Vous êtes sur la page 1sur 12

654

IEEE Transactions on Power Apparatus and Systems, Vol. PAS-102, No. 3, March 1983

THE ECONOMICS OF DIRECT CONTROL OF RESIDENTIAL LOADS ON THE DESIGN AND OPERATION OF THE DISTRIBUTION SYSTEM PART 11 LOAD CHARACTERISTICS
MURRAY W. DAVIS, Senior Member IEEE The Detroit Edison Company Detroit, Michigan THEODORE J. KRUPA, Member IEEE The Detroit Edison Company Detroit, Michigan
MATTHEW J. DIEDZIC, Member IEEE The Detroit Edison Company Detroit, Michigan

Abstract - This paper is the second of a series of three papers which addresses the economics and effects of controlling central air conditioning, water heating, and service voltage on the design and operation of the energy delivery system. Various load control strategies are described which benefit the system generation and T&D system. A unique approach is developed to measure the load reductions and recovery loads of central air conditioning (A/C) and the test data collected over a four-year period is displayed for four different control periods and five different ambient temperature ranges. The thermal response of customers' homes is quantified for the A/C control periods. This paper also summarizes the electric water heater controlled and uncontrolled load characteristics for the many control strategies tested. Circuit and customer voltage profiles are portrayed for the summer and winter seasons along with the percent changes in real and reactive power for changes in customer service voltage.

COMPOSITION OF 1977 CIRCUIT PEAK LOAD MAX. 97F 7-20-77


1.0

, Circuit Load
//
SytmLa
/-

.9
.8

.7
.6

INTRODUCTION As stated in the Introduction to Part I', the economic benefit of directly controlling customer loads depends on the magnitude of the load reduction, the percentage of energy payback, the cost of control equipment and the length of time the load can be controlled during periods of high fuel costs or during periods of system generation and T&D capacity shortages. The purpose of this paper is to define the characteristics of the controllable residential load and show how these characteristics are modified by using direct load control strategies. A number of strategies involving the radio control of water heaters (W/H), residential central air conditioners (A/C) and service voltage (SVC) was investigated. Each strategy or combination of strategies was tested to see how the modified characteristics of the load reflected back through the distribution network, and ultimately (Part III) how these modified characteristics affect the distribution design criteria and the selection of an economic system design.
Development of Load Control Strategies The purpose of previous direct load control experiments and studies has been to defer high cost generation capacity and to save high cost fuel. However, there may be pitfalls in ignoring the distribution system when designing load control strategies which benefit only the system generation. This is especially true when there is a time shift between system and distribution peak loads. Figure 1 shows the relationship between Detroit Edison's system, residential distribution circuit, and major appliance daily load profiles on a summer day when the outdoor temperature reached a maximum of 97F. Detroit Edison's system load normally peaks in the early afternoon (2 pm to 3 pm), whereas the residential distribution circuit peak load occurs in the evening (6 pm to 8 pm). The early afternoon system peak is attributable to a large component of industrial load which peaks early in the day while the early evening distribution peak results from the large component of residential central air conditioning load which peaks at from 5 pm to

CL cL

.5

.4
.3 .2
_\
I

J
\

,~~~~~~~~~~~~

.1'

Electric Range L\oad I'PI I%l


%

U-

f~~~~~~~~~~~~~~~~~~
2
4
am

I'

W/H Load

10 12

14

16 18
pm

20

22

4---JD.t-46tt
24 2

4 am

Time-hrs

Fig. 1. A Comparison of System, Distribution Circuit and Major Appliance Loads on a Hot Summer Day.
6
pm. This implies that a load control strategy aimed at reducing system peak loads may produce recovery loads which create higher

able for printing May 20, 1982.

than normal distribution circuit peaks. To further explain the significance of this discussion, strategies designed to benefit only system generation and to benefit both generation and distribution facilities are given in Figure 2. The strategy benefitting only the generation consisted of interrupting the A/C from 1 pm to 5 pm with 15 minutes of off-time and interrupting the W/H from 2 pm to 6 pm. Even though the circuit load was reduced by 12 percent at the time of system peak, the recovery load created a distribution peak 16.6 percent higher than its uncontrolled peak load. However, strategies can be devised to reduce both the system and distribution peaks, such as a 1 pm to 10 pm A/C control and a 7 pm to 10 pm W/H control. In this case, the load reduction at the time of the system peak was slightly less than before, but the peak load on the distribution circuit was 4.3 percent less than the uncontrolled profile. The art of developing load control strategies involves obtaining as much load reduction as possible while still meeting the needs of the customer, and having the recovery loads appear during a desired or off-peak period. Over a four year period, a number of control strategies 82 SM 440-6 A paper recommended and approved by the IEEE were tested on the Hickory distribution circuit. From these tests, load Power Systems Engineering Committee of the IEEE Power reductions, recovery loads, energy consumption, thermal responses of Engineering Society for presentation at the IEEE PES 1982 customers' homes and distribution facilities, distribution network Summer Meeting, San Francisco, California, July 18-23, energy losses and voltage profiles were measured. The cost-benefit of 1982. Manuscript submitted September 14, 1981; made avail- each strategy was then evaluated.
0018-9510/83/0002-0654$01.00
1983 IEEE

