Human-centered sensor-based Bayesian control: Increased energy efficiency and user satisfaction in commercial lighting

The need for sustainable, efficient energy systems is the motivation that drove this research, which targeted the design of an intelligent commercial lighting system. Lighting in commercial buildings consumes approximately 13% of all the electricity generated in the US. Advanced lighting controls1 intended for use in commercial office spaces have proven to save up to 45% in electricity consumption. However, they currently comprise only a fraction of the market share, resulting in a missed opportunity to conserve energy. The research goals driving this dissertation relate directly to barriers hindering widespread adoption---increase user satisfaction, and provide increased energy savings through more sophisticated control.; To satisfy these goals an influence diagram was developed to perform daylighting actuation. This algorithm was designed to balance the potentially conflicting lighting preferences of building occupants, with the efficiency desires of building facilities management. A supervisory control policy was designed to implement load shedding under a demand response tariff. Such tariffs offer incentives for customers to reduce their consumption during periods of peak demand, trough price reductions.; In developing the value function occupant user testing was conducted to determine that computer and paper tasks require different illuminance levels, and that user preferences are sufficiently consistent to attain statistical significance. Approximately ten facilities managers were also interviewed and surveyed to isolate their lighting preferences with respect to measures of lighting quality and energy savings. Results from both simulation and physical implementation and user testing indicate that the intelligent controller can increase occupant satisfaction, efficiency, cost savings, and management satisfaction, with respect to existing commercial daylighting systems.; Several important contributions were realized by satisfying the research goals. A general model of a daylighted environment was designed, and a practical means of user preference identification was defined. Further, a set of general procedures were identified for the design of human-centered sensor-based decision-analytic systems, and for the identification of the allowable uncertainty in nodes of interest. To confirm generality, a vehicle health monitoring problem was defined and solved using these two procedures.; 1'Daylighting' systems use sensors to determine room occupancy and available sunlight, and automatically dim the lights in response.