Sei sulla pagina 1di 46

Uncertainty of Voltage Control in Active Distribution Networks

Nokhum Markushevich

Contents
Introduction .................................................................................................................................................. 3
Uncertainties of Situational Awareness ........................................................................................................ 5
Direct measurements................................................................................................................................ 5
Accuracy ................................................................................................................................................ 5
Latency.................................................................................................................................................. 6
Representation of the situation ................................................................................................................ 8
Imbalance .............................................................................................................................................. 8
Selection of reference measurements for state estimation (SCADA). ................................................. 8
Selection of voltage critical points ...................................................................................................... 12
Load/energy sensitivity to voltage ...................................................................................................... 15
Sensitivity of components of the objective function to controls ....................................................... 28
Uncertainty of Decision-Making ................................................................................................................. 30
Uncertainty of power flow based models............................................................................................... 30
Uncertainty of nodal load models ...................................................................................................... 31
Uncertainty of secondary equivalent models ..................................................................................... 31
Uncertainty of DER/microgrid models ................................................................................................ 32
Uncertainty of objective function and optimization processes .............................................................. 33
Uncertainty of Control (power equipment properties) .............................................................................. 33
Uncertainty of the impact of LTCs .......................................................................................................... 33
Uncertainty of the impact of capacitors ................................................................................................. 39
Uncertainty of the impact of DERs.......................................................................................................... 42
Conclusions ................................................................................................................................................. 43
Acknowledment. ......................................................................................................................................... 43
References and further reading .................................................................................................................. 43
Figures

Figure 1. Histogram of standard deviations of voltages during 15 min interval .......................................... 5


Figure 2. Histogram of variation coefficients of kW during 15-minute interval ........................................... 6
Figure 3, Sample circuit ................................................................................................................................. 9
Figure 4. Errors in primary voltage models ................................................................................................. 10
Figure 5. Histograms of errors in the secondary voltage models ............................................................... 10
Figure 6. Errors of the models of cumulative losses in the primary feeder ............................................... 11
Figure 7. Histograms of errors in the models of load losses in distribution transformers ......................... 12
Figure 8. Sample circuit ............................................................................................................................... 14
Figure 9. Weekdays load shape of the sample household ......................................................................... 20
Figure 10. Composition of load for day 1.................................................................................................... 21
Figure 11. Composition of load for day 2.................................................................................................... 22
Figure 12. LTVs for day 1 and day 2 ............................................................................................................ 23
Figure 13. Composition of five loads .......................................................................................................... 24
Figure 14. LTV of five loads ......................................................................................................................... 24
Figure 15. Coefficients of variation of LTVs for individual loads and for sum of five loads ........................ 25
Figure 16. Heater’s load under different voltages during one heater cycle ............................................... 26
Figure 17. Total customer load under different voltages during one heater cycle .................................... 26
Figure 18. Change of load due to reduced voltage for one customer ........................................................ 27
Figure 19. Change of five loads due to reduced voltage for a group of customers .................................... 27
Figure 20. Sensitivities of components of the objective function to the bus voltage. DER kW=100%....... 29
Figure 21. Sensitivities of components of the objective function to the bus voltage. DER kW=0%........... 30
Figure 22. Symmetrical unregulated voltage deviations. Bandwidth = 2*LTC step (0.625%) .................... 34
Figure 23. Histogram of voltages within the bandwidth for the case in Figure 2....................................... 35
Figure 24. Symmetrical unregulated voltage deviations. Bandwidth = 2% (LTC step=0.625%) ................. 36
Figure 25. Histogram of voltages within the bandwidth for the case in Figure 4....................................... 37
Figure 26. Asymmetrical unregulated voltage deviations. Bandwidth = 2%, (LTC step =0.625%) ............. 38
Figure 27. Histogram of voltages within the bandwidth for the case in Figure 26..................................... 39
Figure 28. Status of capacitor when the operating point is within the bandwidth. Initial status of
capacitor – ON, voltage change from low to high. – operating point, capacitor is ON ....................... 40
Figure 29. Status of capacitor when the operating point is within the bandwidth. Initial status of
capacitor – ON, Voltage change from high to low. . – operating point, capacitor is OFF ................... 41
Figure 30. Status of capacitor when the operating point is within the bandwidth. Initial status of
capacitor – OFF, Voltage change from low to high. – operating point, capacitor is ON ...................... 42
Figure 31. Status of capacitor when the operating point is within the bandwidth. Initial status of
capacitor – OFF, voltage change from high to low. – operating point, capacitor is OFF ........................ 42

Tables

Table 1 Correlation vs latency of voltage measurements at a residential customer ................................... 6


Table 2. Standard deviations of differences in voltage measurements due to latency at a residential
customer ....................................................................................................................................................... 7
Table 3. Correlation vs latency of load measurements at a residential customer ....................................... 7
Table 4. Standard deviations of differences in load measurements of residential customers due to latency 7
Table 5. Correlation vs latency of load measurements at a substation bus ................................................... 7
Table 6. Standard deviations of differences in load measurements at a substation bus due to latency ........ 7
Table 7. Errors in nodal load models ............................................................................................................ 9
Table 8. Errors in models of the power flow in segment of the feeder ....................................................... 9
Table 9. Errors in the models of voltages in secondaries ........................................................................... 11
Table 10. Errors in the models of load losses in distribution transformers ............................................... 12
Table 11. Low voltage critical points under different bus voltages at different times of day ................... 14
Table 12. Low voltage critical points under different bus voltages and different control conditions ....... 15
Table 13. Illustration of uncertainty of CVR-factors ................................................................................... 16
Table 14. LTV and CVR factors at different times [17] ................................................................................ 17
Table 15. Correlation between load changes at the test and reference buses .......................................... 18
Table 16.Assumed LTV-kW factor and loads of appliances ....................................................................... 19
Table 17. Impact of load errors on voltage drop in distribution transformers........................................... 31

Introduction

There are two basic methodologies of Volt/var optimization in distribution networks:

 One based on detailed power flow models [1] – [8]


 One based on regression models and sensitivity factors [9]-[10].

In the recent years, popular objectives of the Volt/var control applications of the Distribution
Management Systems (DMS) became the short-term load reduction via voltage reduction based
on short-term Load-To-Voltage dependencies – (i.e. LTV-factors), and the energy conservation
based on the long-term CVR-factors. Load and energy losses are considered here as a part of the
objective function. The optimization is limited mostly by the voltage quality constraints at the
customer terminals and by loading constraints of the power elements in both distribution and
transmission systems. Other constraints may include additional voltage constraints at
distribution and transmission buses, operating reserves of real and reactive power in the
transmission and distribution systems, power factors at particular buses, etc.

While the typical voltage tolerance at the customer terminals is 10% (±5%), the actual room for
voltage optimization is much smaller. For each individual customer terminal, it is the distance
from the initial voltage to the target optimal voltage. For instance, if the initial voltage is the
nominal one (0% deviation), and the target voltage is -5%, then the room for optimization is 5%.
If the individual customer can control the voltage at this terminal, the full 5% can be utilized.
However, if the means for voltage control are common to a group of customers, the room for
voltage control is defined by the distances of the initial extreme voltages within the group to the
target voltage and by the sensitivities of the voltages to the means of voltage control. For
instance, assume two capacitors are connected to a primary feeder – one in the middle of the
feeder and another, which is twice the size of the first, is connected close to the end of the same
feeder. The initial critical voltage at the terminals of the customers connected upstream from the
capacitors (e.g., through a heavily loaded distribution transformer and secondaries) is -4%. It
means that the room for lowering the voltage for these customers is only 1%. Assume also that
the smallest room at the customer terminal close to the end of the feeder is 2%. Assume also that,
by switching OFF the first capacitor, the voltage in the upstream node reduces by 0.5%, and the
voltage in the second node reduces by 1%. Hence, switching OFF the first capacitor is acceptable
for both nodes. However, switching OFF the second capacitor instead of the first reduces the
voltages by 1% in the first node and by 2.5% in the second one. This is not acceptable for the
second node. Hence, the second node is the voltage-critical one for the control of capacitors.
If in the same example, the voltage can be controlled only by a feeder voltage regulator at the
head of the feeder, the room for voltage reduction is defined by the first node. If the only voltage
controlling devices is a substation LTC, the room for voltage control is defined by the lowest
voltages among all customer terminals connected to the substation bus. Because of the great
diversity of the properties of nodal loads and power elements, the actual room for voltage
optimization is in many cases in the range of a single-digit percentage, with the lower values
applicable at times of heavy load.

