter what happens to other modules, the parallel system still functions. Each module contains the intelligence it needs to be a functioning member of the group.

Each UPS synchronises to the bypass source, common to all modules. If one loses by-pass feed, it autonomously and intrinsically synchronises to another’s bypass. This is made possible by advances in control software and the invention of mega fast microprocessors. With this type of system there is no potential single-point-of-failure as there is no distribution of synchronisation signals.

Load sharing control algorithms in the UPS software maintain load balance by constantly making adjustments in response to variations in output power requirements. Each module is programmed to conform and thus they avoid conflict – true wireless paralleling.

Selective tripping is dealt with in new parallel systems by software algorithms within each module assessing information such as current/voltage and a running record so it can continually compare present waveforms with previously recorded ones. Using high-speed calculations, the unit can detect a fault even before hardware sensors have detected it. Should a module detect itself to be faulty, it turns off its inverter IGBT transistors within milliseconds. The result is a selective trip that instantly isolates the faulty unit from the system.

EATON’s soluation is designed for high-density environments in which space is at a premium. It takes up less room because no tie cabinet is required. Paralleling is accomplished through a plug-and-play bus structure that mounts easily in the back of the equipment rack. Each UPS module operates independently and is not reliant on an external master controller or a complex web of inter-module control wiring. No added circuitry or components are required for additional UPS modules to be switched in to operate in parallel, which is perfect for incremental system expansion.

How can we help?