From Forecast to Action: How Utilities and Traders Are Operationalizing Weather Intelligence for Winter Reliability

Jim Robinson (Meteomatics):

Good afternoon and evening—thanks for joining us. This session is From Forecast to Action: How Utilities and Traders Are Operationalizing Weather Intelligence for Winter Reliability. I’m Jim Robinson from Meteomatics. We’ll dig into how data and weather intelligence are reshaping the energy landscape. I’m joined by Sharon Abbas and Candice Thompson. Candice will also run a live demo of MetX later.

A couple of quick notes before we start: if you’re watching without registering, please register so we can share the slides and recap afterward. And drop questions in the chat on LinkedIn, YouTube, or our platform—we want this to be a two-way conversation.

Sharon, let’s kick things off. We’ve got Ben Perry from Yes Energy and Matt Sargent from NorthWestern Energy. We’ll start with Ben.

Shifting Demand with Rising Temperatures

Sharon Abbas:
Ben is unpacking how rising temperatures are shifting electricity demand in regions like the Pacific Northwest—places that used to coast through summer. He’ll show how cooling loads are changing operations and trading strategies. Great to have you, Ben.

Ben Perry (Yes Energy):
Thanks for having me. I’m on the product team at Yes Energy, focused on forecasting. We produce load forecasts for North America, Europe, and APAC. I’ve spent a lot of time on how temperature impacts demand—and how that response has changed as cooling penetration evolves across regions.

I’ll cover three things:

1. how temperature affects power demand,

2. a technique we used to quantify summer temperature sensitivity over time, and

3. what we’re seeing across different regions.

First, the classic relationship: demand is high at very cold and very hot temperatures, and lowest around a mild “comfort zone” (roughly 60–65°F). Looking at ERCOT’s north-central region (Dallas area), you can see that bottoming-out point and how the response differs by hour and season—max daily temps in the middle of the day push that inflection warmer.

Now compare with Seattle City Light (via CAISO’s Western EIM data). Seattle has a cooler climate, a lower overall AC penetration, and a gentler warm-side slope—but the “turn-on” for cooling happens at a lower temperature. Normalize the curves and you see it clearly: for the same 75°F day, Seattle’s proportional response kicks in faster than ERCOT’s, despite far fewer homes with AC. That early sensitivity matters.

Two drivers are changing these relationships over time:

• AC penetration: Regions like Texas have been near 100% for decades, so year-to-year changes are small. In places like Washington and Vermont, penetration has been much lower and is rising. Small absolute increases can materially change demand sensitivity.

• Warming: Using 30-year rolling averages to smooth noise, we see July temps rising ~1°F or more in Dallas and Burlington over the last ~15 years. In Seattle, peak July temperatures are up almost 2°F, which is notable for peak demand.

To quantify changing sensitivity, we used archived versions of our short-term statistical load models (May 2023 vs May 2025 vintages). For each, we ran temperature-step scenarios (e.g., +1°F, +4°F, +7°F vs “normal”) for July 2026. Holding the same meteorology path fixed isolates structural changes in response rather than just “it was hotter.”

In Seattle City Light, the 2025 model shows a larger load uptick than the 2023 model for the same temperature steps—especially at +7°F—indicating rising cooling sensitivity (from more AC and behavior changes). Doing this across Texas (ERCOT North Central), Vermont, and Seattle for all months of 2026 shows:

• Texas: Minimal differences except in Jun–Aug, and even then generally <1% (already saturated AC).

• Vermont: Stronger summer differences (Jun–Aug), consistent with increasing AC penetration.

• Seattle: Broadly higher percentage impacts across all hours in Jul–Aug, with the largest summertime increases.

One nuance in Vermont: behind-the-meter solar can mask midday demand in summer, but the net effect still trends up because more AC is present overall.

Takeaways:

• Temperature–demand relationships change over time; they’re not static.

• Regions with historically low AC are moving fastest on the warm side of the curve.

• Keep models fresh: recalibrate variables, reassess weather stations and representativeness, and incorporate new demand classes (EVs, data centers) where appropriate.

Q (Sharon): How should utilities and market participants adapt their models to these trends?

Ben:
Update models frequently with the latest data; recalibrate your temperature variables and station selection. Track emerging demand types (EVs, data centers) and decide if they’re weather-sensitive or not. Don’t be afraid to reconfigure inputs or add new data sources if last year’s setup isn’t capturing current dynamics.

Taking Wildfire Risk Head-On (NorthWestern Energy)

Jim:
Let’s bring in Matt Sargent, meteorologist at NorthWestern Energy. Matt, you’ve been busy—walk us through your wildfire program and how MetX fits.

