Taking Advantage of Amperon’s Net Demand Forecast on Snowflake

In modern electricity, net demand is an increasingly valuable concept for grid operators and market participants to understand and focus on. Net demand is needed to predict, plan, and manage the amount of supply required to meet the remaining demand after accounting for non-dispatchable generation, such as wind and solar. The data is particularly important across regions with high renewable energy penetration, like CAISO and ERCOT, as it helps identify how much additional capacity is needed from traditional sources to ensure reliability.  

Although net demand is a fundamentally simple concept to understand (load minus renewables), do not conflate simple with easy! Modeling, tracking, and leveraging this number is a different story. Its factors are increasingly in flux. For example, weather variables such as temperature, wind, and irradiation are inherently localized across time and space, and climate-driven weather events are becoming more unpredictable. Newly installed renewable energy capacity is growing, but distributed energy resources such as behind-the-meter solar are not well monitored or reported. Renewable generation is not only impacted by weather variability, but also subject to transmission constraints and economic and regulatory curtailment, and the complexity of real-time energy storage optimization is difficult to follow.  

Given this context, net demand is used to drive operational efficiency and strategic decision-making, but the use case is evolving. Market participants may use net demand to predict price movements, inform bidding strategies, optimize assets, manage risk, meet regulatory compliance, and strategize long-term investments. This is all to say that advancements in data analytics, market research, and rigorous modeling techniques are crucial to an accurate and reliable forecast. Accomplishing any number of these tasks requires integrating both forward and historical net demand data throughout your operations.  

Amperon recently made it easier to track and follow these predictions by revamping and sharing our net demand dataset through our partnership with Snowflake. Why? Because customer appetite for data is growing. We wanted to make it easy for customers to have a comprehensive set of forecast information, allowing them to fully leverage Amperon’s forecasts in their own analytic processes and workflows. We run an hourly fifteen-day forecast updated every hour to keep up with the fluctuations in demand, wind, and solar. This means there are 360 “prints” per forecast, which is defined by the unique number of regions and weather vendors we model. Today, Amperon’s coverage spans throughout North America and Europe, with zonal-level forecasts available in ERCOT, PJM, MISO, and SPP (with plans to continue growing).

Snowflake’s compute power and diverse toolset complements Amperon’s comprehensive views of net demand, where our community of users can analyze net demand, demand, wind, and solar predictions (individually or altogether), across every region and load zone available, without having to worry about complex SQL.  

We work with customers to engineer views by use case, with additional columns that make sense and add value. We have much more than the forecast history – we created columns for Actuals, Absolute Error, a Boolean indicator for Peak Net Demand Hours, time segments for Market Hour Ending and Beginning, the ISO’s Zone name, and weather variables.  We also include every prediction time for demand, wind, and solar separately, with predictions truncated to the hour, both the market and UTC timezone, upload timestamps, and “Days Before Market Date”, which help analyze our accuracy by days back.

Below are a couple of charts to illustrate how easy it is to analyze large amounts of data.  

In the first chart below, we see how the Absolute Error decreases, on average, as we approach the forecasted market date in ERCOT for 2024.  

This next chart shows the day-ahead net demand nMAE broken out by hour ending in ERCOT for 2024 so far. ERCOT has a notably higher error during HE 10, and overall, Amperon’s net demand forecasts are more accurate compared to the ISO for all hours.  

As a leader in electricity forecasting, Amperon’s data scientists are in a great position to accurately predict net demand. Amperon trains, updates, and monitors both its machine learning algorithms and fundamental models against a constant stream of big data, which spans thousands of weather points from our premium weather vendors and various market sources. We pride our accuracy on weather granularity, weightings, and performance optimization.

Whether you’re looking to analyze Amperon’s accuracy against sudden drops in renewable generation, check our improvement over the ISO, or simply pull our renewable forecast accuracy as of a certain time in the past, Amperon is here to support your journey. It’s our mission to provide you with a trustworthy forecast that is delivered reliably with data and explanations readily available, as well as offer service that is prompt, friendly, and insightful. We look forward to working with you!  

To get started using Amperon’s net demand data on Snowflake, contact us here.