Anticipating energy demand is science — and art

Providing electricity to a community is a complex business, and forecasting energy demand may be one of the most crucial tasks a utility faces to keep reliability high and rates affordable.

It’s a complex science, with an element of art. Accurate forecasting is critical to projecting revenues, which impacts not just CPS Energy, but it’s owner, the city of San Antonio, which receives nearly $300 million annually.

Because there is no large-scale storage of energy, that which is needed must be produced and delivered in almost near-instant time frames. If the system gets out of balance, the entire energy grid could collapse. That’s why determining demand — for the next day, the next year and the next quarter century — is so critical.

The consequences of poor planning can be immediate; without enough power CPS Energy would have to buy it on the open market, where it can cost exponentially more than if the utility had generated it. And, consequences can be long-term, if not enough capacity or infrastructure is built to meet demand.

Long-term forecasting —  using the data from monthly and even hourly variables over 35 years —  is crucial to planning when the next resource for generating power will be needed  to determine which neighborhoods will need new poles, wires or substations. Utilities must also anticipate how much energy customers will use each year.

To calculate the electricity demand of each year, CPS Energy has analysts on staff and on contract who take existing data to build a statistical model and then project energy use. That data includes:

  • Past and projected use by household customers, taking into account everything from the types of homes they live in (old, new, efficient, inefficient) to the types and age of appliances found in the homes (refrigerators, water heaters, air conditioners, etc.)
  • Actual past weather, implications for future weather, and the ever changing relationship between weather and consumer demand for energy
  • Past and projected economic data about our community
  • Past and projected prices for electricity.

It’s a little easier to project what household use of electricity might look like, since homes use energy in similar ways and neighborhoods or areas of the community have similar types of housing. Estimating commercial energy use is more challenging since building types for companies vary widely and have different purposes — office to warehouse, manufacturing to retail.

Weather is the single greatest driver of energy use, and for a utility, it divides the energy use into cooling degree days (when you run the AC because it’s hot), heating degree days (when you run the heater because it’s cold) and other (any other kind of energy use, like water heating or lighting). Regional weather data is gathered from the National Oceanic and Atmospheric Administration. These days, weather can be modeled fairly well 7 to 10 days in advance, but seasonal forecasts are less certain. Averaging typical weather over the last 15 years or so and combining it with projections of future weather results in projected patterns of energy use.

Finally, economic data — how the economy is doing — is important to incorporate, because it impacts how people use energy. During times of economic stress, customers are more sensitive to price and strive to use less. This is particularly true of commercial customers, who typically manage expenses carefully.

Ultimately, this data is built into a set of models that forecast how the energy will be used, where it will be used, and whether it’s used on a hot day, a cool day, or for a non-weather-related purpose.  For the mathematically oriented, the Residential Statistically Adjusted End-Use modeling Framework looks like this:

Monthly Residential Use Per Customer = a+bc x XCoolm + bh x XHeatm + bo x XOtherm + em

We can test how the model works by looking backward — that is, how well does the model explain what happened in the past?

(Image) The blue line in the chart above is the model’s predicted outcome, and the red is what actually happened. With a few exceptions, in 2002 and 2003, the model almost overlays actual energy use.
The blue line in the chart above is the model’s predicted outcome, and the red is what actually happened. With a few exceptions, in 2002 and 2003, the model almost overlays actual energy use.

As CPS Energy evolves to focus more on energy efficiency and managing energy demand, our model is changing.  Today, we have the ability to reach out to more than 75,000 residential customers during periods of high demand (summer afternoons) and ask that they change their AC settings.

We also partner with commercial customers to shift their energy load from summer afternoons between 3 and 6 p.m. (the peak) to other parts of the day. Both of these efforts reduce energy demand during peak times, which lowers customers’ costs and slows the need to add more power plants.

Modeling will never be completely accurate. For example, revenues this year are lower than projected. That means less money for CPS Energy to pay the bills, and less money to the city of San Antonio, which relies on that funding to pay for public safety and infrastructure.

But in large part, CPS Energy’s demand forecasting has been accurate, and the utility will continue to assess future energy demand with precision and rigor. Our customers deserve nothing less.

Tracy Idell Hamilton

Tracy Idell Hamilton was part of the Corporate Communications team at CPS Energy.

Leave a Reply

Your email address will not be published. Required fields are marked *