SECTION 4. BUILDING PERFORMANCE AND FINDINGS FROM THE IMDS

Section 4 begins with a brief description of the building. This is followed by a discussion of historical energy use data and the development of a baseline model against which to evaluate future changes in energy use that result from findings from use of the IMDS. The last section discusses the findings from the IMDS during the first few weeks of use, which includes a discussion of energy savings opportunities identified by the research team. We note that these are preliminary findings, subject to change upon further inspection of the IMDS data.

Pilot Site Characteristics and Historical Baseline Data

As mentioned, the building selected for the demonstration is a 100,000 sqft office building at 160 Sansome Street in San Francisco, also known as the Hong Kong Bank Building. The building is about 30 years old, with two 200-ton chillers that are also 30 years old. Additional facts about the building characteristics and operating patterns are described in Table 4–1 below in reference to changes in energy use over time.

Table 4–1. Building Characteristics and Features

In order to understand the current energy use and potential savings it is necessary to obtain historical energy use data. We also collected other whole-building energy use intensities to compare this building with others of its type. The historical data are needed to develop a baseline to evaluate changes in energy use that may result from the use of the IMDS. The data collected for the baseline analysis include:

• Described below

• Utility bills (energy and costs) for electricity (plus monthly peak demand) and steam use between 1991 and 1997
• One year of half-hourly electricity data

• Average daily temperatures for the city of San Francisco (Kissock, 1998)
• Hourly weather data

• Minute data for the entire suite of 90 points from May, 1998 to August 1, 1998


The data collected and presented are limited to energy use data at the request of the property manager. We do not include energy costs or rates.


Whole-Building Energy Use Comparison Data

 
Figure 4–1 shows that the site annual energy use intensity (EUI) is fairly typical compared with related benchmarks. The building used 90 kBtu/sqft-yr in 1996, which consisted of 64 kBtu/sqft-yr for electricity and 24 kBtu/sqft-yr for purchased steam. The first of the comparison data sets is the EUI for a 100,000 sqft large office building from a Northern California simulation prototype developed from energy analysis of 74 similar buildings (labeled CEC No.Cal, Akbari et al. 1993). The second EUI is the west-coast large office building average from the US Department of Energy’s Commercial Building Energy Consumption Survey (labeled CBECS-West, CBECS, EIA 1995). The third, and most similar, is the average EUI for San Francisco office buildings from BOMA (labeled BOMA-SF, Energy User News 1995).
 
 


Figure 4–1. Annual Site Energy Use (kBtu/sqft-yr) of Demonstration Site
and Comparison Buildings

Energy Use Data
 
Whole-building energy use has varied from slightly under 80 kBtu-sqft-yr to 90 kBtu/sqft-yr over the last six years (Figure 4–2). This variation is related to changes in building occupancy during system upgrades. In 1994, three floors were not in use due to construction and asbestos abatement, resulting in a lower energy use. In 1996, new tenants moved in with a large number of employees and computers, increasing plug and cooling loads. Over the past few years, lighting retrofits and conversion from constant volume (CV) to variable air volume (VAV) have occurred on about half of the floors. Table 4–2 outlines some of the changes in occupancy and construction activities in the recent operational history.
 
Table 4–2. 160 Sansome History
 

Year	Occupancy*	Construction Floors**	Other Comments
1991	600- 800	2
1992
1993
1994	200	12, 9, 5	Lower occupancy after construction. Relatively cool weather year.
1995	300-400	6
1996	400	17,18	Large kitchen removed
1997	500-600
* Occupancy figures are general estimates made by the building operator. ** Construction involved complete gutting of interior and asbestos removal.  Lighting retrofits were also completed, replacing F40/T12 lamps with T8.

 
Energy consumption declined once again in 1997, due to completed retrofits, increased automation, and the installation of motion sensors and hallway tenant switches. Similar trends can be seen in the annual maximum peak demand data (Figure 4–3). The recent peak is slightly over 6 W/sqft.

