FELLSMERE ELEMENTARY SCHOOL

The third project in which occupancy sensors were installed is an elementary school in Central Florida which is serving as a pilot project to demonstrate energy savings in public buildings similar to that achieved by the Texas LOANSTAR project (Verdict et al. 1990). Termed FLASTAR (Florida Alliance for Saving Taxes and Resources), the project has entailed the comprehensive metering of a Florida elementary school with which to demonstrate energy savings potential. Over twenty channels of weather and sub-metered energy data has been collected since April 12, 1995.

The facility is composed of the main school building, with an attached new wing and various portable classroom areas. All school lighting circuits are individually sub-metered so that this end-use can be separated. Figure 4 details the proportions of the sub-metered end uses from electricity consumption data from April 12 to December 4, 1995 prior to the installation of the occupancy sensors. Metered lighting energy use has averaged about 17% of total facility energy consumption.

The large "other" end-use category represents refrigeration, kitchen cooking loads and miscellaneous end-uses such as computers, office equipment, and water coolers. Measured electricity consumption has totaled approximately 2,200 kWh on school days and 1,300 kWh on non-school days. On this basis, annual estimated energy consumption for the 35,000 square foot facility is approximately 75 kBtu/ft2. During the summer of 1995, the first retrofit, replacement of aging chillers was completed with an estimated 10% reduction to measured cooling energy use at the facility (Sherwin et al. 1996).

The interior lighting system is predominantly from fluorescent fixtures. Two-lamp fixtures based on the T-12 F34CW lamp with magnetic ballasts are most common with 513 of this type and 133 of mixed one, three, and four-tube fixtures. As audited, the connected lighting load is 59.0 kW or about 2.0 W/ft2. This compares to 1.4 W/ft2 for more contemporary efficiency lighting systems for schools (McIvaine et al. 1994) and suggests potential for improved controls. Audited classroom desk-top illuminance levels were from 76 to 85 foot candles; each room is outfitted with two wall switches that control one half of the classroom electrical lighting.

Schedules strongly affect lighting energy consumption. The last day of regular school occupancy for the Spring semester at Fellsmere Elementary was on June 6, 1995. However, during the summer, some of the faculty and secretaries were present from Monday to Thursday from 8 AM to 3 PM. Custodians were also on site from Monday to Thursday from 6:30 AM to 3:30 PM. Summer school was not held in the the portion of the building metered in the project. The Fall school schedule resumed on August 21 and continued until December 15 and faculty and staff remained until December 22. Spring session commenced on January 3, 1996.

Since metered data showed lighting accounts for about 17% of electrical end use at this facility, an occupancy sensor retrofit appeared to possess considerable promise. The school staff appears to make efforts to turn off lights after hours; however, there are numerous data to show lights being inadvertently left on after hours and on weekends (Sherwin 1996). A previous technical assistance report (TAR) and analysis for the Institutional Buildings Program (IBP) had estimated a savings for the retrofit of 25,960 kWh per year based on an assumed 20% reduction in daily lighting hours at the facility (Bosek, Gibson and Associates 1995). Estimated project cost was $10,192 with a 4.1 year simple payback.

The occupancy sensors were installed on December 15th. A total of 59 controls were installed in the facility; 39 ceiling-mounted PIR sensors were placed in classrooms and the 20 wall-mounted units were installed in office and administrative locations. The total cost for the sensors and hardware was similar to that at Northwest Elementary $5,154 (or $87/control). However, the cost of labor for installation was much higher at $9,365. The labor cost for the installation is difficult to reconcile since the estimate shown by R.S. Means Mechanical Estimator is only 3.5 hours per sensor installation-- an allowance which already seems liberal given our experience at Northwest Elementary. The TAR estimate for the retrofit labor was $3,803. The large disparity in labor costs for the installation are currently unexplained.

The first analysis of the measured lighting load profile for school days showed an increase of lighting electricity consumption of approximately 27% from 16.70 kWh/Day to 21.2 kWh per day, as shown in Figure 5. On the other hand consumption on non-school days dropped by 20% from 6.91 kWh/Day to 5.53 kWh. Based on previous installation experience we suspected that the sensors were poorly installed or improperly adjusted.

On February 22, 1996 the occupancy sensors were tuned in an effort to increase the energy savings. Tuning consisted of reducing the time delay from 12 minutes to approximately 7 minutes in most areas and changing the program selection. The program dictates which technology (ultrasonic and/or infrared) is used to initially turn on the lights and which technology is used to keep the lights on. Prior to tuning, either ultrasonic or infrared would turn the lights on. This was changed to a setting where both technologies must detect movement in order for the lights to come on. As shown in Figure 5, this resulted in an improvement in performance, but still did not produce effective savings.

Although the tuning reduced the light energy use, usage was still greater after the sensors were installed and tuned than with manual switching. We suspect this is due to false positives occurring and inadvertent tripping of the sensors when occupants enter the space momentarily. The reasons for the poor initial performance seem to be a combination of factors recently described by the county energy coordinator (Aiken 1996). The specific controls installed were obtained through a procurement process in which the lowest bidder was selected. The acquired equipment was found to possess characteristics which may have compromised performance. Based on examination of the data, it appears as if a number of the ultrasonic sensors are falsely triggering during evening hours, increasing consumption. Another cited complaint was the long "strike time" of the sensors; once lights were turned off, they would not turn back on for some 11 seconds. This led the installation crew to alter the sensor set time delay in some locations to the maximum available (15 minutes). As described above, both in our studies and those performed by PNL (Richman et al. 1994), proper setting of the device time delay is crucial to achieving potential energy savings.

A further reduction to potential savings at the facility may be behavioral (LaPointe 1996). Prior to installation of the control sensors, all facility staff punctually turned off lights when leaving unoccupied spaces. However, now staff leaves all occupancy sensor switches with the room lights to be triggered on when an occupant enters spaces. Based on observation by facility staff, lights are now on more of the time in the average classroom than they were prior to the retrofit since the typical space is left on for 7-minutes after it is vacated until the occupancy sensor turns off the lighting. Also, even a momentary visit by a single individual to a room or rooms in this configuration will result in the lights being on for 7 minutes, whereas they would likely not be powered at all in this instance under manual control. Regardless, the failure in this case of the addition of occupancy sensors to produce savings as installed, points to the importance of proper specification of equipment, a careful installation and setup, and adequate instruction to users. Such commission is critical to achieving expected energy savings.