Load Forecasting

Predicting Electricity Demand Supports Grid Reliability & Efficiency

Predicting the demand for electricity helps ensure that consumers have a reliable supply of power today and in the years ahead. Making these predictions – called load forecasting – is a job PJM does routinely.

Load forecasting helps PJM make decisions about how to operate the bulk electric system and plan for upgrades. The goal is to ensure an adequate supply of power for reliable service at the most reasonable cost. PJM members also use forecasts to make decisions about competing in power markets in the near term and about investing in new plants and equipment over the long term.

Some factors in load forecasting are predictable, such as power usage in large data centers that operate around the clock with consistent usage patterns. Other factors require forecasters to adjust their models to reflect unexpected changes in human behavior, as with the reduction in demand caused by the COVID-19 pandemic – or the dramatic increase in demand caused by the rapid growth of data centers and artificial intelligence in recent years.

A New Era of Electricity Demand

For much of the past decade, PJM's peak demand forecasts were declining. Shifts in the economy, more energy-efficient appliances and the spread of rooftop solar panels all reduced the amount of electricity consumers drew from the grid.

That trend has reversed.

PJM's 2026 Long-Term Load Forecast – released in January 2026 – projects average annual summer peak growth of 3.6 percent over the next 10 years, compared to roughly 0.3 percent in 2021. In absolute terms, the forecast projects the region's summer peak demand rising from approximately 154,000 megawatts today to more than 222,000 megawatts by 2036 and more than 253,000 megawatts by 2046 – an increase of nearly 100,000 megawatts within 20 years.

What is driving the increase? The primary driver is data centers – the large computing facilities that power internet services, cloud computing and artificial intelligence. PJM's board of directors has estimated that of the approximately 32,000 megawatts of additional peak demand expected between 2024 and 2030, roughly 30,000 megawatts – more than 90 percent – comes from data centers.

Data centers are highly predictable users of electricity, which is one reason they have historically been easy to model. But their sheer scale has changed. Data center operators have submitted requests to connect tens of thousands of megawatts of new load across the PJM region in recent years. PJM has overhauled how it vets these requests, requiring operators to demonstrate realistic construction timelines and financial commitments before their load is included in the official forecast.

Other contributors to load growth include:

• Building electrification: More homes and businesses are replacing gas-powered heating and water heating with electric heat pumps, adding new demand to the grid – particularly in winter.
• Plug-in electric vehicles: As more consumers switch to battery-powered cars and trucks, their charging requirements place additional demand on the grid. The PJM region is projected to see millions of additional electric vehicles over the next decade.
• Manufacturing growth: New semiconductor, electric vehicle battery and industrial manufacturing facilities are coming online or expanding across the region.

Winter demand is growing especially fast. While PJM has historically seen its highest demand on hot summer days when air conditioners run at full capacity, the gap between summer and winter peaks is narrowing. The region set an all-time winter peak demand record on Jan. 22, 2025, during a period of extreme cold. PJM has raised its reserve requirements in part to account for this growing winter risk.

Long-Term Forecasting

Each year, usually in early January, PJM issues an updated long-term forecast for the PJM region. This report provides a 20-year load projection covering peak usage, energy consumption, load management, and data on distributed solar generation, battery storage and plug-in electric vehicles. Forecasts are provided for individual zones, load deliverability areas and for the RTO overall.

About load deliverability areas: Parts of the PJM region are connected to the rest of the grid by limited transmission lines. These areas – called load deliverability areas – may need to rely more heavily on local power plants to meet demand, which is why PJM forecasts them separately.

The long-term forecast supports two key activities:

Capacity Market: The long-term load forecast is a major input into PJM's Reliability Pricing Model (RPM), more commonly known as the capacity market. Through RPM, PJM holds auctions in which power plants, demand response programs and other resources commit years in advance to be available when the grid needs them most. Ensuring enough capacity is available to meet forecast demand – plus a safety margin – is the fundamental purpose of the auction.

Regional Transmission Expansion Plan: Planners use the long-term forecast to evaluate the need for upgrades or additions to the high-voltage transmission system – the network of large power lines that moves electricity across the region. This planning looks up to 20 years into the future. The 20-year horizon reflects a federal requirement, under FERC Order 1920, that regional grid operators plan transmission to meet projected long-term demand.

