> One problem with simple chargers is they only draw power during a small part of the AC cycle.[5] If too many devices do this, it causes problems for the power company.
I'm mostly familiar with this from UPSes, but I assume it's basically the same problem on a different scale.
Lets say you have a device that requires 110 Watts - 1 Amp at 110 Volts. And you have a UPS that is rated to provide exactly the same.
In an ideal world (spherical cows, etc), and a device with a power factor of 1 (e.g., perfect), this works.
But if you have "non-sinusoidal current" (as his footnotes word it), your device isn't actually pulling 1 Amp. If you graph the voltage and current draw out, you should see the current draw forms a wave that matches the voltage (in shape, not value). If it doesn't, then at some parts of the wave, you're drawing more current than you claim - and at others, you're drawing less. So you're still drawing 1 Amp on average, but at any given instant, you're probably not.
So back to our spherical cow UPS. What looks like a perfect match on paper, goes wrong - 60 times a second, you're drawing more than those 110 Watts and causing an overload condition. And it's an issue that gets horrible when you scale it, because every device is receiving exactly the same wave form - so every device is using more than you think at precisely the same time. Like the trope of people jumping on a bridge at the same time, each consumer causes the same condition in perfect unison with each other.
It's not really that your device draws more than the claimed wattage, it's how the power arrives.
Resistance is proportional to current (I = V/R) so drawing 100W at 240v will result in a lower resistance than 100W at 110v. In an ideal world this doesn't matter but the real world had wires with resistance etc.
This means if you draw your peak current at peak voltage you're using as little current as possible, but if the peak current is offset you'll be drawing more amps than you would if you were in phase.
This causes parasitic losses in the distribution network and makes utilities grumpy, amongst other things.
I think what you're describing is phase shift (typically seen with inductive loads), where the current sine is the right form, but lags behind the voltage sine.
The (fifth) footnote in the article has "The difficulty comes from the nonlinear diode bridge, which charges the input capacitor only at peaks of the AC signal." And "If you're familiar with power factors due to phase shift, this is totally different. The problem is the non-sinusoidal current, not a phase shift."
So the demand looks like inrush current - but inrush at every single cycle. This produces a non-constant load, where the current draw instead graphs more like an ECG and less like a sine wave. If you overlay this ECG-style graph over a perfect sine, you see that to average the same draw, the spike has to peak much higher than the sine - because it's drawing nothing for the rest of the cycle.
The net result is basically the same (which is why both problems come under 'power factor') - you're drawing current in a very inefficient manner - but that's why I'm describing 'drawing more than the claimed wattage', because at the peak of each cycle, you do.
The simple answer is that all the current (and thus all the power) is drawn at the peaks of the sin wave. This means for most of the sin wave no current is being drawn. This is a problem because now the average power rating that a power plant can produce -- where the power is produced for all parts of the wave in roughly equal amounts -- is no longer enough, they have to be able to generate much more power just for the peak of cycle.
The term "load inbalance" is most commonly used in three phase systems where ideally the currents on all three phase/outer/hot conductors should add to zero, having no current on the neutral conductor. Having significantly different loading on the three phases will create changes in voltages on the individual conductors and generally not make efficient use of transmission capacity.
Interestingly, having a lot of switching power supplies with poor power factor correction on a three phase network (e.g. in an office building) will also create currents on the neutral, but with three times the network frequency (and harmonics), as during one cycle, the current for the peaks caused by...
What kind of problem?