DER Integration Is Not An IT Problem Part 3: Things to Know About the Problem Domain

Real time markets and IT do not solve these problems.

Grid architects use reference models to depict problem domains (this is different from enterprise IT, where reference models are functional decompositions). Our reference model is “a bird’s-eye view of a domain or problem space” (Meier 2011). Very broadly, a problem domain is the set of items the grid architect must be concerned with, work with, work around, or modify. In the course of the architecture synthesis process, we create both present state and future state reference models (see purple boxes in Figure 1).

Figure 1. Reference models in the architecture synthesis process.

Rather than present an entire U.S. present state reference model here, let’s look at a number of key items/issues related to DER integration that come from consideration of such a model.

Distribution systems are not structured like transmission grids

• Electrical physics applies across the board but the structure and operation of distribution systems is not the same as for transmission systems. Three-phase distribution feeders are generally unbalanced. Laterals may be one, two, or three phases. Most feeders are simple radials, although in urban and suburban settings variations exist, including dense meshes, partial meshes, open loops, and backfeed inter-ties.

• Many distribution systems have little or no distribution SCADA, and laterals are often fused (although new technology is gradually replacing fuses with more sophisticated devices and newer distribution control devices have built-in line sensing). Traditional distribution voltage regulation, protection, and power flow controls are not made for reverse power flows and so must be upgraded.

• A distribution feeder is not an infinite bus. This point must be taken to heart by those that think it does not matter what is connected to the distribution feeders. Many field trials are misleading in this regard due to low scale.

• A service transformer secondary bus is a dead short across the set of loads it serves. This matters when using inverters to inject power into the grid via a secondary. The same applies to two inverters connected to the same primary feeder phase(s).

• Many distribution feeders are structured as” large-wire/small-wire”, meaning that wire sizes decrease as you move away from the substation. Consequently, putting new large loads at the ends of feeders or attempting to inject large amounts of current near feeder ends is not feasible on these circuits.

• Many distribution feeders are already operating near maximum capacity and cannot take much additional load (like EV chargers) or significant injection without upgrade. Reconductoring is the answer; using residential DER and load turndown or usage time-shifting as non-wires alternatives is not.

• The closer you get to the distribution edge, the less you can benefit from the inherent smoothing of DER aggregation, so power flow volatility increases, in the limit it approaches the volatility of a single device.

Load behavior is complex

Distribution systems, substations, feeders, and feeder sections all have individual load curves that peak at different times. They do not align with each other and more importantly, do not align with the bulk power system demand peak. Whiteboard Flatlanders, take note of this peak demand divergence: trying to control loads or usage in a naïve cluster-coupled manner to accomplish precise system peak time shifting or shaving leaves the grid kinda screwed.

Don’t do this to the grid.

Rooftop solar PV, behind the meter storage, and remotely managed responsive loads induce power flow volatility that propagates into bulk power systems,¹ and also cause apparent load to be less than actual load. Lack of visibility of this effect at the system operator level creates issues in terms of planning ready/spinning reserves and maintaining balance. In addition, forecasting of rooftop solar is, shall we say, iffy. Before this can be considered anywhere near reliable at scale, somebody has to become orders of magnitude better at weather prediction. Nobody does it well enough now.

DER dependability/availability concerns and aggregator woes abound

Private DER and third party DER aggregators do not have the same obligations as bulk power generator owner/operators. Relying upon DER for stable system operation is problematic due to dependability issues. Homeowners can and do opt out of letting their home systems be commanded by aggregators or grid operators and do so without notice. DER operators and aggregators can and have exited their businesses abruptly.

In fact, most DER operators actually have not wanted to provide energy into the bulk system in the U.S. because they did not want to be held to the rigorous contractual, regulatory, and technical requirements of generators. They only want to provide apparent load reduction. As far as supplying grid services goes, the economics are not favorable to the owners of the DER. Grid services is a small pie, presently divided among incumbent suppliers. For DER aggregators to play, they must take market share away from the incumbents and divide it among their subscribers. Given that some DER like solar are not dispatchable, this leaves the DER aggregators at a competitive disadvantage to the incumbents and as the number of DER subscribers increases, their shares of the services pie decrease. Vast numbers of DERs? Well then, in the limit, as the number of DERs involved increases, compensation for each individual DER goes to zero. It is not worthwhile for the DER owners to let their systems and devices be used that way.

But DER aggregators face other hurdles in the U.S. Most aggregators that were tying to use residential DER have abandoned those plans for economic reasons. They have either exited the business altogether or have offered their platforms for sale to distribution network operators and told them that the distribution operators should run DER for themselves. The reason for this is that the aggregators discovered that after they paid to recruit DER subscribers² (essentially marketing costs), paid the installer and maintenance organizations, paid for access to the devices via the manufacturers, and could no longer get government subsidies, there was not enough margin left over for the aggregators to survive.

Consequently, U.S. FERC Order 2222 has largely become irrelevant.³

Had the foregoing not been the case, another problem of an architectural nature would have arisen. It has been a fantasy of the grid chattering class that there would be large numbers of competitive DER aggregators. If they were not to be given exclusive territories, then the interleaving of aggregator subscribers would result in entanglement coupling (a form of hidden coupling, which could change arbitrarily over time). This would have created significant coordination issues with implications for voltage regulation and feeder reliability when power is being injected into feeders from the DERs. This is an example of where the “feeder is not an infinite bus” issue matters. Dispatching active DER in aggregation clusters instead of taking DER electrical location and grid state into account (oh, wait, see Sophistries discussion later in this essay) would, how you say, have a high degree of fail.⁴ And, scalewise, it can get much worse than the diagram below illustrates.

