The University of California at Davis is where the ‘father of the plug-in hybrid & electric vehicle, Professor Andrew Frank, first explained the potential load-leveling synergies between electric cars, the electric grid, and the electricity used at home and work. It was an inspiring vision for the 80’s, 90’s, and 00’s, and it is now actually playing out in certain communities — the City of Davis among the first.
David Phillips, Director of UC Davis Utilities, presented at the OSIsoft User Conference last month. As he describes the role of his utility today:
“Our new mission is data. We’re no longer just the campus source for energy and water. We’re now also the campus operational data center.”
Phillips runs a city-scale water, waste, and energy utility. With strong leverage on both the supply and demand sides of the meter, UC Davis Utilities is uniquely positioned to innovate energy and water operations. Compared to other public utilities, it can act fast in delivering new services that further sustainability goals. The work it is now doing in energy load disaggregation exemplifies just how UC Davis works to make the necessary connections between research, operations and public-facing mobile and web applications. It is using OSIsoft’s PI System as the platform that pulls from a geographic information system (GIS), an asset management system, an enterprise financial system for invoicing, and a SCADA system. It is the umbrella system that ties data from all these systems together and enables high-resolution analysis. It is bringing all the intellect and energy of a top-tier university engineering program to bear in getting things launched too!
In his presentation Phillips details an electrical meter disaggregation process that his team is piloting for greater precision in tracking and reporting energy and water consumption by municipal infrastructure systems. One use case is now looking at the energy consumption trends of a central Davis parking structure, the Gateway Garage. This site was selected for the pilot because it was essentially a city block on one meter, with a range of predictable activities contributing to consumption patterns. The single meter registered lighting, elevators, electric car charging stations and other plug loads such as the ticketing kiosks. Disaggregating the activities is a matter of subtracting known consumption amounts from the usage totals to discover the unique signature of each load. According to Phillips,
“Our goals are to create a feedback loop, providing the data insight that will lead to operating the systems better and designing them better for the future. It is trendy to put in more and more submeters. With this project, we’ve learned the information you can glean from high resolution analysis of one meter, can be more valuable than the ROI of putting in 1000 additional meters.”
With PI, the frequency of data collection can be increased down to every two seconds—a big bump in resolution compared to the standard 15-minute-interval meter data. The higher data frequencies started to reveal the different data signatures to the UC Davis team of data scientists.
“You can write machine-learning rules to trigger alarms, if an expected data signature does not appear in the collected data,” explained Phillips. “For example, the type of power use surge that occurs when the garage elevator is used has an obvious shape. Should that signature disappear at any time during normal operation of the parking structure there could be an elevator outage, which has safety and liability implications. You want to trigger action to fix that problem without waiting for garage attendants or users to call in the need for service.”
Another use case for disaggregation was for the management of electric-car charging stations in the garage. UC Davis has eight different branded systems installed across campus, with samples of each in the Gateway Parking Structure. With the help of a Microsoft Azure Machine Learning expert and US Davis Computer Science students, a public-facing mobile app that lets electric car owners on campus know where they can find an open charger was built and a beta version deployed in a matter of weeks.
One more example of the power of this load disaggregation approach concerns the garage lighting system. With the California Lighting Technology Center on the Davis campus, UC Davis was an early adopter of high-efficiency LED lighting with built-in programmable controls. A first-generation LED induction system was installed in some areas of the Gateway Garage, along with a new-generation system that features mesh network lighting controls in other areas. One differentiator between generations is power meters integrated into LED fixtures that give visibility into status and that send notifications in the event of an outage. Phillips and his team realized that disaggregated load analysis could approximate this feature. Low or depleted power draw indicates lamp outage or near outage, and they could detect that with data collected at 2 second frequency. This was possible with PI, so they wrote machine rules to trigger an alarm message when this signature is detected.
The desired outcome of a disaggregation analyses is a set of such machine rules and apps. Once the algorithms are set, then the dissaggregation can be done automatically via machine-learning. Today the Gateway Garage project includes a web service that does the disaggregation as trained by the researcher’s first set of machine rules. “It’s about 50% right, right now,“ according to Phillips. “The future is open-sourcing the signatures for all these devices.” The goal here will be to accumulate a database of signatures that can be shared across the entire open-source community of developers interested in improving facility operations and energy efficiency.