Load modelling in distributed generation planning

Class-18: Load modelling in
distributed generation planning
Prof. (Dr.) Pravat Kumar Rout
Department of EEE
ITER
Siksha ‘O’ Anusandhan (Deemed to be University),
Bhubaneswar, Odisha, India
1

Introduction
 Load models can be classified into two broad categories: (1) Static
models; (2) Dynamic models
 There are two types of approaches to identify the model
parameters: (1) measurement based; (2) component based
 Load modelling is essential to power system analysis, planning, and
control.
 It is necessary to understand the characteristics of the modern
loads with emerging smart grid technologies such as distributed
generators (DGs), electric vehicles (EVs), and demand side
management (DSM)
2

Continue…
 The uncertainty and difficulty of load modelling comes from
the large number of diverse load components, time-varying
and weather-dependent compositions, and the lack of
measurements and detailed load information.
 The goal of load modelling is to develop simple mathematical
models to approximate load behaviours.
 Load modelling consists of two main steps: 1) selecting a
load model structure, and 2) identifying the load model
parameters using component or measurement-based
approaches.
3

Continue…
 Component-based or physically-based modelling is based on the
knowledge of physical behaviours of loads and mathematical
relations that describe the functioning of load devices. However,
obtaining such information is not always possible, which motivates
which motivates the research in measurement based modelling.
 Measurement-based modelling collects measurements from data
acquisition equipment to derive load characteristics. The main
advantage is that this approach directly obtains the data from the
actual network, and can be applied to any load. However, a
developed model at one network location may not be applicable
to other locations.
4

Continue…
 The parameters of the load models are estimated by fitting the
acquired data to a load model structure using identification and
estimation techniques. Artificial Neural Network (ANN) is also used
to model loads by mapping the input data set to the output.
5

Load Modelling Terminology
 Load Model: A load model is defined as an analytical, mathematical,
equivalent circuit based, physical component based, or any other suitable
representation of electrical equipment (or group of these), which describes
changes in the electrical characteristics of the modelled equipment as a
function of variations in the relevant supply conditions (e.g. supply system
voltage).
 Electrical characteristics: A set of parameters which characterise the
electrical behaviour and response of the modelled load for various changes
in system operating and loading conditions. Generally, the two most
important load characteristics are the variations of active and reactive power
demands given as a function of supply system voltage.
6

The need for improved load models
 New load models must be developed when new loads are introduced to
the power system in large numbers.
 Regardless of the introduction of new loads, there is always a need to
regularly review and update existing load models to ensure that they are
representative of the loads which are currently used within the power
system.
 Changes in technological trends, or the introduction of performance
legislation, are both likely to alter the electrical characteristics of loads.
7

Continue…
 The growing volume of installed DG and the possible implementation of
DSM schemes are only two areas which are of interest to the research
community.
 The potential widespread adoption of electric vehicles (EV) may
significantly increase the demand within LV residential networks, and will
require increased levels of control in the power system.
 These above examples suggest that electrical power system components
must become more sophisticated and more flexible; and this should be
reflected in power system simulation tools.
8

 Static load models: It describes the electrical characteristics of the
modelled load as function of the known or specified system
parameters. Static load models use only the present state supply
conditions to determine the required load characteristics, i.e. it is
assumed that the load response to changes in the operating
conditions is instantaneous.
 Dynamic load models: describe changes in the load characteristics
as a function of the previous and current state of the supply
conditions, e.g. in response to time-dependent transient changes in
the supply conditions.
9

 Aggregate load model : A load model can represent either individual
loads or aggregate loads and the term is often used interchangeably.
However, as both individual load models and aggregate load models
are covered in the main body of this thesis, it is important to make a
clear distinction between the two. In this thesis, the term load model
refers to the model of an individual load, while the term aggregate
load model refers to a load model which represents the electrical
characteristics of a group of individual loads.
10

