Multi-Sensor Single-Actuator Control of HVAC Systems

Abstract

It is common to control several rooms in a\nbuilding with a single sensor in one of the rooms and a\nsingle actuator driving just one control element such\nas an air damper. New, low-cost, wireless sensor\ntechnology now offers the opportunity to replace the\nsingle sensor in one room with a network of sensors\nhaving at least one sensor per room. This paper\naddresses this multi-sensor, single-actuator control\nproblem. We used computer simulations and\noptimization to study the problem. We designed a\ncomputer simulation of the heat transfer behavior of a\nsection of a building that accounted for the effects of\nweather, building materials, ventilation, and loads\nfrom occupants and equipment. We considered ad hoc\nmethods (such as averaging) of using information\nfrom multiple sensors. We also developed a new,\nmodel-free method of using information from multiple\nsensors that is based on a simple optimization\nprocedure. The optimization procedure can be\nconfigured to optimize comfort or to optimize energy\nunder comfort constraints. We compared the\nperformance of the single-sensor strategy with the ad\nhoc strategies and optimized strategies using annual\nsimulations of a four-room, perimeter section of a\nbuilding and weather data from Sacramento,\nCalifornia. We report heating and cooling energy\nperformance along with two comfort metrics, the\naverage number of rooms within the ASHRAE\ncomfort zone and the Predicted Percentage\nDissatisfied (PDD). The results show that most of the\nmulti-sensor control strategies do better than the\nsingle-sensor strategy on the basis of both energy\nperformance and comfort. The energy-optimal\nstrategy reduces energy consumption by 17% while\nreducing PDD from 30% to 24%. The\ncomfort-optimal strategy reduces energy consumption\nby 4% while reducing PPD from 30% to 20%. The\nperformance improvements occur primarily when the\naverage load among all rooms is nearly zero, with\nsome rooms requiring heating while others require\ncooling. Under these conditions, the single-sensor\nstrategy either overcools or overheats, whereas the\nmulti-sensor strategies use almost no energy.