Analytical Methods for Application to Sensor Networks for Lighting Control

Abstract

This paper describes a data processing and evaluation framework for application to a lighting control sensor network. Many buildings include systems to detect occupancy and control building services. Current systems use single measurement points to detect occupancy, and there can be significant uncertainty associated with the measurement of occupancy. More effective occupancy measurement and control are provided by sensor networks. A sensor network reduces uncertainty because data from more than one detector provides converging information concerning space occupancy. While a network of sensors reduces the uncertainty associated with individual sensor measurements, the utility of a sensor network for control depends on the analysis techniques applied to the data stream, and the ability of these techniques to produce results that better correspond to occupancy than current systems. Eight data processing algorithms are described: logical functions (OR, AND, & MAJORITY); moving average; rule-based reasoning; Bayesian belief network; least squares estimation, and; artificial neural networks. Three metrics that can be applied to evaluate the effectiveness of the data fusion methods are also described: the total occupied time measured by single vs. multiple sensors; the ϕ coefficient, and; the number of times that a controller using the associated method would have taken an inappropriate action (that is, switching the lights off in an occupied space [a false-off] or switching the lights on in a vacant space [a false-on]).

Keywords:

• Light control Analytic Methods
• Programming
• Networks
• Sensors.