IMDS Use and Findings Logsheet

 
We have asked the building operator and the on-site technical manager to complete logsheets detailing findings made using the IMDS, shown in the appendix. They have been asked to record information about graphs that find most useful, and to print the graphs for our records. These logsheets will help us to learn the following the types of problems and characteristics of the building’s energy use that the operators find most interesting. We want to learn about how they use the IMDS, addressing specific questions such as what time increments are most useful, and what are preferred formats for viewing data. This information will help in both the technology transfer and the automation of diagnostics described in Section 5.
Figure 4–2. Total Annual Energy Use from 1991-1996
(Elec. Plus steam kBtu/sqft-yr)


Figure 4–3. Annual Maximum Peak Electricity Usage (kW)

Figures 4–4 and 4–5 show seasonal trends in energy consumption from historical data (not the IMDS). Figure 4–4 shows six years of data for each month of the year. Figure 4–5 shows the average monthly energy use over the six-year period. We see very little change in electricity usage over the year. One might expect an increase in electricity use during warmer months from cooling energy. We do, however, see some temperature sensitivity of electricity use at the daily level, as further discussed below. Steam follows the pattern one expects, with increased usage during colder months when more heat is required.



Figure 4–4 Monthly Energy Use (1991-1996).
(The first 6 bars are monthly energy use for 6 Januarys from 1991 through 1996,
followed by 6 bars for Februarys, etc.)

Figure 4–5. Average Monthly Whole-Building Energy Use
(Elec. Plus Steam in kBtu/sqft-yr for 1991-96)

Figure 4–6 shows the hourly electric load profiles for about three months (June 19 through September 30, 1997). The load profiles show that the building is extremely regular in its usage pattern. Nighttime energy use is extremely low. All HVAC systems and most equipment tend to be off at night, with HVAC coming on at about 6AM. Although we do not yet have end-use data, there appear to be four distinctive day-types that can be easily identified. First, weekends and holidays are days with low power similar to nighttime power. (There are few nighttime and weekend occupants; after-hour HVAC services are available at a relatively high price.) Next, there appear to be typical workdays that are those when the chillers are not needed. The next higher load shape represents days when one chiller was used. Finally, the highest power days are those when both chillers are used. These days correspond to the periods with the warmest weather.
 

Figure not available
 
Figure 4–6. Hourly Electric Load Profile for 160 Sansome Street.

The highly regular and well-controlled building systems suggest that basic equipment scheduling will not be where we will find energy savings. Rather, we expect that the IMDS can be used to improve chiller and cooling tower control. We will only explore these changes after we first give the on-site staff time to use the system without our intervention. The current outdated EMCS, unlike most for this type of building, does not provide any information about the chilled water supply temperature or condensing water temperature. We also expect that the overall cooling plant has poor efficiency (high kW/ton). We provide some examples here of the opportunities for improving the cooling tower performance. The cooling towers are blow-through towers with centrifugal fans, which are inherently inefficient. We will consider the savings possible with a variable frequency drive for the tower fans. We will examine the general conditions of the cooling tower, such as the fill water treatment and airflow rate. We will consider alternatives to the current cooling tower operation, such as changing the fill or water treatment, or perhaps increasing the louver area. Another possibility might be to increase the condenser flow by removing obstructions (such as the strainer, globe and balancing valve, and orifice plates, etc.) and possibly running two pumps to one chiller.
 
Weather Data and Regression Analysis
 
The relationship between electricity consumption and outside air temperature is a useful characteristic of the building’s energy use profile, and will serve as a model for evaluating changes in energy use from use of the IMDS. Daily average outside air temperature data were obtained from a nearby National Weather Service weather station through a public domain web site (Kissock, 1998). We condensed three months of half-hour data into daily averages. A regression of hourly temperature data against energy use is problematic because it is affected by autocorrelation. That is, the temperature in one hour is heavily dependent on the temperature during the previous hour. There is some auto-correlation in daily temperature data as well, but it is not as significant. See Piette et al, 1997 and Ruch et al, 199x for further discussion of this issue.

