Enhancing energy efficiency in commercial buildings via behavioral interventions – a comprehensive analysis


Buildings account for almost half the electricity consumption and a third of CO2 emissions worldwide.  Demand for building electricity is projected to rise by 15 % in 2050 [1]. This projected increase poses a serious risk toward sustainable development and the environment. Thus, it is imperative to reduce energy consumption while maintaining requisite levels of building operations through improvements in energy efficiency.

Energy consumption in buildings primarily stems from three sources: Heating, Ventilation, and Air Conditioning (HVAC); Lighting; and Plug and Process Loads (PPLs). Of these, several efforts have focused on efficient HVAC and Lighting solutions, which have further been optimized through Demand Response (DR) programs. However, there’s a dearth of studies on PPLs despite being recognized as the next big hurdle in improving energy efficiency. To this end, researchers from Carnegie Mellon University and NASA Ames Research Center performed a rigorous study on plugload energy consumption and related behavioral interventions [2]. The team, comprising Dr. Chaitanya Poolla, Dr. Abraham Ishihara, Mr. Dan Liddell, Prof. Steven Rosenberg, and Dr. Rodney Martin, conducted field experiments in both university and commercial building environments. Their work offers key insights into improving building energy efficiency, and is published in the scientific journal, Sustainable Cities and Society.

The literature investigating plugload energy consumption is relatively scarce and most of the published works either lack well-designed experimentation or statistical modeling. To address this gap, the authors designed effective experiments and modeled the effects of interventions on plugload consumption. These experiments were conducted in commercial buildings, including a government office building at the NASA Sustainability Base and a university building at Carnegie Mellon University’s Silicon Valley campus. An important feature of the experiments is in employing matched pairs design to strengthen the causal connection between plugload consumption and the interventions. These interventions involved monetary incentives, dashboard-based visual feedback, or a combination of both. The researchers combined techniques from design of experiments, behavioral psychology, human-computer interaction, statistical inference, and energy efficiency to estimate the impact of visual feedback and monetary incentives on plugload consumption.

The study revealed several key insights. First, the effects of interventions on plugload consumption were found to be significant, both statistically and practically. In the office environment, dashboard feedback resulted in a reduction of 9.52% in plugload energy consumption. On the other hand, dashboard feedback in the university setting led to a tangibly higher reduction of 21.61%, which increased to 24.22% when augmented with monetary incentives. These results underscore the effectiveness of behavioral interventions on occupant plugload consumption. Second, the study pioneers statistical time-series models to predict the effects of interventions on plugload energy consumption. The accuracy (RMS) of the behavioral predictions was found to be 75-80% using autoregressive models with exogenous inputs (interventions). These prediction models are central to integrating occupant plugload consumption into building optimization frameworks.

By demonstrating the effectiveness of interventions and statistical models, the study offers informs policy design for sustainable urban development. The findings by Dr. Poolla and colleagues significantly advance the current scientific understanding of plugload consumption in commercial buildings. However, the authors also acknowledge the need for subsequent studies with larger sample sizes and longer study duration to investigate the washout or persistence of effects.

In conclusion, the researchers conducted comprehensive experiments to study the effects of behavioral interventions on plugload consumption in commercial buildings. Among the interventions employed, dashboard feedback is found to be both statistically and practically significant in reducing plugload consumption. Further, the authors pioneer the development of models that predict plugload consumption as a function of exogenous interventions. These models are central to integrating occupant behavior within building optimization frameworks, enabling a holistic approach to improving building energy efficiency. As we progress toward a greener future, this study offers valuable insights for policymakers, engineers, and researchers striving to enhance energy efficiency in buildings and mitigate their environmental impact.

About the author

Chaitanya Poolla works as a software research scientist at Intel Corporation on problems related to computer performance and power. He received a B.Tech (Honors) degree in Aerospace Engineering from the Indian Institute of Technology – Kharagpur in 2011, and MS and PhD degrees in Electrical and Computer Engineering from Carnegie Mellon University in December 2016.

He has contributed to more than ten scientific publications in internationally recognized journals and conferences. He is also a recipient of the NASA honor award for group achievement, BOEING fellowships, the IIT silver medal for best graduating student in Aerospace engineering, and the gold medal in the Indian national mathematics olympiad.

He is interested in topics such as machine learning, optimization, design of experiments, statistical inference, decision sciences, dynamics, optimal & adaptive control. His work is primarily concerned with addressing challenges faced by the computing, energy, and transportation industries. Apart from his professional activity, he is interested in philosophical topics such as ontology, epistemology, and metaphysics.

About the author

Abraham K. Ishihara is currently a senior engineering manager at Aurora Innovations and adjunct faculty at Carnegie Mellon University.  His research interests are in Autonomous Systems, UAS and Upper E Air Traffic Management; Edge Computing; Modeling and Simulation of Aircraft; Autonomous UAS swarms; Optimal Control and Optimization Methods; Nonlinear, Adaptive and Neural Network Control; Distributed Parameter Control; Stochastic Control; Control of Delay Systems; Photovoltaic (PV and CPV) Modeling, Diagnostics, and Optimization; Intelligent Building Control.

About the author

Dan Liddell is a senior software engineer for KBR, working at NASA Ames Research Center.


[1] ExxonMobil, Energy demand: Three drivers, 2023

Go to Energy demand: Three drivers

[2] Chaitanya Poolla, Abraham K. Ishihara, Dan Liddell, Rodney Martin, Steven Rosenberg.  Occupant plugload management for demand response in commercial buildings: Field experimentation and statistical characterization. Sustainable Cities and Society 84 (2022) 103984

Go to Sustainable Cities and Society

Check Also

Unveiling the Global Decline in Lake Water Storage: Causes, Consequences, and Solutions