 Load type: A group of individual loads with the same intended end-use, e.g.
TVs, which may have similar or different electrical characteristics. This is
generally how load-use statistics are presented, e.g. in government level
reports, where loads are typically grouped into the following broad load
types: consumer electronics, information and communication technology
(ICT) equipment, lighting, wet, cold, cooking and space and water heating.
 Load category : A group of electrical devices which, for the purpose of load
modelling, have the same, or similar, electrical characteristics. This may
consist of devices from more than one load type. A load category may be
further divided into several subcategories due to variations in electrical
circuits.
11

 Load sectors : are defined as ’places where similar activities are performed’.
Therefore, there are inherent similarities in the load composition, i.e. load types
and load categories present in the aggregate load, and also in the load-use
patterns. The most common classification of load sectors is residential,
commercial and industrial , but more extensive definitions may include
subsectors, e.g bank office, hospital, hotel, factory and retail, as subsectors of
the commercial sector . This may also be referred to as the load class.
 Load composition : is the percentage contribution of different load categories
to the total active power demand of a group of loads connected at one point
of delivery for end-use consumption of electricity. This may also be referred to
as the load mix.
12

Types of load models
 Load modelling refers to the mathematical representation of the
relationship between the power and voltage in a load bus.
 Load models can be classified into two main categories: Static and
dynamic models.
13
Load models currently used in the industry for (a) steady-state analysis and
dynamic studies (b) active power and (b)reactive power

Static load models: 1: ZIP Model
 ZIP model is commonly used in
both steady-state and dynamic
studies.
 This model represents the
relationship between the voltage
magnitude and power in a
polynomial equation that
combines constant impedance
(Z), current (I), and power (P
components.
14
where P and Q are the active
and reactive powers at
operating voltage; P0 and Q0
are the active and reactive
powers at rated voltage (V0);
Zp, Ip, and Pp are the ZIP
coefficients for active power;
and Zq, Iq, and Pq are the ZIP
coefficients for reactive power.

Static load models: 2: Exponential Model
 The exponential model relates the power and the voltage at a
load bus by exponential equations.
 This model has fewer parameters and is usually used to represent
mixed loads .
 More components with different exponents can be included in
these equations.
 For example, by using three exponential components, the
exponential model can be converted to a ZIP model.
16

Static load models: 3: Frequency Dependent
Model
 This model is derived by multiplying the exponential or ZIP model by a
factor that depends on the bus frequency.
 The factor can be represented as follows.
where f is the frequency of the bus voltage, is the nominal
frequency, and is the frequency sensitivity parameter. Adding the
frequency term to the ZIP model has no physical meaning, since the
component related to the constant impedance becomes dependent
on the frequency.
18
)](1[ 0ffaFactor f −+=
fa
0f

Static load models: 4: Electric Power Research
Institute (EPRI) LOADSYS Model
 This model is used in the ( EPRI) LOADSYN computational program and
Extended Transient Midterm Stability Program (ETMSP) for dynamic
studies .
 The model combines ZIP, exponential, and frequency-dependent
models.
19

20
where P0 and Q0 are the power consumed at the rated voltage
V0 of a device, if the model is used to represent a specific device.
If it models the aggregate load at a bus, V0, P0, and Q0 are initial
operating conditions. The active power is represented by
frequency dependent and independent components. The
reactive power is composed of two terms. The first represents the
reactive power consumption of the load, and the second
approximates the effect of the reactive consumption minus
compensation, which finds the initial reactive power flow at a bus.

Continue…
Pa1 is the frequency-dependent fraction of active power, Qa1 is the reactive load
coefficient representing the ratio of uncompensated reactive load to active power,
Kpv1 and Kpv2 are voltage exponents for frequency dependent and independent
active power, respectively. Kqv1 and Kqv1 are voltage exponents for the reactive
power without and with compensation, respectively. Kpf1 and Kqf1 are the frequency
sensitivity coefficients for active and uncompensated reactive power load,
respectively. Kqf2 is the frequency sensitivity coefficient for reactive power
compensation.
21

Dynamic Load Models
 Studies in voltage stability require the use of dynamic load models
for accurate representation.
 Dynamic load models (DLM) are therefore necessary for dynamics
studies, i.e., the analysis of power system behaviour following small
or large disturbances.
 Dynamic models express the active and reactive powers as a
function of voltage and time.
22