1.2
1.1 1.0

SUMMER CONTROLLED CI RCU IT LOADS


n Control

gy

~ ~~

^
\

16.6%
L4.3% \ System &

stribution
k

.9 .8

ner \

~~~

J Ii
,,f I

~7(

ODistribution~~~Control

~~Strategy

655 The W/H bimodal daily load profiles are primarily a function of inlet water temperature and the differences between summer and winter living habits, therefore the W/H units were interrupted during each of the four seasons for most of the 17 control periods shown in Table II, with the off-time varying from one to five hours. A total of 383 W/H tests were conducted. Simultaneous control of A/C, W/H and SVC was also tested. Thirty tests were conducted using combinations of A/C and W/H control strategies and 55 tests were conducted using A/C, W/H and SVC strategies. Tests involving only SVC were conducted on 178 days during the four seasons.
Load Characteristics The economic benefits of direct load control are related to: (1) the magnitude of the controllable load; (2) the degree to which the load can be controlled; (3) the amount of time during the year the load is available for control (annual load factor); and (4) the cost to control the load. A knowledge of uncontrolled and controlled load characteristics therefore, is paramount to evaluating the cost-effectiveness of direct load control strategies. Given in Figure 3 are the annual load duration curves for both the Hickory distribution circuit and the A/C component of the circuit load which is comprised of 270 central A/C units. An arbitrary intersection, such as point "A," shows that 90 percent of the time the circuit demand is less than 46 percent of the annual peak load. Although the A/C peak is large, or 46 percent of the circuit peak load, it is only available for control about 20 to 30 percent of the time. Also, the air conditioning load is not usually interrupted but instead is cycled from 25 to 50 percent (15 to 30 minutes of off-time during any one hour of control). Therefore, the degree to which A/C load can be controlled is much less than the degree to which W/H load can be controlled (100 percent) during any one hour of control. On the Hickory circuit, a 25 percent A/C cycling time would reduce the annual circuit peak load by only 11.5 percent.
100r
90

.7
.6
.5
.4

A/C and W/H Control Strategies ------ 1-10 pm A/C, 15 min 7-1 0 pm W/H 1-5 pm A/C, 15 min 2-6 pm W/H
I

I_

.3

I I

I I I I

I I

6
am

10

12 14 16 18 20 22 24 pm Time - hrs

4
am

Fig. 2. System and Distribution Load Control Strategies. Control Strategies Tested Three different methods of direct load control were investigated: (1) A/C cycling, (2) W/H interruption and (3) service voltage control (SVC). Tables I and II display those tests conducted where a significant amount of data was collected. Since the A/C demands are a function of the ambient air temperature and other meteorological parameters, test results were categorized under five different daily maximum temperature ranges (see Table I). Four different control periods were selected to benefit the generation or distribution facilities. A total of 145 A/C tests were made with the nominal off-time varying 10 and 20 minutes during each hour of the control period.
TABLE I Air Conditioning Load Control Strategies MaximiLur Control Periods 2-6 pm 6-10 pm 1-10 pm Ambie at 1-5 pm
77

CUMULATIVE LOAD DURATION - 1978 DC 1597 CIRCUIT & CENTRAL AIR CONDITIONING

-I

80
0 -j
.

70

60

0.

Tempersiture Range :-:950]F


90-940DF

' 50

78

77
x x

78

79

80
x

77

78
x

77

79

80
X

x
x x x

x
x x x x x x x x

x
x
x

x
x x
x

40
30

Total Circuit

85-890 F 80-840DF 75-790'F

x x
x

x
x

20 10

TABLE II Water Heating Load Control Strategies


Control Periods
pm-pm

10

20

30

40

50 60 % of Time

70

80

90

100

1-4 2-3 2-4 2-5 2-6 2-7 3-4 3-5 3-6 4-5 4-6 4-7 4-8 5-6 5-7 7-10 8-11

Fall Summer Winter Spring 77 78 79 80 77 78 79 80 77 78 79 80 77 78 79 80


x
x x x x

x x
x x

x x
x

x x x x
x
x x

X
X

x
-

x x x
x

x x x x x x x x x x x x

x
x

x
x
x

x
x x x x x x x

x x
x

Fig. 3. Cumulative Annual Load Duration Curves for Distribution Circuit and A/C. Figure 4 shows the ratio between the total circuit A/C load and the A/C load interrupted for a 15 minute off-time. All loads shown are at the time of circuit peak. The uncontrolled circuit load curve represents an upper boundary of an envelope of possible demands; these are representative of days which are clear or slightly overcast, but no rain. The presence of rain during the day results in a significant reduction in circuit peak demand.. A temperature sensitive load characteristic begins to develop for A/C load when ambient temperatures are as low as 68F. At this temperature, the A/C load is detectable, but by no means controllable. The circuit demand increases by an average of 29 kW/OF in the 7074F maximum temperature range. The slope of the temperature sensitive load curve increases rapidly to 57 kW/IF for the 80-84F range and reaches a maximum of 72 kW/F for the 90-94F range. Demand comparisons made between clear sky 97F and 102F days, indicate

656
5.0
CONTROLLED AND UNCONTROLLED A/C UNIT (3 TON)

! 00

E
a

at Circuit Peak 270 Units Control Period 1 -10 pm min ~~~~~15

A/C Load Reduction

H Load Reduction
1000
do

Controllable at C tat Circuit Peak


\ 2OTz,

at Circuit Peak 128 Units

Base Load

W!H Load at

Circuit Peak

60

65

70

75

80

85

90

95

100

Maximum Daily Temperature OF Fig. 4. Circuit Peak Loads and Load Reductions as a Function of Maximum Daily Ambient Temperature. circuit demands become saturated at 97F, and further increases in demand are no longer the result of increases in A/C load but are simply a result of increases in base loads such as refrigerators, fans and other appliance loads. The 270 A/C's on the circuit account for the rapid rise in circuit demand on days with an 80-100F range of maximum temperatures. This translates into an increase of 259 watts per F in diversified demand for each A/C unit. From the viewpoint of distribution released capacity, this load may seem attractive to control, but since its annual load factor and hence its load management effectiveness factor2 is low, the benefit for the generation system in terms of deferring base or intermediate loaded units may be quite small. In general, this is representative of conditions which exist in northern climates since there are relatively few days, usually ten or less, when the maximum ambient temperature exceeds 90F. Central Air Conditioning Load Characteristics: The A/C unit hourly diversified demand profiles for the five maximum ambient temperature ranges are shown in Figure 5. These demands are based on the metered A/C units on the Hickory circuit whose average connected nameplate rating is 6.37 kW and capacity is 4.07 tons. The measured average full load running power for these units was 5.72 kW, or 10 percent lower than the manufacturers' nameplate data. This discrepancy was due to both a lower service voltage on hot days when the circuit is heavily loaded and manufacturing tolerances. It is not uncommon to have service voltages five percent below nominal as will be described in the section on voltage profiles. The measured full load running current is important to determine the actual maximum demands, demand reductions and recovery loads.

2 4 6 8 10 12 14 16 18 20 22 24 2 4 6 am pm am Time - hrs