Two linear regressions, one for weekends and the other for weekdays, were developed using an analysis tool from researchers at Texas A&M known as Emodel (Kissock et al, 1994). (Weekday and weekend energy use are dramatically different, as shown in Figure 4–6 above.) EModel was designed to integrate data processing, graphing, and modeling of building energy use data to determine baseline energy consumption and calculate retrofit savings, supporting simple linear and change-point regression models. Results are shown in Figure 4–7. The upper regression line represents weekday usage and the lower regression line represents weekend usage. The regression statistics are as follows:
 

Fraction of the variance explained by the model R²=0.57
Adjusted R² adjR²=0.57
Measure of deviation from the model RMSE=299.35
Non-dimensional measure of deviation CV-RMSE=6.0%
Autocorrelation coefficient of the residuals p=0.49
Domain = 48 - 78 (deg F)
Range = 1646 - 6573 (kWh/day)
   
Figure not available
  Figure 4–7. Electric Power vs. Outside Air Temperature (7/15/97 – 7/14/98)

An R2=0.57 represents a reasonable correlation between weather and electricity usage. Even more important, however, is the low RMSE and CV-RMSE. We are not concerned with predicting energy use at a given temperature; rather, we are concerned with establishing a baseline model to estimate changes in energy use over time, which are likely to be energy savings from the IMDS. Studies suggest that in this case the RMSE is a better indicator of goodness of fit than R². The baseline model will be evaluated further in the Phase 3 effort.

We attempted to determine a long-term historical relationship between weather and energy use by comparing monthly temperature averages to our monthly energy use data. We saw little or no correlation between electricity and monthly temperature averages. (Note there is also little variation in this data – San Francisco monthly temperature averages between 1991 and 1996 ranged approximately between 50 and 65 degrees Fahrenheit.) A heating slope was observed in steam consumption data.
 
Operational Findings from Initial IMDS Data
 
In this section we review results from the first few days of data collected in early May 1998. These graphs were developed by the project team and were not initially shared with the on-site staff because we were examining how the staff would use the IMDS on their own. These results will be shared with them later in Phase 3. The graphs are screen shots from Electric Eye, the data visualization software used by the on-site operator.
 
Whole Building and Major End-Uses. The whole-building and major end-use data provide an overview of major operating trends (Figure 4–8). The graph shows whole-building power, total cooling (chillers, pumps, towers, and air handlers), total lighting, and plug loads, all in area normalized units (W/sqft). The remainder is additional miscellaneous loads such as elevators, plus zone fans such as variable-air-volume boxes. This most dramatic pattern is the sharp drop off of power each day at 6:00 PM. This reflects good tracking of tenant schedules and needs. Most of this reduction is due to effective HVAC management. Another interesting observation is that the lighting load does not exceed 1.1 W/sqft, and the plug load is less than 0.7 W/sqft. This is valuable information for design of new buildings or ducting systems where HVAC systems are typically sized to meet combined lighting and plug loads greater than 5 W/sqft.

Click for Figure 4-8

Figure 4–8. Preliminary Whole Building and Major End Use Data
 
Cooling System. The cooling system is the largest electric load for days that require cooling. Most buildings do not have measurement systems to evaluate the cooling system thermal load or electricity use. The IMDS allows for precise measurements of these loads. During early May the cooling system requires 0.8 W/sqft when the cooling system is operating without the chillers (Figure 4–9). This represents about 25% of the total load of the building. When the cooling system is operating with two chillers the load is 1.75 W/sqft accounting for approximately 40% of the whole-building electric load (Figure 4–10). In both cases cooling energy is equal or greater than the lighting load.
 