Dynamic Load Models: 1: Induction Motor (IM )
 In dynamic models, the active and reactive power is represented as a
function of the past and present voltage magnitude and frequency of the
load bus. This type of model is commonly derived from the equivalent circuit
of an induction motor, shown in Fig. below. Where Rs an Rr are the static and
rotor resistances respectively, Xs, Xr and Xm are the static, rotor and
magnetizing reactance, respectively, and s is the rotor slip. The induction
motor model is considered as a physically-based model.
23

Dynamic Load Models: 2: Exponential
Recovery Load Model (ERL)
 The exponential recovery load model represents active and
reactive power responses to step disturbances of the bus
voltage.
 This model is commonly used for representing loads that slowly
recover over a time period, which ranges from several
seconds to tens of minutes.
 ERL is also used to model on load tap changers (OLTCs)
which restore the nominal supply voltage after a disturbance.
24

where xp and xq are state variables related to active and
reactive power dynamics, Tp and Tq are time constants of the
exponential recovery response, Nps and Nqs are exponents
related to the steady-state load response, Npt and Nqt are
exponents related to the transient load response.
25
The model is developed as a nonlinear first-order equation to
represent the load response, as shown below.

26 The ERL is further extended as an adaptive dynamic model. The model
has the same characteristics as the exponential recovery model, but
with the power being a function of the voltage multiplied by the state
variable.

Dynamic Load Models: 3: Composite Load
Models (CLM)27
Dynamic
Load Models
: 3:
Composite
Load Models
(CLM)
ZIP+IM
Complex
Load Model
(CLOD)
Western
Electricity
Coordinating
Council
(WECC) CLM

Composite Load Models (CLM): 1: ZIP+ IM
 The composite load model consisting of
ZIP and an induction motor used in
industry for dynamic studies.
 Several composite load models were
considered including ZIP+IM and
Exponential+IM.
 The report concluded that the ZIP+IM
structure is able to model loads with
various conditions, locations, and
compositions.
 The equivalent circuit of the ZIP+IM
model is shown.
28

Composite Load Models (CLM): 2: Complex Load
Model (CLOD)
CLOD is an aggregate dynamic model of large and small motors, non-linear
models of discharge lighting, transformer saturation effects, constant MVA, shunt
capacitors, and a series impedance and tap ratio to represent the effect of
intervening sub-transmission and distribution elements, as shown in figure.
29

Composite Load Models (CLM): 3:Western
Electricity Coordinating Council (WECC) CLM30

Composite Load Models (CLM): 3:Western
Electricity Coordinating Council (WECC) CLM
 The model is assumed to have 80% static loads and 20% dynamic ones.
 The static part is represented by existing data, and the dynamic part is an
induction motor model.
 The model structure is shown, which includes an electrical representation of a
distribution system with a substation transformer, shunt reactance, and a feeder
equivalent.
 At the consumer side, the load model includes a static load model, one power
electronics model, and four types of motor models.
 Although CLM provides a detailed modelling, it is hard to implement as there
are 131 parameters to be identified.
31

4: Artificial Neural Network based Modelling
 ANN-based load modelling matches observed system behaviours without
using a physical form to obtain the output, i.e., it has no physical meaning and
purely relies on measurement data.
 An ANN is composed by a set of processing units interconnected by weights.
The ANNs are trained using a succession of input and output patterns, resulting
in the final values of the connection weights that determine the load model.
 Although ANN is powerful in representing complex nonlinear systems, obtaining
enough data over a wide range of operation conditions is challenging.
 In addition, ANN-based models must be updated periodically when new
measurement datasets are available.
32

5: Low-Voltage (LV) Load Models
 LV networks are usually represented by lumped models . However, the
integration of renewable DGs and the implementation of demand-side
management highlights the need for more detailed modelling of LV
loads.
 Most of existing research focuses on characterizing consumption
profiles of LV residential loads. There are a few studies on developing
physical models to represent their electrical characteristics. ZIP and
exponential models are most commonly used to represent LV loads.
33