Fig. 6. Comparison of Demands for an A/C Unit on Controlled and Uncontrolled Days. The A/C demands are extremely temperature sensitive, however, other environmental factors, particularly rain, also have a significant effect. Any amount of rain during the daylight hours tends to lower and distort the hourly A/C demand patterns. The demands shown in both Figure 5 and Table III are for clear and partially overcast days only. Rainy days have been excluded; this increases the diversified demands, but makes them more representative of circumstances when load control would be implemented. The number shown in parentheses on Figure 5 indicates the number of weekdays in the Detroit Edison service area which are likely to fall within each maximum ambient temperature range specified. Since there are very few hot days and many tests to run, a unique method was developed to conduct multiple tests on the same day.
100 _

Ta9a59F>
90

'\Tao Max 970

CENTRAL AIR CONDITIONING LOADS SUMMER SEASONS 1977, 1978, 1979


:

85

Ta Ma

95X

5.0

r',

co
E 4.0 *' oD .20
:t 0
m
*-

0~~~~~~~~
0 75

80/

cm X

3.0

O0

c 2

2.0

am

10 12 14 16 18 20 22 24 2
pm Time - hrs

4
am

Fig. 5. Hourly Diversified A/C Demands.

Fig. 7. Outdoor(Tao) Temperature Profiles on Controlled and Uncontrolled Days.

TABLE III Uncontrolled Central Air Conditioning Load Characteristics for Summer Seasons 1977, 1978, 1979

Maximum Ambient Temperature Range

a95TF

90-94F 85-89F 80-84F 75-79F

Daily Peak (kW) 5.277 3.816 2.782 1.332 1.021

Time of Peak 4 pm 5 pm 7 pm 6 pm 6 pm

Daily Energy Consumption (kWh) 6 am-6 am 79.347 48.256 34.676 14.626 10.912

Daily Load Factor 6 am-6 am 0.633 0.527 0.519 0.457 0.445

Methods of Calculating Central Air Conditioning Load Reductions and Recovery Loads: A number of investigators have determined A/C load reductions and recovery loads by subtracting the hourly loads on controlled days from the hourly loads during the control period on

657 uncontrolled days. In some cases little concern has been exercised in grouping controlled and uncontrolled days which have similar daily ambient temperature profiles. Periods of rain and other differences in weather conditions represented in the controlled and uncontrolled days affect the amount of A/C load available for control. In Figures 6 and 7, although the maximum ambient temperature for the two days was 95F or greater, the temperature profiles for each day were significantly different, such that the computed load reduction would be less than actual. In general, when statistical comparisons are made between controlled and uncontrolled loads without concern for differences in weather, the load reductions are either much smaller or much larger than actual and the standard errors of the estimate are sometimes as great as the load reductions themselves. To alleviate this problem, the Hickory project used two circuit laterals which have approximately the same number of customers, connected kW of A/C load, the same number of controlled A/C's and about the same connected kW of controlled A/C. (See Table I, Part I, and compare the statistics for laterals #1 and #2.) This feature allowed controlled and uncontrolled data, based on the same weather, to be collected on the same day.

DAILY TEMPERATURE PROFILES

IF

II

II

.1

6
am

10

14

18
pm

22

6
am

10

14

18
pm

22

6
am

10
-

14

18
pm

22

6
am

10

14

18
pm

22

6
am

10

14

18 22 pm

Time hrs

Fig. 8. Daily Ambient Temperature Profiles for the First and Third Temperature Buildup Weeks.

658
EFFECT OF TEMPERATURE BUILDUP
500

EFFECT OF PREVIOUS HOT DAY


7-8-77 - e--e
LATERAL #2

LATERAL #1

400

.300

6-10 pm 15 min W/H Control 2-6 pm 7-21-77 Uncontrolled


7-19-77
@----

A/C Control

So0
400

300
200

A/C Control
200

A, B, C,
6-1 0 pm
15

min

100

7-20-77

100

Uncontrolled

6
am

10

12

14

16

18
PM

20

22

24

4
am

6
am

1012 1416

-1.820 2224.2
PM

I0
4
am

Time

hrs

Time

hrs

Fig.

9.

(a)

&

(b) Comparison
Profiles

of Lateral #1

and Lateral

#2 Load

A number of examples will now be given to illustrate why air conditioning load reductions and recovery loads cannot be obtained by comparing two different days' load data. To make this comparison two sets of figures are referred to: (1) the daily ambient temperature profiles in Figure 8, and (2) the corresponding lateral load profi'les in Figure 9. At the top of Figure 8 are the hourly ambient temperatures for the first temperature buildup week, and at the bottom of this figure are the hourly ambient temperatures for the third temperature buildup week. These buildup periods represent the two hottest weeks during

Figure 9b. Both days, July 8, 1977 and July 21, 1977 had the
temperature and
bient temperature
average

same

peak

amount of cloud cover, but followed different am-

the average temperature

buildup periods, are different days of the week and 1977 was 30F higher than the on July 21, temperature on July 8, 1977 making the load profiles signifi'c-

antly different. Example (c): A comparison of the controlled load profiles (Figure 9c) on July 19, 1977 and July 5, 1977, beginning at 7 pm and ending
at

11I

pm, shows that the average difference between these two

the

summer.

is 21.8 kW for the 6 pm to 10 pm A/C control

profiles period. Since there are 21


average

Example (a): A comparison of controlled and uncontrolled loads on days which have the same daily peak ambient temperature and are within the same ambient temperature buildup period is made in Figure 9a. This figure shows the controlled and uncontrolled load profiles on lateral for two days where the peak ambient temperatures are the same, namely 970F. The uncontrolled load on July 20, 1977 was significantly higher than the controlled profile on July 19, 1977 up to the time of the control period. The reason for this difference is because July 20, 1977 was the third day of an ambient temperature buildup period and the average temperature was 30F higher on July 20, 1977 than on July 19, 1977. If the July 20, 1977 load profile were adjusted downward to match the hourly demands before the control period on July 19, 1977, this would result in small or negative load reductions.- Although this comlarison was made with lateral load profiles, the same would be true for individual A/C units. Example (b): A comparison of controlled and uncontrolled loads on days which have the same daily peak ambient temperatures and are from different ambient temperature buildup periods show even larger differences in load before the control period. This example is given in
EFFECT OF

controlled A/C units

on

lateral # 1, the difference in load per controlled


As will be
seen or

A/C

customer is 1.03

later, the

hourly load

reduction is 1.02 kW per A/C unit, units

21.42 kW for the 21 controlled

period on days where the higher. The difference between the hourly controlled loads on these two days was as large as the load reduction itself, even though both days were cloudy until 10 am; the day of the week was the same; and the maximum and average temperature for each day was approximately equal. One of the reasons why the load profile on July 19, 1977 was lower than the profi'le of July 5, 1977 was bec'ause heavy rain (1. 16 inches) was recorded for the morning of July 19, 1977. Example (d): Ideally, the lateral load profiles should compare favorably if (l) two clear days are selected where the maximum and average temperature was approximately equal, (2) the days chosen are the same day of the week, and (3) both days are the third and maximum ambient temperature day of a thermal buildup period. Such a case is given in Figure 9d, for lateral #2, and in Figure 8 for days July 6, 1977 and July 20, 1977. The only difference is that July 6, 1977 is the third
a

during

6 pm to

10 pm control

maximum ambient temperature is 950F

or

RAIN

FALL

CONTROLLED & UNCONTROLLED GROUP

5%,00
400

500 LATERAL #1
LATERAL #1 & LATERAL #2

A/C Control

UNCONTROLLED

A, B,C, D
6-1 0 pm 15

400

min
300

300

200

F'
Lat #2 7-5-77
.---

200

7-6-77
7-20-77

Lat #2 Lat #1

100

7-20-77-e--

100

