Explaining how the Council forecasts load growth for the Pacific Northwest power system

When you need to see in the dark, which will illuminate the terrain better – a flashlight or an aerial flare? This spring, Council Power Division staff will be using the latter approach to produce a new 20-year load forecast for the future of the Pacific Northwest’s power system.

At March’s Council meeting, Senior Energy Forecasting Analyst Steve Simmons and Senior Power Analyst Tomás Morrissey explained their analytical approach and methodology to load forecasting (read presentation | watch video). “We’re going to try to illuminate a pretty large area,” Simmons said.

This was the first of a multi-part discussion on load forecasting this spring, which is a key component of the Ninth Northwest Power Plan. At a meeting on April 29 that will be hosted online, staff will present comprehensive results from the new load forecast for the Northwest.

The Council’s approach to load forecasting

As the chart above shows, growth in electricity demand does not travel upward on a linear trajectory. It ebbs and flows over time. This is true for both annual energy and peak demand. This demonstrates the need for greater flexibility in forecasting, and to capture a range of possible futures rather than one future plotted precisely on a graph. This is why the Council uses forecast ranges instead of an exact "best guess" number. This has been the Council’s approach ever since its first Power Plan in 1983. This method, innovative for the time, was developed at a moment when the Northwest’s power system had veered badly off course due to errors in load forecasting in the 1960s and 1970s, resulting in disastrous over-building of the region’s electricity generation resources.

While load forecasting computer modeling systems, analysis, and methodologies have all advanced tremendously since 1983, electricity demand growth still ebbs and flows over time in similar patterns. For the Ninth Plan, the Council's forecasts of the Northwest’s power demand will consider several possible trajectories, which capture and reflect a range of future uncertainties for how much electricity demand materializes on the Northwest’s power system, and by when. This range of uncertainty is core to successfully planning for the future.

Improvement in computer modeling, growing data complexity

At March’s meeting, Simmons noted that the goal is to create an accurate and comprehensive forecast of demand for electricity in the region across 20 years. To do that, staff needs to analyze the region’s current and historic energy use, which helps to gain an understanding of what might drive changes to future demand. That requires building a computer model. The demand forecast is an output from this model, which is highly input data driven and is getting more complex for the Ninth Plan.

In producing the 2021 Power Plan, staff continually bumped up against limitations to their old models’ ability to do long-term load forecasting, among other essential tasks in power planning. In 2023 staff contracted with Itron, a company offering energy forecasting software tools, to upgrade the Council’s long-term load forecasting.

Power system analysts now have the capability to produce annual, monthly, and hourly forecasts of load across 20+ years for the Northwest as well as for 13 individual utilities’ balancing authorities. They’ll also be able to include data from 27 weather stations in Oregon, Washington, Idaho, and Montana, which will allow staff to forecast changes related to weather conditions. Forecasts will represent residential, commercial, and industrial sectors, as well as for electric vehicles, data centers, electrification, and rooftop solar. Those added capabilities have made data inputs and management more complex and challenging, Simmons said. Staff is taking care to monitor quality and check for accuracy for all inputs going into the model to develop the load forecast.

Simmons reviewed data sources staff is working with for this load forecast:

Building stock – new and existing by type

• Units
• Square feet

End use technology, such as space heating or cooling in buildings

• Fuel type
• Unit saturation
• Energy efficiency
• Load shape

Economic conditions

• Population
• Employment income

Quantitative and qualitative data and analysis on industries and the tech sector, including data centers and chip fabrication facilities

Future weather

Electric Vehicles

• Registration & Sales
• Usage
• Load shape

Rooftop Solar

• Installations and shape

Load shapes and future demand trajectories

A vital part of the Ninth Power Plan will be to evaluate cost-effective energy efficiency and demand response potential and compare and contrast it with other resource options to meet future energy needs in the Northwest. Therefore, the initial load forecast will freeze efficiency at today’s levels and assume no demand response. This will result in a forecast that might be higher than actual long-term loads, or have larger peaks. For example, unmanaged electric-vehicle charging that often occurs in after-work hours can coincide with other peak hour power needs. Utilities pursuing demand response could manage the charging in several different ways that have less impact on power system peaks – such as after midnight. The initial load forecast will assume unmanaged charging, leaving the demand response potential of managed charging as an option to the model.

Near the end of the power planning cycle, once the Council has made decisions on how much cost-effective efficiency, rooftop solar, and demand response should be included, staff will re-run the load forecast to get the final version to include in the Ninth Plan.

Staff will also be analyzing three demand characteristics – magnitude, timing, and shape – for six key futures: weather, economic growth or stagnation, electric vehicles, data centers, building electrification, and hydrogen production.

• Future weather affects summer loads’ peaks and the timing will occur throughout the Ninth Plan’s 20 year horizon.
• Electric vehicles affect residential loads’ peaks and will have a large impact mid-way through the 20-year period. EVs will be a significant source of demand in some zones, such as Western Washington and Western Oregon, while not as much in others.
• Data centers will be single large loads that will come on early in the plan period. The profile will be flat. It will be significant in some zones but not others.
• Building electrification will affect winter loads’ peaks, but will have a larger impact late in the 20-year period.
• Hydrogen production will be a single large load with a flat profile that will also be late in the 20-year period.