Odd triplen harmonic currents and neutral overloading go hand-in-hand.

A facility manager of a data center was informed by the Information Systems Department that additional file servers would be installed and would be powered by an existing panelboard. Therefore, an analysis of the panelboard, its feeder, and loading need to be done.

To find out if a problem would occur with the proposed load addition, the facility manager asked the maintenance electrician to review the annual testing and maintenance reports on the feeder of the existing panelboard. The feeder in question consisted of three 500kcmil phase conductors, a 1/0 AWG grounded (neutral) conductor, and a 1/0 AWG isolated grounding conductor. All conductors had 75 [degrees] C insulation. The data center's feeders were installed with the practices of time, which included reducing the neutral conductor in the feeder. The annual testing and maintenance report indicated loads of 99A, 130A, and 77A respectively for the phase conductors and 130A for the neutral conductor. The load on the feeder after the addition of the new file servers was estimated to be 132A, 182A, and 101A respectively for the phase conductors and 197A for the neutral conductor.

The electrician noted that an overload problem would occur, not with the phase conductors but with the neutral conductor. The phase conductors would be loaded to approximately 60% of their ampacity (B phase as worst case) while the neutral would be overloaded to 131% of its ampacity.

Detailing the problem

The problem was more than an imbalance on the phase conductors. Using the following equation, the anticipated (normally expected) neutral current ([I.sub.N]) is 46A for the present load and 71A for the estimated revised load.

[I.sub.N] = [square root of ([I.sub.[A.sup.2]] + [I.sub.[B.sup.2]] + [I.sub.[C.sup.2]] - [I.sub.A][I.sub.B] - [I.sub.B][I.sub.C] - [I.sub.C][I.sub.A])]

The measured neutral current of 130A is obviously more than the 46A calculated.

The cause of this problem was harmonics. The load on the existing panelboard was line-to-neutral connected nonlinear loads. This type of load consists primarily of odd triplen harmonic currents (e.g. 3rd, 9th, 15th, 21th, etc.). These currents add in the neutral conductor and can cause measured neutral currents as much as twice the phase conductor currents.

How can neutral currents be twice as great as phase conductor currents? The rules change on power systems having odd triplen harmonic currents. Let's investigate this further.

Nonsinusoidal currents

The neutral current consists of the imbalance of the phase conductor currents ([I.sub.A], [I.sub.B], and [I.sub.C]). If these currents consist only of the fundamental (60 Hz) current, there is a 120 [degrees] phase shift between each phase and the summation of these currents at every instant in time is zero.

Graphically, this is shown in Fig. 1 below. Here, three phase currents are superimposed on a graph. By stopping time at the noted instant, the amplitudes of each of the currents when added together will equal zero. As a result, the neutral current is zero.

If the phase conductor currents contain both fundamental and odd triplen harmonic currents, the result is very different. Odd triplen currents are zero sequence currents in that they are in phase and will add in the neutral conductor. The harmonic current with the largest profile in the odd triplen harmonic currents is usually the third harmonic. Fig. 2 shows us the sinusoidal waveforms of each of the phase currents, along with that of the third harmonic for all three phases. While the fundamental (60 Hz) currents of the phases cancel each other, the third harmonic (180 Hz) currents of each of the phasesadd together. Fig. 3 shows how these currents flow in a 3-phase, 4-wire schematic.

When the third harmonic is present, a distorted voltage waveform will result and will provide a characteristic signature of the nonlinear load. Fig. 4 shows how the combination of sinusoidal voltage waveforms and third harmonic waveforms create harmonically distorted waveforms.

Possible solutions

In relation to our potentially overloaded neutral conductor, three options were developed to address this problem.

* Replace the feeder with one that has full size neutral.

* Use a neutral filter to reduce the anticipated neutral current within the rating of the conductor.

* Relabel the isolated grounding conductor and reconnect it as a parallel grounded conductor.

The first option was not considered due to the problems and time involved in replacing the feeder in this particular application. Financially, the data center could not afford the extended outage required to replace the feeder.

The second option was seriously considered. The neutral filter uses a zig-zag transformer design that reduces odd triplen harmonic currents by a factor of 7.5 or greater. It uses the phase shift characteristics of a transformer to cause cancellation of the odd triplen harmonics. This would reduce the anticipated neutral current to a value less than the current rating of the conductor. The filter was reasonably priced and could be installed with a little downtime of the feeder. The disadvantage of the filter was that it had to be installed adjacent to the affected panelboard. This took up floor space and meant additional cooling load to the center's cooling system.

Though the third option was a late entry, it proved to be the winner. The isolated grounding conductor (same size as the neutral conductor) was never used as intended. Other feeders to the data center did not have grounding conductors; instead, the metal raceway system was used as a ground. After discussions with the electrical inspector and the facility manager, the decision was made to relabel the isolated grounding conductor, disconnect it, and reconnect it as a parallel grounded conductor. This doubled the capacity of the neutral. The consensus was that this situation was no worse than other feeders in the system. And, the work could be accomplished with minimum downtime on the feeder. The anticipated neutral current would now be approximately 65% of the capacity of the conductor.

James Moravek, P.E. is Vice President of Hammel Green & Abrahamson, an architectural and engineering firm in Minneapolis, Minn.