~~~~~~~I
4
am

I'
18 20 22 24

0
4
am

10

12

14

16

4
am

10

12

14

16

18
PM

IPM
Time
hrs

-20

22

24

4
am

Time

hrs

Fig.

9.

(c)

&

(d)

Compari'son
Profiles

of Lateral #1

and Lateral

#2 Load

659

day of the first thermal buildup period for the summer season and July 20, 1977 is the third day of the season's third thermal buildup period. For this ideal case, the average difference between the loads on lateral #2 for these two days during the period of interest (2 pm to 11 pm if A/C units were controlled from 1 pm to 10 pm) is still 16.1 kW or .77 kW per controlled A/C unit. Selecting controlled and uncontrolled data in this manner will reduce the error in calculating the load reductions and recovery loads, but there are probably only one or two days during the summer season where these ideal conditions exist. Since this approach seriously limits the amount of data for analysis, a more suitable method is to compare the controlled and uncontrolled load data from laterals # 1 and #2 on the same day. Since the uncontrolled load profiles for laterals # 1 and #2 are nearly the same (see Figure 9d, July 20, 1977) this method results in' three advantages. First, selecting two days with similar weather conditions is no longer necessary; second, all days when the load is controlled during the summer season are available for analysis, and third, the standard error of the estimate for calculated load reductions and recovery loads is reduced to approximately half that obtained from the ideal case outlined above. Some researchers3 4 have attempted to circumvent the problem of comparing loads on two different days by selecting controlled and uncontrolled groups of air conditioners in their service area. With this method, controlled and uncontrolled loads may be compared on the same day, but if the two groups are not in close proximity or localized, rain patterns and temperature profiles could still be significantly different. With the Hickory circuit, all the data was collected within a one-quarter square mile area and thus all customers in the experiment experienced the same weather conditions. Central Air Conditioning Load Reduction and Recovery Loads: The load reduction and recovery loads obtained from laterals # 1 and #2 for the control periods and ambient temperature ranges listed in Table I are displayed in Figure 10, and are also tabulated in Tables IV and V. It should be noted that the first hour of load reduction begins one hour after the control strategy is initiated. The three control periods selected for the benefit of system generation were 1 pm to 5 pm, 2 pm to 6 pm, and 6 pm to 10 pm, the latter period being used for pumped storage hydro. The fourth, or the 1 pm to 10 pm control period was the only one which benefited both generation and distribution facilities. By observing the data in Table IV and Figure 10, the hourly load reductions tended to increase to a maximum value in the third or fourth hour of control for all control periods. This maximum value was then followed by a rapid decline in load reductions toward the end of the control period. The reason for the increase during the first three to four hours could be attributable to a period of synchronization. With the aid of Figure 11 this synchronization period is explained as follows. For the first hour of a control period, some A/C units would not have been running during the off-time period, whether or not they had been interrupted; also some other A/C units would have been running during the off-time if they had not been interrupted. On the Hickory circuit the probability of an A/C unit running during the hour of maximum demand if it had not been interrupted in any preceding hour P(ON) is shown for an undersized and an oversized unit in Figure 12. The data shows that P(ON) is about .75 for an average size unit when the peak outdoor temperature (Tao) is 90F. During the second, third, and sometimes fourth hour of control, P(ON) increases, because more and more units that were idle during the previous off-times desire to run when switched off and the original diversity among the units is lost. This demand accumulation continues, depending on unit size, temperature, etc., until all the units are synchronized in time and the load reductions reach a maximum.

