A utility’s portfolio dispatcher is heading into the third consecutive day of extreme temperatures. Another event may be needed, but the portfolio's programs have a limited number of dispatch days remaining.
Should they call today? If so, which programs and technologies within the portfolio should they dispatch? Given the hourly weather and system needs, how much load reduction will actually show up, how persistent will it be, during which hours, and where?
Typically, portfolio managers answer these questions with a modeled estimate that's never actually been checked against what the meter recorded. That’s a real gamble, given the costs and reliability at stake.
The Challenge with Modeled Estimates
Estimates usually come from approximating reduction by technology, averaging past performance, or relying on historic settlement results, none of which reliably matches what shows up at the meter. Program administrators, procurement teams, grid operators, and regulators all notice the gap, eroding trust in the forecasts, the programs themselves, and whether committed capacity will actually show up when needed. The result is often an over-dispatch of resources or over-procurement capacity to compensate. It's a real enough problem that CAISO's own market monitor has flagged demand response resources delivering barely half of what was scheduled on the state's highest-load days.
For utilities that rely on a DERMS/VPP partner to run dispatch, there's an added layer of skepticism: the forecast utilities are planning around is coming from the same party whose performance it's judging.
Programs with strong measurement have already solved the retrospective side of this by proving what happened after an event. What's missing is the forward-looking side, built with the same rigor. Whether a utility calls its own events or works through a partner, the real question doesn't change: how much can you trust the number in front of you before the event happens?
Forecasting, Built on the Same Foundation as Measurement
Recurve’s FLEX Forecasting performance from the measured results at the meter. It shows expected load and reduction for every hour before, during, and after an event, segmented across programs, technologies, vendors, geographies, and distribution nodes. That level of insight informs how and when to dispatch, and whether resources should be stacked to meet demand.
With FLEX Forecasting, program teams are able to:
- See a 14-day forward view of expected event performance, to ensure a program's limited event calls get spent on the highest-potential impact days
- Analyze forecasted performance hour by hour, including two hours of pre-event behavior and three hours of snapback, to further optimize dispatch strategy
- Break forecasts out by vendor, technology, geography, distribution node, feeder, substation, and more, so dispatch can be targeted to the needs of the event
- Test dispatch strategies by comparing different start times and durations
- Check any forecast against a meter-based measurement, whether it's generated in-house or coming from a DERMS/VPP partner platform
Now the portfolio dispatcher can determine whether today is the right day to call, and select the resources most likely to produce the required load shape. After the event, stakeholders can compare the forecast directly with measured results, building confidence in both future forecasts and the resources behind them, and giving procurement teams the validation they need for more efficient, cost-effective resource planning.
How It Works
FLEX Forecasting runs on the same inputs that make FLEX's measurement trustworthy: historical measured events, AMI interval data, enrollment and service point data, and the specific technologies and providers behind each enrolled resource, layered with weather forecasts and solar irradiance (GHI) data by participant location.
The baseline behind each forecast uses Recurve's meter-based OpenDSM methodology to build a forward-looking counterfactual: what a participant would be expected to consume under forecasted conditions if no event were called. Forecasts update daily, and because they sit on the same foundation as measurement, every forecast can be checked against what actually happened. That comparison feeds back into the model, sharpening the next one: a continuous loop of forecast, dispatch, deliver.
How FLEX Forecasting is Different
That continuous feedback between forecast and outcome is the foundation of FLEX Forecasting. It also sets itself apart by being:
- Meter-based, not modeled, so it’s rooted in how participants actually showed up
- Checked against real outcomes every time, not untested like most forecasts
- Independent, whether it's run in-house or through a partner
- Granular enough to inform dynamic dispatches, instead of portfolio-wide
That's what turns a forecast into a number people can plan around, and transforms the dispatch decision into one backed by actual data.
See It In Action In Your Own Portfolio
If your team is still calling events off an estimate that's not checked against the meter, it's worth seeing what a forecast built on real meter data looks like. Talk to our team and we'll walk through what it would show for your program.