Last week in 'Forecasting 101,' we explored foundational concepts in a world of relatively predictable patterns, or 'known knowns.' While those fundamentals remain crucial, the landscape is rapidly shifting. For many regions, the new reality involves navigating unprecedented challenges, where extreme weather events and massive demand surges are becoming increasingly common, placing immense stress on both the grid and forecasting teams.

Texas stands as a prime example of this volatile new era, having weathered several energy crises in recent years. In fact, even as I write this, ERCOT is bracing for potential record demand due to an incoming heat wave – a stark reminder that these extreme occurrences are the new normal. It’s clear this isn't just a fleeting trend. Heat waves, winter storms, and unprecedented demand spikes driven by factors like data centers and EV adoption are reshaping the energy profile. Welcome to the new peak – it’s not just higher, but fundamentally different from past surges.

In today’s post, we’ll examine the characteristics of these events and the evolving demand profiles, exploring how forecasters can adapt their strategies to improve prediction accuracy and overcome the heightened challenges this new environment presents

Defining the "New Peak"

The term "peak load" has always been central to energy planning, but its very definition is undergoing a seismic shift. Yesterday's peaks, often predictable and following established seasonal patterns, are being replaced by a "new peak" – one that is not just incrementally higher but qualitatively different. 

Drivers of the New Peak:

• Persistent climate change impacts: The scientific consensus is clear, and the energy sector is on the front lines. We are now consistently observing increased frequency, intensity, and duration of extreme weather events. Sustained, record-shattering heat domes in summer and severe, prolonged cold snaps in winter are no longer statistical outliers but recurrent phenomena that directly drive energy demand to unprecedented levels.
• Accelerating electrification: The global push towards decarbonisation is leading to a rapid electrification of sectors previously reliant on fossil fuels. The burgeoning adoption of EVs introduces significant new charging demand, often concentrated in evening hours. Similarly, the switch to electric heat pumps for residential and commercial heating adds substantial winter load. This "demand coupling", where heating, cooling, and transport all rely on the electric grid, fundamentally alters load profiles.
• Shifting load profiles and system dynamics: The confluence of extreme weather and electrification is creating entirely new demand shapes. For instance, summer heatwaves now often see evening peaks exacerbated as diminishing solar generation coincides with returning commuters plugging in EVs and sustained air conditioning use. This can lead to sharper ramp rates (the speed at which demand increases), higher overall energy consumption during peak periods, and a greater strain on system flexibility compared to previous, more rounded peak demand curves.

Characteristics of the new peak:

• Unprecedented magnitude: We are consistently seeing historical demand records being broken across various grids, pushing systems beyond previously modelled limits.
• Extended duration: Peaks are often no longer short, sharp events lasting a few hours. Heatwaves or cold spells can sustain exceptionally high demand for days, testing the endurance of generation fleets and grid infrastructure.
• Heightened volatility & uncertainty: The precise timing and scale of these new peaks are becoming significantly harder to predict, driven by the erratic nature of extreme weather and the evolving charging habits of a growing EV fleet. This reduces the reliability of purely historical-based forecasting models.
• Amplified compound effects: The true danger of the new peak lies in the increased probability of coincident extreme events. Imagine a scenario with record-high temperatures, unexpectedly low wind and solar generation due to specific atmospheric conditions, battery storage assets already deployed, and a handful of thermal generators on forced outage – such compound events, previously considered highly improbable, are becoming a key concern for system operators and market participants alike, demanding a more sophisticated approach to risk assessment and forecasting.

The ERCOT crucible: A case study in extreme event impact

Perhaps no other market in the developed world highlights the challenges of forecasting and managing these "new peaks" as starkly as ERCOT. Its unique characteristics – a highly liberal market design, its effective electrical isolation from the rest of the U.S., and its diverse, rapidly evolving generation mix – make it a veritable crucible for extreme event impact and a critical learning ground for the global energy industry.

Winter Storm Uri in February of 2021 remains a defining moment. While prolonged extreme cold was the trigger, the crisis exposed systemic vulnerabilities across the fuel supply and power generation spectrum. From a forecasting perspective, the sheer scale of demand miscalculation (both in terms of actual weather-driven load and the failure to anticipate mass thermal outages due to frozen equipment and gas curtailments) led to devastating physical and financial consequences, including extreme scarcity pricing that bankrupted several market participants.

The years following Uri have not offered much respite, particularly during summer. The summers of 2022, 2023, and even last year, 2024, saw ERCOT issue multiple conservation appeals. Prolonged, intense heatwaves pushed demand to successive all-time highs year over year. While widespread blackouts were averted, these periods were characterised by periods of extremely tight reserve margins especially when renewable generation wasn’t available. They highlighted the critical, yet sometimes precarious, reliance on record solar output during daylight hours and the challenges posed when wind generation underperformed expectations during evening net peak periods.

ERCOT challenges magnified by extremes:

The ERCOT market presents unique forecasting and risk management challenges during these extreme events:

• Forecasting net load volatility: With one of the largest renewable penetrations globally, accurately forecasting net load is paramount. Extreme weather can drastically impact renewable output (e.g., low wind during heat domes, reduced solar efficiency at very high temperatures, or winter icing on turbines/panels), making net load forecasting exceptionally challenging.
• Predicting scarcity pricing and ORDC Impacts: ERCOT's energy-only market relies on scarcity pricing mechanisms, chiefly the Operating Reserve Demand Curve (ORDC), to incentivize generation during tight conditions. Extreme events frequently trigger these mechanisms, leading to prices rocketing from typical levels to thousands of dollars per MWh. Forecasting the probability, duration, and magnitude of these scarcity events is vital.
• Managing extreme collateral and credit risk: The price volatility inherent in scarcity events translates directly into massive financial risk. Retailers, in particular, face huge swings in their cost of goods sold, necessitating robust forecasting to manage collateral calls from ERCOT and overall credit exposure.
• Localized impacts of transmission congestion: Even if system-wide generation is adequate, localized extreme weather can exacerbate transmission congestion, leading to basis blowouts (significant price differences between trading hubs) and further complicating procurement and hedging strategies for load in specific zones.