6.0

CONTROLLED AND UNCONTROLLED LOAD CHARACTERISTICS a 950 F

5.0
E 4.0
.0

, 3.0_

2.0
6.0

2-6pm Control Period

_5.0_
84- 0 E

44.0

13.0

N
2.0
-

6.0 5.0
-

4. 0C

3.0 c
am

1.0 12. 13 14 15 16 17 18 19 20 21 22 23 24 pm Time - hrs

am

Fig. 10. AIC Demand Reductions and Recovery Loads for Four Different Control Periods. The synchronization period is followed by a period of residual thermal effects. During this period there is carryover of thermal energy from each controlled hour to subsequent hours - or thermal bunching. The A/C unit is attempting to deplete this gain in thermal energy and thus would create higher demands than its uncontrolled profile if allowed to run during the off-time. By subtracting the controlled profile from the uncontrolled profile small load reductions result. From the load reduction test data in Figure 1 1, it is apparent that only
a portion of the thermal energy is carried forward. Some fraction of the

LE IV Diversified Central Air Conditioning Load Reductions and Recovery Loads


17.5 Minute Off-Time, Clear Days Maximum Ambient Temperature 95TF Hourly Load Reductions (-) and Recovery Loads (+), kW

Control Period 1-5 pm 2-6 pm 6-10 pm 1-10 pm

2 pm -1.10

3 pm -1.22 -1.20
-1.26

4 pm -1.31 -1.28 -1.33

5 pm -1.21 -1.45

6 pm -1.15 -1.20

7 pm +0.80
-1.13

-1.05
-1.19

-1.16

-1.42

-1.25

8 pm +0.46 +0.84 -1.15 -1.08

+0.76 -1.20 -0.90

+0.59

9 pm

10 pm + 0.34

11 pm

12 pm

I am

2 am

3 am

+0.36
-0.97 -0.68

-0.72 -0.45

+0.95 +1.29

+0.26 +1.02

+0.18 +1.13

+0.36

660
TABLE V Central Air Conditioning Average Load Reductions and Recovery Loads by Ambient Temperature Range
Net Energy Loss kWh
Load and Energy Summary Average Percent Load Reduction Energy kW Payback 950F 1.20 36.56 1.25 31.31 1.02 27.31 1.07 35.45
PROBABILTY OF AN A/C UNIT RUNNING DURING THE HOUR OF MAXIMUM DEMAND

1.0

Control Period
1-5 pm 2-6 pm 6-10 pm 1-10 pm

Peak Recovery Load kW

.9
.8

Undersized

3.80 4.30 3.70 6.92 2.86 3.36 2.77 5.64


2.44 2.60 2.22 4.45

.80 .84 .95 1.29

.7
Oversized Unit

1-5 pm 2-6 pm 6-10 pm 1-10 pm


1-5 pm 2-6 pm 6-10 pm 1-10 pm
1-5 pm 2-6 pm 6-10 pm 1-10 pm

.90 1.25 1.00 1.00

90-94F 31.9 23.3 26.7 24.1 85-89F 14.4 16.4 12.3 8.4 80-84F

.84 .88 .76 .74

.74 .60 .83 1.03 .27 .35 .31 .41

.6

2 o

.5

.57 .62 .51 .49


.18 .25 .20 .10

.4 .3 .2
.1
0

*Negligible thermal buildup is lost, hence a loss of energy consumption. The dashed line represents the theoretical maximum demand reduction on an hourly basis without considering carryover of energy; i.e., if each hour were the first hour of control. This represents an upper boundary of possible load reductions, based on the diversified hourly demands for r950F temperatures, and in this case 17.5 minutes of control. The demand reductions for a 30 percent energy carryover were computed from the same demand curve by assuming that 30 percent of the load reduction during a controlled hour was carried forward to the next hour, and that the remaining 70 percent was thus transmitted to the structure and eventually lost to the surroundings later in the day. The heavy solid line represents the measured test data. It is evident that the residual thermal effect was not a constant 30 percent as assumed over the entire control period, but varied with ambient temperature and the number of consecutive hours of control. During a 10 hour control period on a hot day, the residual thermal effect caused a sharp decline in load reductions during the last four hours of control. This implies that by this time, the air conditioner was running at a much higher demand during the 42.5 minute time period than if it were not controlled earlier in the day. Therefore, it can be said that the level of controlled load rises with respect to the uncontrolled curve. If so, it is reasonable to expect that the lower level of load reductions during a control period can result from two entirely different phenomena. One, the level of controllable load is low so there is very little load available to control; and two, the residual thermal effect causes air conditioners to run much harder in the latter stages of a control period, thus raising
2.0 _
-

70

80 90 100 Maximum Daily Outdoor Ambient Temperature, (Tao) F

110

Fig. 12. Probability of an A/C Unit Running During the Hour of Maximum Demand.

CENTRAL AIR CONDITIONING UNIT DIVERSIFIED DEMAND REDUCTIONS 1-10 pm CONTROL PERIOD, 17.5 MIN

1.5
R., C
0

,'X

,x-

Theoretical Maximum Demand Reduction

. 1.0

CD
0

Ambient Temperature -Range >. 950F


.

\0
Increase

Syncronization Residual Thermal Effecta

Carryover * Measured Test \ Data


Decrease

O30% Energy

Control Period Recovery Period 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 am PM Time - hrs

Fig. 11. Comparison of Theoretical Maximum and Measured Diversified Demand Reductions for A/C.

