Smart grid architecture: How hidden couplings could sabotage the smart grid

1

Jeff Taft is chief smart grid architect at Cisco Systems. He has previously held similar roles at Accenture and IBM. He has developed some important insights into the problems that will arise when we take the smart grid to scale. And into ways we can forestall those issues. I asked him to share his insights with you. – Jesse Berst

By Jeff Taft

We often talk about the similarities between the smart grid and other digital networks such as telecommunications and cable. Those comparisons are certainly valid.

But there are also differences – differences that could torpedo the launch of the smart grid unless we plan ahead. One of the most troublesome is a phenomenon I call hidden coupling.

The hidden dangers of modularity

The development world favors loosely coupled software systems. And for good reason. In most circumstances, it makes sense to develop programs as separate modules, then to bolt them together. It is well established in software engineering that decoupling of software modules yields good implementations.

The presumption that we can do this is good in most circumstances... but not with the electric grid. The physics of electric power create hidden couplings and interactions even if we decouple the application software. There is no way to keep subsystems in their separate boxes.

I call these couplings "hidden" because:

·         They are not readily apparent to application programmers. Such hidden coupling does not occur in many other application domains and so can easily be overlooked when developing smart grid applications; especially since it requires power system domain expertise to recognize them.

·         They don't matter at pilot size. They don't do any real damage until the system reaches scale..

Demand response (DR) and volt/VAR optimization provide an example. Developers may act as if they can build those two functions independently. At scale, however, they interact in unfortunate ways. Those interactions may result in less DR than expected (up to 15% less), in voltage violations, or even in feeder level circuit breaker trips.

That is a simple "bilateral" example. In real life, hidden couplings are "multilateral." That is, each subsystem interacts with all the others. Put bluntly, we don't even know yet all the hidden couplings that could occur in the smart grid. But if you start to think about multiple interactions between DR and volt/VAR and distributed generation and electric vehicles and intermittent renewables and… well, you can see how complicated it may become. It is not an exaggeration to say that these multiple hidden couplings have the potential to create wide area blackouts if they occur during times of low stability.

Two strategies

Two strategies have the potential to help. The first is "observability." We must architect our grids to make it easy for each subsystem to see what's going on with all the others. In more technical terms, applications must be able to access all relevant grid state elements in real time.

The second is "federated controls." Just as we like to modularize our subsystems, we like to modularize their controls as well. But if all the subsystems are operated from their own separate silos, they will end up fighting and canceling each other out. Instead, they must be conjoined to account for hidden couplings and conflicting goals.

The smart grid has similarities to other digital networks. And much of the work around smart grid architecture has been derived from "standard" enterprise methods. But the smart grid also has crucial differences – differences so fundamental that we must architect our systems in different ways.

You may also be interested in ...

Interoperability and security for converged smart grid networks

Smart grid: Kurt Yeager on why we need one and what it's gonna take