In addition, there is a degree of uncertainty in any volt/var control methodology. Consequently,


in order to minimize the probability of violation of the standard voltage limits, the standard
voltage tolerances should be reduced to allow for the probable errors in the assessment and of
volt/var control.

With high penetration of DERs with volt/var control capabilities, there will be a significant
increase in the diversity of the properties and behavior of local load/generation. This may
increase the uncertainty of volt/var optimization. However, the presence of multiple local var
resources distributed along the distribution circuits (often in the secondaries) may increase the
room for voltage optimization by flattening the voltage profiles along the circuits.

Taking into account the uncertainty of volt/var optimization by reducing the operational voltage
tolerances reduces the benefits of the optimization. In some cases, it may result in no benefits at
all.

Therefore, it is important to assess the uncertainties and take measures to minimize them.

The issues of the volt/var optimization uncertainties were addressed in [11] - [15]. This paper
continues this discussion with more emphasis on the conditions of Active Distribution Networks.
The process of the volt/var optimization for both of the methodologies mentioned earlier consists
of the following stages each of which introduces a certain degree of uncertainty:
 Situational awareness
 Decision making
 Control
Uncertainties of Situational Awareness

The situational awareness of the initial operating conditions is based on information from the
situational awareness in the initial operating conditions is based on information from different
resources. The major resources are discussed below.

Direct measurements
Accuracy
The decision-making process of volt/var optimization is based on knowledge of the initial
operating conditions. These conditions are defined, firstly, by using direct measurements from
selected sensors. The main sensors for this task are the SCADA devices and the AMI meters.
However, while the SCADA measurements are, typically, momentary readings, the AMI data are
interval averages. This discrepancy introduces an additional uncertainty in defining the initial
situation.

An example of voltage and kW variations during a 15-minute interval for a secondary bus of a
distribution transformer is presented in Figure 1 and Figure 2.

60%

50%

40%

30%
Frequency

20%

10%

0%
0.0% 0.2% 0.4% 0.6% 0.8% 1.0% 1.2%

-10%
Standard deviation of voltages during 15 min interval, % of nominal voltage

Figure 1. Histogram of standard deviations of voltages during 15 min interval


40%

35%

30%

25%

20%
Frequency

15%

10%

5%

0%
0% 10% 20% 30% 40% 50% 60%

-5%
Variation coefficient of kW during 15 min interval

Figure 2. Histogram of variation coefficients of kW during 15-minute interval

Latency
Another uncertainty related to measurements is due to the difference in values at the time of the
measurements compared with the time of the actions (the latency effect [11]). Examples of the
uncertainties introduced by the latency effect are presented in Table 1 through Table 6Error!
Reference source not found..

As seen in the tables, the effect of latency may be significant. For instance, if at the time of the
voltage measurements there room for a 2% voltage reduction, but at the time of implementing
this reduction 15 minutes later, the voltage is already lower by 1.4% (See two standard
deviations in Table 2), the voltage at the customer terminal will be below the standard limit.
Table 1 Correlation vs latency of voltage measurements at a residential customer

Latency Correlation coefficient


0 1
15min 0.88
30min 0.82
45min 0.82
60 min 0.78
Table 2. Standard deviations of differences in voltage measurements due to latency at a residential customer

Latency 15min 30min 60min


Standard Deviation, % of nominal voltage 0.70% 0.86% 0.94%

Table 3 through Table 6 present examples of the latency effect on real loads. As seen in the
tables, the difference in loads of an individual residential customer is significant even for the 15-
minute latency. The difference becomes smaller for the sum of loads of more customers. In
Table 4, the numbers for more than one customer are theoretically modeled based on the number
of samples, assuming these customers are identical and independent of each other. In reality, the
differences can be greater (see Table 6).
Table 3. Correlation vs latency of load measurements at a residential customer

Latency Correlation coefficient


0min 1
15min 0.83
30min 0.64
45min 0.46
60min 0.34

Table 4. Standard deviations of differences in load measurements of residential customers due to latency

Latency 15min 30min 45min 60min


Standard Deviation for one customer, % of actual load 38% 62% 87% 105%
5 identical customers (Pole-top distribution transformer) 17.0% 27.9% 39.0% 47.0%
10 identical customers (A distribution transformer) 12.0% 19.7% 27.6% 33.2%
100 identical customers (An apartment building) 3.8% 6.2% 8.7% 10.5%
1000 identical customers (A feeder) 1.2% 2.0% 2.8% 3.3%

Table 5. Correlation vs latency of load measurements at a substation bus

Latency Correlation coefficient


0min 1
30min 0.99
60min 0.96

Table 6. Standard deviations of differences in load measurements at a substation bus due to latency

Latency 30min 60min

Standard Deviation 5% 10%


Representation of the situation
Not all direct measurements are used for the representation of the initial operating conditions. In
practice, the situational awareness is based either on selected direct measurements, or on values
derived from selected measurements. Some examples are presented below:

Imbalance.
In many cases, information about the primaries is represented by measurements of voltage and/or
load in one phase, or by average phase values. The actual phase values are, typically, different
from such a representation. For example, measurements taken at a distribution substation bus
show a standard deviation of phase voltages from the average voltage equal 0.2% of the nominal
voltage, and the standard deviation of phase voltages from each other equal 0.3%. For the phase
currents, the standard deviations are 5% and 8% respectively. The differences can be much larger
in the secondary circuits.)

Selection of reference measurements for state estimation (SCADA).

The SCADA measurements are used in models based on the power flow for the reference bus
voltages. The power and/or current measurements at the feeder heads and at selected points
along the feeders are used for state estimation (basically, for balancing the distribution
transformer loads with the near real time SCADA measurements [2], [5]. [7], [16].

The number of sensors along the feeders and their location affect the accuracy of the load models
of distribution transformers.

Figure 3 through Figure 7 and Table 7 through Table 10 present a simplified example of voltage
model errors dependent on the number and placement of measurements along the primary feeder.
In this example, the initial errors of the distribution transformer load models range within ±20%.
The voltage drop along the primary feeder is about 6%.
B A A B

Nodes: 1 … 10 … …… 20 ………..… 40 ….…….. 47

Figure 3, Sample circuit

Table 7. Errors in nodal load models

No SA One point three points A Three points B


Average -3% 0% 0% 0%
StDev 9% 10% 5% 7%

Table 8. Errors in models of the power flow in segment of the feeder

No SA One point three points A Three points B


Average -9% -5% 0% -1%
StDev 3% 3% 1% 2%
0.6%

0.5%

Errors. % of nominal voltage 0.4%

0.3%

0.2%

0.1%

0.0%
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55
Nodes

No SA Primary -one Primary-three-A Primary-three-b

Figure 4. Errors in primary voltage models

LV no SA LV one point
30.0% 30.0%

20.0%
20.0%
Frequency
Frequency

10.0%
10.0%
0.0%
0.0% -1.0% 0.0% 1.0% 2.0%
-1.0% 0.0% 1.0% 2.0% -10.0%
-10.0% Errors, % of nominal voltage
Errors, % of nominal voltage

LV three points-A
30.0% LV three points-B
30.0%

20.0% 20.0%
Frequency

Frequency

10.0% 10.0%

0.0% 0.0%
-1.0% 0.0% 1.0% 2.0% -1.0% 0.0% 1.0% 2.0%
-10.0% -10.0%
Errors, % of nominal voltage Errors, % of nominal voltage