The experiences within ERCOT offer a stark preview of the complexities that other grids may increasingly face. Mastering the art and science of forecasting under such duress is no longer a niche skill but a core competency for survival and success.

Why traditional forecasting falters in the face of extremes

• Limitations of historical data: Extreme events are often, by definition, outside the range of typical historical data, making models trained on averages unreliable.
• Non-linear relationships: The impact of an extra degree or two of temperature at 95°F  is different from an extra degree at 110°F . Simple linear models fail here.
• Model assumptions: Many models assume normal distributions, while extreme events represent "fat tails."
• Inability to capture all factors: Difficulty in modelling the simultaneous occurrence of multiple extreme conditions (e.g., high demand, low renewable output, and unexpected generator outages).

Advanced strategies for forecasting the new peak

One of the key things we emphasized in last week’s post was that a good forecast depends on data. It’s hardly surprising that you need to have weather data, market data, customer data, and so on in order to get a more accurate forecast. So it shouldn’t come as a surprise that forecasting for the new peak is much the same. More data.

But not just any data. To overcome the varied factors that contribute to record demand, forecasters need more granular and diverse data inputs. The more weather forecasts that are available, the closer you can get to the real conditions on the ground. But even more useful are hyperlocal weather forecasts (beyond simple temperature – including humidity, wind speed at hub, solar irradiance, etc.). 

More advanced modelling techniques will also be necessary. Machine learning models, probabilistic forecasting, stress testing and scenario analysis: every additional tool in a forecaster’s arsenal is going to come in handy.

On the demand side, a big boost will be the growing amount of detailed customer-level data from smart meters, which will assist retailers in understanding how different households respond to different conditions. Machine learning will play a big role here by enabling dynamic clustering of similar households, automatically grouping those with similar consumption and adjusting those groups as new data comes in.

However, it’s not just about knowing demand levels. The critical variable becomes net load: the portion of demand that must be met by dispatchable generation after accounting for intermittent renewable output. Accurately predicting net load is paramount because extreme weather can create a "double whammy": simultaneously driving up overall energy consumption while potentially diminishing generation.

Beyond energy itself, a stable grid relies on a suite of Ancillary Services (AS) – such as frequency regulation, responsive reserves, and voltage support. During periods of system stress the requirement for these services often escalates to maintain grid reliability. This heightened demand, coupled with potential scarcity of available resources, can cause AS prices to spike. For retailers who often bear exposure to these costs, the ability to accurately forecast both the volumetric need and the likely clearing prices for key ancillary services during tight conditions is crucial.

Gorilla’s approach: Adaptability for any situation

The only certainty around the new peak is that there is uncertainty. Committing towards one path forwards just exposes you to risk that even more change is on the horizon. The safer approach is to work on core competencies that will allow a forecasting team to adapt rapidly to new conditions. 

Gorilla forecasting models are highly adaptable, able to be quickly changed depending on the desires of the team. The platform allows users to implement their own forecasting models, with infrastructure to either run a pre-trained model or re-train a specific algorithm with specific parameter configurations. Machine learning capabilities add the cherry on top, with the ability to perform dynamic clustering of existing groups and to enhance forecasting models through continuous learning.

For retailers in markets like ERCOT, enhancing short-term load and net load forecasts during extreme weather isn't optional—it's a fundamental requirement for survival. While some might attempt to persist with legacy systems and brute-force resource allocation, such approaches are rarely sustainable and offer little defence when faced with the ferocity of events like Winter Storm Uri, where even slight forecasting errors can prove catastrophic.

The Gorilla platform offers a decisive shift from this reactive posture. Engineered for speed and efficiency, it empowers retailers to achieve superior forecast accuracy without demanding exhaustive resources or relying on error-prone manual inputs. Our automated processes scale seamlessly to any required forecast volume. We support a variety of intervals, down to hourly or even 5-minute ERCOT SCED runs if available, and can scale down to the meter-level or up to the portfolio level. 

Critically, Gorilla enables you to conduct multiple scenario analyses at the frequency your business strategy dictates—not constrained by outdated technology. Retailers can predict financial risk in extreme weather conditions, modeling for a wide range of temperatures. This capability is vital for proactively understanding and quantifying your potential exposure to extreme price events, allowing you to regain strategic control over your market position, even when the weather is at its most unpredictable.

Conclusion

The "new peak" is a significant challenge that requires advanced capabilities. Retailers can no longer rely solely on past patterns; they must adopt more sophisticated, data-driven forecasting methods.

There are a variety of factors driving the new peak, including climate change, electrification, and changing demand profiles. Different regions will likely see different key drivers and the disruption that a region like ERCOT has suffered might not be the same in the rest of the US, or other parts of the world. Nonetheless, many of the underlying factors are globally relevant and it is likely that every electricity network and energy retailer will need to think about their approach to crisis periods.

Regardless of how a demand crisis manifests, with the right tools and approach, it's possible to improve preparedness and resilience even in the face of extreme market conditions. Forecasting will be essential, but there are common issues shared by forecasting teams around the world: limited historical data, outdated tools & models, and a failure to adapt to the onrush of new data. Gorilla's forecasting suite has been designed to enable flexibility, granularity, and automation of processes to ensure that teams can adapt quickly to any situation. Get in touch today to see how easy it can be to summit the new peak with Gorilla.