Short-Term Forecasting

PJM regularly prepares a short-term projection of power consumption. Forecasters generally focus on the upcoming week, aiming to keep their projections to within a range of 3 percent above or below actual usage. PJM prepares two primary short-term products:

Hourly Forecast: Looks seven days ahead. It is often used by members in planning their bidding strategies in the Day-Ahead and Real-Time power markets.

Five-Minute Forecast: Looks six hours ahead. This forecast is used by PJM's Security Constrained Economic Dispatch tool – a computerized system that selects the most economical mix of power plants to meet demand from moment to moment, while respecting the physical limits of the transmission grid. The five-minute forecast helps grid operators continuously balance the supply of electricity with the constantly changing needs of consumers.

PJM applies machine-learning technology to its short-term forecasting process – one of the first areas in which PJM has adopted artificial intelligence to improve grid operations.

Key Factors

To predict how much electricity consumers will need, PJM engineers study data from many sources.

Day of the Week: The load forecast can differ significantly between a weekday – when many people are at work or school and electricity usage is high – and a weekend day or holiday, when many businesses are closed and usage is typically lower.

Weather: Temperature, humidity, cloudiness and wind conditions are among the most significant factors in the short-term load forecast. A heat wave spurs consumers to run air conditioners, driving up demand for power. Extreme cold prompts heating systems to work harder. Moderate weather in spring and fall tends to reduce demand. PJM draws weather data from approximately 30 stations across the region.

Economic Trends: In preparing the long-term forecast, planners examine the state of the economy. The amount of electricity needed by commercial and industrial users is a major factor in overall demand. In a vigorous economy, manufacturers with electricity-intensive machinery will likely consume more power than when the economy is slow.

End-Use Trends: PJM planners examine the volume and efficiency of electric-powered equipment that consumers have in service and plan to install. This includes central air conditioning, heat pumps, lighting and major appliances – water heaters, refrigerators, freezers, washers and dryers.

Data Centers: PJM engineers now track requests from data center operators across the region, evaluating the scale, location, timeline and financial commitments behind each project before factoring the load into the forecast. Data centers tend to operate around the clock and consume power at a steady, predictable rate – but their aggregate size has grown large enough to reshape the entire forecast.

Building Electrification: As homes and businesses switch from fossil fuels to electricity for heating, water heating and cooking, total electricity demand grows –and the seasonal pattern shifts, with more demand occurring in winter.

Rooftop Solar: Solar panels and other types of generation installed on the customer's side of the electric meter can reduce the amount of electricity the customer draws from the grid. PJM estimates this "behind-the-meter" capacity and subtracts it from projected demand to develop accurate net-load forecasts.

Battery Storage: Customers and businesses are increasingly installing battery systems that store electricity – often solar-generated – and discharge it during high-demand periods, reducing what they draw from the grid. PJM now accounts for this in the same way it handles rooftop solar.

Plug-In Electric Vehicles: Charging a battery-powered vehicle requires a significant amount of electricity – in some cases equivalent to half the power needed for an entire home. As consumers purchase more plug-in electric vehicles, their combined impact on the grid grows. PJM planners factor projected EV growth into their forecasts.

How PJM Builds the Forecast

PJM's long-term load forecast uses 24 separate mathematical models – one for each hour of the day, for each of the 19 zones in the PJM region. Each model draws on historical load data, weather records, and projections from specialized firms that track economic trends, end-use equipment efficiency and clean-energy adoption.

To account for weather uncertainty, PJM runs the model using more than 30 years of historical weather data. This produces a range of possible outcomes – a median forecast, plus high and low bands reflecting hot or cold weather years. The peak demand figures PJM publishes represent the amount of electricity the region would need absent any reductions from demand response programs, rooftop solar or battery storage. These programs and resources are then subtracted from the gross figure to arrive at the net demand that the grid is expected to serve.

Why Forecast Accuracy Matters

Load forecasting is complex, dynamic and an important part of PJM's mission to supply reliable electricity to more than 67 million people across 13 states and the District of Columbia.

Accurate forecasts support cost-effective grid operations – a forecast that is too high could lead to keeping unnecessary power plants on standby, while one that is too low could leave the grid without an adequate reserve of power during emergencies. As the region enters a period of electricity demand growth not seen in decades, the stakes of getting the forecast right have never been higher.

The National Electric Reliability Corporation, which oversees grid reliability across North America, has identified growing demand and the pace of new resource development as the top reliability challenges facing the PJM region in the years ahead.