ABSTRACT

     Electronic timers were installed in 1998 to control 365 mercury vapor and high pressure sodium lamps in a large tool and die shop. The energy savings were 71,200 kWh per month. An economic analysis based on average 1997 power costs and ignoring lamp life indicated a savings of $5553 per month. When the impact of peak demand and incremental cost was considered, the savings dropped to $3169 per month. Because of peak demand charges, 58% of the savings initially came from 15% of the lamps; the ones which were shut off during the daytime. In the year following the project, electric power was deregulated in Pennsylvania. The average power cost came down, but the incremental (last bracket) power cost went up. The project savings increased to $3345, but the impact of peak demand dropped by 59%. The return on investment plus any savings calculations which might be used in a performance contract depend a great deal on the regulatory environment as well as how and when the lighting is controlled.
     When the authors tried to take lamp life into account, they were surprised at the lack any of published data for HID lamp life for continuous operation. Data was collected for a full year, and the results indicated that the life is surprisingly long, up to 60,000 hours if the lamps are left on. Turning the lamps on and off once a day reduced the lamp life to figures that are more consistent with published data.
 

INTRODUCTION

     What could be easier than analyzing a lighting retrofit? You can either reduce the wattage (a retrofit) or reduce the hours of use (timers, occupancy sensors, etc.). You then divide the savings by the installed cost. The same analysis works for anything from an occupancy sensor to a variable speed drive to a full-blown energy management system or does it?
     In the simple case the savings are the reduction in wattage multiplied by the number of hours of operation times the energy cost. The payback is the initial cost divided by the savings. If a T-8 lighting fixture costs $100 and it saves you 100 watts (0.1 kW) for ten hours a day for six days a week at a cost of eight cents per kilowatt-hour, the simple payback is:

$100 / (10 hours/day x 6 days/week x 52 weeks/yr x 0.1 kw x $.08 per kwh) =
$100 / 312 kwh x $.08 /kWh  =  4.0 years

     An increase in demand means that you must buy more of the first tier (high cost) energy in addition to paying the higher demand charge. Under the old regulated environment, the demand charge was $12.76, but the impact of the pricing tiers pushed the cost of an extra kilowatt of demand to $25.23 (excluding tax) on the HT rate. The energy cost was 8.29 cents per kilowatt hour for the first tier price, and 2.74 cents for the third (lowest) tier price.(1) A sample calculation showing the pricing tiers and the impact of an increase in demand is shown in Figure 1. A 10-megawatt HT customer with 500 "hours of use" ended up paying an average cost of 7.8 cents per kilowatt-hour for power.
     Suppose a PECO HT customer shuts off a one kilowatt (kW) lamp for eight hours a day for 30 days. This would save 240 kilowatt-hours of energy. If this is done on the night shift, it would save 2.74 cents per kilowatt-hour (the third tier energy price), or $6.57 for the month. When this is done on the day shift, it also avoids adding one kW to the peak demand. In addition to the $6.57 of incremental energy savings, there is a savings of $25.23 in demand charges. Instead of a savings of $6.57, there is a savings of $31.80, which is five times the incremental energy cost.  Conventional analysis says that shutting off the lamp would save the average cost of energy, which is the 7.8 cents per kilowatt hour amount calculated in Figure 1. The savings for 240 kilowatt-hours would be $18.72. The savings figure you calculate depends on when you turn the lights off and whether you use the average energy cost or the incremental cost.
 

THE IMPACT OF DEREGULATION

     During the pilot program a competitive energy cost was about the same as the third tier price, or about 2.74 cents. There were separate charges for "Transmission Charges," "Distribution Charges," and "Transition Charges" (stranded cost recovery). Each of these three items had a demand charge, and the distribution and transition charges also had energy pricing tiers. The net effect was a cost of $18.66 for an extra kW of demand. The average cost dropped to 7.5 cents per kilowatt-hour. The incremental cost of additional energy, if there is no increase in monthly peak demand, is the competitive price paid for energy. In this example, that is 2.74 cents.
     For 1999, transmission charges are included in the competitive price offering. The competitive rate for combined energy and transmission charges for an account of this size is about four cents, and in most cases that price is independent of demand. (The load profile may impact on the competitive bid offering.) The impact of an extra kilowatt of demand is $l0.30, and the average cost of energy is 6.5 cents. The incremental cost of energy is the competitive price of four cents. That is, the cost of using an extra kilowatt-hour during off-peak times is four cents (Figure 2). When you are purchasing energy at a flat rate per kilowatt-hour, there is a tendency to use that price as the savings and ignore the demand charges. This may or may not be correct, depending on when the lighting is turned off. It is interesting to note that, in this case, two-thirds of the local power company's revenue comes from stranded cost recovery, and even that is demand sensitive. For someone not buying power competitively, the impact of an extra kilowatt of demand is $21.58, the average cost is 6.75 cents, and the incremental cost of energy is 2.43 cents. These calculations become extremely complex, since the capacity (including transmission), distribution, and transition charges all have demand charges plus three energy price tiers which are demand sensitive.
     Clearly both the pricing scenario and the regulatory environment have a dramatic impact on the savings calculation This is important for the ROI (return on investment) analysis required by corporate financial managers before an investment decision is made. It would also dramatically impact the shared savings calculations that go along with performance contracting.
 