Click for Figure 4-9
 
Figure 4–9. Preliminary Cooling System Data – No Chiller Operation

Click for Figure 4-10

Figure 4–10. Preliminary Cooling System Data – Chillers in Operation

Cooling System and Chillers. The chillers are the heart of the cooling system. Ensuring that they run at peak performance is critical to optimal energy use. The two chillers in the 160 Sansome building are over 20 years old and are rated at 225 tons each. During these few days the chillers meet their specified efficiency. However, the chillers are less efficient when they operate below 100 tons, which is less than 45% of their rated capacity (Figure 4–11). During the first few days of monitoring (during May, 1998), the chilled water loads were well below the rated capacity of the chillers, requiring 1 to 2 kW/ton for the majority of the time. A good practice benchmark of 0.45 kW/ton is shown for reference. It is also important to examine the combined plant efficiency. The efficiency of the combined system is twice that of the chiller, or the chiller accounts for about half of the power of the cooling system. The tower, pumps and air handling fans account for the other half. Any efforts to improve efficiency of the whole cooling system will have to include those components of the system. Again a benchmark of 0.6 kW/ton for the total plant efficiency is shown for reference. (Note that the cooling plant efficiency capping at 3.0 is an error due to limits on the individual points, and has been corrected.)

Click for Figure 4-11

Figure 4–11. Preliminary Cooling System Operation – Graph 1

One dramatic finding from the monitoring was the chiller turn on time. It appears that the chiller was coming on each morning at 7:30am for 15 minutes and then staying off for the rest of the day. This was the result of a recent control programming change. In fact on those days, the chiller should have never have come on at all (Figure 4–12). The chiller was coming on without any load, which could have resulted in a major failure. The other problems shown in Figure 4–12 are further described below.

Click for Figure 4-12
 
Figure 4–12. Preliminary Cooling System Operation – Graph 2

The IMDS also showed a problem with controlling water flow though the chillers. When chiller 1 was off, the chilled water temperature responded to chiller 2 operation (Figure 4–13). This may be the result of chilled water was flowing through chiller 1 and that the back flow preventer or check valve was not operating correctly. This could be corrected by service of the check valve. This problem results in a low temperature boost across the evaporator. The chillers end up working harder (higher kW) for the same amount of cooling (tons), resulting in a poorer efficiency (high kW/ton). This also cuts the capacity of the chillers so the second chiller is brought on-line sooner than needed, which also increases chilled water and condenser water pumping energy.

Click for Figure 4-13

Figure 4–13. Preliminary Cooling System Operation – Graph 3

Cooling Towers. Cooling towers use a small portion of the total energy of a cooling plant. However, their operation can have a large impact on chiller energy use. It is common to operate towers to produce condenser water 10°F warmer than what is optimal. For each degree that the condenser water temperature is too high, there is a 1.2% degradation in chiller efficiency. Therefore the 10° F can be translated into 12% efficiency improvement in the chillers. At 160 Sansome the tower is controlled to supply approximately 75° F condenser water. On the day displayed in Figure 4–14, the wet bulb temperature averaged about 55° F, while the tower was supplying 75° F condenser water. The chiller should be able to operate at condenser water temperatures as low as 55 degrees (this should be re-verified with the manufacturer). This can be corrected through changes to the central plant controls.

Click for Figure 4-14

Figure 4–14. Preliminary Cooling Tower Operation – Graph 1

In addition to the high condenser water temperatures, the towers tend to cycle excessively, creating additional wear on the tower fan motors (Figure 4–15). One remedy for this problem would be to use variable speed drives on the tower fans. Operators could also adjust control algorithms to increase the on/off temperature band and should definitely use both cells with one chiller to improve the energy efficiency of the cooling plant.

