Development of prediction models for thermostat override behavior in direct load control events

Direct load control (DLC) is a prominent solution for demand-side energy management, where utility providers adjust consumers' temperature setpoints through smart thermostats. However, occupant overrides can have a detrimental impact on the overall efficacy of DLC. This study focuses on understanding override mechanisms and occupant responses during DLC. Real-world data from the Ecobee Donate Your Data (DYD) program is used to analyze user interactions with smart thermostats. Through a comprehensive analysis of four types of variables, a decision tree algorithm was found capable of accurately classifying smart thermostat users in terms of their level of compliance with the DLC program. This classification was based on general information accessible through smart thermostats enabling the evaluation of users' potential engagement in DLC without prior DLC experience. Additionally, a clustering algorithm was used to identify three distinct types of DLC participants characterized by their individual thermal comfort preferences and general habits. A predictive model for thermostat override behavior is developed based on the clustering results. These findings contribute to designing targeted and less disruptive DLC strategies that better align with participants' needs and enhance overall DLC effectiveness.

Automated summary

This study analyzes the override mechanisms and occupant responses during direct load control (DLC) using real-world data from the Ecobee Donate Your Data program. The decision tree algorithm accurately classifies smart thermostat users based on their compliance level with the DLC program, and a clustering algorithm identifies three distinct types of DLC participants. A predictive model for thermostat override behavior is developed, contributing to the design of targeted and less disruptive DLC strategies.