Menu

• Main
• Research
• Writings
• Work Highlights
• Contact

3 Questions About Virtual Power Plants

Wyatt Makedonski

Published: December 6, 2024

In the DER Task Force, the following three questions were posed:

1. How far can VPPs get with status quo technology, regulation, policy?
2. What are the most promising avenues for VPP capacity growth?
3. What will it take for VPP capacity to serve 20% of peak demand and deliver distribution grid benefits on top of bulk grid benefits?

Below are some of my brief thoughts on this. I may expand these over time as I only put these together in a matter of hours.

General

On a general note, I have observed that most discussions about virtual power plants and distributed energy resources, including those impacting regulation and technology, are harmed by the fact that there is minimal consensus of what a VPP is and its properties. The term itself is vague. The DOE’s definition is loose and essentially calls any aggregation of DERs a VPP. This lack of shared understanding results in use of abstract terms to cover a wide scope of VPP concepts. For example, a “VPP” of rooftop solar panels producing power in an uncontrolled manner working under the local utility’s policies is very different from a VPP of stationary batteries distributed throughout a TSO’s load zone and participating in one of the TSO’s programs on a daily basis. Further, imagine a VPP composed entirely of electric vehicles participating in the local utility’s demand response program; given the exact same constituents of the VPP with the same purpose, the VPP behaves very differently depending on whether some of the constituent EVs are discharging or not (my company has done the math and run the simulations to prove this). When talking about VPPs, we have to be detailed and understand the properties of each and every aggregation of DERs given that the wide disparity in types and uses of VPPs leads to very different behaviors. My research overlaps with many of these topics. I intend to release two publications in 2025 proposing solutions to handling the ambiguity and diversity of VPPs and their behaviors; one is more abstract and intended for a more general audience, and the other is rigorous and technical (lots of fun math). You can read more here: https://makedon.ski/research. I am looking for collaborators, so, if anyone reading this is interested, I will happily partner with appropriate people conducting research in academia or industry. My contact is in my Slack profile and on my website.

1: How far can VPPs get with status quo technology, regulation, policy?

With respect to technology: For almost every VPP today, the technology is very basic. The VPPs are often pass-through entities, and the more “sophisticated” configurations are more or less just timed on and off switches. This leads to many VPPs exhibiting undesirable load curves (malformed, jagged, low reliability). The technological immaturity regarding the vast majority of VPPs harms adoption; this will become more evident in the coming years as VPP capacity grows. Jagged, unreliable, and unpredictable VPP load curves do not serve the grid well. Regulators are not wrong to punish VPPs and DER aggregations with undesirable loads, and it is their job to penalize unreliability. The status quo technology of VPPs, in terms of software, will not carry us for much longer. We will hit a wall when all of the VPPs’ aggregate capacity passes a certain threshold. Right now, current VPP managers are cruising by on the fact that there is still a relatively low penetration of DERs. The constraints posed by current technology will not just be evident at the aggregate level but will first become more apparent at the fundamental level, i.e. the distribution grid level. In more concrete terms, with enough DER capacity connected to a certain feeder at the distribution level to harm the physical properties of the feeder, there will be more pushback from DSOs/utilities. As I am seeing first hand, this is already evident with utilities upgrading transformers at a certain size and oversizing them. This is a mere band-aid on the problem, kicking the can down the road. The good thing with this large barrier being a software problem is that it is much easier to solve than a hardware problem (i.e. math and models are easier to create than finding a new viable battery chemistry). I propose solutions in my research to the main problems of VPPs that function in the short term and scale to the long term time horizon. I intend to freely release these solutions as research publications in the coming months. I talk more about them in #3.

