The accuracy of load forecasts is of paramount importance to every utility. A single percentage point increase (or decrease) in load forecast accuracy can be worth millions of dollars. Applying the appropriate analytics tools and methods is essential to high quality data. Given the role of analytics and the direct impact on a utility's bottom line, I reached out to Alyssa Farrell, global manager energy and sustainability solutions at SAS to gain her perspective on the challenges, opportunities, and analytics best practices in the load forecasting arena.
The load forecasting challenges that rose to the top of the list were:
- Understanding the current margin of error within short term forecasts and deciding how much error is acceptable given the direct relationship that exists to utility profitability.
- The complexity and quantity of forecasts are placing increasing pressure on forecasting resources. Farrell explained that "a typical utility produces as many as 14 different short term forecasts within a typical 10 hour working day. A few examples include: 1) day-ahead forecasts for trading usually done twice per day, 2) current day forecasts, repeated roughly five times per day to determine if there is a need to ramp up peak generating facilities to manage the load, 3) day ahead forecasts to predict how many renewable resources can be expected to be online tomorrow and 4) multiple weather forecasts.
- Many utility forecasters are nearing retirement. "Forecasters have the authority to override or change a forecast, given their experience with a particular dataset or upcoming event. These manual tweaks are not sustainable given the growing complexity and quantity of forecasts. Utilities are looking to technology for part of the answer. Many are replacing more simplistic tools and black box solutions with technologies that use enterprise data quality and that are able to produce repeatable and auditable results" said Farrell.
In addition, public utility commissions are more heavily scrutinizing the forecasts presented to them, especially in connection with rate cases. This puts more pressure on forecasters to produce accurate, transparent, and defensible forecasts.
A discussion about the challenges would not be complete without reviewing the opportunities. The primary load forecasting-related opportunities that she identified were:
- More sophisticated analytics, in combination with more granular smart meter data, will improve the accuracy and defensibility of forecasts. Farrell explained that by defensibility she means "the ability to consistently reproduce the forecast, and to go in front of a regulator and present the forecast with confidence."
- Improved data visualization tools will enable more effective communication about the results of very complex, statistical, and predictive model-based forecasts. These forecasts contain key pieces of information that can be used by many areas of the business from traders to system planners to utility executives. "You are not providing just a number anymore. You are providing information that enables a dialogue about potential growth opportunities in the commercial sector or perhaps about a geographic area experiencing flat or declining residential sales. Offering the different audiences the ability to visualize the forecast and break it down in ways that make sense to them will help utilities to capture the value that reside in these forecasts" said Farrell.
- There are also opportunities on the horizon to improve renewable resources forecasting as well as price forecasting, the ability to predict the market price of power tomorrow. More accurate forecasts in these areas would be invaluable.
After we covered the challenges and opportunities, I asked Farrell for her perspective on load forecasting best practices and she stated the following:
- Employ a broad set of forecasting techniques. When it comes to forecasting methods and models, there's no such thing as "one size fits all". Volatility, whether as a result of energy prices, weather, or the availability of the commodity itself, makes it incumbent on utilities to use multiple forecasting techniques in order to produce more accurate forecasts. Applying the right forecasting methodologies to the right data and the right problem will reduce the forecasting error.
- Participate in a community of forecasting experts and challenge yourself to learn from others.
- Understand and quantify the impact of a percentage point improvement in forecasting accuracy on your bottom line. It is imperative to know this information when attempting to gain approval of investments in the modernization of your forecasting technologies.
- Choose a technology partner with a strong track record and whose core competency is forecasting.
The next big thing
Farrell shared her perspective about energy forecasting's future:
- The presentation of the results of a forecast on an iPad or via 3D models will become more common. Innovative ways of visualizing and consuming data will provide greater insight across the board, and load forecasting is no exception.
- Increased confidence in wind and solar forecasting will support better planning decisions related to renewables integration.
- Unstructured data may make an appearance. While the value of tweets, posts, and call center transcripts to improve outage management is evident, they may also be useful to forecasting. "I think we could conceivably find value if we are able to capture unstructured data and analyze it in a way that is meaningful for forecasting" said Farrell.
It certainly seems that forecasting is a hot topic again for utilities. New ways to analyze, visualize, and consume data are impacting every utility business process including load forecasting. Amidst the hive of activity are some recognizable challenges such as data accuracy and latency, growing demands for actionable information, key resources nearing retirement, and the increasing expectations of regulators. Even so, there are exciting new opportunities ahead and from what I can see from here, the future looks bright!
Kim Gaddy is a senior analyst for the Utility Analytics Institute and can be reached at email@example.com