Skip to content Skip to navigation

Stanford Energy Services Platform

ESP functional architecture
While encouraging user's proactive involvement in reducing their energy use is an important goal, a complementary strategy looks at how automation can improve efficiency with minimal user input.
ESP functional architecture

Many behavioral researchers in energy design and test feedback interfaces, often building the various parts from scratch. Valuable time and resources are wasted, and researchers are often confined to running very small experiments. An open platform could aid researchers, utilities and third party developers in this area. This project built the Stanford Energy Services Platform (SESP) to collect, clean and store data from sensors, such as smart meters, online interaction and surveys. It provides useful analytics (e.g., establishing baseline energy consumption, comparison with other consumers and disaggregation), graphing, a recommendation system, participant registration and assignment, and front-end display and email notifications suitable for performing experimental manipulations. Five behavioral interventions in the Stanford Energy Behavior Initiative utilized the platform: PowerHouse, the three Facebook applications (Kidogo, Powerbar and PowerTower), and the appliance calculator. We built SESP to be modular, extensible, scalable and secure, with the goal of inviting other researchers and commercial entities to utilize and grow the platform.

The back end contains several logical data stores: electric and gas smart meter data; participant data; website/application activity; project-specific data (e.g., experiment details, investigator names for data access); external data (e.g., weather, regional specific information like fuel type penetration or socio-demographic characteristics), and recommendations, including energy-saving actions and also energy efficient appliance recommendations.

Energy Platform Software website screen shot

We created a web collector to pull users' energy usage data from their accounts on a utility website, so long as the participant has a relevant account and provided us permission, username and password. Sensor-based data are challenging due to large volume and pre-processing needs such as data vetting and cleaning. Participant data includes registration, consent, responses to a customizable survey, experiment assignment, and collection of website activity data. Systematic participant assignment to different conditions is critical for performing experiments and for evaluation so that utilities can determine the impact of a program. Website and application activity data, (accessing a page, expanding text, or clicking a link) enable analysis of engagement, dose effects and mediator and moderator variables.

For research and evaluation purposes, researchers using SESP across a broad range of information technology capabilities can make front-end changes to the applications easily and see the results, so they can continuously iterate. In addition, to create different conditions for their experiments, researchers can create variations on a parent web page, wherein changes to the basic parent page get passed down to any variations easily. They also can integrate freely the back-end services described above.

SESP is flexible, expandable and secure. A straight-forward interface lets researchers extract various data ranges from different experiments. In general, researchers want to analyze data with the statistical packages of their choice, so the platform provides access to data from experiments through raw data dumps. Experimenters and application developers can extend the database and add their own custom data elements. The platform has several analytic tools that researchers can use easily. For example, we compute a single user's historic energy usage baseline and also aggregate a community's energy usage. The system is also set up to allow for weather adjustments. And, because appliance-specific consumption information facilitates energy savings, we are interfacing Bidgely Inc.'s commercially available disaggregation system with SESP.

Several services were developed to group data in ways that could motivate behavior change. For example, we compare a user's energy consumption to their baseline, or to a groups' energy use. The energy data is graphed, and a wrapper allows the graph widget to be embedded within different presentation technologies. SESP supports automated emails that contain statistics personalized by user, as well as customized by the experiment for different experimental conditions. The services layer allows for applications to build on existing services thus extending them to accommodate new types of functionality.

Developers can control the presentation using various technologies, depending on their skill level and desire for customization. A software developer kit for users with programming knowledge allows near complete flexibility in designing and developing applications (e.g., mobile, website, Facebook), while accessing databases or services on the platform. Mash-ups are suitable for those with limited programming knowledge. Users with no programming knowledge can create applications by using pre-made modules or "widgets" in systems such as iGoogle, Google sites, and iWeb. Also, various presentation oriented packages can interface with our platform, such as Drupal or Joomla.

SESP's security is accomplished through physical security, cloud security and storage security. The data center utilizes multi-factor access authentication and closed-circuit television. Cloud security is accomplished through operating system hardening, firewalls and network security removing the threat of IP spoofing, port scanning, etc. Both bucket- and object-level access controls, with defaults that permit authenticated access by the bucket and/or object creator only, provide storage security.

Future work

A phase two effort involves further packaging and documentation for outside consumption. Also, SESP's advanced segmentation algorithms and methodologies need revised graphing modules, recommendation systems and potentially other interface features to fully incorporate lifestyle segmentation algorithms based on hourly or 15-minute data. The further development and integration is included in the plus-up funding.