Most larger facilities receive 3 Phase Power from the power utility company. This consists of 3 hot power legs or phases, a neutral which carries the return current for all three hot legs, and a safety ground. In such facilities it is common for the 120 volt power circuits used to power most of the equipment and lighting in the building to be derived from the 3 of the incoming power legs, some from each leg. The electrical engineer tries to balance the loads so equal amounts of current are drawn on each of the three power legs. There are a number of reasons why such load balancing is desirable, but one reason is that it often results in the least current flowing in the Neutral conductor.
For many years it was taught that in a perfectly balanced 3 phase system, the current in the Neutral would always completely cancel out leaving no neutral current. If the loads are all resistive then this is true.
For example, consider this diagram:
Each of the three colors represents the voltage on one of the 3 hot legs of a 3 phase power service with time as the horizontal axis. If the loads are resistive or motor, then these three colors could also represent the current drawn on each leg. Note that when red is at the positive peak, both blue and green are at 50% negative, therefore the sum of all three currents is zero. You can pick any point in time, and as long as the loads are resistive and equal to each other in magnitude, they will perfectly cancel out in the neutral.
As a result of assuming that this cancellation would always happen it was common to use thinner wire with less current carrying capability for the neutral conductor. While this was very common in the past it is not safe. For example what if the loads on the green and blue legs were shut off leaving only the loads connected to the red leg in use? In this case the Neutral would have to carry the same current as the red leg, and therefore can't be smaller without a fire danger.
However modern electrical loads make matters even worse. Few loads in a modern building actually act resistive drawing current in proportion to the voltage
Note that in this example there is no cancellation of current in the Neutral, and the Neutral must carry the sum of the current drawn on each leg. Also note that the frequency of the current in the Neutral is three times the mains frequency. In other words if the mains frequency is 60 Hz like it is here in the USA, then the current in the Neutral will be mostly 180 Hz.
Now with real world loads things are not likely to be quite as extreme as the diagram above, but they can come close.
In 1983 I was working for Broadway Video in NYC. We were expanding our facility and were concerned about the amount of power available on the 100 Amp 3 phase power service for our central equipment room. I took a current probe and a meter and checked the loading on each leg of the incoming power service to the room. The first thing I noticed was that the power panel was hot. I quickly discovered that the conduit carrying the incoming power mains was so hot I could not keep my hand on it. My first thought was that we had loose connections, but they were all tight. I then used the current probe on each of the incoming mains conductors. Each leg had between 85 and 95 Amps of current flowing which was under what the panel and circuit was rated for, so that should not have been a problem. I next put the probe on the Neutral expecting to find a small current on the order of 10 Amps. What I measured was 180 Amps in this Neutral that was 2 sizes smaller than each of the hot conductors.
Trying to understand what was happening I checked again with a different current probe and found the same thing. I then used a scope instead of a digital voltmeter and discovered that the Neutral current was an almost perfect 180 Hz. This puzzled the engineering staff as it went against everything we had been taught. We checked with the electrician who flatly stated it was impossible. Eventually I checked with John Chester (one of the most innovative audio and computer engineers in NYC) and he helped me understand what was happening.
The electrician had to install new larger conduit since the wires had melted into the existing conduit. It was a huge job, but we were lucky we did not have a fire.
Fast forward to the present and this issue is now acknowledged in the National Electrical Code which recommends that Neutrals be upsized when modern loads are driven. Many touring sound and video companies use power systems with dual Neutral connections so two wires can be used together as Neutral. Fluke makes power quality test instruments designed to look for these sorts of problems, and it is common knowledge that the so called "triplen" power current harmonics (in other words currents that are 3, 6, 9, etc times the mains frequency) will not cancel in the Neutral of a 3 phase power system.
Thanks to Dale Shirk for producing the above diagrams to illustrate this issue.