Aggregator entanglement coupling creates messy coordination problems that the competing aggregators themselves cannot solve. Interleaving defeats aggregation.

If aggregators were to be given exclusive territories or if they managed to absorb other aggregators (shakeouts would inevitably occur), then they could develop market power. Also not awesome.

DER communication and device management problems persist

Dynamic management of DER on a large scale for support of grid operations is problematic. Communication connectivity (generally internet) to edge devices varies greatly in availability, capability, and reliability. Interfaces vary widely and when API changes cause devices to drop offline, device owners are often not able to deal with the necessary upgrades. In the U.S., control access for solar and storage inverters and home smart thermostats generally must go through the manufacturers’ clouds.⁵ Since the two main manufacturers do not make data available from the devices, this cuts off valuable sources of information that could have been used to support granular DER management, so as to deal with some of the technical issues listed above.

Sophistries of transactive energy, DLMP, and prices-to-devices schemes revealed

There are some technical challenges associated with transactive and other price-based schemes for DER management. Communications and interfaces limitations have already been addressed above. These technical challenges are real but are probably solvable in time. A bigger issue is that to make these schemes workable requires knowledge of complete grid connectivity and grid state everywhere on the distribution system but for many distribution systems, this is not available. This limitation is in the hands of the distribution network operators; the DER aggregators cannot change the deficiencies that exist in distribution grid observability. Distribution locational marginal prices (DLMP) schemes fail for this and fairness reasons. Single price-to-devices schemes (price broadcast as a signal) suffer from cluster coupling, which cannot take local conditions into account.

But the real problem for transactive and price-based DER management is that the underlying premise upon which these schemes depend is a fallacy. No rational consumer is willing to bid on electricity for home consumption every five minutes, so these schemes depend on consumers being able to know and express their marginal utility curves - in the worst case, for every electricity-using device they own. Automated approaches expect the devices to be able to have these curves to bid for electricity, and in some cases, even provide the curves to a centralized transactive or other market clearing system on demand.⁶

News flash: consumers do not know what electricity marginal utility curves are or how to figure out what their own are or should be. And never mind that these curves are time-varying.

Imagine consumers being asked to specify forward-looking marginal utility curves for their various appliances and passive loads, and detailing how the curves change with time and circumstances.

Riiiiiight - that could happen.

One transactive energy field trial attempted to solve this problem by providing home controllers that had a simple three-position manual selector knob, with positions for “Economical”, “Normal”, and “Comfort” operational options. Each option had a pre-programmed marginal utility function that had been created by the project designers. In other words, the researchers imposed marginal utility functions on the consumers that the researchers had made up. The consumers had no idea what these curves did or how they affected what the consumers paid for electricity. On every cycle, each home controller had to upload its currently-selected curve to a central clearing process which then solved an optimization problem across all the curves to determine a clearing price that was supposed to induce a particular DER response for the grid.

None of that was based on basic principles; the researchers created something that “worked” even though the hack that made it operate was just the project developers imposing their own preferred functions on the consumer systems in place of the unknowable consumer marginal utility curves. In one field trial, the transactive process proved to exhibit oscillating prices and had to be modified in an ad hoc manner to dampen the oscillations. No wonder nobody uses it.

Scaling issues emerge

Oh, did “optimization” come up in the conversation? Optimization invariably raises its ugly head in the context of managing massive numbers of DERs. If DER penetration is to become as large as the dreamers envision, scaling is a significant issue. Not just from the standpoint of transporting all the data (Claude Shannon can help with this), but also in terms of computation. Depending on the actual optimization methods used, computation times can suddenly break drastically upward as a function of the number of elements being optimized. In the world of DER imagineers, the numbers of elements is projected to be orders of magnitude greater than what is optimized for the grid today.

The breakpoint can be very sharp.

Plus, optimization can be brittle.

Final comment

The U.S. power grid is a shared public good, not just a big Tinkertoy for researchers who like to solve nail puzzles or would-be entrepreneurs who think that they will use the latest techno-fad to “disrupt” the electric power industry and so make themselves rich. Deep understanding of how things work and why they work is necessary to make grid changes, but this understanding has all to often been neglected in the rush to push poorly thought-out processes and gimmicks onto the U.S. electricity-consuming public.

1. Inspired by the well-known ”duck curve” issue for systems that have large amounts of solar generation, Lorenzo Kristov has dubbed active DER “ducklings.”

2. Some aggregators tried to get the distribution network operators to do the DER recruitment for them, but the network operators saw no reason to take on that cost.

3. This is both ironic and fortunate since FERC decided to mandate a very poor hybrid structure for enabling DER to participate in bulk wholesale electricity markets.

4. Having the aggregators disaggregate the DER data kind of undoes the point of aggregation, no? If the aggregators are to be independent competing entities, having them collaborate to solve to coordination issue seems problematic from a market perspective, so the coordination problem must be removed to a higher layer, thus invalidating a major aspect of aggregation.

5. In Australia, the creation and mandate of an open standard called CSIP-AUS has addressed the device access problem for new rooftop solar inverters, although this does not apply to inverters installed prior to the availability of the standard. Ensuring compliance with the new standard introduces an additional level of administrative effort.

6. I once asked a researcher how appliances were going to get the ability to do this transactive energy stuff and he replied that IoT (which in itself is not transactive) would take care of that. I told a friend this - she scoffed and said her fancy refrigerator cannot even make ice properly.