|
This is more or less where the system is heading by default, unless something big changes. But the evolution seems less intentional than a matter of path dependence and lack of holistic planning.
Kristov, De Martini, and Taft worry that Grand Central is not the right model — that it will ultimately increase the cost and complexity of integrating more renewable energy and DERs.
The details get can get technical, but there are two basic problems with Grand Central.
The first is that DERs more and more often serve two masters. They have a relationship with the TSO that bypasses the DSO, in the form of wholesale-market commitments. They also have a relationship with the DSO; it must manage them in the name of distribution-grid stability and reliability.
As DERs and their aggregations grow more numerous and larger, the risk arises that large chunks of the system will receive dueling instructions. The paper’s authors call this “tier bypassing, which occurs when two or more system components have multiple structural relationships with conflicting control objectives.”
The second problem is simply complexity. DERs are still at a fairly nascent level of development, but they are set to explode in coming years, as rooftop panels, electric vehicles, home batteries, and smart meters become more common. Soon there will be all kinds of combinations and aggregations, at all levels, across every one of hundreds of LDAs.
Wholesale markets could go from having dozens of participants to having hundreds, or thousands, or hundreds of thousands.
That’s going to be a lot for a TSO to track — a thicket of new rules, new enforcement mechanisms, and sheer computational bulk. “Under this model,” Kristov, De Martini, and Taft write, “the TSO needs detailed information and visibility into all levels of the system, from the balancing authority area [i.e., the TSO level] down through the distribution system to the meters on end-use customers and distribution-connected devices.”
TSOs would have to track and manage all this information while working alongside, and attempting to coordinate with, dozens of DSOs maintaining local reliability.
Already some TSOs are complaining to FERC that state energy policies are distorting their wholesale markets. Imagine when those federally run markets involve thousands of DER participants, all of which are also subject to a variety of state energy policies and all of which are also constrained by DSO reliability requirements.
These are the kinds of thoughts that give FERC commissioners migraines. Balancing the interests of TSOs against the interests of dozens of DSOs will be an unending hassle.
Some economists like to think that if each energy source and service were priced properly, based on its real-time, location-specific value, the market would allocate electricity with perfect efficiency. Just get the right pricing algorithms in place and let ’er rip.
But there are reasons to doubt that distribution systems, filled with quirky and unpredictable human behaviors, can be adequately guided by the invisible hand alone. They need a more personal touch.
Kristov, De Martini, and Taft take no stand in the paper on whether the Grand Central model is possible, but when I asked De Martini directly, he was frank. “I don’t think the grand centralization model will work at scale,” he said, “as there are too many dynamic, random variables [in distribution systems] involving both machines and humans.”
“As I think about a TSO trying to have full awareness of what’s going on in a distribution system, bringing that together in a simultaneous optimization with the transmission grid, it just doesn’t make sense,” Kristov told me. “It seems needlessly complex. But if you don’t have that, then you need the DSO to step up to some higher-level responsibilities.”
Which brings us to the alternative to Grand Central.
A new, bottom-up architecture for the grid
The alternative grid architecture that the study’s authors propose solves these problems in an elegant way. It is called ... hang on to your hats ... a “decentralized, layered-decomposition optimization structure.” Whee!
Let’s translate that into English. (Side note: Layered or “laminar” structure is a familiar concept in telecoms and software architecture. It is somewhat newer to power systems.)
In the Grand Central model, the TSO optimizes everything in one place, not only power plants at the transmission level, but thousands of DERs and aggregations at the distribution level, in service of wholesale markets and transmission system reliability, while having sufficient real-time visibility into the distribution system to avoid conflicts with local reliability needs.
In Kristov, De Martini, and Taft’s proposed model — which I’m going to call LDO, for layered decentralized optimization, because I don’t want to type all those words again — each layer, the transmission layer and the distribution layer, would be responsible for its own optimization and its own reliability.
Remember tier bypassing? The LDO model would prevent that by effectively sealing the layers off from one another, except at their electrical interface points. The only point of communication and coordination between the transmission layer and the distribution layer beneath it would be at the TD interface (the substations). Everything below the TD interface would be managed and optimized by the DSO.
Responsibility “decomposes” to the layer beneath — that’s what “layered-decomposition” refers to.
The DSO would balance supply and demand within a local distribution area (LDA) using, to the extent possible, local DERs. It would then aggregate all remaining supply or demand into a single bid to wholesale markets (either a purchase or a power offer).
| 
