Demand Response Baselines

Demand Response programs allow electricity consumers to reduce or shift their electricity usage during peak demand hours or grid emergencies. In return, participants receive incentives for event performance and/or capacity provided. Measuring performance for generation resources can be done simply by metering the energy they export to the grid; for demand response resources, however, performance is determined by curtailed load. This cannot be directly measured, so program administrators must instead compare the actual facility load with an estimated load that the facility would have had in the absence of a demand response event - this estimated load is called a baseline, also known as a Customer Baseline Load.
 

Core Principles of Baselines

Baselines strive to adhere to three key principles:

• Accuracy: the baseline methodology must accurately estimate what the load would have been in the absence of a DR event. Customers should receive credit for no more and no less than the curtailment they actually provide.
• Integrity: the baseline methodology must not encourage irregular consumption, and irregular consumption should not influence baseline calculations.
• Simplicity: the baseline methodology should be easy for all stakeholders to understand and calculate, including end-user customers.

Types of Baseline Methodologies

There are numerous types of baseline methodologies that follow different measurement philosophies and aim to improve accuracy for demand response resources with different load patterns. These methodologies include but are not limited to:

• Day Matching: uses interval meter data from the most recent “like” days (excluding exceptions such as previous event days).
• Weather Matching: uses interval meter data from non-event days with the most similar weather conditions to estimate the baseline for the event day.
• Control Groups: measures performance by comparing event participants against an un-dispatched control group with nearly identical load patterns and weather patterns.

We will focus on day matching baselines, the most widely adopted methodology for demand response programs across North American ISOs.

How Day Matching Baselines Work

Day matching baseline methodologies generally follows these steps:

1. Identify eligible baseline days that occurred prior to an event. These commonly exclude previous event days unless there are not enough eligible days. Matching day types (e.g. weekday, weekend) are also often considered.
2. Average hourly loads across the baseline days to generate a baseline. Most methodologies use a simple average, but some may use a weighted average that increases the weight for more recent days.
3. Calculate the load reduction as the difference between the baseline and the actual electricity use for each event hour.

There are many attributes in the calculation that vary between ISOs and utilities, intended to optimize for different event types and durations. One of the more important attributes is the same-day adjustment, which accounts for day-of conditions that might not be captured in the baseline. Adjustments can be scalar (multiplicative) or additive and typically look at the 2-4 hour span prior to event start, with a 1-2 hour buffer right before the event. Suppose that in the adjustment window before an event, the average baseline load was calculated as 100kW and the actual average load is measured as 120kW. A scalar adjustment would scale each hour in the baseline up by 20%, whereas an additive adjustment would add the difference—20kW—to each hour in the baseline. To limit the magnitude of adjustments, methodologies often implement a cap between ±20% to 40%.
 

Baseline Calculation Example

To illustrate the full process, we will calculate the day matching baseline for an event using NYISO’s Average Day CBL method, including the weather-sensitive adjustment (example adapted from the NYISO Emergency Demand Response Program Manual). Suppose that a weekday event is called between 12PM (HB 12) and 4PM (HB 15) and ten eligible baseline days are identified with the following hourly interval data. The event hours are highlighted in yellow. HB stands for hour beginning. All load values are in MWh.

The Average Day CBL methodology filters for the top five eligible baseline days by average event period usage. This is intended to prevent the baseline from being dragged down by uncharacteristic low-usage days.

The CBL is then calculated by averaging the load at each event hour:

With the weather-sensitive adjustment, the above CBL will be adjusted based on the actual usage in the two hours beginning four hours prior to the event start. This is done by calculating the average load in the hours beginning 8 and 9 over the five baseline days.

Now, suppose the actual load on the event day is as follows:

The average usage in the adjustment window (HB 8 and 9) is 4.5. The gross adjustment factor is 4.5/4.2 or 1.07. The CBL is therefore adjusted upward by 7%. The following table shows the final weather-adjusted CBL and the computed load reduction for the event period.

Day Matching Baselines in ETB

ETB Developer can model and visualize day matching baselines for supported grid services programs. If your proposal has a program that uses a day matching baseline (e.g. DLRP), the ETB Analytics modal will include a Grid Services Program Baselines graph below the Grid Services Program Events graph.

The event and baseline graphs visualizing a DLRP Immediate event.

The baseline graph shows the hourly baseline values for a program as a dashed line. The baseline reduction in each 15-minute interval is calculated as the difference between the baseline value and the net demand and is shown as a green or red bar to indicate positive or negative load reduction, respectively. Note that by default, the bars show the net load reduction for all concurrent events in a 15-minute interval. To isolate the load reduction for a specific event type or types, click the legend items to toggle event types on and off.

If the baseline method for a program does not include exports or does not consider negative load reduction, these are accounted for in the graph. However, the shown baseline reduction may not represent the program’s full calculations, which may be more complex. For example, DLRP Immediate events that start between 8AM and 6PM use the average hourly load reduction in the best four of the first six hours of the event for performance factor and reservation payment calculations.