by John Rossi, Comverge
For decades, residential demand response programs provided peak load reduction by remotely controlling HVAC systems and other high energy-use appliances. The incentive for customers was most often a fixed payment in exchange for giving utilities the ability to control their appliances for a predefined number of hours per year. Utilities installed at customers’ homes remotely controllable devices-either switches or thermostats-that provided predictable load reduction with a known load shape for the duration of an event.
During the past few years, customers have started buying Wi-Fi-connected thermostats that offer convenient Web programming and potential opportunities for heating and cooling bill savings. The Wi-Fi-connected devices also provide electric utilities with a resource they can use for demand response programs. This could reshape the program economics because participant recruitment and device installation costs are removed, making implementation significantly less expensive than the traditional model. Generally these consumer-owned devices, however, come with some use restrictions not found in earlier utility-sponsored programs.
We call the approach of using these customer-supplied devices for demand response “bring your own device,” or BYOD. Let’s quantify the factors that affect the value of third-party devices to fairly compensate for the load these resources provide. This will enable electric utilities to use the connected devices cost-effectively in their service territories.
What Drives Demand Response Value?
The starting point for this quantification should be the value that the devices bring to utilities or independent system operators. Several factors drive this value:
1. Program load reduction must be predictable.
2. Program load must be reliable.
3. Program load should be available for the maximum practical number of hours per day and year.
Predictable Load Reduction
Because electrical power must meet demand in real time, utilities and ISOs continuously balance supply and demand. For demand-side resources to contribute to balancing, utilities require insight into the quantity of load available from resources at any time. It’s also important to consider the amount of lead time required to dispatch the resource.
Resources with less lead time are more flexible and more valuable than long lead time options. Also, if the lead time is short enough (e.g., less than 10 minutes), the resource could provide ancillary services benefits in addition to capacity and energy benefits. With predictability comes the shape of the load during the period of use and recovery after an event when load snapback occurs. For example, a demand resource that exhibits a net load reduction over some time period but does so with periods of high reduction combined with periods of low or even increased usage would not be as valuable as a flat load shed throughout the event because these variations must be offset using fast-responding regulation or similar high-value assets. Also, some devices or offers to customers might present restrictions that hinder load shape predictability. A thermostat program that precools a home before an event or has temperature rise restrictions (e.g., temperature will rise only 2 degrees) affects the delivered load shape. The questions are: How can we use this type of resource to our benefit; and what is a fair valuation for this resource?
Finally, automated load dispatch is more predictable than load that relies on residential customers’ taking manual action at a certain time. This is a major reason why markets such as PJM require automated load for capacity programs.
After a resource is scheduled, a utility must be able to rely on the resource to deliver the quantity and load shape committed by the resource.
In the words of one utility control room operator, “I’d rather not have the resource at all if I can’t count on the delivery. If I didn’t have it, I’d procure an alternative in advance. But if it doesn’t show up, I have to procure an alternative in real time.”
Another factor in reliability is the ability of a customer to opt out of an event. It’s a marketing plus if opt outs are available to customers, but a long history of performance is required to quantify and account for this opt out probability.
Maximum Practical Availability
In addition to dispatch lead time, the other time parameter that influences value is the number of available hours provided by the resource. In the world of residential HVAC control, the availability is generally the peak summer period.
The variables include:
▪ Number of event hours per day;
▪ Number of successive days; and
▪ Total hours per season.
Each factor influences the real value that the resource provides. Given all this, how can we put a fair value on these parameters of predictability, reliability and availability? In supporting residential demand response programs for decades, both as a supplier of devices and control software to utilities and as a provider of pay-for-performance programs, Comverge has amassed a database that can be used to put relative values on the parameters of interest.
In the residential demand response space, there are two predominant methods of assessing load drop provided during an event. The first of these is a sampling methodology, where two similar groups are formed and one is exercised during an event, while the other serves as a control group. The load drop is then the difference between these two groups multiplied by the population of participants. This approach works well for a population where similar control methods and similar program rules exist for the entire population (e.g., a program where the population in controlled by switches or thermostats that use the same predefined algorithm). Unfortunately, this technique is not well-suited to assessing load from multiple types of devices that may use different control methods and whose customers may have different options (e.g., some may be able to opt out easily, others may not).
Therefore, we are left with the second method, which compares a customer’s usage to his own prior usage on similar days. There have been numerous attempts at defining a baseline methodology that gives an accurate estimate of demand reduction for the population during events. They entail averaging usage over preceding days and adjusting based on the current day’s usage in the hours prior to an event. This adjustment can be either up or down. But in general, it is a method of compensating for higher temperatures on an event day. This scaling feature is important to accurately assess the load drop contribution.