Figure 5. Histograms of errors in the secondary voltage models


Table 9. Errors in the models of voltages in secondaries

LV no SA LV one point LV three points-A LV three points-B


Average 0.45% 0.15% 0.01% 0.06%
STDev 0.51% 0.45% 0.16% 0.27%

0.00%
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55
Errors in models of accumulated primary losses , % of actual losses

-2.00%

-4.00%

-6.00%

-8.00%

-10.00%

-12.00%

-14.00%

-16.00%
Nodes

No SA One point Three points A Three points B

Figure 6. Errors of the models of cumulative losses in the primary feeder


No SA One point
25.0% 25%
Frequency 20.0% 20%

Frequency
15.0% 15%
10.0% 10%
5.0% 5%
0.0% 0%
-60% -40% -20% 0% 20% 40% 60% -60% -40% -20% 0% 20% 40% 60%
Errors, % of actual loss in distribution Errors, % of actual loss in distribution
transformer transformer

three points A Three points B


25% 25%
20% 20%

Frequency
Frequency

15% 15%
10%
10%
5%
5% 0%
0% -60% -40% -20% 0% 20% 40% 60%
-60% -40% -20% 0% 20% 40% 60% Errors, % of actual loss in distribution
Errors, % of actual loss in distribution transformer
transformer

Figure 7. Histograms of errors in the models of load losses in distribution transformers

Table 10. Errors in the models of load losses in distribution transformers

No SA One point Three points A Three points B


Average -6.4% 1.4% 0.0% 0.5%
StDEv 19% 20% 9% 15%

As seen in the tables and figures above, the state estimation significantly reduces the uncertainty
of the parameters of the power flow based models, and the number and placement of the primary
measurements used for the state estimation significantly affects the accuracy of the models.

Selection of voltage critical points


Let us designate as “voltage-critical points” the terminals whose voltages are the first to reach
the voltage limits after some particular control is applied. Then, the room for voltage
optimization is limited by the voltages at these “voltage-critical points”. It may also be limited by
the currents in some circuit elements where the current is the first to reach the current limits.
These are the “current-critical segments”. There can be many such voltage-critical points and
current-critical segments, depending on the particular operating situations and control.
The values at the critical points are not necessarily at the limits or very close to them. The
criterion for a value to be a critical one is for it to be the extreme value for a particular control
before this control is applied.

In the detailed power flow model, all these critical values are known m in practice in near real
time. When volt/var optimization is applied to the power flow model, the impact of the
optimization on the critical values is also known.

In the regression/sensitivity models, the critical points are defined based on analyses of available
AMI readings and by using predefined sensitivities of the critical values to the intended controls.
Ideally, a large amount of near real time readings should be analyzed to adequately determine the
critical values for a near real time optimization. Such an approach presents a challenge due to the
possible mass of information to be collected in near real time. Therefore, analyses of the data
collected from the past time, e.g., a day before, may be used to reduce the number of monitoring
sites to a preselected set of bellwether points. These points can be updated based on the new past
performance analyses. This latency can introduce another uncertainty in the situational
awareness of the conditions for volt/var optimization.

An illustrative example of the dynamics of critical points is presented below.

The sample circuit (Figure 8) is a distribution feeder with ten nodal loads connected to secondary
circuits fed from distribution transformers. The loads belong to different categories: residential
(R), small commercial (SC), and medium commercial (MC). The power factor of the natural load
(not net load) ranges from 0.85 through 0.95. Aggregated photovoltaic distributed generators
(PV) are connected to an equivalent point of on the secondaries. The penetration of the
renewables is 30%. All PVs are capable of autonomous volt/var control. They are set to support
the voltages at their terminals within the range of -1% through -2% of nominal voltage. The var
capabilities of the PVs are limited by the currents through their inverters, based on the rated
power factor of 0.9.

The independent variables (the ones available for central control) are the substation bus voltage
and the three capacitors connected along the primary feeder.

The volt/var control of the PVs is dependent on the local voltages determined by the loading of
the circuits and by the centrally applied control. Therefore, from the standpoint of the central
control by other means, the PV volt/var values are dependent variables, unless the settings of
their volt/var control are also centrally controlled.
G G G G G G G G G

Nodes: 1 2 3 4 5 6 7 8 9 10

R SC R R MC R MC R SC R

Figure 8. Sample circuit

The following operating conditions were considered in this example:


A. Time – 1 PM, PV’s real power generation close to 100%
B. Time – 8PM , PV’s real power generation close to 0%.
In both cases, the bus voltage deviations were changed in the range from 2% through 10%, and
the capacitors were controlled to keep the voltages at their terminals within the range of 0%
through 2% above the nominal voltage. The capacitor controls were applied after the PVs
finished their adjustment to the initial voltage, because the VPs have a smaller response time
than the capacitor control. After the capacitor controls were complete, the PV volt/var values
automatically adjusted to the voltage changes.

Table 11 presents the low voltage critical points for different times and for different bus voltages.
As seen in the table, most of the critical points are different at different times, and some are
different for different bus voltages.
Table 11. Low voltage critical points under different bus voltages at different times of day

Bus
2 3 4 5 6 7 8 9 10
voltage, %
Node 1
Node 2
Node 3
Node 4 8PM
Node 5 1PM 1PM 1PM 1PM 8PM
Node 6
Node 7 1PM 1PM 1PM 1PM 1PM 1PM 1PM & 1PM 1PM &
8PM 8PM
Node 8 8PM 8PM 8PM 8PM 8PM 8PM
Node 9
Node 10 8PM 8PM 8PM 8PM 8PM 8PM 8PM 8PM 8PM
We then considered another condition where, instead of the capacitors being the control variable,
they were kept ON the full time.

Table 12 presents the low voltage critical points for both conditions: a) when the capacitors are
controllable (Cond. 1), and b) when the capacitors are not controllable (Cond. 2). As seen in the
table, the sets of critical points are significantly different for these two conditions.
Table 12. Low voltage critical points under different bus voltages and different control conditions

Bus 2 3 4 5 6 7 8 9 10
Voltage
Node 1 Cond. 2
Node 2
Node 3 Cond. 2
Node 4 Cond. 1
Node 5 Cond. 2 Cond. 1
Node 6 Cond. 2 Cond. 2
Node 7 Cond. 2 Cond. 1 Cond. 2 Cond. 1
& 2 & 2
Node 8 Cond. 1 Cond. 1 Cond. 1 Cond. 1 Cond. 1 Cond. 1
& 2 & 2 & 2 & 2
Node 9
Node 10 Cond. 1 Cond. 1 Cond. 1 Cond. 1 Cond. 1 Cond. 1 Cond. 1 Cond. 1 Cond. 1
& 2 & 2 & 2 & 2 & 2 & 2 & 2 & 2 & 2

The high voltage critical points can be analyzed similarly.

As follows from the above example, for different operating and control conditions, different
points should be monitored and different sensitivities should be determined.

Load/energy sensitivity to voltage


Other components of the situational awareness of the volt/var control conditions are the short-
term real and reactive load dependencies on voltage (LTV-factors) and the corresponding energy
dependencies on voltage (CVR-factors).

A methodology for assessing these factors based on field tests was introduced in 1990s and used
in a number of utilities, [17]-[19]. This methodology is based on an active experiment, where the
load and voltage changes at a test bus or feeder and at a reference (control) bus or feeder are
included in a regression analysis. A number of conditions should be satisfied for the experiment.
Some of them are listed below:
 The correlation coefficient between the load changes before the test at the test bus and
at the reference bus for the same times of different days should be close to 1
 The correlation between the voltage change at the test bus and the load change at the
reference bus should be close to 0.
 The voltage at the test bus should be changed every other day at about the same time
of the day
 Other means of volt/var control along the distribution feeders, e.g., capacitors, should
not change their states during the test times. If this is impossible, some additional
processing of the collected data may be needed [19].

The uncertainties of determining the LTV and CVR factors are minimized by closely satisfying
these conditions. However, even under the best test conditions, a significant degree of
uncertainty remains. The significance of the uncertainty depends on the purpose for which these
factors are used.

