The procedure for this work starts with applying the statistics from the accumulated and formatted occupancy data for the office room set to DOE-2 models for a typical office building. The occupancy schedules in the DOE-2 models are then varied, to determine the effects on the model predictions for energy use, peak demand and heating requirements.

A previously developed DOE-2D file, listed in Appendix A, is used as the baseline model. This file describes a 31 story office tower, with typical dimensions, materials, density and equipment. The lighting power density is set to 1.3 Watts per square foot to avoid overestimating the potential savings. This value is consistent with the "Energy Guidelines for Commercial and High-Rise Residential Buildings in Colorado" [COEC, 1997] for unit power densities in offices.

The baseline building uses "full" occupancy and lighting schedules, listed in Appendix B, covering the period from 8 AM to 5 PM. This scheduling combination is identified as Base Occupancy Base Lighting (BOBL). The same schedule is used for each weekday over the year, and a different schedule is used for weekends and holidays.

The other runs are identical to the baseline except for changes made to the weekday occupancy and lighting schedules. These changes are to the weekday schedules for the period from 8 AM to 5 PM. The changes are either made to the occupancy schedule alone or to the occupancy and lighting schedules together. The alternate schedules are also listed in Appendix B.

From the data for all the runs in the Set Size procedure, the annual hourly average occupancy was established and compiled into an occupancy profile. This profile has an average occupancy rate of 0.483 and a peak of 0.537 between 10 and 11 AM, as shown in the Basic Schedules provided in Table 3.1. This average schedule for occupancy is used in the run Average Occupancy Base Lighting (AOBL), applied to the occupancy schedule in the model, also for each weekday of the year. The average schedule is applied to both the occupancy and lighting schedule in the Average Occupancy Average Lighting (AOAL) run, modeling the control of the lighting by the occupancy sensors, also for each weekday of the year.

The peak schedule consists of two parts, one for four days of the week and the other for the "peakday", also shown in the Basic Schedules in Table 3.1. The four-day portion is an adjusted average schedule. This is calculated so that in combination with the "peakday" schedule, the weekly averages would exactly match the average occupancy schedule discussed above. The "peakday" schedule is from the averaged peak values from the Set Size procedure. For the adjusted average schedule the hourly values were reduced so that the composite "peakday"-and-adjusted-average would be equal to the corresponding hour and day of the average schedule. The equation for this adjustment is:

  Eq 3.1: Adjusted Average = 1.25 * (Average - (0.2*Peak)

The adjustment is applied to each hour's values individually. This composite schedule is used for each week over the year. For this schedule the overall average is 0.483 while the peak is 0.741 between 10 and 11 AM on all "peakdays".

The composite adjusted-average-and-peak schedules are applied as the occupancy schedule in the runs Peak Occupancy Base Lighting (POBL), and as the occupancy and lighting schedules in the runs Peak Occupancy Peak Lighting (POPL). These schedules are presented in Appendix B.

The assignment of the "peakday" to a particular day of the workweek could produce unexpected variations in the model's results. To investigate this, the POPL models for the Denver location were run with each day of the week as "peakday" in turn. The results of this investigation are summarized in Table 3.2, which shows the comparisons of the peak demand magnitude and day-hour for each of the five different weekdays as "peakday". The table shows the electric demand for the base case (BOBL) run and the average occupancy and lighting schedule (AOAL) and the five peak occupancy and lighting (POPL) runs. The column headed "Amt" is the peak demand value for that run and the column headed "Day/Hr" is the day of the month and the hour of the day when the peak demand occurs. At the bottom are comparisons of the year's peak (maximum among twelve months) for each run and the average monthly peak (average over the twelve months). The "dev Avg" entry for each of the weekdays shows the difference between that run's value and the average value for all the weekdays. In addition, these values are compared to the corresponding values for the BOBL run, showing each run's value as a percentage of the BOBL value.

The results presented in Table 3.2 show that whatever day of the week is used as the "peakdays", the change in the peak demand with a peak-and-adjusted-average schedule is only two-thirds of the change with an average-only schedule. As the "peakday" changes, the day and hour when the peak occurs also changes, and is typically on one of the "peakdays" in the month. The small differences in the magnitude of the peak demand between the different "peakdays" are always less than the difference between any of the "peakdays" and the AOAL or BOBL run results.