Payment Parameters for BYOD Providers
The guiding principle is to pay only for the delivered performance of the resource. This is especially important if the resource is a capacity resource where system reliability depends on delivery as promised. The actual valuation of each parameter is a function of the specific market in question.
The following explains the rows in Figure 1:
a. Pay for performance. Criteria should be based on load shed measured against a baseline scaled for the day of event temperature (see Calculating Baselines subhead).
b. Availability.Recognition that quick-responding assets (such as residential air conditioning) offer more optionality than slower resources.
c. Capacity payment. If the resources are eligible for a capacity payment, then delivery as contracted is necessary to help ensure system reliability. A common structure to enforce this delivery is a sliding scale of payments. As an example, resources could be paid on a linear scale when delivering between 90 and 110 percent of the quantity bid. At levels between 75 and 90 percent, the resource is paid half of the bid price; below 75 percent, no capacity payment is made.
d. Load shape. A measure of how flat a profile the asset provides throughout an event. A resource that deviates from a flat profile during an event by providing more energy than anticipated at some times, less than anticipated at other times, stresses other resources to compensate for the variation. This requirement is intended to incentivize a flat load shape on at least an hourly basis.
e. Portfolio.A list of active end customers must be provided and locked down prior to the delivery month. The performance will be judged based on the aggregate performance of this defined population compared to the aggregate baseline for the same population. No additions or deletions from the list can be made during the performance period.
f. Energy payment/penalty. Compensates device providers at the contract amount for load drop during an event. The amount of the load drop is measured against the baseline for the aggregate population. If the load increases against the baseline during an event, the device provider will pay as a penalty twice the per kilowatt-hour energy price for usage above the baseline. This requirement compensates the utility for having to procure replacement energy at the market price for the gap in delivered energy.
Bring Your Own Device Aggregator
Taking full advantage of existing connected devices adds considerable complexity to the operation of a utility demand response program. Historically, demand response programs used switches or thermostats that acted the same and cycled load to provide a predictable, flat load throughout an event.
With these new BYOD-connected devices, the program rules must be developed and administered for multiple device providers, each of which may have different performance characteristics or restrictions. Customers for each device provider must be registered and a performance baseline by customer constructed daily. Events must be dispatched given the constraints agreed to with each device provider, and post-event settlements must be computed and disbursed. This is much different than when a utility went to a computer, scheduled an event and pushed a button-and this is the easy part. The complicated part is blending the resources from multiple sources to provide the desired load shape while using the most cost-effective combination of resources.
So how can utilities cope with this complexity and still cost-effectively use this new third-party load? The answer is to work with a BYOD aggregator that understands the complexities of managing two types of device providers (called directly controllable and bulk provider here).
Directly controllable devices. A directly controllable device provider is one that allows the BYOD aggregator real-time access and control over the load control devices in customers’ homes and access to device status and settings for setting a baseline, controlling during an event and analyzing performance after an event.
In this directly controllable case, the BYOD aggregator would bear the responsibility of bidding load into utility programs and ensuring device performance during an event. The directly controllable provider still would have access to the device and the ability to offer other programs to customers, and the device would carry the provider’s brand name, etc.
Bulk provider. A bulk provider would contribute load on a set quantity basis. The bulk provider would offer a defined amount of load and be responsible for estimating and delivering this load from the defined population. The BYOD aggregator would register the participants, calculate adjusted baselines from meter data and evaluate the performance after an event. The contract between the device provider and the BYOD aggregator would mirror a contract between a large industrial customer and a curtailment service provider, where the performance risk is the responsibility of the industrial customer.
Because many bulk providers lack experience or expertise in estimating load availability for a given control scenario, a skilled BYOD aggregator would help provide this input. Also, the BYOD aggregator function is critical in blending the offers from multiple bulk providers with the directly controllable load to provide the desired load shape at the required times.
A capable BYOD aggregator also can attract bulk providers by structuring a risk profile that provides income potential with minimal risk of customer backlash. One could imagine that a cable provider that sells thermostats, for example, would not wish to risk customer attrition for the energy payments available; however, a BYOD aggregator could structure a low-impact approach that provides some revenue but little risk of losing high-value customers. The details of working out these individual arrangements with bulk providers and with directly controllable device providers are a natural role for an experienced BYOD aggregator.
The trend to customer-provided communicating devices with the potential for demand response offers an opportunity to lower the cost of utility and ISO demand-side management programs. These devices, however, must receive compensation commensurate with their value. Proper baseline implementation and pay for performance rules, which include penalties for underperformance, are needed to ensure that these resources are compensated fairly for their contributions to capacity and energy markets.
Finally, the complexity of understanding load potential and choosing the most cost-effective combination of resources for a given event puts a premium on finding a BYOD aggregator with the background and experience to perform this task.