According to a 2012 report by the Edison Foundation, more than 65 million residential smart meters will be deployed by 2015 in the US. These smart meters record electricity used at least every hour, generating thousands of data points everyday that can be disaggregated to reveal consumer operating routines and behaviors while using energy appliances.
Aligning this data with measurable household characteristics (e.g. age, income, family size), it is possible to effectively predict the likelihood of a household enrolling in a residential energy efficiency program. However, measurable household data is hard to come by and not many residents are willing to open up their homes to analysis. Consequently, households are not being recruited for energy efficiency programs in a targeted manner, which leads to a low prediction enrollment percentage (less than 50%) and high volume of wasted marketing resources.
At the Fraunhofer Center for Sustainable Energy Systems CSE, researchers have developed a new patent-pending method to predict a household’s probability to enroll in an energy efficiency program using only twelve months of smart meter data. . Through the use of advanced machine learning algorithms, the prediction enrollment percentage jumps up to 90% (level of accuracy obtained in our study of a US West Coast behavior-based residential program). Due to this high prediction rate, households can be more efficiently targeted for energy saving programs, thereby increasing the cost-effectiveness and efficacy of marketing activities and, ideally, increasing enrollment rates.
To learn more about the System and Method of Prediction of Household Enrollment in Energy Saving Program read the 2014 IEEE International Conference on Systems, Man, and Cybernetics paper.