The results with Friday as "peakday" are closest to the average among the weekdays, for both the year's peak (dev Avg = 3.2) and the average monthly peak (dev Avg = 22.1). Based on the results of this investigation, Fridays were chosen for the "peakday" of the composite peak-and-adjusted-average schedule.

The set of five runs was made for each of three locations, Denver, Grand Junction, and Chicago, for a total of fifteen DOE-2 runs. In addition, three runs were made for New York. For each run, the monthly building energy use was recorded. The energy use was as electrical energy, electrical demand and natural gas for heating fuel.

Using the DOE-2 runs' data for checking, the energy costs were computed. The energy rates used were the Public Service Company of Colorado (PSC) electrical rate for commercial customers, and a $1.00 per 100,000 BTU natural gas charge. The PSC rate is called secondary general (SG) and is typically applied to commercial service. This rate has a monthly service charge of $15.30, an energy charge of $0.01787 / kWh and monthly demand charge of $12.55 / kW-month, but not more than $0.16 / kW used. Although this maximum electrical charge is specified, it does not pertain in these runs, since the demand portion of the electrical charges is not extremely large.

The annual energy information is used with the SG rates to calculate the annual energy costs for electrical energy, electrical demand and natural gas. This is checked against the DOE-2 economics and matches exactly.

For each location, the different occupancy and lighting schedules are compared to the Base Occupancy Base Lighting (BOBL). The results are summarized in Tables 3.3 through 3.6 (3.3, 3.4, 3.5, 3.6). The difference between the BOBL and each other run is considered the potential savings for that run. The difference and the percentage with respect to the BOBL run are shown for each of the five cost calculations. These percentages are also the percentage differences in the energy use or demand. In addition, the percentage of the electric charges and the total charges that are due to the electric demand charges are provided.

The savings for the runs with peak schedules are compared to the savings for the runs with average schedules, for occupancy only or occupancy and lighting configurations. The differences and the percentage with respect to the peak schedule are provided in Table 3.3 through Table 3.6 (3.3, 3.4, 3.5, 3.6), and again these percentages describe the energy use or demand as well.

The difference in predicted savings modeled for peak vs. average schedules for occupancy-controlled lighting is calculated as (BOBL-POPL) - (BOBL-AOAL). The percentage difference with respect to the peak schedule is also calculated as (POPL-AOAL) / (BOBL-POPL).

The value obtained from this calculation can be described as the error in predicting the savings due to using an average schedule instead of a peak schedule. Inspection of this value shows that the results are sensitive to the magnitude of both energy and demand charges. For each location, this percentage difference is calculated for demand charges of 5, 10 and 15 $/kW-month and for energy charges from 0.01 to 0.10 $/kW, with constant natural gas charge of $1.00 per therm. These results are illustrated in Figures 3.1 through 3.4 (3.1, 3.2, 3.3, 3.4).

The results shown in Tables 3.3 through 3.6 (3.3, 3.4, 3.5, 3.6) are for each of the four locations, using the SG rate from PSC. The use of average schedules overestimates the savings of electrical energy and demand, while underestimating the reduction in natural gas energy. This is true for Denver, Grand Junction and Chicago, for occupancy only (Base Ltg) and occupancy and lighting combined (Scheduled Ltg) schedules. The percentages are consistent for all locations, but do vary depending on whether the lighting is also controlled.

For the Base Ltg schedules (AOBL and POBL), the overall differences are very small or non-existent. While the basic trends still hold, the significance is quite small. The overall effect is very dependent on the relative balance of energy costs.

For the Scheduled Ltg results (AOAL and POPL), the trends are evident and more consistent. The estimated savings from models with peak schedules are consistently less than those using average schedules.

As shown in Tables 3.3 through 3.6 (3.3, 3.4, 3.5, 3.6), across the four different locations, the difference between the prediction based on the peak schedule to the prediction based on the average schedule is 9% to 12% for electrical energy, 24% to 35% for electrical demand and -50% to -54% for natural gas. This indicates that the average schedule overestimates the electrical energy savings and demand reduction by applying occupancy sensing controls, and underestimates the increase in natural gas. The differences are consistent and significant.

