I am at DistribuTECH this week in Orlando. This is North America's largest electric power utility event. I don't know the number of attendees this year, but last year it was close to 10,000.
The first talk I heard was a riveting talk that hit on a key theme of this year's conference, big data and analytics. The talk was given by Lee Krevat of Sempra US Gas and Power and Tim Fairchild of SAS and was entitled "What can a regulated electric utility learn from Moneyball ?" Moneyball is a book by Michael Lewis about baseball and how the Oakland A's applied big data and analytics to become one of the top teams but without having the deep pockets of the richest teams.
Moneyball for transformers
As an example of the relevance of Moneyball is a utility that took every bit of data that could potentially be relevant to the lifecycle of a transformer, some 80 variables in all, and ran correlations with transformer failure data - for 1500 transformers. They found that just 10 of the variables predicted 91% of transformer failures. This compares with the utility's traditional way of forecasting failure which only predicted 15% of the failures. If you ask an experienced power engineer what the most important factor in determining the lifetime of a transformer, the historical load, and especially the overload, on the transformer would be at the top of the list. What the utility actually found was that the correlation analysis showed that load was only the eighth most important factor. The analysis showed that the most important factor was the number of meters attached to the transformer - which probably wouldn't have been on anyone's list.
Lee and Tim presented other analogies where analytics play a key role in both baseball and electric power.
The good face
Baseball scouts often profile players by the look of their face, body, and a "baseball look", not always on their actual statistics. Similarly utilities often profile circuits to prioritize capacity investment decisions, often based on "gut" feeling. Now they are able to leverage data to make better investments.
Age before beauty
Scouts show a preference for younger players even though baseball players with better statistics, but who are older, are also often available without long term contracts and at a significantly lower salary. By analogy large utility transformers have traditionally been replaced based on age. But now statistics on many transformers are being monitored and the transformers only replaced if their statistics are in decline and predict failure.
Lee and Tim suggested other practical applications of this approach which include using machine learning to predict failures of wind turbines and to determine why solar farms often generate less electricity than expected.