For instance, if the objectives are just load or energy consumption reduction, and the knowledge
of the amount of reduction is not of high priority, then a wide range of the LTV/CVR factors will
serve the purpose. However, if one needs to know how the distribution and/or the transmission
losses or the currents in the circuits will change when the voltage is reduced by different means,
then the LTC/CVR factors and load power factors should be known, along with other operating
parameters, and a number other operating parameters should be known. Under some
combinations of all these factors, an error in the LTV/CVR factors may lead to the opposite of
the expected result.

Table 13 and Table 1 present a comparison of test results of CVR and LTV factors derived from
data published in [17] – [18].
Table 13. Illustration of uncertainty of CVR-factors

Test Site CVR-W, % of Coefficient of Correlation


average Variation coefficient between
changes of load at
the test and the
reference bus
Utility Bus 1 141% 5.84% 0.89
1 Bus 2 133% 3.10% 0.93
Utility Bus 3 Spring 62% 0.94
2 Bus 3 Summer 76%
Utility Feeder 1 74% 7.64%
3 Feeder 2 129% 2.40%
Feeder 3 82% 6.25%
Feeder 4 104% 3.47%
Feeder 5 122% 3.39%
Feeder 6 111% 3.24%
Bus 4 98% 2.63%
Utility Bus 5 103% 7.00%
4 Bus 6 90% 5.75%
Bus 7 77% 3.33%
Bus 8 97% 2.13%

Table 14. LTV and CVR factors at different times [17]

Short-term CVR-W factors


Time intervals LTV factors 15:45 - 22:45 - 5:30 5:45 - 10:30 10:45 - 15:15
for Bus 3 22:30
Bus 3, Spring 103% 71% 39% 54% 75%

Bus 3, Summer 113% 71% 92% 66% 47%

Time intervals 10:30 - 18:30 - 23:15 23:30 – 6:30


for Bus 4 18:15
Bus 4 86% 100% 97%

The values in the tables are percentages of the average CVR-factor for all tests presented in the
table. The coefficients of variation present the uncertainty of the CVR-factor determined in each
test. In addition, the CVR-factors are different for different buses, feeders, and times even for the
same utility. If either one of the CVR factors, or if an average factor is used for a long period, an
additional uncertainty is introduced.

The CVR and LTV factors derived from the tests presented in [17] – [18] represented the
sensitivities of the load and energy of the corresponding bus or feeder (including losses). These
values can be used for assessing the effect of bus voltage change on the total load of the
corresponding bus or feeder. However, the CVR/LTV factors of the nodal loads along the
feeders differ from the total bus or feeder sensitivities. Applying these average values across all
nodal loads introduces an additional uncertainty in the situational awareness of the operating
conditions. For instance, in the example above, the load of bus 3 includes a portion of electric
heating load. In the nodes or feeders where heating load is prevalent, the CVR factor may be
much smaller than the bus average, even negative. This difference means a different result of
the power flow models for some segments of the circuit, e.g., an increase of current due to
voltage reduction instead of the expected decrease.

Moving the field tests closer to the nodal loads, which seems possible with the advance of AMI,
presents another challenge. Even for loads of similar nature, in addition to the changes of the
load due to common factors, there are more changes due to different individual factors. The
greater the individual changes, the more they mask the changes due to common factors.

Table 15 presents an analytical example of the effect of individual changes on the correlation
between a test load and a reference load. The loads are of similar households with about the
same average consumption, but with different individual changes. The Coefficients of Variations
(CV) presented in the table are predominantly due to the individual changes. The correlation
coefficient between the same time changes of these loads is 0.19.

Four cases were considered in this illustration. The cases differ in their average loads, simulating
different levels of aggregation of similar loads.
A common factor proportional to the average load for each case was added to the sample loads to
simulate the aggregation of the loads. The correlation factor between the same-time load changes
is presented in the third line of the table. As seen in the table, the smaller the fluctuations of
individual changes, the better the reference load represents the test load.
Table 15. Correlation between load changes at the test and reference buses

Case 1 Case 2 Case 3 Case 4

CV1 65% 13% 7% 3%

CV2 69% 14% 13% 3%

Correlation 0.31 0.65 0.83 0.97


factor
CV of LTV
factor for ±2%
of voltage 694% 19% 11% 5%
change
CV of LTV
factor for ±4%
of voltage 117% 10% 6% 3%
change

Another challenge for the field tests of the CVR/LTV factors in active distribution networks is
the advent of distributed generation. The loads at the test and reference buses are now net loads,
which are highly dependent on the behavior of the distributed generation. It makes it more
difficult to find an adequate reference bus because distributed generation occurs with different
degrees of penetration and has different behavior at different buses. In addition, the current
methodology expresses the CVR/LTV factors in terms of a percentage change of load per
percentage change of voltage. The percentage change of net load does not represent the
CVR/LTV factors [20]. It means that to apply the existing methodology for CVR/LTV tests, the
natural load should be separated from the net load, which is a challenge by itself and a cause for
additional uncertainty.

It is worth considering a new methodology for determining the CVR/LTV factors based on local
factors, such as the characteristics of the components of customers loads and the patterns of load
behavior. We believe that artificial intelligence algorithms can be developed to assess the
CVR/LTV factors of nodal loads based on the capabilities of the AMI, the information made
available in the utility Customer Information Systems, and the individual or group characteristics
of electric devices.

Below is an illustrative example of a very simplified approach to an assessment of LTV/CVR


factors based on the patterns of nodal load shapes. The first example is based on assumed LTV
factors and loads for some household appliances, as shown in Table 16.
Table 16.Assumed LTV-kW factor and loads of appliances

Appliance LTV Load, kW


Refrigerator 1 0.18

Light 1.6 0.2

TV 1.6 0.15

Electric cooking 2 2.5

Laundry washing 1 1.5

Laundry dryer 2 3

Heating 1.8 2

Dishwasher 1 1.5
Figure 9 depicts the load shapes of the sample household for a winter month
8

5
kW

0
9:00
0:00
0:30
1:00
1:30
2:00
2:30
3:00
3:30
4:00
4:30
5:00
5:30
6:00
6:30
7:00
7:30
8:00
8:30

9:30

12:30

16:00
10:00
10:30
11:00
11:30
12:00

13:00
13:30
14:00
14:30
15:00
15:30

16:30
17:00
17:30
18:00
18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30
Mon1 Tue1 We1 Thu1 Fri1 Mon2 Tue2 We2
Thu2 Fri2 Mon3 Tue3 We3 Thu3 Fri3 Mon4
Tue4 We4 Thu4 Fri4 Mon5 Tue5 Average

Figure 9. Weekdays load shape of the sample household

As seen in the figure, the load shapes differ significantly for “similar” days, which demonstrate
the impact of the “individual” factors.

Two “similar” days were selected for analyzing the load shapes and assessing the likely
composition of participation of different appliances in the total load. The assessment was based
on the knowledge of the following factors:
 Type of customer (residential with electric heating and cooking)
 Season (winter)
 Ambient conditions
 Time of day
 Load of appliances
 Occupation of residents (retirees).
kW

0
1
2
3
4
5

0.5
1.5
2.5
3.5
4.5

0:00
0:45
1:30
2:15
3:00

Refrigirator
3:45
4:30
5:15
6:00

Dishwater
6:45
7:30
8:15

Light
9:00
9:45

TV
10:30
11:15
12:00
12:45
Heating 13:30
14:15
15:00
Figure 10. Composition of load for day 1
Cooking