the level of the controlled load. A period of recovery always follows the control period. On days when the maximum temperature exceeds 95F, recovery loads were evident for three or more hours. Table V gives a summary of the load reductions, recovery loads and energy for the four control periods tested. During the first season of tests, it was found that the radio receivers had a random and inoperative failure rate of about 15 percent. Therefore, before and after each season a receiver reliability test was conducted by transmitting a specified number of test tones and noting whether each receiver responded properly (by observing the receiver operation counters). Those receivers which either responded randomly or failed to respond were replaced. The load reductions in Table V may be 15 to 20 percent higher than some researchers' data because the reliability of the receivers was very high (99 percent), the data was collected on clear days, and the average unit size was about 4 ton. The average load reductions were the highest for the 2 pm to 6 pm control period for each temperature range, because the controllable loads were higher during this period. The average load reduction for the 1 pm to 10 pm control was the lowest primarily due to the residual thermal effect of the longer control period. The net energy loss was the highest for the 1 pm to 10 pm control with an overall energy payback of 35.45 percent on days when the ambient temperature was 950F. On lower ambient temperature days, the residual thermal effect was still noticeable, but to a lesser degree. In this case, the process of recovery begins during the final stages of control. This makes the load reductions and recovery loads smaller. For the lower temperature ranges, the pattern of hourly load reductions retains its characteristic shape but declines rapidly for the 80 to 84F range. The load reductions during these lower ambient temperatures occur primarily during the synchronization period. Once synchronization is complete, the load reductions for subsequent hours decline rapidly to zero. Figure 10 illustrates the importance of controlling late into the evening hours to eliminate distribution peak recovery loads. For the 1 pm to 5 pm and 2 pm to 6 pm control periods, the peak recovery load is actually greater than the uncontrolled case. Therefore, in those instances where T&D capacity is a consideration, A/C load control should be continued until 10 pm or later to mitigate the effect of a late evening recovery. Indoor Temperature Response and Customer Comfort: Of the A/C control periods tested, it was found that the maximum indoorttemper-

661
CONTROLLED AND UNCONTROLLED A/C UNIT (3 ton) kW
SEASONAL WATER HEATER LOADS 1978 2.0

z11

-- -

-A-

Spring

xt 'A
uI~m

Fall |/ t\

Summer
Winter

c 1.5 n
*-

C
._

W ,w 4

-c

1.0

U-

o 0 X

.5

6 8 am

10 12 14 16 18 20 22 24 2 pm Time - hrs

4 6 am

Fig. 15. Water Heating Unit Hourly Diversified Demands.


TABLE VI Seasonal Uncontrolled Water Heating Load Characteristics for 1978 (87 gal. avg.) Daily Energy Time Consumption Daily Load Daily Factor kWh of Peak 6 am-6 am 6 am-6 am Peak (kW) 0.549 25.62 1 9 am 1.945
2.011 1.424 1.738 9 am 8 am 9 am 26.377 17.379 22.293 0.547 0.509 0.534

6 8 10 12 14 16 18 20 22 24 2 4 am pm am Time-hrs

Fig. 13. A/C Unit Demand and lndoor(Tai) Temperature Profiles on Controlled and Uncontrolled Days.
ature rise above the uncontrolled temperature was only 3.3F with an average of less than 2F. It was concluded that other variables, such as the difference in outdoor temperature profiles on controlled and uncontrolled days, tended to mask the effect of A/C control. This may only be true in northern climates where there is typically a 200 to 30F difference between the maximum and minimum outdoor temperatures. Such an example is given in Figures 13 and 14 where the indoor temperature (Tai) rise above the uncontrolled value was only 1F for a 2 pm to 6 pm control on a 95F day. This 2400 square foot, two-story colonial-type home was considered to have an undersized A/C unit of
CONTROLLED AND UNCONTROLLED A/C UNIT (3 ton)
100

Profile Winter

Spring
Summer Fall

90 X
88

95
90/

Tao

max

95F x

86
84

85

7-20-80
/

//
/ /

Tao

max=

9
82 80
78

80
i

8-7-79

75
70
65-

.--

Indoor Tempere 2-6 pm Control


7-20-80

76

60

\<8-7-79

Uncontrolled

Indoor Temperature (Tai)

74
72

Ir2

4 6
am

l, ............. 18 20 22 8 10 12 14
16
pm

24 2

4
am

Time-hrs

Fig. 14. Relationship between Indoor (Tai) and Outdoor (Tao) Temperatures.

3 ton or 4.5 kW. It is thought that the reason the Tai rise was so small may be due to the relatively few hours Tao was above 80F and that the home was well insulated for the winter heating season. Since the ASHRAE comfort standards were maintained for all the control periods tested, 96 percent of the customers accepted the slightly higher indoor temperatures. Water Heating Load Characteristics: The diversified W/H unit hourly demand profiles for the four seasons are given in Figure 15. These demands are based on an average capacity of 87 gallons which is higher than the system average. As seen in Table VI the higher peak demands in the winter-spring seasons are the result of low inlet water temperatures. The lower peak demands in the summer are due to higher inlet water temperatures and lower usage in the evening hours. Water Heating Load Reductions and Recovery Loads: Table II lists the W/H control periods investigated. The Detroit Edison Company interrupts approximately 180,000 electric water heaters primarily for fuel savings, however, system demand reductions have occasionally been used in times of system generation capacity shortages. For this reason most of the control periods selected are in the afternoon or evening. Several of the common strategies for the winter and summer seasons, and their results are summarized in Figures 16 and 17 and Table VII. This data may not be typical for small capacity water heaters. Some of the general conclusions derived from the tests are given below. In viewing the two figures and the table, when the water heaters are controlled on peak, the recovery loads are higher, the recovery period is longer, and the percent energy paybacks are higher than when the units are controlled during the mid-day valley load hours. This is true in both the summer and winter. The first hour of recovery load is approximately the same during summer and winter. Since the uncontrolled loads are lower in the summer, the percent energy paybacks are higher because most of the payback is achieved in the first hour of recovery. Another reason why the paybacks could be lower in the winter is because lower temperature water is used during the control period and thus less percentage of payback energy is needed. The W/H load is more desirable to control than A/C load because W/H load reductions of from .5 kW to 1.0 kW can be achieved yearround, while the same range of A/C load reductions can only be