With respect to regulation and policy: VPP regulation varies drastically by country, state, municipality, TSO/ISO, and even DSO/utility, so it is hard to have a single answer that fits all situations. Of the entities that have some sort of VPP regulation, it is generally rudimentary and a temporary solution to just prolong future decisions. Pretty much all current VPP regulation works under the assumption that load from DERs is a relatively small part of aggregate load at any given time. This assumption only lasts for a few years, depending on location. More dangerously, most regulations assume that DER loads are stable and uniform. This is not the case in the majority of the examples I list in the general note; instability and non uniformity increase with cross pollination of DERs in a single VPP and the intended use of the VPP. The lack of a shared understanding of VPPs, like I mention in the general note, hampers regulations that in turn harms DER adoption. Regulation of VPPs and DERs has to become more targeted and rigorous in order to be more positively impactful. It may seem contradictory, but I also think that regulation of VPPs and DERs could benefit from being more axiomatic and principled on the condition that those principles are well defined and built upon a rigorous set of mathematical axioms and tools.

2: What are the most promising avenues for VPP capacity growth?

PV, stationary batteries, and EVs are all growing quite rapidly. It is important to note that these growth rates are impacted by government policy at federal, state, and local levels over the coming years. There are a variety of more technical publications with forecasts and models predicting DER growth. I’m not familiar with their models or methodology, so I will not comment on them. The hardware for VPP & DER capacity growth is important as those physical devices are what is physically storing and moving electrons. One thing few people talk about though is the importance of software. Without proper software and management of DER loads, this will likely be a barrier to adoption. This is because of the undesirable loads that I discuss in #1. It can be argued that this is already true and obvious using California’s NEM 3 as an example; in this example, the solution is a mix of both hardware and software. I explain some of the reasons why software is important to VPPs in #3.

3: What will it take for VPP capacity to serve 20% of peak demand and deliver distribution grid benefits on top of bulk grid benefits?

Given the growth rate of DERs relative to the overall growth rate of all electricity capacity in the US, DERs are well on their way to reaching 20% of peak demand. Software will be key in this. Without the right software, there will be barriers put up in the name of reliability. The right software can significantly increase the capacity of DERs without installing an additional kilowatt of generation. That may sound strange as it requires an inversion in thinking. When people think of DERs, they think of “conventional” DERs like PV, batteries, and EVs; in other terms, they think of things that can produce power. Things that can consume power should also be viewed as DERs (with the caveat that they need some degree of flexibility). I call these “virtual” DERs. If we expand our thinking, then many more devices become DERs, i.e. flexible assets that can benefit the grid. The main example of a virtual DER, and the most promising with its massive untapped capacity, is building HVAC systems (the heating and cooling systems of buildings). I wrote more about them here: https://www.indermediate.com/i/136492740/distributed-energy. I include heat pumps as part of HVAC systems. A similar example of virtual DERs are water heaters. This example only scratches the surface. I explain that the aggregate loads of computers in data centers are virtual DERs here: https://makedon.ski/w/ai-data-center-load-growth. Another example are refrigeration units, both residential and commercial. Virtual DERs are not theoretical; at my first company, I turned HVAC systems into DERs and optimized them using machine learning aka AI (the field of AI is large and much more than just chatbots). When I say software, that includes optimization and AI which are critical to DERs at both a nodal and aggregate level, much more than most realize.

In #1, I explained that most VPPs have undesirable loads. That will cause increasing problems as DERs scale. Grid operators target smooth, stable loads that are predictable and deterministic. Software, including optimization, is critical in improving VPP load quality, impacting adoption. I solve this in my publication “Dynamic Optimizing Virtual Power Plants” where I propose methods, algorithms, and AI models for smoothing and improving VPP loads. As I discuss in #1, DERs are already starting to overload certain distribution transformers. This is a critical problem that will increase in importance. To my knowledge, no VPP today is topologically-aware of distribution constraints. I solve this problem and much more in another publication I intend to release in 2025. You can find more about my research here: https://makedon.ski/research. As I stated before, I am more than happy to collaborate on my research with qualified people. If they are interested, my contact is in my Slack profile and on my website.