15:45
16:30
17:15
18:00
Washing

18:45
19:30
20:15
21:00
Drying

21:45
22:30
23:15
kW

0
1
2
3
4
5
6

0:00
0:45
1:30
2:15
3:00

Refrigirator
3:45
4:30
5:15
6:00
6:45

Dishwater
7:30
8:15

Light
9:00
9:45

TV
10:30
11:15
12:00
12:45

Heating
13:30
14:15
15:00
Figure 11. Composition of load for day 2 15:45
Cooking

16:30
17:15
18:00
Washing

18:45
19:30
20:15
21:00
Drying

21:45
22:30
23:15
LTV, %kW/%Vo;lt

1
2

0.8
1.2
1.4
1.6
1.8

0:00
0:45
1:30
2:15
3:00
3:45
4:30
5:15
6:00
6:45
7:30
8:15
9:00
9:45

Day1
10:30
11:15
12:00

Day2
12:45
13:30
14:15

Figure 12. LTVs for day 1 and day 2


15:00
15:45
16:30
17:15
18:00
18:45
19:30
20:15
21:00
21:45
22:30
23:15
16

14

12

10

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95

Refrigirator Dishwater Light TV Heating Cooking Washing Drying

Figure 13. Composition of five loads

1.8

1.6
LTV, %kW/%Volt

1.4

1.2

0.8
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94
5 loads First Second Third Fourth Fifth

Figure 14. LTV of five loads


0.25

0.2
Coeffcients of Variation

0.15

0.1

0.05

0
LTV1 LTV2 LTV3 LTV4 LTV5 LTV of five loads
CV-Night CV-Day

Figure 15. Coefficients of variation of LTVs for individual loads and for sum of five loads

The LTV-factors presented above are short-term factors. After a time interval following the
change of voltage, the adjustment of load starts counteracting the immediate change of load. The
length of this time interval and the degree of adjustment are different for different types of load.

Electric heating load has a high degree of adjustment. Illustrative examples of possible
adjustment based on the composition of load for day one are presented in Figure 16 through
Figure 18. The figures illustrate the effect of a voltage reduction of three percent. As seen in the
figures, at the times when the heaters were ON under initial voltage, the load is reduced
according to the short-term LTV factors. However, due to the lower voltage, the heaters stay ON
a longer time, which adds load to the non-heater loads. This addition significantly increases the
loads of individual customers at some time intervals and reduces the diversification for a group
of loads.
0.05

0.04

0.04

0.03

0.03
kWh

0.02

0.02

0.01

0.01

0.00
10:30 10:31 10:32 10:33 10:34 10:35 10:36 10:37 10:38 10:39 10:40 10:41 10:42 10:43
Initial voltage Lower voltage

Figure 16. Heater’s load under different voltages during one heater cycle

0.060

0.050

0.040
kWh

0.030

0.020

0.010

0.000
19:15 19:16 19:17 19:18 19:19 19:20 19:21 19:22 19:23 19:24 19:25 19:26 19:27 19:28 19:29
Initial voltage Lower voltage

Figure 17. Total customer load under different voltages during one heater cycle
3.000

2.500

2.000

1.500
kW

1.000

0.500

0.000
17:30
17:34
17:38
17:42
17:46
17:50
17:54
17:58
18:02
18:06
18:10
18:14
18:18
18:22
18:26
18:30
18:34
18:38
18:42
18:46
18:50
18:54
18:58
19:02
19:06
19:10
19:14
19:18
19:22
19:26
19:30
-0.500

Figure 18. Change of load due to reduced voltage for one customer

12.000

10.000

8.000

6.000
kW

4.000

2.000

0.000
17:30 17:40 17:50 18:00 18:10 18:20 18:30 18:40 18:50 19:00 19:10 19:20 19:30

-2.000
Load 1 Load 2 Load 3 Load 4 Load 5

Figure 19. Change of five loads due to reduced voltage for a group of customers
As seen in the figures, most of the time the load is reduced. However, toward the end of the time
when the heaters are ON the load increases. When considering energy conservation, these load
increases may override the load reduction (depending on the level of penetration of the
heating/cooking loads leading to an increase in energy consumption under lower voltage.

Equation (1) represents in a very simplified manner the ratio of the energy consumption of an
electric heater with its voltage reduced compared to the consumption with the voltage unchanged
[21] .

E(V)/E(Vinit) ≈ LN[1-(Thigh-Tlow)/(Lim0 × V2-Tlow)] ×

[1+LTV × (V-1)]/LN[1-(Thigh-Tlow)/(Lim0-Tlow)] (1)

Where

Thigh and Tlow – are thermostat settings,

Lim0 – is the temperature limit which the heater can reach under the initial voltage Vinit,

V – is the changed voltage, pu

As follows from the equation, when E(V) < E(Vinit), the E(V)/E(Vinit) ratio is greater than one for
all practical values of the variables in the equation.

In our example, the penetration of the heating/cooking load is about 70%, and the ratio
E(V)/E(Vinit) = 1.08.

When the voltage is reduced for a group of customers, the diversification of the load is reduced
due to the higher probability of load overlapping. Figure 19 illustrates the change of the load of a
group of customers due to lower voltage. As seen in the figure, in some cases the load of
individual customers increases at the same time resulting in significant increases of the group
load.

As follows from the above discussion, the LTV and CVR factors may differ significantly from
each other, for different days, times of day, and seasons

Sensitivity of components of the objective function to controls


Volt/var control in the distribution system affects a number of operational components that
should be taken into account in the objective function and constraints of the control. These
components may include the following:

 Voltage deviations in the voltage-critical points


 Currents in the current-critical segments
 Load and/or energy consumption
 Losses
 Power factor for given buses
 Cost of energy
 Cost of control

In the volt/var control methodology based on the detailed power flow model, the impacts of the
controlling measures on the components of the objective function are modeled during the
optimization procedure, given that the model includes all independent and dependent variables.
In active distribution networks, the dependent variables include the locally controlled DERs.

The regression/sensitivity methodology implies the use of partial sensitivities determined by


running a power flow model of a simplified distribution system [10]. However, if the model of
the distribution system does not include all significant dependent variables that are present at the
time of decision-making, the partial sensitivities will not represent the actual effect of the
controls on the components of the objective function.

Illustrative examples of the dependencies of some components of the volt/var objective function
on the controls of a voltage regulator are presented in Figure 20 and Figure 21.

125.00

120.00 4.0

Voltage deviation in critical point, %


Pload = 0.7308Vbus + 96.558
115.00
Percent of value under Vbus=5%

R² = 0.9726 2.0
110.00

105.00 0.0

100.00
-2.0
95.00

90.00
Vcrit = 0.6039Vbus- 6.6029 -4.0
R² = 0.9291
85.00

80.00 -6.0
0.00 2.00 4.00 6.00 8.00 10.00 12.00
Deviation of bus voltage, %

Real load Losses-kW


Feeder head Amps Feeder head PF
Voltage in critical point Linear (Real load )

Figure 20. Sensitivities of components of the objective function to the bus voltage. DER kW=100%

In Figure 20, the DERs with reactive control capabilities are generating 100% of their kW which
means that their reactive power capabilities are limited due to the presence of the active current
component. It is assumed in the examples that the DERs locally control the voltage at their
terminals.
140.0 5.0

4.0
130.0
3.0

Voltage deviation in critical point, %


Percent of value under Vbus=5%

2.0
120.0
Pload = 0.0714Vbus+ 99.73
1.0
R² = 0.9148
110.0 0.0

-1.0
100.0
-2.0
Vcrit = 0.0933Vbus - 4.8113
-3.0
90.0 R² = 0.8203
-4.0

80.0 -5.0
0.00 2.00 4.00 6.00 8.00 10.00 12.00
Deviation of bus voltage. %

Real load Losses-kW


Feeder head Amps Feeder head PF

Figure 21. Sensitivities of components of the objective function to the bus voltage. DER kW=0%

In Figure 20, the DERs are not generating kW which means that their reactive power capabilities
are much greater than in the previous case.

As seen in the figures, in the first case, the actual linearized sensitivity of the voltage at the
critical point is 0.6% per percent change of the bus voltage, and the linearized load sensitivity is
0.74% per percent change of the bus voltage.

In the second case, the sensitivities are 0.09% and 0.07% respectively.

Similarly, significant differences apply to other components of the objective function.

Hence, in active distribution networks with a high penetration of smart inverters, the impact of
the DER cannot be ignored in the base models used for the decision-making process. This
suggests that a detailed power flow model should be the base model used for volt/var control
optimization in active distribution networks.