662
TABLE VII Electric Water Heating Load Reduction and Recovery Loads Load and Energy Summary Peak Average Net Percent Load Load Energy Energy Reduction Reduction kWh Payback kW kW Winter -1.16 61 1.12 0.99 -0.74 60 0.97 0.93 -1.42 64 1.11 0.99 -0.33 63 0.89 0.89 -0.63 79 1.11 1.00 -0.34 66 1.00 1.00 -0.74 65 1.11 1.06 +0.07 106 1.11 1.11 -0.45 90 1.57 1.48 Summer -0.36 83 0.75 0.70 0.00 100 0.68 0.71 -0.06 98 0.85 0.71 +0.92 163 0.85 0.74 +0.81 147 0.89 0.87 +0.10 104 0.88 0.78 +0.06 103 0.73 0.73
CONTROLLED AND UNCONTROLLED W/H UNIT (87 gal. avg.) SUMMER SEASON
CONTROLLED AND UNCONTROLLED W/H UNIT (87 gal. avg.) WINTER SEASON

Control Period pm-pm

1st Hour Recovery Load kW


1.46 0.95 1.63 0.56 1.41 0.70 0.69 0.89 1.65

1-4 2-4 2-6 3-4 3-6 4-5 4-6 5-6 7-10


1-4 2-4 2-6 4-6 5-7 7-10 8-11

1.29 1.09 1.57 0.97 1.73 1.70 1.50

2.5

2.0

---- 2-4 pm Control

Uncontrolled

1.51

1.0

.5

LI
I

6 8 am

10 12 14 16 18 20 22 24 pm Time-hrs

4 6 am

6
am

10 12

14 16 18 pm Time-hrs

Fig. 16. Summer Water Heater Distribution Circuit Off-Peak and On-Peak Load Reductions and Recovery Patterns.

Fig. 17. Winter Water Heater Distribution Circuit Off-Peak and On-Peak Load Reductions and Recovery Patterns. obtained when the ambient temperature is 90F or greater. Also the percent energy paybacks for the water heater are two to three times more than the 30 percent paybacks of A/C control in a northern climate. Customer and Circuit Voltage Profiles Most utility distribution systems have been designed to provide normal service voltages within a bandwidth of Range A, 114 to 126 V, as specified in ANSI Standard C84. 1. Detroit Edison has used voltage reduction to ensure system security since 1948; and has used radio control receivers on 760 distribution station bus and line regulators to lower the voltage five percent since 19685. (See Figure 3, Part I companion paper.) Before using voltage control to reduce demands or energy, utilities should measure the voltage profiles throughout the distribution network for at least one year to avoid subjecting customers to voltages below, 114 V. When the temperature is greater than 900F, the circuit becomes heavily loaded, as seen in Figure 4, and it is not uncommon for the service voltage to drop 5 to 10 volts between the early morning and the peak load time. See Figure 18. For the 1978 summer season, the voltage regulators on laterals # 1 and #2 were set to regulate at a nominal 125 V and voltage histograms were constructed in .5 V increments for the laterals, transformers, and cus-

SUMMER SEASON 1978


LATERAL #4 9-8-78 >950 F

250

200

cn 4-

150
100
50

X '

125 127

100i<

TRANSFORMER #15 9-8-78

U-

10- 50 X1
40 30
20
10

126 -80
125 -60
>)

\,

124 -40 X 123 I- 20

X =

663 service voltage occurs on the heaviest loaded transformer and/or at the customer with the highest loads and longest secondary and service drop. Figures 19(b) and 19(c) illustrate that the first customer from the feedpoint with the highest load and served from a lightly loaded transformer had the lowest service voltage on the circuit, or 115.2 V. Figure 19 clearly illustrates the difficulty in maintaining Range A voltages during hot days if the service voltage is intentionally lowered to obtain demand or energy reductions. Furthermore, the Hickory circuit is compact and serves only a quarter square mile area; while many rural circuits may serve an area of 40 square miles. As a result, the voltage control tests were scheduled on the Hickory circuit during those days in the summer when the temperature s-90'F. Both the lateral voltage regulators and the distribution station bus regulator were used to obtain the desired voltage reduction. Voltage control tests were scheduled throughout the fall, winter and spring seasons, because the load on the circuit was much lower than in the summer. During the winter the lateral voltage regulators were set to three distinct voltage levels on a weekly basis to mitigate the effects of weather differences. The voltage settings were 124 V (normal). 128 V (raise) and 120 V (lower). The regulators were in close electrical proximity and the interaction caused a .5 V difference in the voltage profiles on laterals #1 and #4 as shown in Figures 20(a) and 20(f). The voltage control during the winter produced very poor customer voltage regulation; the wider bandwidths and flatter voltage profiles being intentionally created by the three lateral voltage settings spanning almost 10 volts. The widest bandwidth of 13.5 volts occurred in Figure 20(j) for the last customer on lateral #4. In actual practice the wide bandwidths such as those in Figures 20(c), 20(h), 20(e), and 20(j) would not occur since the feedpoint voltage would be set at a nominal value of 124 volts; with voltage reduction only, the bandwidths would be narrowed to approximately 9 volts or a range of 123 to 114 volts. Because the high voltages (126 to 129 V) occurred during the light load early morning hours, this was deemed acceptable for test purposes.

0 6
5

J9

2 1

Real and Reactive Power Changes with Voltage Control The results of the winter and summer voltage tests are listed in Table VIII. A one percent change in voltage resulted in about a one percent (.96) change in real power for high load days where the daily peak load was greater than 90 percent of the winter peak. This compared to only .76 change when the daily peak load was 76 to 100 percent of the winter peak. This is to be expected since one of the largest components of load in the winter was electric water heating and for the higher load days the voltage drop throughout the circuit is greater. The percent AQ/AV in the winter was more than twice the percent AP/AV. During the' summer, voltage control tests could not be conducted when temperatures exceeded 90'F because some customers would have had voltages below the 114 V minimum. Therefore SVC tests were conducted in the 80-90F temperature range only. The percent change in real power for this temperature range was 1.26 while the change in reactive power was 4.66 for each percent change in voltage. The summer and winter voltage control tests therefore indicated that AQ/AV increased with an increase in temperature. This appears appropriate in that the A/C induction motor load becomes predominate at about 80' to 85'F.
TABLE VIII Source Voltage Control Tests
19.78, 1979 Winter (January, February)

am

Time - hrs

pm

Fig. 18. Typical Lateral, Transformer, and Customer Voltage, Power Factor, and Load Profiles on a >'95F Day. tomer service points. The smooth curves in Figure 19 represent the voltage in percent time for all summer days and the vertical bars consist of data for only those days where the temperature exceeded 90'F. The vertical dashed lines show the highest and lowest observed voltages for each location. From this figure it should be noted that the lowest customer voltages are almost independent of location on the circuit. This fact makes it somewhat difficult to select circuits where voltage reduction can be implemented. Figures 19(c) and 19(h) are the first customers' voltage histograms and Figures 19(e) and 19(i) are the last customers' voltage histograms on laterals #1 and #4, respectively. The lowest customer

