The review of the DOE-2 runs indicates that the use of average occupancy schedules can overestimate the savings available in occupant sensing lighting controls. The use of peak occupancy and lighting schedules is evidently preferable for such models. When the lighting is not controlled by the occupancy sensors, the difference in the predictions is negligible.

The details of the building in the model, the location and the utility rates can all affect the specific results. However, the building in this model does have a ratio of energy charges to demand charges around 65%, which corresponds to the average PSC customer with SG rates. This, and the relative normality of the building model, indicates that this building can be considered as representative of office facilities. The consistency of the results across the different locations indicates that the results are not especially sensitive to location effects. If any location effects were to be expected, it would be related to heating dominated climates, where the increase in natural gas charges would more completely offset the electrical savings. In cooling dominated climates, the expected savings would tend to increase, but the difference between the peak schedule model and the average schedule model would continue to vary as the significance of the demand charges varies.

The rate structures and economic data used in this work are quite basic, with the simple fixed cost of natural gas and the constant energy and demand charges. The electrical rate details are considered in the results shown in Figures 3.1 through 3.6 (3.1, 3.2, 3.3, 3.4, 3.5, 3.6) and are consistent with the understanding that the difference between peak and average schedule model persists over a broad range of demand and energy charges. Clearly, as the significance of the demand charges increases, the difference between the models increases.

Certainly there would be differences in the economics when rate schedules with time-of-use, ratchet or real-time pricing are used. However, the nine hour period of workdays which is the arena of this work is within the on-peak period for most rate structures. As shown in the DOE-2 runs, the peak demand occurs typically in Hr15 or Hr16, while the occupancy and lighting schedules used are identical for the periods outside of 8 AM to 5 PM on weekdays. It is likely that the use of more sophisticated rate structures will affect the details, but not the overall results.

From analysis of the DOE-2 models and the results of applying average or peak-and-adjusted-average schedules for occupancy and lighting, it is evident that for adequate economic evaluation of occupancy-sensing lighting controls, peak occupancy data should be part of the model. In order to predict the peak occupancy rates for an appropriate range of set size and average occupancy rates, a set of curve fits were made and a software routine developed to assist in the calculations. This provides a simple means to estimate the peak occupancy information necessary to predict the potential savings due to occupancy sensing lighting controls.

Comparisons of the predictions of the software routine and the recorded data show good agreement. Comparisons with the limited available peak and average occupancy data from other facilities shows that the original NCAR data and the predictions based on that data are reasonable.

The current predictions are based on a data set which only extends to 50 rooms. The result is that the effective limit at present for this tool is 50 offices. This is significantly smaller than many facilities, but can be directly applied to those at or under this size.

The use of peak occupancy data in building simulation models to determine the potential savings due to occupancy-sensing lighting controls is necessary, in order to avoid errors in predicting the effect on peak demand. As the demand charges become a greater portion of the total electrical charges, as is currently the trend, the significance of this issue increases.

The lack of available peak occupancy data has been addressed by the development of a simplified prediction tool which uses the number of rooms and the average occupancy of those rooms to calculate the associated peak occupancy rates. Since these two parameters are known or can be estimated for most lighting control design projects, this prediction tool allows greater accuracy in developing the information which is the basis for evaluating the usefulness of occupancy-sensing controls. As is shown in the work presented, these controls can substantially reduce both the energy use and peak demand of the facilities where they are applied.

The work shown here is a only a beginning in the investigation of the significance of occupancy data in building models and design decision support. There are many associated issues which deserve further consideration.

First, additional variations in the DOE-2 models would be informative. The interaction of lighting and HVAC systems deserves more analysis, particularly for the locally controlled spaces which are the typical domain of occupancy-sensing controls. Additional work with substantially different buildings, locations, base schedules, HVAC systems and building plants would provide more information about the useful limits of the data developed here. In addition, there is the opportunity to investigate the effects of combinations of lighting controls, using the work presented here to refine the occupancy sensor portion.

As mentioned above, the application of various rate structures would demonstrate the sensitivity of the results to rates and rate structures. The wide and increasing range of utility rate structures means there is a wide and increasing field for refinement of this work.

As for the prediction tool, certainly further curve fit approaches should be considered. Non-linear equations using the data already developed would be a reasonable next step. Incorporating additional data and revising the tool is another logical step. Also, it may be that there are additional independent variables, beyond number of rooms and average occupancy, which can help produce a better prediction tool. However, the relationships between set size, average occupancy and peak occupancy, as well as the successful validation of the simplified prediction tool, indicate that these two independent variables are currently the best available information for such predictions.

The prediction tool could be extended above the current limit of fifty rooms by revising the procedure for selecting which rooms are in a set. The procedure used in this work does not allow duplication of any room within a set, by rejecting rooms which have been selected at random but are already part of the set. The procedure could be revised to permit duplication. Such random selection would proceed until the desired set size is achieved. Since this modification would necessarily include rooms being counted twice or more in a set, the variance among the sets may increase. This procedure would allow for any set size, with the recognition that the sets can only be composed of some combination of the 99 offices in the original data. This would extend the applicable range of the prediction tool.

Perhaps the most important future work would be to acquire additional data from other facilities and evaluate the correspondence across facilities, for both the data and the prediction tool. It is certainly feasible that the NCAR facility has inherent idiosyncrasies which make the basic data atypical, and the only means for determining this issue is a thorough comparison with other facilities.

Finally, future work is to perform the same occupancy data analysis on the remaining NCAR data, for the remaining 88 rooms for which data is available. In particular, this should be done for rooms which are not in the office or office service categories but are typically found in office buildings, such as conference rooms, circulation spaces, filing or reading rooms. This could extend this work, done on office and office service spaces, to cover a range of office buildings and related facilities.

References

Table of Contents