Matt Sargent (NorthWestern Energy):
We serve a vast territory (most of Montana, plus parts of South Dakota) with a relatively small customer base. Fire risk is concentrated in western Montana’s complex terrain, with rapid growth in the WUI (wildland-urban interface).

We’ve adopted a consequence-based approach. We segmented our system into Wildfire Weather Monitoring Zones—down to the circuit and division level—and modeled “what if” ignitions every 200 meters along our network, letting fires burn unsuppressed for 24 hours to assess consequence (structures, timber, agriculture, watersheds). Combined with asset health (age, splices, rebuilds) and daily Severe Fire Danger Index (SFDI) overlays, we can prioritize where to apply EPSS (enhanced protective settings) or, in rare cases, a PSPS (public safety power shutoff).

Scale is the challenge: 649 zones need continuous weather awareness. As a one-person meteorology team, producing bespoke forecasts every six hours is impossible. That’s where Meteomatics MetX comes in.

We integrated all zones into MetX’s visualization and alerting. I analyzed 15 years of 3-km hourly data per zone to define local percentiles for wind gust, wind speed, and relative humidity. We fed those thresholds into MetX alerting. When any forecast hour from now to Day 7 exceeds a percentile threshold for 6+ consecutive hours, MetX texts and emails me. That reduces 649 zones to a short, actionable list.

From there, MetX dashboards make it easy to assess context: wind, gusts, RH, recent precip, snow depth—everything operations needs for a go/no-go on EPSS or targeted shutdowns. It’s also executive-friendly; I can brief non-meteorologists quickly with intuitive visuals.

A recent event (Oct 31–Nov 1): MetX issued 16 alerts with an average lead time of 46.3 hours. 14 of 16 verified against our custom percentiles. That gave us time to pre-position crews and apply EPSS—practical, measurable value.

We’re also expanding our sensing stack: wildfire cameras with AI smoke detection and ~200 new weather stations across MT/SD and Yellowstone by end of next year. MetX ties the weather piece together and scales my bandwidth.

Jim:
What’s changed most this year?

Matt:
From 15-year hourly analyses: relative humidity is trending lower and wind gusts higher across much of Montana in recent years. That combination elevates fire weather risk. Integrating MetX alerts has effectively doubled our ability to anticipate and act—data-driven decisions instead of ad-hoc judgments.

Building It in MetX (Live Demo)

Candice Thompson (Meteomatics):
You saw the finished product in NorthWestern’s dashboards. Let me show how to build something similar in MetX.

1. Add assets (GeoJSON)—transmission lines, pipelines, rivers, whatever matters.

2. Style & stack layers for context (city labels, boundaries).

3. Define thresholds for key parameters (e.g., wind gusts in mph at 35/45/55/65). Order them most-severe on top and color-code (red → orange → yellow → green).

4. Use advanced thresholds combining multiple parameters (e.g., high wind gusts AND low RH), plus optional rate-of-change rules (e.g., temperature drop over six hours).

5. Animate the forecast to see when and where assets cross thresholds.

6. Add concise tables/plots alongside the map so operations can see timings and magnitudes at a glance.

All of this feeds into alerting so teams get notified with enough lead time to act.

Jim:
Seeing thresholds colorize along actual networks is a game-changer—fast, clear, and operational.

Seasonal Outlook (Heating Season)

Chris Hyde (Meteomatics):
For November through March, the backdrop is a weak La Niña, possibly starting near ENSO-neutral. Expect:

• Temperatures: Persistent warmth in the East/ERCO T early in the season; cooler risks building Pacific Northwest/Upper Midwest mid-winter.

• Precipitation: Above normal in the Pacific Northwest/Columbia Basin and Great Lakes (lake-effect uptick), below normal across much of the Southeast and portions of the Southwest/ERCO T.

Volatility is the theme: a roller-coaster with frequent swings. Specific risks we’re watching:

• New York City (Nov): Leaning warmer than normal overall.

• Chicago / Western PJM / MISO (late Dec): Cooler risk vs baseline.

• Grand Coulee / Columbia Basin (late Dec–early Feb): Colder than normal risk plus above-normal precip → increased snowpack potential.

• ERCO T: Dryness enhances warmth, especially for highs, if the pattern cooperates.

We’ll keep monitoring and update as needed.

Jim Robinson:
Thanks to Ben, Matt, Candice, and Chris—and thanks to everyone who joined. If you haven’t registered yet, please do so we can send the slides and recap. Follow us on LinkedIn and YouTube, and reach out if you want a deeper, industry-specific MetX demo. Have a great evening!