The evolution of the smart electrical grid and demand response

GS: Back in the day, I looked at the concept of a smart grid and thought it just makes sense to be able to have an ecosystem of energy assets operating in unison to balance supply and demand. Over the years I have simplified my view of this smart grid. At its heart I believe all it is, is the development and integration of a virtual power plant that can look, feel and act like a traditional resource to a system operator. With that said, a VPP needs to be flexible enough to support a very dynamic environment, so from a Hilo point of view, are you interested in specific technologies or is it more of a capacity game for you?

DSG: I think that really depends on what needs you’re trying to address, right? I don’t think it’s just about capacity, but more about how you orchestrate all the different hardware components you have under control to achieve the utility’s objectives.

In our case, the first product we tackled is winter peak shaving, where we have to follow a very precise control curve over a certain period of time. The shape of the curve, and how precisely we follow it, are as important as the raw curtailing capacity. Before and after curtailment is as important as what happens during the control period.

This a winter peak period product, so that led us to control line voltage thermostats in the residential space, HVAC in buildings and partner up with SensorSuite for centrally controlled rental building heating. So, we had to build out control models for each technology that would allow us to fit the control curve as tightly as possible.

Now, we’re already looking past that space to widen the control possibilities. We’re adding water heaters and smart car chargers to the mix, and we’re working on central/24V thermostats, battery systems and a wide range of devices in the commercial and institutional space to further add to the control capacity. The thing is, some of these technologies are not ready for prime time from the cost perspective.

Take batteries as an example. In my mind, there’s three large valuation categories for what they bring to the grid. You can limit costs by doing frequency control, support trading forecasts, support production forecasts, or peak power management for customers. You can defer investments by doing voltage control, freeing up reserve, through peak power, or do price arbitrage and help the trading desk.

In Quebec, the economics of batteries makes a lot of sense for large projects, but doesn’t make any sense right now for residential projects. We don’t have the same high energy prices as California or Australia that you know well Glen, and we have snow! So solar/battery combinations don’t contribute as much to the ROI of the projects. That means the utility has to foot 100% of the bill, which doesn’t make sense yet for our market.

To get back to products we’re introducing, as I said we started with line voltage thermostats, and we’re expanding into water heaters and car chargers as they free up capacity year-long, not just during the winter. We can introduce a bunch of other control products using these assets, which we can’t with heating.

There are also more practical aspects to consider. For instance, smart water heaters have a natural life cycle. They need to be replaced every 10 years. This creates an opportunity to introduce new technology and it unlocks other products in our virtual power plant, like frequency control, or mechanical inertia simulation to compensate for solar and other assets.

So I don’t think it’s just about capacity, it’s about the richness of your control options.

GS: The Quebec market is very interesting to me, a very clean one in terms of hydro availability and one that appears to be swimming in capacity. With that said, the grid we see today is going to look very different by the end of this decade. As an example, internal combustion engine vehicles are slated to be banned from sale in 2030. The large volumes of EVs taking their place bring with them their own unique demand curve and are likely to compound already constrained or stressed grid regions. When you add in the electrification of HVAC, process heating and, well, of everything, how do you keep that grid in balance? What challenges do you see as the electrification machine ramps up in Quebec?

DSG: To understand that, I think you need to step back and take a look at the two main parallel goals. The first one for me is how do you decarbonize society as quickly as possible—in Quebec’s case that means supporting the transition to electric vehicles, but also using hydro power plants to support the introduction of spikier renewables like wind and solar. And secondly, how do you maximize revenue and minimize investments during that transition. Solving the spikiness problem is key to answering that question.

You need consumption to be efficient, predictable and as linear as possible. That means you need flexibility in consumption and production habits, and you need to orchestrate behaviours to statistically balance out that production and consumption. In other words, you need a bunch of different assets with different electrical signatures that you can pull in for the different scenarios that you’re trying to control for.

In my mind, tools like virtual power plants are going to become mainstream so that you can pull from all the different attributes of connected assets to provide grid stability, and that’s while being as transparent to the customer as possible. We don’t have a choice, connecting all these EVs and renewables to the grid is going to invert the traditional distribution architecture, so that implies that you have to orchestrate all these new inputs.

So to answer your question more directly, the challenges are both technological and behavioural. People have to understand and accept the fact that modulating energy consumption in real time is the way to a cleaner future. That means all of the underlying technology needs to work smoothly—we need efficient software interconnections between aggregators and utilities, standard ways of communicating, forecasting, doing bid/ask, data streaming technologies, reliability and cybersecurity. . . I could go on and on.

