Short-Term Load Forecasting Experiences

The experiences shared on this month’s Press Release whereby I engaged heavily in the hand-on training of the engineers on the use of ANNSTLF (Artificial Neural Networks Short Term Load Forecaster) taught me some important lessons! Basically, the goal of the training initially was to make the engineers vary familiar with this remarkable forecasting system. I have done a lot of work on forecasting in the past always with questions on my mind. The forecasts would work under one set of data but fail on another.

With ANNSTLF I though we still had some problems of the same sort. But my trusted software engineer, Dejan Paravan, showed me how various "additional" tools work with ANNSTLF and that was a "Eureka" moment for me. Finally, I can "play" with the system and train engineers to become true "forecasters" rather than computer button pushers.

So for a full intensive one week in Rio de Janeiro at ONS, Brazil and two days the following week at the California ISO, I was able to have real fun with the forecaster. In fact I remember telling the class that although we may be delivering a fine car like a Maserati, they still needed to learn how to become superior racecar drivers. ANNSTLF is a Maserati, in my opinion, but training on it effectively would transform the engineer into a true forecaster. This involves several key "tricks" of the trade. In the attachment to the press release we demonstrate how special events data can be handled. Many more issues were dealt with. The final result on forecaster performance was well beyond my expectations!

Stay tuned till next month. In the meantime, I appreciate your direct feedback – adebs@dsipower.com