Uncertainty of Decision-Making

Uncertainty of power flow based models


The power flow models [2], [5], [7], [8], [12], [22]-[25] are based on detailed information about
the distribution circuit topology and facilities including the area from the feeding transmission-
to-distribution substation to the secondary buses of distribution transformers, the equivalent
models of the secondary circuits, and the relevant equivalent of the transmission system. The
load flow parameters are derived from the load models at the secondary buses of distribution
transformers, and from selected near-real-time reference measurements used in a procedure
similar to a state estimation.

Such models can be used for different objective functions of different optimization
methodologies, including different independent and dependent variables.

Let us consider the components of the uncertainties in the power flow based models.

Uncertainty of nodal load models


In the pre-smart-grid distribution power flow, the nodal loads were represented by “typical”
nodal load shapes for several load categories. With the advances of distributed generation and
demand response, the nodal loads do not have a typical representation in many cases. In order to
adequately represent the nodal load in the power flow models used for volt/var optimization, data
collected from the AMI system, DER controllers, and relevant DMS and external databases, as
well as from model processors should be used (Note: are all these “ands” correct?). These
approaches are discussed in [11] – [13], [25]. As shown in [11], the impact of the errors of load
models on the voltage drop in distribution transformers are as presented in Table 17.
Table 17. Impact of load errors on voltage drop in distribution transformers

Not to exceed VD error= 0.5% Not to exceed VD error= 0.3%


Error in kW, % in Error in kvar,% of Error in kW, % in Error in kvar,% of
DT KVA DT kVA DT KVA DT kVA
3-phase ±5 ±10 ±5 ±5
DT ±25 ±5 -25 +10
+25 -10
1-phase ±25 ±15 ±5 ±15
DT ±10 ±25 ±15 ±10

When the nodal loads are balanced with the near-real-time load measurements at the beginning
of the feeder, and/or in some segments of the feeder, the errors in voltage drop in the distribution
primaries can be within 0.2% of the nominal voltage [11].

Uncertainty of secondary equivalent models


The voltage drop in the secondary can be a significant portion of the overall voltage drop in the
distribution system. So far, the representation of the secondary circuits in the utility databases is
inadequate for a direct modeling of the voltage drop in the secondaries. However, by using the
information available from the AMI systems in combination with the detailed power flow down
to the distribution transformer secondaries, plausible secondary circuit equivalents can be
modeled [11]. Such equivalent may keep the error of the voltage drop in the secondaries around
0.3% of nominal voltage [11].

Therefore, the end result error of the voltages at the customer terminals can be assessed as a
combination of the errors in measurements of the reference voltage (e.g., at the feeding
substation bus), and the errors of modeling the voltage drops in the primaries, distributing
transformers, and in the secondaries. If the errors are ±0.5%, ±0.2%, ±0.5%, and ±0.3%
respectively, and they are independent of each other, then the resulting overall error will be
±0.8%. If the errors are ±0.3%, ±0.2%, ±0.3%, and ±0.3%, the resulting overall error will be
±0.6%.

Some results of a comparison of voltage models with field measurements as presented in [5] are
as follows:

 The errors in the voltage model at the primary buses of distribution transformers
are within ±0.5% with about 95% confidence
 The errors in voltage model in the secondaries of distribution transformers are
within ±1% with about 95% confidence.
These results were obtained before the AMI data were available. With the support of AMI, the
accuracy of voltage modeling can be significantly improved.

Uncertainty of DER/microgrid models


There will be many DERs of different types in the active distribution networks, such as

 Stand-alone DERs with or without reactive power capabilities


o Monitored and centrally controlled
o Monitored and locally controlled
o Non-monitored and locally controlled
o Non-monitored and not actively controlled
 DERs in microgrids and in other composite customer entities with a customer energy
management system
 DERs embedded in small customer premises

All of these DERs should be represented in the detailed power flow model in a way that is
consistent with their impact on the components of the objective function.

The sources of information for modeling the different types of DER are as follows:

 Direct data acquisition from monitored sources


 Data derived from DER Data Management Systems, containing the DER’s metadata,
such as DER rating data, data on contracts, relevant historic information, as well as other
data relevant to the management of the DER (metadata)
 Models developed by DER model processors, as suggested in [12] and [25], able of
creating adaptive near real time and short-term look-ahead models of DER. The model
processors derive the models from the combination of data obtainable from the DER
controllers, if monitored, from the DER Data Management System, GIS, AMI Data
Management System, DER providers’ databases, historic measurements and external
databases by using pattern recognition or a learning like algorithm The final models
developed by the model processor should also assess the degree of uncertainty of the
models [12]. This assessment can be based on statistics of previous performance.
.

Uncertainty of objective function and optimization processes

Typically, approximations are necessary when defining the objective function of a complex
process. Some of the approximations are inherent in the methodology of the optimum search
mechanism, which is typically an iterative process, others are introduced in the representation of
the constraints, e.g., when soft constraints like penalty functions are used.

Uncertainty of Control (power equipment properties)

Uncertainty of the impact of LTCs

Substation transformers with under-load tap changers (LTC), as well as voltage regulators,
control voltage by steps. The voltage controllers of these devices have Up and Down settings.
The difference between them is the controller’s bandwidth. The size of the bandwidth should be
greater than the step of the LTC plus the measurement error of the controller to prevent
“hunting”. Some utilities increase the bandwidth to reduce the number of operations in order to
reduce the cost of maintenance.

After the voltage optimization procedure comes up with an optimal voltage value for a specific,
the implementation of this value is executed with the uncertainty defined by the voltage
controller bandwidth. The greater the bandwidth, the greater the uncertainty.

Consider illustrative examples of the errors of voltage control by a step-wise LTC with a
bandwidth of different sizes.

Figure 22 illustrates a case where the uncontrolled voltage (the voltage propagated from the
high-voltage side before the LTC operates) changes from a higher than the desired voltage to a
lower than the desired voltage, and the bandwidth of the controller is two times the step of
control.

As seen in the figure, the controlled voltage deviates from the band-center setting by up to one
step when the uncontrolled voltage is on the higher level, and by down to one step when it is on
the lower level Most of the time, the controlled voltage is close to the target voltage, and is
essentially uniformly distributed within the bandwidth (see the histogram in Figure 23).

1.06

1.04

1.02
Voltages, p.u.

0.98

0.96

0.94
0 10 20 30 40 50 60 70 80
Time intervals during a load cycle

Vmin Vmax Not regulated Limited by LTC Bandcenter Average Voltage

Figure 22. Symmetrical unregulated voltage deviations. Bandwidth = 2*LTC step (0.625%)
16%

14%

12%

10%
Frequency, %

8%

6%

4%

2%

0%

Voltage within bandwidth, p.u.

Figure 23. Histogram of voltages within the bandwidth for the case in Figure 2.

Figure 24 illustrates a case where the bandwidth is four times the size of the step of control. In
this case, the systemic bias is much more noticeable, and the probability that the controlled
voltage is close to the target is low (see the histogram in Figure 25).
1.06

1.04

1.02
Voltages, p.u.

0.98

0.96

0.94
0 10 20 30 40 50 60 70 80
Time intervals during a load cycle

Vmin Vmax Not regulated Limited by LTC Bandcenter Average Voltage

Figure 24. Symmetrical unregulated voltage deviations. Bandwidth = 2% (LTC step=0.625%)


14%

12%

10%
Frequency, %

8%

6%

4%

2%

0%

Voltage within bandwidth, p.u.

Figure 25. Histogram of voltages within the bandwidth for the case in Figure 4.

As seen in the above figures, when the uncontrolled voltage is above the desired voltage (which
means that the LTC moves the steps to a lower voltage), the bias is on the upper side of the
bandwidth, and when the uncontrolled voltage is below the desired (the LTC moves the steps to
a higher voltage), the bias is on the lower side of the bandwidth.