What’s certain is that it will involve a lot of different actors and we’ll need to find ways of involving society to get there.

GS: That sounds like the blueprint of a modern grid for sure. With that said, I don’t see a lot of room for variability in that overview. “Efficient,” “predictable,” “linear” are all words that imply reliability is at its core. We all know that DR [demand response] has a somewhat checkered history, specifically behavioural DR. The reality is that when you ask someone to curtail load, they are making a trade-off, comfort for cash if it’s residential HVAC loads, and productivity for cash if it’s commercial loads. And they will decide based on their own self-interest rather than those of the grid. The point being there is someone you are relying on to activate the resource and they can be quite unpredictable. With the proliferation of IoT-enabled devices, technical solutions are becoming more and more prevalent. From your perspective, what is the better resource, behavioural or technical DR?

DSG: Let me just step back quickly in case the distinction between behavioural and technical isn’t clear to the people listening in.

In behavioural, users respond to external incentives, such as price signals to motivate themselves to change behaviour. That means they are creating the different programs and automations to minimize their costs—or they are acting out of goodwill because they understand the importance of minimizing energy consumption.

In technical demand response, those behaviours are automated or assisted by software or hardware, like through the use of IoT devices which can respond to commands from a cloud ecosystem.

For me, there’s also a third category, which bridges the gap between the behavioural and technical, and that’s coaching systems. We can learn from the user and the population as a whole and then use that knowledge to try to assist users in consuming more efficiently or detect faults or recommend changes to habits or equipment. Machine learning can help a lot with this scenario.

I think both behavioural and technical DR are complementary, but the technically driven aspects deliver a lot more value. Behavioural demand response can’t follow control curves, and it can generate secondary peaks. It also can’t orchestrate across a wide range of scenarios and it depends on goodwill. So all in all, that means these systems are more fragile and less predictable.

Maybe the last aspect is that technical DR can be built into a larger virtual power plant framework and that means that you can control resources which are less obvious to the user, like water heaters, batteries and vehicle-to-house or vehicle-to-grid. That’s a lot more complex than programmable thermostats.

GS: I like the coaching idea—opens up some interesting concepts. Gamification is one that I’ve often thought could improve engagement—you could loosely compare Ontario’s ICI program to a gamification style system. Having done my time in GA (global adjustment) busting I can tell you some of my customers thought it was a game, only problem was that they were playing the house and as we all know the house always wins. . . With that said, the program curtails some 1400 MW of capacity, clearly a sizable and valuable chunk of capacity for Ontario’s IESO (independent electricity system operator). In your experience, how does a utility value the DR resources you are bringing to market?

DSG: That’s a tricky one! My experience is that a lot of utilities often haven’t done the full end-to-end analysis of their comparative cost streams, and so have trouble putting a monetary value on DR, let alone frequency control or other more complex products. And that’s because they are often large entities, split into silos, and you need end-to-end data to get that to happen.

In general, I think there’s three large value categories for utilities, but in practice there’s a lot more than that. You have deferred investments, that applies to transport and distribution, and that’s mostly about right-sizing your equipment. In our case, we can reduce seasonal peaks and that means not having to scale your infrastructure to support a few dozen days of bad cold spells, reducing energy imports or maximizing exports. To defer investments, you can also apply control strategies to more specific geographies, like a neighbourhood, to manage peaks in that area. That can mean delaying replacing aging or undersized infrastructure in the distribution network.

You can also better manage coming back from power cuts by avoiding the huge synchronized loads that happen then—this’ll especially be true when everyone charges their cars. This strategy can reduce damage to equipment, or delay the need to deploy larger equipment.

So first one is deferred investment. The second one is freeing up contingency reserves and that’s done through better frequency management or very-short-term assets like batteries. That allows you to sell that power elsewhere. This one is tricky in that you really have to trust your DR system to be reliable, otherwise you can crash your grid.

The third one is about arbitrage and maximizing your revenue, which basically means providing control products that a trading desk can pull from. I think this is analogous to what trading desks did in the ’90s and 2000s. They built up low latency interconnects between financial hubs to make trading as efficient as possible. In my mind, virtual power plants provide the same type of breakthrough and they will be a game changer to leverage spot price differences across the energy landscape.

Just thought of a fourth one as I was speaking. I think there’s competitive advantages as well that can be drawn from better grid management. If you manage your grid more efficiently, and that means not having to build a new plant to provide peak demand, you can pass those savings on to consumers and be more competitive in the market.

As I said, there’s a lot more ways to value this, but these are probably the main ones.