Load Model Parameter Identification
 Load model identification methods can be classified into two categories:
component-based and measurement-based.
 The component-based methods aggregate models of individual electrical
components to form an aggregated load model. This approach requires
knowledge on the load composition, i.e., the percentage of load
consumed by each type of load components.
 Measurement-based approaches leverage data from devices such as
PMUs, smart meters, etc. A model structure is selected and its parameters
are derived using computation techniques such as statistics, artificial
intelligence, and pattern recognition. Component-based methods start
from the individual components, while measurement-based ones start from
the measurement data .
35

Component based approach
 The measurement-based load modelling approach obtains the input data for load model
parameters from measurement devices installed in the actual power system. These devices will
measure voltage and/or frequency variations and the resulting changes in load
characteristics to either staged or naturally occurring disturbances, which are then fitted to
the selected load model form.
 Load is commonly divided into industrial, commercial, and residential classes.
 The approach requires three datasets: 1) models of individual components, 2) component
composition, i.e., the percentage of load consumed by each load component, and 3) class
composition, i.e., the percentage contribution of each load class to the aggregate load.
 The individual load components can be represented using static or dynamic models.
 The parameters for individual component models can be obtained through laboratory
experiments.
36

37
Component based modelling approach

Measurement based approach
 If the measurement-based load modelling approach can be considered a
’top-down’ technique, the component-based load modelling approach can
be described as ’bottom-up’ approach, as the aggregate load model is
constructed from the models of the individual load components within the
aggregate load.
 The main advantage of this modelling approach is that no measurements of
the power system are required to develop the aggregate load model.
 However, this must be compensated for by gathering large amounts of data
to determine the load composition of the modelled aggregate load.
38

39 Measurements based modelling
approach

Comparison of Measurement and Component
based approaches: Measurement based
Advantages
 Based on data from the actual
systems
 Provide accurate models for
measured locations and time
 A generic method that can be
applied for various models
 No need to have deep knowledge
of loads
Disadvantages
 Unable to account for different
operation conditions
 Models are developed using
data measured in certain
periods at specific locations,
which lacks generalizability
 Measurements with large
disturbances are hard to obtain
40

Comparison of Measurement and Component
based approaches: Component based based
Advantages
 Field measurement is not required
 Physical representation of end-use
devices
 Can be applied to different
operation conditions
 Demand side measurement is
considered
Disadvantages
 Requires characteristics of
individual load components
 Accurate and comprehensive
load composition information is
hard to obtain
 Low adaptability to the
integration of new loads
41

Conclusion
 Although load modelling has been extensively studied, more research is
imperative to update existing load models and understand characteristics
of modern loads with emerging technologies.
 For load model structure development, more sophisticated models that
balance flexibility and complexity are needed.
 Conventional load modelling methods using measurement data in a
certain period may not be able to capture time-varying load behaviours,
and lack generalizability.
 More research is needed to develop advanced algorithms to perform
online load modelling using the real-time data.
 The increasing penetration of DGs and the implementation of demand-side
management poses additional challenges to load modelling.
42

Conclusion
 Future research on parameter estimation algorithms should be able to
process data from existing and emerging measurement devices with
different resolutions, such as smart meters, PMUs, and SCADA.
 The algorithms should be robust to bad data, missing measurements,
changes in the voltage regulation scheme, and noises.
43

References
 Arif, Anmar, Zhaoyu Wang, Jianhui Wang, Barry Mather, Hugo
Bashualdo, and Dongbo Zhao. "Load modeling—A
review." IEEE Transactions on Smart Grid 9, no. 6 (2017): 5986-
5999.
 Collin, Adam John. "Advanced load modelling for power
system studies." (2013).
44

Questions
 What are the two “Load Model Parameter Identification”
methods? Differentiate between these methods.
 What are the various types of load models?
45

Load modelling in distributed generation planning

More Related Content

What's hot (20)

Similar to Load modelling in distributed generation planning (20)

More from Siksha 'O' Anusandhan (Deemed to be University ) (20)

Recently uploaded (20)

Load modelling in distributed generation planning