If the position of the uncontrolled voltage relative to LTC settings is known, the likely
uncertainty of the voltage control can be expressed as follows:

Vactual = Vopt + [(Bandwidth/Step)-1]/2 × Step ± Step/2, when the uncontrolled voltage is above
the upper boundary of the bandwidth,

and

Vactual = Vopt - [(Bandwidth/Step)-1]/2 × Step ± Step/2, when the uncontrolled voltage is below
the lower boundary of the bandwidth.

For example, if the bandwidth is 2.5%, the step is 0.625%, and the optimal voltage is 5%, Vactual
ranges from 5.625% to 6.25%, when the uncontrolled voltage is above the upper boundary of the
bandwidth, and from 3.75% to 4.375%, when the uncontrolled voltage is below the lower
boundary of the bandwidth.
If the position of the uncontrolled voltage is unknown, or the uncontrolled voltage is not beyond
the voltage controller’s bandwidth (see e.g., Figure 26 and Figure 27), the uncertainty of voltage
execution is expressed as follows:

Vactual = Vopt ± Bandwidth/2.

In our numerical example it would be

Vactual = 5% ± 2.5%/2 = from 3.75% to 6.25%.

1.02

1.01

0.99
Voltage, p.u.

0.98

0.97

0.96

0.95

0.94
0 10 20 30 40 50 60 70 80
Time instances during a load cycle

Vmin Vmax Not regulated Limited by LTC Bandcenter Average Voltage

Figure 26. Asymmetrical unregulated voltage deviations. Bandwidth = 2%, (LTC step =0.625%)
30

25

20
Frequency, %

15

10

Voltage within bandwidth, p.u

Figure 27. Histogram of voltages within the bandwidth for the case in Figure 26.

As shown in the above examples, if, in addition to the sizes of the steps of control and of the
bandwidth, the position of the uncontrolled voltage relative to the control settings is known, the
uncertainty of the control can be assessed more accurately.

Uncertainty of the impact of capacitors

The initial operational model for the optimization of the volt/var conditions should include all
significant dependent variables One of them is the status of locally controlled capacitors (from
the standpoint of central control). If the statuses of such capacitors are not monitored, their
statuses should be assessed by the state estimation procedure based mostly on the reactive power
balances. These balances include the reactive loads, the DERs’ reactive power, and the
capacitors’ reactive power. In order to reduce the uncertainty of the assessment of state of the
capacitor, additional information should be used. For instance, such information may include
previously recorded step-wise changes of reactive power measurements consistent with the steps
of relevant capacitors, or previous assessments of the position of the reference operating
parameters relative to the control settings.

The biggest uncertainty of the assessment of the statuses of the capacitors occurs when the initial
power flow/state estimation model estimates that the reference parameter at the terminals of the
capacitors is within the bandwidth of the controller.
Figure 28 through Figure 31 show illustrative examples of the dependencies of the capacitor
statuses on the initial status of capacitors before the change of the reference parameters (it is just
voltage in this example) and the previous position of the voltage relative to the control settings.

In these examples, the initial operating point is within the bandwidth of the voltage controller.

1.03

1.02

1.01
Voltages, p.u.

0.99

0.98

0.97
0 20 40 60 80 100 120 140 160 180
Time intervals during a load cycle

Vmin Vmax Not regulated Capacitor effect


i

Figure 28. Status of capacitor when the operating point is within the bandwidth. Initial status of capacitor – ON, voltage
change from low to high. – operating point, capacitor is ON
1.03

1.02

1.01
Voltages, p.u.

0.99

0.98

0.97
0 20 40 60 80 100 120 140 160 180
Time intervals during a load cycle

Vmin Vmax Not regulated Capacitor effect

Figure 29. Status of capacitor when the operating point is within the bandwidth. Initial status of capacitor – ON, Voltage
change from high to low. . – operating point, capacitor is OFF

1.03

1.02

1.01
Voltages, p.u.

0.99

0.98

0.97
0 20 40 60 80 100 120 140 160 180
Time intervals during a load cycle

Vmin Vmax Not regulated Capacitor effect


Figure 30. Status of capacitor when the operating point is within the bandwidth. Initial status of capacitor – OFF, Voltage
change from low to high. – operating point, capacitor is ON

1.03

1.02

1.01
Voltages, p.u.

0.99

0.98

0.97
0 20 40 60 80 100 120 140 160 180
Time intervals during a load cycle

Vmin Vmax Not regulated Capacitor effect

Figure 31. Status of capacitor when the operating point is within the bandwidth. Initial status of capacitor – OFF, voltage
change from high to low. – operating point, capacitor is OFF

As seen in the figures, the statuses of the capacitors when the voltage is within the bandwidth are
different depending on the previous status of capacitors and the position of the previous voltage
relative to the voltage settings.

Uncertainty of the impact of DERs

DERs in active distribution networks can be active actors controlling voltages and vars. The
DERs with reactive power capabilities can provide volt/var control either by changing the
reactive power injection under independent kW injection, or by active changing of both the real
and reactive power injections [26]. The DERs without reactive power capabilities can still
control voltage by curtailing the kW, e.g., to prevent overvoltage. (Note, add period)

If the DERs are centrally controlled, individually or by groups, the uncertainty of the execution
depends on the accuracy of the DER models, including the availability of the controls, the
operational limits the interactions of the combination of the related DERs, the actual kW
injections, and other local parameters. It also depends on the local settings of the individual DER
controllers, including the droop and the bandwidth of the controllers, as well as on the energy
management systems of the composite customers with multiple DERs [27-32].

The uncertainty of the DER controls significantly increases in times of rapid intermittency. The
assessment of the expected uncertainty of DER controls can be based on historic statistics and
current ambient conditions.

Conclusions

1. The voltage tolerances for volt/var optimization under normal conditions are narrow.
Hence, the accuracy of voltage and var control is critical for preventing violations of
standard voltage quality.
2. As any other complex control process, volt/var optimization in active distribution
networks has some degree of uncertainty.
3. Measures to reduce the degree of uncertainty should be considered when choosing the
methodology and the properties of the control process to ensure that power quality
remains within standard limits and to obtain the expected benefits.
4. The probable uncertainties of the process should be assessed and taken into account in the
decision-making procedure.
5. The following groups of the sources of uncertainties of the volt/var optimization process
in the active distribution network are discussed in the paper:
 Situational awareness
 Decision making
 Control
6. Some recommendations for the reduction of the uncertainties are suggested in the paper.

Acknowledment.

Author thanks Mr. Martin Delson for editing the paper.

