The sort of demand charge cited in that article is fairly rare. The typical demand charge simply looks at the peak demand of the customer, regardless of when that occurs. Managing demand charges is a key business proposition for "Smart Grid" applications. For example, one of
Comverge's offerings coordinates when HVAC compresses turn on, sequencing them to eliminate spikes.
The challenge in trying to manage the sort of "coincident peak" demand charges described in the article is to predict exactly when the system peak demand will occur. Under one plan (now replaced) in the UK, all of the transmission charges for the month were allocated based on a customer's use during the top three, non-consecutive hours. This rule spawned a service business of forecasting those hours; companies even guaranteed that they could hit two of the three. Of course, the more people who attempt to shift usage from a peak hour to an adjacent hour, the more likely it is that the peak itself will move.
Having on-site storage combined with good software to control energy usage could certainly save commercial customers a lot of money. If residential customers were ever charged for their power use on a sensible tariff design, then they could, too.