Smart building temperature control using occupant feedback

Abstract

In the first part of this work we incorporate active occupant feedback to min- imize aggregate user discomfort and total energy cost. User feedback is used to estimate the users comfort range, taking into account possible inaccuracies in the feedback. The control algorithm takes the energy cost into account, trading it off optimally with the aggregate user discomfort. A lumped heat transfer model based on thermal resistance and capacitance is used to model a multi - zone building. We provide a stability analysis and establish convergence of the proposed solution to a desired temperature that minimizes the sum of energy cost and aggregate user discomfort. However, for convergence to the optimal, sufficient separation between the user feedback frequency and the dynamics of the system is necessary; otherwise, the user feedback provided do not correctly reflect the effect of current control input value on user discomfort. The algorithm is further extended using singular per- turbation theory to determine the minimum time between successive user feedback solicitations. Under sufficient time scale separation, we establish convergence of the proposed solution. Simulation study and experimental runs on the Watervliet based test facility demonstrates performance of the algorithm and validates the model.

Finally, we present an end-to-end framework designed for enabling occupant feedback collection and incorporating the feedback data towards energy efficient operation of a building. We have designed a mobile application that occupants can use on their smart phones to provide their thermal preference feedback. When relaying the occupant feedback to the central server the mobile application also uses indoor localization techniques to tie the occupant preference to their current thermal zone. Texas Instruments sensortags are used for real time zonal temperature readings. The mobile application relays the occupant preference along with the location to a central server that also hosts our learning algorithm to learn the environment and using occupant feedback calculates the optimal temperature set point. The entire process is triggered upon change of occupancy, environmental conditions, and or occupant preference. The learning algorithm is scheduled to run at regular intervals to respond dynamically to environmental and occupancy changes. We describe results from experimental studies in two different settings: a single family residential home setting and in a university based laboratory space setting.

In the third part we present a game-theoretic (auction) mechanism, that re- quires occupants to “purchase” their individualized comfort levels beyond what is provided by default by the building operator. The comfort pricing policy, de- rived as an extension of Vickrey-Clarke-Groves (VCG) pricing, ensures incentive- compatibility of the mechanism, i.e., an occupant acting in self-interest cannot ben- efit from declaring their comfort function untruthfully, irrespective of the choices made by other occupants. The declared (or estimated) occupant comfort ranges (functions) are then utilized by the building operator (HVAC system operator) – along with the energy cost information – to set the environment controls to optimally balance the aggregate discomfort of the occupants and the energy cost of the building operator. We use realistic building model and parameters based on our test facility to demonstrate the convergence of the actual temperatures in different zones to the desired temperatures, and provide insight to the pricing structure necessary for truthful comfort feedback from the occupants.

In the second part we develop a consensus algorithm for attaining a common temperature set-point that is agreeable to all occupants of a zone in a typical multi- occupant space. The information on the comfort range functions is indeed held privately by each occupant. Using occupant differentiated dynamically adjusted prices as feedback signals, we propose a distributed solution, which ensures that a consensus is attained among all occupants upon convergence, irrespective of their temperature preferences being in coherence or conflicting. Occupants are only as- sumed to be rational, in that they choose their own temperature set-points so as to minimize their individual energy cost plus discomfort. We use Alternating Direction Method of Multipliers (ADMM) to solve our consensus problem. We further establish the convergence of the proposed algorithm to the optimal thermal set point values that minimize the sum of the energy cost and the aggregate discomfort of all occupants in a multi-zone building. For simulating our consensus algorithm we use realistic building parameters based on the Watervliet test facility. The simulation study based on real world building parameters establish the validity of our theoretical model and provide insights on the dynamics of the system with a mobile user population.

This work was motivated by the problem of computing optimal commonly-agreeable thermal settings in spaces with multiple occupants. Most of the current studies and solutions developed for building thermal control have been designed independent of the occupant feedback. An acceptable temperature range for the occupancy level is estimated, and control input is designed to maintain temperature within that range during occupancy hours. Consider office floors with cubicles, conference rooms, student dorms, homes, and other multi-occupant spaces where temperature set-points on thermostats are chosen irrespective of the number of occupants and their individual preferences. This existing approach is not only non user-centric but also sub-optimal from both energy consumption and occupant satisfaction/productivity perspectives. It is thus highly desirable for such multi-occupant spaces to have a mechanism that would take into account each occupant’s individual comfort preference and the energy cost, to come up with optimal thermal setting. Individual occupant’s feedback and preference can be obtained through wearable sensors or smart phone applications. In this work we propose algorithms that take into account each occupant’s preferences along with the thermal correlations between different zones in a building, to arrive at optimal thermal settings for all zones of the build- ing in a coordinated manner. This approach is also more performance efficient than controlling the zonal temperature set-points of the building in an isolated manner.