Click for Figure 4-15

Figure 4–15. Preliminary Cooling Tower Operation – Graph 2
 
Pumps. Several interesting problems were found with the pumps. On the chilled water system, the chilled water flow exceeded design flow by 15% (Figure 4–16). Due to the cube law of pumping, decreasing chilled water flow to the design rate would result in a 34% reduction in pumping energy (100[1-1/(1.15)³]). It would be simple to trim the impellers on the pump to accomplish this. Another problem on the chilled water system was in the pressure drop across the chiller evaporator. The design pressure drop is 5 psi while the measured pressure drop is 8 psi (Figure 4–17). The expected pressure drop is 6.6 psi (5 * (1.15)²), indicating a possible problem with the tubes in the evaporator; cleaning should improve this. (Notice the design and measured pressure drop for the water through the condenser is much closer).
 
Click for Figure 4-16
 
Figure 4–16. Preliminary Pump Operation – Graph 1
 
Click for Figure 4-17
 
Figure 4–17. Preliminary Pump Operation – Graph 2

A more dramatic problem was found in the condenser water system. In chiller 2 the design flow is 615 gpm. The measured flow was 60% below the design flow (Figure 4–18). While increasing the condenser water flow will consume more pumping energy, it will make the chillers operate more efficiently. More flow and better heat exchange in the condenser of the chiller should improve the chiller efficiency and reduce the energy consumed by the chiller.

 
Click for Figure 4-18
 
Figure 4–18. Preliminary Pump Operation – Graph 3
 
One additional comment on the pumps is that air was found in the pipes. Thus, when the second chiller was brought on, the flow actually was reduced. This can be seen in Figure 4–12. This problem is under investigation for Phase 3.
 
Summary of Findings from Initial Use of IMDS
 
The key findings are summarized below. During Phase 3 we will be developing estimates of the energy and other impacts of these issues. The chiller start-up problem was by far the most severe problem, and was actually identified on the first day of operation of the IMDS. The building operator saw Figure 4–12 in the first training session and took immediate action to correct the problem. This problem could have caused a major chiller failure. The operator is also aware of the tower cycling, but has not yet corrected the problem.
 
Chiller Issues

• Start-up peak due to control sequence modifications.

Implications: Could have resulted in major failure of chillers from no load

• Back flow through one chiller when other chiller off.

Implications: Energy waste, chillers used more than needed. causes low delta T through chillers

• High condensing water temperature, plus only one tower cell at a time

Implications: Higher than optimal chiller energy use

• Chilled water flow 15% greater than design flow

Implications: Higher pumping energy use.

• High pressure drop on water side of the evaporator

Implications: Greater pumping energy.

• Low condenser flow (60% below design)

Implications: Poor energy efficiency
 
Longer-Term Energy Saving Opportunities – In addition to the lessons from the first few days of data, we have observed several other issues in the building’s performance. The availability of high-quality building operations data is useful to identify concepts for additional low-cost modifications and retrofit opportunities. One of the most significant opportunities for the cooling plant is to make better use of the cooling tower system. There appears to be a significant opportunity to use the towers more, and run them during mild weather in place of, or before the chillers. The towers could be used to pre-cool the building, which should result in peak demand savings on warm days. It may be cost effective to add pre-cooling coils or use the chillers as heat exchangers. It also appears that the air-handler silencers are larger than needed. The project team has observed that the fans are very quiet. The silencers increase the fan pressure drop resulting in higher energy costs. Lowering the pressure drop would also reduce the fan heat load, expanding the opportunities for free cooling. Finally, there are additional opportunities for further reductions in the lighting power densities, which will have the added savings from reducing the cooling load.

Section 1. Project Overview
Section 2. Pilot Site Selection and Technology Innovation Findings
Section 3. IMDS Description and Accuracy
Section 4. Building Performance and Findings from the IMDS
Section 5. Automation of Diagnostics
Section 6. Economic Issues and Related Technology
Section 7. Conclusions and Future Plans
Section 8. References

Appendix A. Web-Based Performance Analysis Tools
Appendix B. Diagnostic Plots
Appendix C. IMDS Points, Sensors, and Data Production Systems
Appendix D. IMDS Findings Report Log