- Posted on March 19, 2014
- 4 views
I recently had the pleasure of visiting with David Hamilton, load forecasting manager at Old Dominion Electric Cooperative (ODEC). As native Texans, Hamilton and I are unique. He is part of another exclusive group, long-time energy forecasters.
Energy forecasting description
David described energy forecasting like this:
"Forecasting energy and peak demand is not the same as forecasting widgets or computer sales. Forecasting the real time delivery of electrons that people demand is a study of consumer behavior and similar to real-time point-of-sale forecasting."
A bit about forecasters
Forecasters provide the starting point that drives utility cost models and how much revenue the utility can expect going forward. Hamilton said "Forecasters are at the cusp of helping utilities plan for tomorrow, next year, and a decade from now. Building a new power plant can cost over one billion dollars. You want to ensure that the new plant you forecasted 10 years earlier will be needed."
Much of the code used today was written by Hamilton's generation. "Guys like me learned to do forecast models using the old SAS code. Many programs were developed 20 to 25 years ago when mainframe computers were the norm. The majority of this code will need to be re-written as utilities switch to newer technologies capable of handling much more data.
In addition to ODEC's own forecasting needs, ODEC builds and files forecasts for each of its 11 member cooperatives with the Rural Utility Service. These forecasts are used for system planning and to determine borrowing needs given the projected demand growth.
When Hamilton joined ODEC, an off-the-shelf software tool and Microsoft Excel were used for energy forecasting. The inefficiencies and the data validation challenges were obvious. The approach opened the company up to potential criticism and second guessing from regulators and auditors.
"We needed a system that was easy-to-use, open, and auditable. I recommended SAS. I had used and programmed in SAS for years and did not see anything else that could compete. I've found other software lacking in data management and the ability to house, analyze and report on large volumes of data" said Hamilton. Investor Owned Utilities with large forecasting staffs can sometimes work with lesser software. It is more important for smaller companies like ODEC to invest in an efficient and comprehensive solution that will stand the test of time.
ODEC management was reluctant to spend the amount of money that Hamilton recommended. Within a couple of years, ODEC senior VPs were saying "this is the best money we have every spent in our lives" according to Hamilton. "Anyone can come in and look at our process and our reports and how the data is structured. They can examine how we build our models and forecasts and see that we use best forecasting practices" said Hamilton.
The next generation of forecasters
It is time to hire and train the next generation of forecasters. Hamilton said "I'm on my last rodeo. Colleges are not teaching students to be energy forecasters. They need to learn the business just like I did. It will be challenging to teach new hires how to rewrite the code and to learn how to use the new software tools, construct the data, and generate the reliable forecasts that utilities require."
Better analytics tools and automation can help ease the transition. ODEC purchased a forecasting bundle that contains the entire SAS business analytics suite, and has been slowly gravitating to more automation of the forecasting process. Ultimately, Hamilton expects to automate:
- Control of the data using information maps inside SAS
- Safety checks
- Data management and validation including identification and resolution of data quality issues
- Generation of reports, graphics and presentation materials to enable senior management to make quicker and better informed decisions.
One no longer needs to be a programmer to run forecasting processes from start to finish. "If you can think about how you want your project to work, SAS will write the code for you, and it is repeatable. The code can replicate the process and produce the same output each time. One person can accomplish the work of five or seven resources just a few years ago" said Hamilton. Automation of the data management function will save up to 90% of the work that forecasters do today according to Hamilton.
"When I am gone, our system is automated enough that it will continue running until someone new can be trained. If you have a black box and you lose your forecasting guru, you are dead in the water. While you may spend a bit more now, investing in better forecasting tools is the right decision over the long term" said Hamilton.
A week ago a young analyst at another Generation and Transmission (G&T) Cooperative called Hamilton for advice. This G&T is looking at SAS and other forecasting tools. Much like ODEC, the G&T has only two forecasting resources. Hamilton asked the analyst "How much experience do you have?" and "Do you know how to forecast energy?" The analyst indicated that he is a recent graduate and just learning. David said "if you go with Excel or a black box option, you're on your own. You need to consider how you are going build those models and will need to learn what is important. Excel was designed as a desktop application and it handles a limited amount of data plus you cannot analyze your data if you use a black-box system. The result will be less accurate forecasts that will not satisfy senior executives. Some may question if you know what you are doing." Hamilton offered the following recommendations to this young analyst and all utilities:
- Examine your forecasting needs over the next 10 to 20 years when selecting a forecasting tool
- Don't be penny wise and pound foolish when attempting to build out a forecasting system
- Avoid black box solutions and ensure you can analyze and validate your data
- Purchase a solution that is scalable, transportable and auditable
Kim Gaddy is a Senior Analytst with the Utility Analytics Institute. Kim can be reached at email@example.com