Demand-controlled ventilation (DCV) systems have someadvantages over other building ventilation strategies, as theycan maintain acceptable indoor air quality (IAQ) of a spacewhile reducing the overall energy consumption of buildingventilation. When selecting sensors for DCV, it is important toconsider their performance and cost. Over the past severalyears, there has been an increase in the availability of low-costsensors. However, the feasibility of using the currently availablelow-cost sensors within DCV is an area that requiresfurther investigation. The focus of the work presented in thispaper is to evaluate the feasibility of current low-cost carbondioxide (CO2) sensors for use in DCV using a controlled environment.The performance of three low-cost CO2 sensormodels was verified to determine their suitability in the controlof DCV. Preliminary testing revealed unacceptable inaccuracyin one of the three sensors. This sensor uses micro-hotplatetechnology for gas sensing and was excluded from detailedtesting. The two sensors tested in detail use nondispersiveinfrared (NDIR) technology.

The accuracy of the first NDIR sensor (model A) was satisfactory;some of the sensor measurements deviated from thedosed concentration by more than 100 ppm, but remainedwithin 150 ppm. The nonlinearity of sensor model A wasgreater than model B but was acceptable—the maximum deviationfrom the linear line of best fit ranged between 55 and92 ppm. The repeatability of sensor model A was acceptable;sensor measurements during the three days of testing werealways within 100 ppm when measuring the same CO2 concentration.In fact, the maximum recorded nonrepeatability was73 ppm. The hysteresis of sensor model A was satisfactory;most sensor measurements were within 100 ppm and all werewithin 150 ppm when measuring the same CO2 concentrationwhen approached from varying directions. However, sensormodel A had a tendency to underreport CO2 concentrations,which could reduce IAQ if the sensors were not calibrated orif the tendency to underreport was not considered in the controlalgorithm. Sensor model A consistently underreporteddecreasing CO2 at a larger magnitude, which would likelycause the DCV system to turn off sooner than desired, potentiallynegatively impacting IAQ.

The accuracy of the second NDIR sensor (model B) wasfound to be better than model A and was deemed acceptable. Thesensor measurements were always within 100 ppm of the dosedconcentration. Sensor model B did have a tendency to overreportCO2 concentrations, which could result in more energy consumptionthan ideal; however, the impact is expected to be low due tothe better accuracy of the sensor. The nonlinearity of sensor modelB was satisfactory; the maximum deviation from the linear line ofbest fit ranged between 30 and 55 ppm. The repeatability of sensormodel B was satisfactory; all but one of the sensor measurementsduring the three days of testing were within 100 ppm whenmeasuring the same CO2 concentration. The hysteresis of sensormodel B was exceptional; sensor measurements were typicallywithin 20 ppm when measuring the same CO2 concentrationwhen approached from either high or low concentrations andwere always within 60 ppm. Sensor model B was found to be suitablefor use in DCV.

This work shows that ultra low-cost CO2 sensors in thearea of $27 CAD ($20 USD) each may not be suitable for DCV,but that low-cost CO2 sensors in the area of $80 CAD($60 USD) could be suitable for developing a low-cost controllerand sensor package for managing indoor CO2 concentrations.