BULB LIFE

     The US Department of Energy sponsors the "Green Lights" program. At a "Green Lights" workshop participants are taught that turning a lamp on and off shortens its life, but it also results in fewer hours of use over the course of a year.(2) For example, turning a bulb off once during a ten-hour workday might cut its life from 20,000 hours to 15,000 hours. However, if this also resulted in a reduction in the hours of use from 2500 hours a year to 1875 hours a year, it results in an eight year life either way. The assumption that it ends up as a "wash" is a good one for most situations - or is it?
     Lamp life is usually specified on the package and in most supply catalogs.(3) The standard figure is based on ten hours of use per cycle. In a typical office the lamps go on at eight in the morning and are turned off at six in the evening. The life is a statistical average. Every lamp is different. For a T-8 Fluorescent light, for example, the impact of cycling is well documented. (Figure 3) If a lamp is left on continuously, it will last about 37,000 hours. If the lamp is run an average of three hours every time it is turned on, one can expect 15 to 20,000 hours.
     For a high intensity discharge (HID) lamp, the lamp life is well documented for operation of ten hours per cycle or less. The ten-hour life can be found in most industrial supply catalogs. There does not appear to be any published information for periods of over ten hours. This is quite surprising. A 400-watt sodium lamp is extremely common. Sources checked included GE, Phillips, Sylvania, the "Green Lights" program, and the Lighting Research Institute.(4) In addition, one of the authors spent a day walking around at Light Fair in New York asking about HID life for continuous operation. Although there was no hard data, several people offered an opinion that a lamp would last 10 to 25 percent longer if it was left on continuously. For a ballast, the information was even sketchier. GE did offer that a ballast should last 60,000 to 100,000 hours. However, a sodium ballast will tend to continue firing the igniter if the lamp burns out. This will destroy the igniter in a 6 to 12-month period. With mercury and metal halide fixtures, igniter burnout is not normally considered to be a problem. (4)
 

REPLACEMENT COST

     As with many other things in life today, it is now often cheaper to throw away a complete fixture and replace it rather than try and replace only the ballast or the igniter. Today the cost for a replacement fixture for a 400-watt sodium lamp, including ballast, is about $200.  The cost of the lamp is about $27. In addition, there is a labor cost involved. In our case study, these lamps are located 30 feet up in the ceiling. The only access is by "cherry picker." Replacing a bulb safely requires two people, and the high lift must be maneuvered around the equipment and work-in-process on the shop floor. The Budd Company's cost figures are proprietary, but if a typical manufacturing labor cost is $35 per hour, including benefits, the cost of labor to replace a bulb or fixture is about $20.
 

THE LIGHTING TIMER PROGRAM

     Unit 6 has three "Bays." Each bay is about the size of two football fields laid end to end. B&C-Bays are used for die machining and tryout work. A-Bay is used for die storage. Each die set is roughly a nine-foot steel cube weighing 20 to 30 tons. A drawing which shows the layout of the shop, the location of the lighting fixtures, and a sample of the data collected appears in Figure 4.
     Each bay has three lighting panels. A decision was made to install an electronic programmable timer to control each panel. Each lamp, with the exception of the cross aisles, was turned off from 12:30 a.m. to 6:30 a.m. and all day on Sundays. In addition, lamps were turned off in A-Bay during the day shift. A-Bay has a significant amount of natural lighting with large window areas down the length of one side, in addition to skylights overhead and windows on each end. Also, A-Bay is primarily used for die storage and not for machining work. Each timer has a manual override button. On unusually dark days or on occasions when more light is needed, a Unit 6 employee would push the override button. These occasions were relatively rare.
     Normally lamp replacement is done by going down through the bay with a "cherry picker" and a two-man crew at regular intervals - about every four months. For this project, data was collected for six months before and six months after the installation of the timers, without any relamping. By the end of this period, about 15 percent of the lamps had failed. There was enough design margin in the lighting layout that this reduced level of lighting did not cause problems. The number of lamps that burned out over this period is shown in Figure 5. In addition, a "lighting logger," a photocell attached to a timer, was used to monitor the number of hours of lamp operation.
 