Applying the PSC SG rates and combining the costs into total energy charges, the overall change from the base to either of the Base Ltg runs (AOBL and POBL) is under one percent and considered negligible. For the Scheduled Ltg runs, the differences between the baseline (BOBL) and either Scheduled Ltg (AOAL or POPL) run is much greater. The overall predicted savings is between 7.2% and 7.8% for the average schedule (BOBL-AOAL) and 3.0% to 4.5% for the peak schedule (BOBL-POPL).

When the difference in the estimated savings is calculated, the results are consistent with the trends described above. The Base Ltg runs show small differences which are quite sensitive to the electricity rates applied. However, for the Scheduled Ltg runs the differences in the predicted savings persist. The peak schedules predicts total annual savings that are between $40,000 and $52,000 less than the savings predicted by the average schedules. Using the peak schedule predicted savings as the "true" estimate, this means the average schedule predictions are 73% to 140% too high. 

The same percentage difference for the Scheduled Ltg runs ((AOAL-POPL) / (BOBL-POPL)) is calculated for a range of demand and energy charges as shown in Figures 3.1 through 3.4 (3.1, 3.2, 3.3, 3.4). These figures indicate that the difference is consistent among the four locations. The differences are quite sensitive to the demand rates when the energy charges are comparatively low and converge to around 25% as the energy rates increase. Each of these figures shows the locus of the PSC SG rate difference, corresponding to the data in Tables 3.3 through 3.6 (3.3, 3.4, 3.5, 3.6).

To evaluate if the results were dependent on the values of the average occupancy, the average and peak schedules were increased. This was done by separately computing the standard deviations associated with the Set Size occupancy values, for each hour, for both the average and peak occupancy rates. The hourly average and hourly peak occupancy rates are each increased by the appropriate value of the standard deviation. The adjusted average profile was computed for the revised profiles, producing the set of profiles labeled as "Increased by One Standard Deviation (Denver +1 StDev)" in Table 3.1 and the DOE-2 occupancy and lighting schedules shown on Appendix B. These profiles were applied to the AOAL and POPL runs for Denver, and the results are provided in Table 3.7 and Figure 3.5.

The same procedure was performed for an increase of two values of the standard deviations. These profiles are shown in Table 3.1 as "Increased by Two Standard Deviations (Denver + 2 StDev)" and in Appendix B. The results of the DOE-2 models are summarized in Table 3.8 and Figure 3.6.

Compared to the results shown in Figure 3.1, the two runs with +1 or +2 StDev occupancy rates have increasing "steepness" in Figures 3.5 and 3.6. This trend can be considered a progression from 0 to 1 to 2 increments of standard deviations added to both occupancy rates. As the difference between the average and peak schedules increases, the difference in potential savings between using average or peak occupancy schedules also increases.

Finally, a tabulation of the occurrences of the peak demands in the DOE-2 runs shows that the timing of the peak demand is affected by the peak schedules (POPL). Tables 3.9 through 3.12 (3.9, 3.10, 3.11, 3.12) show, for each month for each location, the dates and times of the electrical peak demand. For the Denver runs with schedules increased by one or two standard deviations, the same information is provided in Tables 3.13 (Denver +1 StDev) and 3.14 (Denver +2 StDev). For any single location, the day and time of the peaks are almost the same for all the runs except for the POPL. The POPL schedule, with the peaks on Fridays, changes the date when the peak occurs for over half the months each year, and the date of the annual peak for each location. These results are presented in the tables, with the days which are "peakday" (i.e. Friday) shown in bold.

The time of day when the peak occurs changes slightly for some of the runs. For all locations, the annual peak occurs in July and for all locations but New York, the annual peak occurs in Hr15 or more likely Hr16. The months with the most cooling loads typically have the monthly peak demand occur in Hr15 or Hr16. This is also true for the "peakdays" comparisons presented in Table 3.2.

Figures & Tables for Chapter 3

Chapter 4

Table of Contents