Forecasting energy demand during holidays can be tricky, as demand typically drops with fewer people working, reducing the usual weekday electricity consumption. But, as we know from years past, there can be exceptions. Trading energy before holidays can also be hectic with traders sometimes placing bids farther in advance to accommodate time off, meaning there is still lots of wiggle room for changes in forecasts. In this article, we will guide you through various demand patterns that could affect the holiday season.
Holiday & Weekend Energy Demand Patterns
In most markets, weekend demand patterns are predictable, typically showing significantly weaker demand than weekdays. This drop is due to lower commercial and industrial energy usage while people are home. On cooling degree days, energy demand often peaks in the evening when temperatures remain high, and households are actively using appliances and air conditioning. Conversely, during winter heating degree days, the demand peaks can often occur in the morning, when temperatures are typically at their lowest, and heating and household appliance needs are at their highest. However, evening winter demand peaks are also possible depending on the ISO.
The chart below shows how weekend demand in PJM can be around 10 GW lower than weekdays. Insights like these are invaluable for energy producers, consumers, and other market participants. For instance, financial traders planning weekend or early-week trades typically act days in advance, instead of waiting until the Day-Ahead market. Accurately predicting these demand patterns is critical to staying ahead.
At Amperon, our forecasts incorporate the latest load and weather data and leverage ML-based models to continuous train and calibrate to detect these diverse demand patterns.
Similarly, national holidays often bring lower demand, as many people are off work. When a holiday falls on a Monday, the preceding Friday also tends to see reduced demand, as people extend their weekends and take this day off.
However, extreme weather events can override all these patterns, driving up demand unexpectedly. This was seen in 2022 with Winter Storm Elliott which hit Christmas day, a Sunday that year, as well as the polar vortex over MLK weekend in 2024. During these times, temperatures reached chilling lows, which led to a spike in energy demand beyond typical holiday levels. Though power outages from Winter Storm Elliott temporarily suppressed demand, energy needs remained high wherever power was available. This trend of extreme weather coinciding with holidays in recent years has offered companies valuable data, expanding their sample sizes for handling such events.
In this chart, the polar vortex that hit Texas over MLK weekend in 2024 was much colder than predicted by about 5-7 degrees F. This caused demand to come in much higher than expected for a typical holiday weekend.
The Duck Curve Explained
One of the most commonly observed demand curves is the “duck curve,” characterized by peaks in the morning and evening, with a midday dip. This midday drop reflects milder temperatures and decreased household appliance usage while people are typically at work. However, some regions experience a deeper midday dip than others, due to behind-the-meter (BTM) solar generation. CAISO is a prime example, where widespread residential and commercial solar installations reduce midday demand significantly, as BTM solar generation satisfies a substantial portion of the energy needs.
The CAISO demand curve has shifted dramatically over the past decade due to increasing BTM generation, as seen in the graphic below from energy.gov. By contrast, PJM still has some distance to go before its midday curve matches the depth seen in CAISO.
Credit: https://www.energy.gov/eere/articles/confronting-duck-curve-how-address-over-generation-solar-energy
Here, you can see how the duck curve affects Indigenous People Day in California.
Lighting Load Demand Pattern
Another interesting demand pattern is “lighting load.” On overcast days, people rely more on artificial lighting, which raises midday energy demand beyond typical levels during peak solar hours when sunlight is not enough for natural lighting. Forecasting this is straightforward when a large weather system dominates a region. However, pop up afternoon thunderstorms impact demand differently. Typically, these storms are correlated with very hot temperatures which fuel these storms. When these afternoon thunderstorms occur, these will weaken demand as they cool off the earth. Overcast days, on the other hand, may lower overall demand due to milder temperatures but increase lighting load from artificial light use.
The chart below demonstrates how lighting load associated with atmospheric weather events can influence the duck curve in CAISO. During midday hours, CAISO typically experiences significant behind-the-meter solar penetration, which reduces grid demand (as illustrated by the duck curves above). Our machine learning models integrate new weather data hourly from multiple vendors to help provide the most accurate forecasts. Accurately predicting the effects of cloud cover on demand days in advance benefits all sectors of the energy industry, particularly energy producers.
For instance, natural gas plants often cycle off during midday when demand is low due to an abundance of solar energy. However, when reduced sunlight and increased lighting load drive up midday demand, these plants remain operational. Knowing this trend well in advance allows plant operators to secure sufficient gas at potentially lower prices, enhancing cost-efficiency and preparedness.
Amperon uses machine learning models to generate our demand forecasts, allowing our systems to learn from every event – expected or not. With a week La Nina expected to occur this winter, weather patterns could shift unexpectedly, and winter conditions may differ from the past few El Nino years. Read Amperon's Winter Grid Outlook to ensure you are equipped with the most accurate demand forecasts for the upcoming holiday season where anything can happen.