DR Baseline Methodologies Explained

A baseline is an estimate of the electricity consumption (load) that would have occurred at a facility during a Demand Response (DR) event if the facility had not curtailed its usage. It's the benchmark against which actual metered load during the event is compared to calculate the load reduction achieved (performance).

Accurate baselines are critical for fair compensation and program compliance. Different programs and markets use various methodologies:

Common Baseline Types:

• Average of Recent Similar Days (X-of-Y Baselines):
  This is the most common approach. It calculates the baseline for the event day by averaging the facility's load during the same hours on a selection of recent, similar non-event days.
  • Example: A "High 4 of 5" baseline (used in PJM) looks at the 5 most recent similar weekdays (excluding event days and holidays) and averages the load during the event hours from the 4 days with the highest consumption.
  • Example: A "10 of 10" baseline (used in CAISO PDR) averages the load during the event hours from the last 10 eligible similar days.
  • Variations: Different programs specify the number of days (Y), how many to select (X - e.g., highest, middle, all), criteria for "similar" days (e.g., weekday vs. weekend, weather conditions), and exclusion rules (holidays, previous event days).
• Adjustments (Weather or Same-Day):
  To improve accuracy, X-of-Y baselines are often adjusted to account for conditions on the actual event day that differ from the historical average days.
  • Weather Adjustment: Uses weather data (like temperature) to scale the historical average baseline up or down based on the event day's weather.
  • Same-Day Adjustment: Compares the facility's actual load *before* the event window starts on the event day to the average baseline load during those pre-event hours. The difference (additive or multiplicative) is then applied to the baseline during the event hours. This helps correct for unusual load patterns on the event day itself. (Common in CAISO, ISO-NE).
• Firm Service Level (FSL) or Contracted Level:
  Instead of calculating a baseline from past usage, the participant commits to maintaining their load *at or below* a specific, pre-agreed maximum kW level during events.
  • Example: California's BIP program uses FSL. A customer might commit to an FSL of 500 kW. Their performance is measured simply by whether their metered load stayed at or below 500 kW during the event.
  • Use Case: Suitable for industrial facilities with predictable processes that can reliably shut down specific equipment to reach a known load level. Offers simplicity in measurement but requires high confidence in meeting the FSL to avoid penalties. Used in some utility interruptible programs and ERCOT ERS.
• Regression Modeling:
  Uses statistical models (regression analysis) that correlate the facility's load with independent variables like time of day, day of week, temperature, humidity, occupancy, or production levels. The baseline for the event period is predicted using the model and the actual conditions during the event.
  • Use Case: Can be more accurate for loads with high variability or strong correlation with external factors, but requires more data and modeling effort. Sometimes used as an alternative baseline option in some programs.
• Real-Time Telemetry / Meter-Before-Meter-After:
  For very fast-responding resources (ancillary services), the baseline is often implicitly the load measured via telemetry *just before* the dispatch signal or frequency trigger. The reduction is the difference between the pre-event load and the load immediately after responding.
  • Example: ERCOT RRS measures the drop from the instantaneous load when frequency dips below the trigger point.
• Control Group (Less Common for C&I DR):
  Uses a statistically similar group of customers who are *not* participating in the DR event as a proxy for what the participating group's load would have been. More common in evaluating residential programs.