Electric utilities generally market demand-response and energy-efficiency programs to very broad groups, such as residential customers and small businesses, without knowing which programs are reducing consumption during periods of very high demand or, in the case of efficiency, eliminating waste. A major promise of smart grid meters is to make such programs more effective, yet using the vast amounts of data from smart meters for this is poorly understood. This project experimented with several methods to process data to inform groups of customers about the most worthwhile efficiency measures to consider, to understand the demand-response potential in particular circumstances, and to help utilities provide energy at minimum cost and with minimum environmental impact.
This project segmented utility customers into groups that exhibit similar energy consumption behaviors, based on a detailed history of energy consumption, plus information about building locations, demographics and weather conditions at the time of consumption. It used data from about 900 Google PowerMeter participants measuring electricity use every 10 minutes and a PG&E sample of approximately 12,000 customers with hourly smart meter data.
Six most frequent load shapes from the dictionary and their average daily kWh using hourly data of 125K customers, August 2010 - July 2011.
Among the findings:
A population of about 1,000 users can be segmented into about 10 groups for demand response, and a scalable clustering technique that uses a statistically meaningful distributional metric was developed.
Using hourly electricity and weather readings to characterize residential customers' temperature-dependent consumption such as air conditioning or heating, segmentation and targeting of users may offer savings twice that of current demand-response programs.
About 900 households users may be described with good conﬁdence by 13 patterns, with certain types of appliances and behaviors related to appliance operation affecting consumption patterns most. Consumption statistics may be used to target residential energy efﬁciency programs to achieve greatest impact in curtailing cost of service.
Inferring occupancy states from consumption time series data showed that temporal patterns in the user's consumption data can predict user consumption at a population level.
The number of refrigerators and entertainment devices (e.g., VCRs) are among the most important determinants of daily minimum consumption, while the number of occupants and high-consumption appliances such as electric water heaters are the most significant determinants of daily maximum consumption.
Acknowledging climate change as a motivation to save energy showed correlation with lower electricity consumption.
Contrary to some previous studies, the researchers observed no significant correlation between electricity consumption and income level, home ownership, or building age.
Some otherwise energy-efficient features such as energy-efficient appliances, programmable thermostats and insulation were correlated with a slight increase in electricity consumption.
As part of Phase II of the Energy Sensors & Behavior Initiative, researchers are developing about three types of analytics using smart meter data, which will support a behavioral program(s), and serve as a foundation for the Segmentation and Baseline Analytics Platform. They will also develop data visualization methods used by Stanford projects, partners, etc. They will also improve segmentation and apply the methods to a behavior program project(s), develop base-lining (for calculating accurate energy savings), and segmentation visualizations and reports (to allow others to evaluate their programs with smart meter data). Finally, they will develop and test intervention(s) based on the team's advanced segmentation and information feedback analytics.