Power Laws: Forecasting Energy Consumption

More accurate forecasts of building energy consumption mean better planning and more efficient energy use. In this challenge, competitors used machine learning to build the most accurate predictions of the future from limited data in the past. #energy

€23,000 in prizes
mar 2018
1,032 joined

alt-text

Each challenge in the Power Laws series explored a different aspect of energy efficiency and management. The winning algorithms were released under an open source license to spread understanding about how energy modeling works and what approaches are most effective.

Why

The ability to forecast a building’s energy consumption plays a critical role in energy efficiency. Good forecasts can help implement energy-saving policies and optimize operations of chillers, boilers and energy storage systems. They also provide a baseline for flagging potentially wasteful discrepancies between expected and actual energy use. Often these forecasts need to be built using limited data and still be as accurate as possible.

The Solution

Using selected “time windows” of data from Schneider Electric, competitors built models to predict future energy consumption across buildings. These predictions were compared with the actual recorded consumption, information that was withheld from the developers. The models that found the most useful signals in the limited past data won prizes for the most accurate estimates.

The Results

The best algorithms used weather, holidays, and synthetic features created from the data to produce the most accurate forecasts. The top algorithm predicted consumption within 0.3% of actual recorded measures, on average, and had an R-squared above 0.99. These approaches will help build increasingly accurate forecasts for building managers and automated management systems seeking to use energy more efficiently.


RESULTS ANNOUNCEMENT + MEET THE WINNERS

WINNING MODELS ON GITHUB

DATASET ON SCHNEIDER ELECTRIC EXCHANGE