Data-driven techniques that extract insights from sensor data reduce the cost of improving system energy performance through fault detection and system health monitoring. To lower cost barriers to widespread deployment, a methodology is proposed that takes advantage of existing sensor data, encodes expert knowledge about the application system to create ‘virtual sensors’ and applies statistical and mathematical methods to reduce the time required for manual configurations. The approach combines sensor data points with encoded expert knowledge that is generic to the application system but independent of a particular deployment, thereby reducing the need to tailor to individual deployments. This paper not only presents a method that detects faults from measured energy data, but also (1) describes an engagement method with experts in the energy system domain to identify data, (2) integrates domain knowledge with the data, (3) automatically selects from among the large pool of potential input data, and (4) uses machine learning to automatically build a data-driven fault detection model. Demonstration on a commercial building chiller plant shows that only a small number of virtual sensors is necessary for fault detection with high accuracy rates. This corresponds to the use of only five out of 52 original sensor data points features. With as few as four features, classification F1 scores exceed 90% on the training set and 80% on the testing set. The results are implementable and realizable using off-the-shelf tools. The goal is to design with domain experts an energy monitoring system that can be configured once and then widely deployed with little additional cost or effort.

Accurate savings estimations are important to promote energy efficiency projects and demonstrate their cost-effectiveness. The increasing presence of advanced metering infrastructure (AMI) in commercial buildings has resulted in a rising availability of high frequency interval data. These data can be used for a variety of energy efficiency applications such as demand response, fault detection and diagnosis, and heating, ventilation, and air conditioning (HVAC) optimization. This large amount of data has also opened the door to the use of advanced statistical learning models, which hold promise for providing accurate building baseline energy consumption predictions, and thus accurate saving estimations. The gradient boosting machine is a powerful machine learning algorithm that is gaining considerable traction in a wide range of data driven applications, such as ecology, computer vision, and biology. In the present work an energy consumption baseline modeling method based on a gradient boosting machine was proposed. To assess the performance of this method, a recently published testing procedure was used on a large dataset of 410 commercial buildings. The model training periods were varied and several prediction accuracy metrics were used to evaluate the model's performance. The results show that using the gradient boosting machine model improved the R‐squared prediction accuracy and the CV(RMSE) in more than 80 percent of the cases, when compared to an industry best practice model that is based on piecewise linear regression, and to a random forest algorithm.

Buildings account for about 40 percent of the total energy consumption in the U.S. and emit approximately one third of greenhouse gas emissions. But they also offer tremendous potential for achieving significant greenhouse gas reductions with the right savings strategies. With an increasing amount of data from buildings and advanced computational and analytical abilities, buildings can be made “smart” to optimize energy consumption and occupant comfort. Smart buildings are often characterized as having a high degree of data and system integration, connectivity and control, as well as the advanced use of data analytics. These “smarts” can enable up to 10–20% savings in a building, and help ensure that they persist over time. In 2009, Microsoft Corporation launched the Energy-Smart Buildings (ESB) program with a vision to improve building operations services, security and accessibility in services, and new tenant applications and services that improve productivity and optimize energy use. The ESB program focused on fault diagnostics, advanced analytics and new organizational processes and practices to support their operational integration. In addition to the ESB program, Microsoft undertook capital improvement projects that made effective use of a utility incentive program and lab consolidations over the same duration. The ESB program began with a pilot at Microsoft's Puget Sound campus that identified significant savings of up to 6–10% in the 13 pilot buildings. The success of the pilot led to a global deployment of the program. Between 2009 and 2015, there was a 23.7% reduction in annual electricity consumption (kWh) at the Puget Sound campus with 18.5% of that resulting from the ESB and lab consolidations. This article provides the results of research conducted to assess the best-practice strategies that Microsoft implemented to achieve these savings, including the fault diagnostic routines that are the foundation of the ESB program and organizational change management practices. It also presents the process that was adopted to scale the ESB program globally. We conclude with recommendations for how these successes can be generalized and replicated by other corporate enterprises.

The after-market labeling of a device by its users often indicates problematic usability, which can affect the device's energy consumption. For example, when people find a lighting control panel difficult to use, they often write instructions on a piece of paper and affix it nearby as a reminder to themselves and to help others. We collected over a hundred examples of these “folk labels” from commercial and residential buildings through an online contest inviting people to upload photos of folk label examples, informal solicitation of colleagues, online searches, and personal observation. Some folk labels offered guidance or reminders (e.g., turn off light before fan on a projector); some provided specific instructions for multi-function controls or addressed problems identifying orientation or direction (right/left, on/off). We categorized these folk labels (e.g., by location, subject, form, etc.), and analyzed them according to usability guidelines and heuristics1 . In addition, we evaluated their potential impact on energy consumption. We found that most folk labels indicated usability issues in three areas: visibility of available options, natural mappings, and consistency. For example, one light switch looked like a simple toggle ON-OFF control, but actually could control various dimming options (violated “visibility of available options” principle). One would naturally turn the light on full power if one did not know how to use the dimming capability. We discovered that folk labeling provides a simple means of identifying usability problems.

A “smart” device will remain efficient only as long as the settings and other parameters allow it to be. Thus, the degree of usability is becoming an element of energy efficiency similar to other physical characteristics. We developed and tested a procedure to quantify the usability of thermostats. The procedure assumes that usability can be represented by a user’s ability to accomplish a set of tasks. Thirty one subjects were tested in their ability to accomplish six essential tasks on programmable thermostats. The tests revealed a wide variation in the subjects’ ability to accomplish the same task on different thermostats. Thus it was possible to discern thermostats that were more effective than others. We created a metric based on data that are easy to collect and unambiguous that appears to reflect the usability of a task. Metrics from different tasks can be added and an overall usability “score” calculated. This approach, as well as the metric, can be applied to other devices where poor usability may impede energy-saving behaviour.