References and further reading

1. Nokhum S. Markushevich, Voltage and Var Control in Automated Distribution Systems; Third
International Symposium on Distribution Automation and Demand Side Management, DA/DSM 93,
January 1993, Palm Springs, California
2. Nokhum S. Markushevich Modeling Distribution Automation, Fourth International Symposium on
Distribution Automation and Demand Side Management, DA//DSM 94, January 1994, Florida
3. Nokhum S. Markushevich Ron E. Nielsen, Update on DA Pilot Project at B.C. Hydro , Fifth DA/DSM
Conference January 1995, San Jose, California
4. Nokhum Markushevich, Ron E. Nielson, Dynamic System Load Control through Use of Optimal Voltage
and Var Control, 1998 Dynamic Modeling Control Applications for Industry Workshop, IEEE Industry
Application Society, 1998, Vancouver, Canada
5. Nokhum S. Markushevich, Aleksandr P. Berman, Charles J. Jensen, and James C. Clemmer,
Implementation Of Advanced Distribution Automation In U.S.A Utilities, , Electricity Distribution,
8PM01. Part 1: Contributions. CIRED. 16th International Conference and Exhibition on (IEE Conf. Publ
No. 482), Amsterdam, 8PM01
6. Nokhum Markushevich and Edward Chan, Integrated Voltage, Var Control and Demand Response in
Distribution Systems , IEEE, March 8PM09, Seattle
7. Nokhum Markushevich, Applications of Advanced Distribution Automation in the Smart Grid
Environment , T&D Online Magazine, January-February 8PM10 issue. Available:
http://www.myvirtualpaper.com/doc/Electric-Energy/january-february-8PM10/8PM10012802/#18
8. Valentina Dabic, Cheong Siew Jamie and Dennis Acebedo, BC Hydro’s Experience on Voltage VAR
Optimization in Distribution System, Presented at the 8PM10 IEEE PES conference, 8PM10
9. Jeff St. John, How Smart Meters Are Helping Utilities With Voltage Management. Available:
https://www.greentechmedia.com/articles/read/smart-grid-snapshot-ami-enabled-voltage-control-on-the-
rise
10. Management of energy demand and energy efficiency savings from voltage optimization on electric
power systems using AMI-based data analysis, US 9325174 B2. Available:
https://www.google.com/patents/US9325174
11. Benefits of Utilizing Advanced Metering Provided Information Support and Control Capabilities in
Distribution Automation Applications, EPRI Product ID 1018984, Technical Update, December 2009.
Available:
http://www.epri.com/abstracts/Pages/ProductAbstract.aspx?ProductId=000000000001018984&Mode=do
wnload
12. Development of Transmission Bus Load Model (TBLM), Available: http://collaborate.nist.gov/twiki-
sggrid/pub/SmartGrid/TnD/TBLMUseCase_V14-03-13-13-posted.pdf
13. Nokhum Markushevich, AMI Information Support of Distribution and Transmission Operations,
Available: http://collaborate.nist.gov/twiki-
sggrid/pub/SmartGrid/TnD/Requirements_for_information_exchange_between_AMI_and_DMS-v1.pdf
14. Nokhum Markushevich, Uncertainties of TBLM and Information Needed to Assess it, Available:
https://collaborate.nist.gov/twiki-
sggrid/pub/SmartGrid/TnD/Uncertainty_of_TBLM_component_models1.pdf
15. Nokhum Markushevich, The Benefits and Challenges of the Integrated Volt/Var Optimization in the
Smart Grid Environment, Presented at IEEE PES GM 8PM11, Detroit
16. Nokhum S. Markushevich, Enhancement of the Authenticity of the Electric Network Power Flow
Calculation, Electrichestvo, # 12, 1979, pp. 1-5. (In Russian)
17. Alf Dwyer, Ron E. Nielsen, Joerg Stangl, Nokhum S.Markushevich, Load to Voltage Dependency Tests
at B.C. Hydro, IEEE Transactions on Power Systems, Vol. 10, No. 2. May 1995
18. N. S. Markushevich, R. E. Nielsen, J. M. Hall, A. K. Nakamura, R. L. Nuelk, Impact of Automated
Voltage/var Control in Distribution on Power System Operations, DA/DSM Conference, 1996
19. Nokhum Markushevich; Aleksandr Berman and Ron Nielsen, Methodologies for Assessment of Actual
Field Results of Distribution Voltage and Var Optimization, presented at IEEE PES 2012 T and D
20. Nokhum Markushevich, New Aspects of IVVO in Active Distribution Networks, Presented at IEEE PES
2012 T and D
21. Nokhum Markushevich, Analysis of electric heating load dependency on voltage. Available:
https://www.scribd.com/document/376566559/Analysis-of-electric-heating-load-
dependency-on-voltage
22. Guidelines for Assisting Understanding and Use of IntelliGrid Architecture Recommendations:
Distribution Operations, Use Case for Distribution Operation Model and Analysis, EPRI Palo Alto, CA:
2006. 1013612. Available: http://www.epri.com/search/Pages/results.aspx?k=1013612
23. IntelliGrid Architecture Development for Distribution Systems. Requirements and Device Information
Models for Integrated Advanced Distribution Automation Applications, EPRI, Palo Alto, CA: 2008,
Available: http://www.epri.com/search/Pages/results.aspx?k=1013843
24. Distributed Energy Resources and Management of Future Distribution, EPRI, Palo Alto, CA: 2010.
1020832. Available:
http://www.epri.com/abstracts/Pages/ProductAbstract.aspx?ProductId=000000000001020832
25. Nokhum Markushevich, Cross-cutting Aspects of Smart Distribution Grid Applications, Presented at
IEEE PES GM 2011, Detroit
26. Nokhum Markushevich, Vars versus Watts from Distributed Energy Resources. Available:
https://www.scribd.com/document/376740872/Vars-versus-Watts-from-Distributed-Energy-Resources
27. Coordination of Volt/var control in Connected Mode under Normal Operating Conditions, Use Case
Description. Available: http://smartgrid.epri.com/Repository/Repository.aspx/
28. Update aggregated at PCC real and reactive load-to-voltage dependencies, Use Case Description.
Available: http://smartgrid.epri.com/Repository/Repository.aspx/
29. Updates of capability curves of the microgrid’s DERs, Use Case Description. Available:
http://smartgrid.epri.com/Repository/Repository.aspx/
30. Updating information on microgrid dispatchable load, Use Case Description. Available:
http://smartgrid.epri.com/Repository/Repository.aspx/
31. Updating dependencies of the microgrid operational model on external conditions, Use Case Description.
Available: http://smartgrid.epri.com/Repository/Repository.aspx/
32. Update aggregated at PCC real and reactive load-to-frequency and load-to-voltage dependencies in the
emergency ranges, Use Case Description. Available:
http://smartgrid.epri.com/Repository/Repository.aspx/
33. Nokhum Markushevich, Who Benefits from Voltage/var Control in Active Distribution Networks?
http://www.energycentral.com/c/iu/who-benefits-voltagevar-control-active-distribution-networks
34. Nokhum Markushevich, Voltage/var Optimization in Active Distribution Networks
https://www.scribd.com/document/376758184/Voltage-var-Optimization-in-Active-Distribution-
Networks
35. Implementing Distribution Performance Optimization Functions that Integrate with Advanced Metering:
AMI for Volt/Var Optimization. EPRI, Palo Alto, CA: 8PM10 108PM091. Available:
http://www.epri.com/abstracts/Pages/ProductAbstract.aspx?ProductId=00000000000108PM091 (for fee)
36. Nokhum Markushevich, Information Support of Operations of Power Systems with Active Distribution
Networks. Available: http://www.energycentral.com/c/iu/information-support-operations-power-systems-
active-distribution-networks
37. Nokhum Markushevich, The Challenges of Load Reduction via Voltage Reduction in Active
Distribution Networks Available: http://www.energycentral.com/c/iu/challenges-load-reduction-voltage-
reduction-active-distribution-networks
38. Nokhum Markushevich, Voltage Reduction Effect in Active Distribution Networks. Available:
https://www.scribd.com/document/376755241/Voltage-Reduction-Effect-in-Active-Distribution-
Networks
39. Nokhum Markushevich, Watts or Vars From DERs for Voltage Support?,
http://www.energycentral.com/c/iu/watts-or-vars-ders-voltage-support
40. Nokhum Markushevich, When to Use DER’s Watts for Voltage Support? Available:
https://www.scribd.com/document/376754060/When-to-Use-DER-s-Watts-for-Voltage-
Support-Watts-for-Volts-scribd
41. Nokhum Markushevich, Reactive Power Capabilities of DER’s Smart Inverters.
http://www.energycentral.com/c/iu/reactive-power-capabilities-der%E2%80%99s-smart-inverters
42. Nokhum Markushevich, Operations of Smart Inverters in Active Distribution Networks. Available:
https://www.scribd.com/document/376627135/Operations-of-Smart-Inverters-in-Active-
Distribution-Networks
43. Nokhum Markushevich, Information Exchange between Advanced Microgrids and Electric Power
Systems. Available: https://www.scribd.com/document/376567099/Information-Exchange-between-
Advanced-Microgrids-and-Electric-Power-Systems
44. Nokhum Markushevich, On the Subject of Mitigating Voltage Fluctuations due to the Variability Of
DER. Available: h https://www.scribd.com/document/376625816/Mitigation-of-Voltage-Fluctuations-
by-Controlling-Reactive-Power-of-DER

Potrebbero piacerti anche