Date of Award
Spring 2023
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Management
First Advisor
Alizamir, Saed
Abstract
As climate change continues to accelerate, extreme weather events have created unprecedented variability in energy demand and consumption, overloading existing infrastructure and causing large-scale system breakdowns. Fundamentally, these adverse events are driven by the supply-demand mismatch within energy markets. Utility firms face fluctuating demand that cannot be precisely estimated, and economically viable storage capabilities remain scarce. Despite the crucial role of weather-related consumption patterns, and the expected increase in the frequency and magnitude of weather anomalies caused by climate change, little is known about how thermostats are used. This dissertation analyzes the impact of weather on optimal household behavior and contrasts real-world behavior with theoretically optimal decisions. This dissertation also asks important policy questions around using price to incentivize households to reduce consumption. Chapter 2 develops an analytical model to establish what optimal thermostat control entails when considering the impact of weather on consumption. A novel component of this chapter is the incorporation of rational inattention into the decision-making process. This concept posits that decision-makers are not necessarily fully informed and rational agents, but instead, pay a cost to acquire and analyze information before taking action. This cost of information processing allows for a natural approach to evaluating the theoretical effectiveness of technology solutions such as smart thermostats. Using a continuous time model for utility and energy consumption, We characterize a household’s optimal control policy by reducing a complex time-dependent problem to a static optimization problem. Chapter 3 is an empirical study of household smart thermostat usage. The study examines how households make decisions about their thermostats in practice. Of particular interest is the impact of weather on household decision making and how this real-world behavior differs from the optimal behavior discussed in Chapter 2. To that end, we empirically analyze a novel data set provided by ecobee Inc.. This data set sheds light on the distinct nature of long-term decisions, represented by the system mode, and short-term decisions, represented by the numeric setpoints. In our study on long-term behavior, we model household movement between modes as a Markov process, with the transition probabilities arising from a multinomial logit model. Underpinning the logit model is a Dynamic Linear Model which captures reactions to both seasonal trends and unanticipated temperature shocks. A key insight is that household responses to the weather can be classified into one of three archetypes based on the frequency with which they use the auto mode. When examining short-term behavior, we focus on the setpoints chosen and how they are adjusted over time. We model the direction of these changes using a multinomial logit that captures time-dependent temperature effects while controlling for preferences and anchoring behavior. Short-term behavior can be categorized along two dimensions: (1) how many pre-programmed schedules a household uses and (2) how frequently a household overrides the pre-programmed setpoint. Using these models, we conduct a set of simulation studies to investigate the impact of climate change on energy consumption and find that cooling requirements in 2050 could be 70% higher than 2019. We also establish a clear link between consumption and thermostat behavior. Users that automate their long-term decisions consume more than those that do not, while households that rely on the thermostat only in the short-term consume less. Chapter 4 considers the implications of the findings from Chapters 2 and 3 on dynamic pricing for energy. Specifically, we examine the burden that time-of-use (TOU) pricing in Ontario has on households whose occupancy pattern does not conform to the traditional ‘9-5’ schedule. By accounting for strategic behavior, we are able to get a better understanding of the impact of TOU pricing on these households. We also consider how ‘shiftable’ tasks, such as doing laundry, are impacted by occupancy schedules and TOU pricing. This chapter highlights the trade-off between fairness, equity, and efficiency in pricing life’s necessities. In summary, this dissertation studies thermostat behavior and the impact of weather on a household’s decision-making process, and how these findings impact the design and implementation of pricing programs.
Recommended Citation
Blair, Michael Ryan, "Understanding Weather-driven Thermostat Behavior and its Implications for Residential Energy Consumption" (2023). Yale Graduate School of Arts and Sciences Dissertations. 909.
https://elischolar.library.yale.edu/gsas_dissertations/909