RESULTS - ENERGY SAVINGS

     Again, The Budd Company's actual cost data is proprietary. However, the catalog price for the timers employed is $220 including tax. The timers are four-channel, seven-day programmable timers manufactured by Intermatic. They are readily available from most electrical supply houses. In addition, if we assume a generous eight hours of installation labor at $35 per hour (each timer required mounting on a column, conduit, and wiring of up to 12 breakers), we come up with a total project cost of about $500 per timer or $4500 for nine timers.
     The lights were turned off for 60 hours per week. (i.e. six hours a night and 24 hours on Sunday). In addition, in A-Bay the lights were turned off during the daylight hours.
     In B-Bay and C-Bay the savings for each of the six timers installed (ignoring lamp life and using 1997 tariff energy costs) are:
     21 sodium lamps per timer x 0.4 kW x 1.15 (15% energy for the ballast)   =
     9.66 kW x 60 hours / week x 4.3 weeks / month                                        =
     2492 kWh / month x $.0274 / kWh                                                            =
     $68.28 / month
Plus
     19 mercury lamps per timer x l.0 kW x l.15 x 00 x 4.3                               =
     5637 kilowatt hours x $.0274IkWh                                                            =
     $154.46 / month.

     The total savings are $222.74 per month. The payback is 2.2 months.

     In A-Bay each of the three timers controls fewer fixtures, but the impact of peak power demand comes in to play. Each timer controlled fourteen 1,000 watt mercury lamps. The lamps were shut off for 108 hours each week. The savings per timer are:
     Energy – 14 x l.0 kW x 1.15 x 108 x 4 .3 x $.0274  =  $204.86 per month.
     Demand – 1 x / kW x l.15 x $ 25.23 per kW            =  $406.20 per month.

     The total savings are $611.06 per month. The payback is less than one month.

     The total annual savings for nine timers is $38,016 per year. Because of the impact of demand, 58 percent of the savings initially came from the three timers in A-Bay, even though A-Bay only contains 15 percent of the lamps that were being cycled.
 

RESULTS - LAMP LIFE

     Before installation of the timers, lamps burned out at the rate of about one every two weeks, or about 26 lamps (seven percent) per year (Figure 5). This is extraordinarily long. It would require about seven years for 50 percent to burn out, which corresponds to a lamp life of approximately 60,000 hours. There are many reasons for arguing that this is not a valid number. The sample is small. (But it is still 365 lamps). The period of time is short (six months is still a long time). Mercury lamps don't really burn out, they just grow dim.(5) (In fact, in this study they went out at a higher rate than the sodium lamps.). The data includes fixture, ballast, and igniter failure. (This means the lamp life is really longer). The environment is hostile. (There is beat, dust, and vibration. Again, this may mean the life is actually longer). We are starting with lamps which have been in service, not with new lamps. (The authors don't know how to assess the impact of this factor).
     After the timers were installed, the lamps were being cycled once a day. The rate of burnout increased to one lamp every 19 days for the sodium lamps (23,000 hour life) and one lamp every 10 days for the mercury lamps (16,000 hour life). These figures are more in line with published date The net result is a two-thirds reduction in lamp life, coupled with a corresponding reduction of one-third in the hours of use. It does not, in this case, "wash out." The result is an extra 25 bulb changes per year. At a cost of $47 per change, this is an extra $1175 per year. However, the total energy savings are so significant that the only impact is an increase in the payback time from 1.2 up to 1.25 months. In this case the impact of lamp burnout is relatively small, even though it amounts to $1175 per year.
     However, suppose that we had installed occupancy sensors that turned the lights on and off four times a day such as during breaks or during periods when there was no activity in a portion of the shop. As shown in Figure 5, this would shorten the life of the lamps to 12,000 hours. Now the lamps would burnout at the rate of 3.5. per week. The result would be an increase in lamp replacement cost of $7300 per year, which reduces the savings by 21 percent.
     The ROI is still attractive, but if this project is done on a performance contract, the impact on the economics becomes quite significant.
 

CONCLUSIONS

     Life is not as simple as we would like it to be - even for a project as straightforward as a lighting retrofit. Lamp life, on-peak and off-peak power rates, the structure of demand charges, and the effect of deregulation all come into play. In this study the lighting was either on or it was off.  If you start varying the level of intensity, or if you are dealing with something like a motor with a variable speed drive, it becomes even more complex. Energy economics are very "site specific." The impact on return on investment (ROI) calculations or performance contracts varies greatly with how a program is implemented as well as with the regulatory environment.  Whether the numbers come from an in house engineer or an outside contractor, it is important to consider how these factors will impact the economics of a project.
 

