It’s time to evolve beyond demand response
When the concept of demand response (DR) first appeared some 40+ years ago, the complexities related to overcoming technical and market obstacles seemed insurmountable. Programs that started small with phone calls (or pages!) made by utility staff to commercial and industrial customers requesting a manual reduction in power consumption laid the groundwork for the automated systems available now.
Technology has improved since those early days and today systems are managed by utilities or third parties and use devices that receive signals to automatically curtail load. Government policy has been a key driver in advancing DR around the globe, and it continues to do so, as pointed out by the IEA in these recent developments:
-
South Australia is mandating certain air conditioners installed after July 1, 2023 to be demand response ready under the new Technical Regulator Guideline.
-
The European Union approved an action plan in October 2022 for digitalizing the energy system, which includes establishing requirements and procedures to facilitate data access for demand response.
-
Korea launched a new pilot program in December 2022, where intelligent appliances automatically respond to demand reduction requests instead of consumers’ manual entries.
DR systems have grown steadily to account for a small but important amount of system capacity. In California, demand response amounted to 3 - 4% of total system resource adequacy capacity (or about 1,760 MW) in July, August, and September 2021. While this is a small percentage, it can make all the difference when the total capacity shortfall is predicted to be 1,700 MW.
And yet, even with all the policy and technology advancements, the fact that in California the capacity only amounts to 3-4% is a little underwhelming. Do we need to do better to get more folks enrolled in DR programs or is it time to give up on DR?
How is demand response defined?
Before we get deeper into the topic, let’s begin with a mutual understanding of what exactly is meant by DR. The purpose of DR is clear: instead of relying on power generators (traditionally utilities) to respond to changes in demand (I.e., increase power generation), the demand is reduced to fit within generation capacity constraints. Reductions are usually achieved through pricing signals (incentives to reduce your energy use and/or penalties if you don’t) and these signals can be proactive, such as with time-of-use pricing, or reactive, such as during grid emergencies. Some folks also include ancillary services under the umbrella of DR, encompassing control of resources to provide non-spinning and spinning reserves and frequency and voltage support.
This is where things start to get muddy. Is providing ancillary services still a DR function if we’re talking strictly about load reduction? And how do resources like battery storage and electric vehicles (EVs) factor into the equation if they can act as both load and generation source? What technology is needed to manage the escalating complexity of this resource mix? At what threshold does an automated DR system start to become inefficient and begin to buckle under the stress of balancing thousands of resources simultaneously within physical network constraints? Wouldn’t a distributed energy resource management system (DERMS) offer a more robust and future-proof solution? So many questions!
Problems with demand response
To be clear, DR has its place and needs to stay. However, when it comes to being a silver bullet to solving the grid stability issues of the 21st century grid, it has several limitations and can’t be the only tool in the load management toolbox.
Availability
The availability of registered DR resources when they’re needed during an event is not guaranteed.
On high load days in the summer, about one-third resource adequacy requirements met by demand response capacity were not available or accessible to the ISO across peak net load hours. California ISO report: Demand Response Issues and Performance 2021
There are many reasons why DR resources may not be available when needed. These include structural limitations such as resources not being enrolled in DR programs on weekends, to issues related to energy market scheduling that make real-time dispatching unlucrative or technically unfeasible. Furthermore, it can be difficult to ascertain ahead of time whether a resource can deliver load curtailment in the amount it’s being paid for if it’s never done so in the past. For these reasons and more, there can be significant discrepancies between what’s recorded as DR capacity and what’s available in the moment it’s needed.
Long-lead times
While long-lead time DR resources are fine for day-ahead planning, they are less helpful during emergency and real-time situations. As the grid becomes more saturated with intermittent renewable resources, the ability to respond in real time to fluctuations is critical.
In September 2021, about 60 percent of supply plan demand response was registered with start-up times of 5 hours or more. California ISO report: Demand Response Issues and Performance 2021
Participation saturation
Going back to the 3-4% statistic, maybe this is the most participation we can hope for. From a residential perspective, programs can be difficult to navigate with negligible financial rewards. Plus, many folks don’t realize there are DR programs they can sign up for. This leaves only the most dedicated folks signing up for DR programs, which tends to be the minority.
Integration of behind-the-meter renewables and EVs
There is a lot of sophisticated modeling that goes on to determine and fulfill the needs of energy markets around the nation. Weather used to be a strong predictor of energy use (higher air conditioning use in summer and heating in winter), and while that still holds true, the impact of homeowner solar PV and battery energy storage means the formula is no longer as straightforward as calculating the anticipated load of a hot summer day. Today’s calculations need to factor in net load, which is the remaining load after solar generation has been subtracted from it. Making these calculations is extremely difficult in markets with a high penetration of solar PV because those systems are typically not under the control of the utility. The utility doesn’t always have an accurate handle on how productive the panels are (if at all) daily, let alone in real time.
EVs are another challenge because they don’t always charge in the same place, again making load estimation challenging. Once vehicle-to-grid (V2G) power flow becomes common place, that’s going to introduce a whole new dimension to net load forecasting.
Solutions that elevate DR to the next level
The driving force behind DR has provided a solid foundation of policies and technologies related to energy efficiency and grid management. Now it’s time to springboard beyond those foundational elements to tackle the challenges associated with a 21st century grid head-on. This means moving past the single dimension of managing load to a multi-dimensional approach that aggregates various technologies and is more sophisticated, efficient, and scalable.
This is where DERMS comes in. A DERMS that is aware of network constraints and manages individual points on the grid (think an entire household, complete with heat pump, pool pump, solar PV and EV) to net load is going to result in the most efficient and effective grid management approach. Such a tool needs to be able to forecast 24 hours in advance and be capable of managing deviations in real time. Many current DR systems issue control commands only once or a few times per day, which is not going to cut it once a large contingent of renewables and EVs cause deviations from the day-ahead predictions. For utilities and grid operators looking to get ahead of the increasingly more complex management issues associated with higher saturation of distributed energy resources, a DERMS is the most prudent long-term investment. The foundations of DR remain part of the strategy, but the strategy evolves to be one component of a well-orchestrated and balanced grid.