Type of Days % AV Average Load 5.4 76-100% High Load 90-100%


5.5

% V Range (lateral)

Source AP % A % AV AV
0.76

-3.4% (120.8V)

0.0% (125.4V)

+2.0% (127.9V)
+ 1.8% (127.8V)

2.32

-3.7% (120.9V) -2.8% (1 19.5V)

0.0% (125.5V)
0.0% (122.9V)

0.96

2.11

80-90F

7.0

191 81 Summer (July, August) (Circuit)

(128.OV)

+4.2%

1.26

4.66

664
SUMMER VOLTAGE PROFILES

LATERAL #1

% of Time
50
-

LATERAL #4

40

_ 30

Iur I L
I

20

_ 10 _ 0

-,r

30 [

20

ak
(g) CUSTOMER
L fI

(b)
LO

tI

CUSTOMER
IN

20
-

NLfI

NI

n l

..

ni

10
0

rdlllV
LD

ai~~~~~1

I LO

(c}
TRANSFORMER 20
10 0

(h)

TRANSFORMER

(d)
aw: L1O
~

>s

CUSTOMER

20 10

(i,

L
124 122 120 Voltage (e) 118
FITT

128

126

0 116

_ ~Ln
128 126

ml

CUSTOMER

CN1Cg
124
122 Voltage 120

n~~~~~~~~~~~~~~~~~~~Ll
118 116

ln

"X"

~~~~~~~~(j)

Fig. 19. Lateral, Transformer and Customer Voltage Histograms During the Summer Season.

CONCLUSIONS Obtaining accurate A/C load reductions and recovery loads is the first step in assessing the economic benefit of controlling A/C loads. The method outlined in this paper had several advantages. First, it permitted an accurate measurement of the load reductions and recovery loads because the controlled and uncontrolled groups of A/C load were confined to a small geographic area, namely one distribution circuit; thus, the effect of weather as a variable was the same for each group. SecQnd, because all the controllable loads (A/C and W/H) were on the`ame distribution circuit, the change in the network energy losses and thermal capacity could be monitored for each control strategy. This could not be achieved by selecting a random sample of controllable loads which are located throughout the entire T&D system. A distribution circuit load management study, such as the one described in this paper, should not replace studies consisting of system-wide samples of customer controllable loads; however, this type of study does complement the others adding understanding as to how changes in customer load profiles reflect back through the system. Also, monitoring distribution circuits is the only approach available to determine demand changes resulting from changes in service voltage. One of the disadvantages of a distribution circuit load management study is that the sample sizes are small and may not be representative

of the system average. Therefore, it is suggested that a blend of the results learned from distribution circuit studies and customer load control studies be used in estimating the system-wide economic implications of service voltage control and direct load control. Part III of this series of papers describes how distribution circuit and customer load studies can be combined to translate sample results into systemwide

impacts.

REFERENCES Murray W. Davis, Theodore J. Krupa, and Matthew J. Diedzic, Jr., "The Economics of Direct Control of Residential Loads on the Design and Operation of the Distribution System, Part I Design of Experiment." Companion paper, manuscript submitted to IEEE, pp. 1-8, January, 1981. 2. Ronald L. McIntyre, James D. Cyrulewski, and James M. Goodrich, "Practical Benefits of Load Modification for The Detroit Edison Company." IEEE-PES Summer Meeting, Minneapolis, Minnesota, 80SM516-5, pp. 1-8, July 13-18, 1980. 3. George W. Brazil, "Arkansas Power and Light Company Air Conditioning Load Management Program." Paper presented at the IEEE-PES Summer Meeting, Vancouver, British Columbia, A79 513-3, pp. 1-7, July 15-20, 1979.
1.

WINTER VOLTAGE PROFILES


LATERAL #1
/

REGULATOR BANDWITHS

% of Ti me
20

TmNd
9

10 F
0

I!

-2

4t
ILj
at

(a) -TRANSFORMER

0'4
LA I

10
II

20
10
0

(b)
CUSTOMER

20

10
0'

tc)

TRANSFORMER

20

10
0

(d)

CUSTOMER

20
~~~~~~~~~~~LO,

10
0

128

126
2

124

122 Voltage
(e)

120

118

.130

128

Fig. 20. Lateral, Transformer, and Customer Voltage Histograms for Three Different Lateral Voltage Settings During the Winter Season.
4. "Positive Load Control Pilot Program." Georgia Power Company, pp. 1-19, Interim Report 1980. 5. B.F. Smith and E.H. Arjeski, "Load Management by RadioControlled Voltage Reduction." Presented to EEI, Electrical System and Equipment Committee, Oklahoma City, Oklahoma, pp. 1-22, October 19, 1971.

k ~ ~ ~ 0_
LATERAL #4
N

665

95
X

(f) TRANSFORMER

'00

--o _~~~~~~~~~~~
iai~~~~~~ II
1

-J

(h)

TRANSFORMER

CUSTOMER

126

124

LI

I,I

Voltage

122

120

118

116

(i')

l4

ACKNOWLEDGEMENTS
The authors wish to express their appreciation for the contributions of Satyendra Basu for developing the computer programs and analyzing the data. The authors also appreciate the assistance of Stanley D. Zalewski and Marla K. Cezon for preparation of the figures and tables and aiding in the data analysis.

Vous aimerez peut-être aussi