References
1. PECO Electric Company, "Electric Service Tariff," December31, 1996, p.47, Revised January
     1, 1999, p.49.
2. US - EPA, "Green Lights Program - Lighting Upgrade Manual, 7th Edition, Workshop Slide
     Hardcopies, December 1994, p.60.
3. Grainger Industrial and Commercial Supply, Catalog No.389, 1998-99, pp.922-1012.
4. McCormick, Charlie, Product Information Coordinator, Lighting Systems Department, General
     Electric, Hendersonville, NC, Private Fax Communications, "Gels Ballast Life," 4/7/97, 5/13/97.
5. Philips Lighting Company, "Guide to High Intensity Discharge Lamps," North American Phillips
     Company, Somerset, NJ, August 1991.
 

About the Authors

James L. Watters is President of Delaware Valley Utility Advisors, (DVUA) a Lansdale, PA firm that specializes in cost reduction programs for industrial and commercial clients including telecommunications, gas, and water & sewer charges as well as electric power. Jim holds a BS in Metallurgical Engineering from Illinois (Urbana), an MS in Materials Science from Stanford, and an MBA from Harvard. His prior experience includes Research Engineer for motor and transformer lamination steels at Bethlehem Steel, Plant Engineering Manager at Corning Glass, and President of Narox, Inc., an industrial gas company in Hopewell, VA. Jim started DVUA in 1991. DVUA is a Green Lights Partner and Jim is a Green Lights Surveyor Ally.

Thomas J. Flaherty is Facilities Manager for the Philadelphia Plant of The Budd Company, a
Thyssen-Budd Automotive Company facility. The company's domestic headquarters are located
in Troy, Michigan. Tom graduated from Villanova University with a BS in Mechanical
Engineering. He joined the Philadelphia Plant in 1989 as a Facilities Engineer.
 

                                                                                    Case I            Case II

Peak Demand - kilowatts                                                   10,000                  10,001
Energy Consumed - kilowatt-hours                               5,000,000            5,000,000

1st Tier Energy Use = 150 x Demand                            1,500,000            1,500,150
2nd Tier Energy Use = 150 x Demand                           1,500,000            1,500,150
3rd Tier Energy Use = All Remaining kWh                    2,000,000            1,999,700

Tariff Rates
Customer Charge                                                         $    286.86           $    286.86
 Demand Charge @ $12.76 per kW                         $ 127,600.00       $ 127,612.76
1st Tier Energy Cost @ $.0829 per kWh                  $ 124,350.00       $ 124,362.44
2nd Tier Energy Cost @ $.0550 per kWh                  $  82,500.00        $  82,508.25
3rd Tier Energy Cost @ $.0274 per kWh                  $  54,800.00        $  54,791.78

                                                  Totals                 $ 389,536.86        $ 389,562.09

Average Cost per Kilowatt-Hour                                                 $ 0.078

Cost of an Extra Kilowatt of Demand                                          $ 25.23
 

                                                                                Case I                Case II

Peak Demand - kilowatts                                               10,000                     10,001
Energy Consumed - kilowatt-hours                           5,000,000                5,000,000

1st Tier Energy Use = 150 x Demand                       1,500,000                1,500,150
2nd Tier Energy Use = 150 x Demand                      1,500,000                1,500,150
3rd Tier Energy Use = All Remaining kWh               2,000,000                1,999,700

Tariff Rates
Customer (Fixed Distribution) Charge                    $      286.86            $       286.86

Variable Distribution Charge
Demand Charge @ $1.66 per kW                         $   16,600.00          $   16,601.66
1st Tier Energy Cost @ $.0088 per kWh               $  13,200.00           $  13,201.32
2nd Tier Energy Cost @ $.0052 per kWh              $    7,800.00           $    7,800.78
3rd Tier Energy Cost @ $.0016 per kWh               $   3,200.00            $    3,199.52

Competitive Transition Charge
Demand Charge @ $3.55 per kW                            $  35,500.00          $  35,503.55
1st Tier Energy Cost @ $.0189 per kWh                 $  28,350.00          $  28,352.84
2nd Tier Energy Cost @ $0112 per kWh                 $  16,800.00          $  16,801.68
 3rd Tier Energy Cost @ $.0035 per kWh                $   7,000.00           $    6,998.95

Energy & Transmission Charge
(At Competitive Rates - Assumed to be                  $ 200,000.00          $ 200,000.00
 $.04 per kWh)
                                     Totals                                 $ 328,736.86          $ 328,747.16
 

Average Cost per Kilowatt-Hour                                            $ 0.065

Cost of an Extra Kilowatt of Demand                                    $ 10.30