Intelligent Control and Protection of Power Systems in the Russian
Cities
Nikolai Voropai, Victor Kurbatsky, Nikita Tomin, Dmitry Efimov and Irina Kolosok
Electric Power System Department, Melentiev Energy Systems Institute, Irkutsk, Russia
Keywords: Power Grid, Smart Cities, Control, Protection, Artificial Intelligence, Russia.
Abstract: A distinctive feature of the energy system development in Russian megalopolises is the need for a
comprehensive approach to the problem of making the network intelligent. The paper presents the following
contributions: (1) intelligent operation and smart emergency protection in Russia including requirements for
new protection systems; (2) a description of smart grid territorial clusters in the interconnected power
systems of Russia. (3) state estimation (SE) techniques as informational support of the intelligent power
grid control including SE with phasor measurements use, dynamic SE, and cyber-physical security issues of
SE; (4) a hybrid Volt/VAr control approach based on AI techniques such as machine learning and multi-
agent systems based models.
1 INTRODUCTION
Cities today are home to more than 50 percent of the
world’s population and by 2050 it is estimated that
2.9 billion people will be living in cities. These cities
and megapolosies will need new and intelligent
infrastructure to meet the needs of their citizens and
businesses (ABB, 2013).
Cities might experience significant concentrations
of electric vehicles and renewables in certain city
districts. Left unmanaged, new loads can
dramatically increase load on the system at certain
times of the day and cause circuit breakers or fuses to
trip with resulting outages. The traditional response
would be to resize substations or strengthen
distribution lines and equipment. Grid automation
can be used to defer some of these upgrades.
While city grids are generally strong enough to
integrate renewables without significant capacity or
voltage challenges, additional power system
protection is required to cope with bi-directional fault
currents. With distributed renewables, at certain
times renewable generation could exceed
consumption, resulting in power flowing from the
customer into the grid. New protection schemes are
required to cope with these situations safely and
isolate only those parts of the grids experiencing
problems.
Large cities and megalopolises of Russia
(Moscow, St.-Petersburg, Yekaterinburg, etc.)
represent the most dynamically developing territories
of Russia with a growing electricity demand after the
economic crisis of 1998. These territories require
special attention when planning the development of
energy infrastructure and have a number of
characteristic features. In the nearest future the
development of the networks in megalopolises and
large industrial centers in Russia will result in the
formation of systems with a complex multi-loop
structure (Voropai, 2016).
A distinctive feature of the energy system
development in Russian megalopolises is the need
for a comprehensive approach to the problem of
making the network intelligent. It is reasonable to
consider the entire process of production,
transmission, distribution and consumption as a
single whole rather than divide the processes
according to the balance inventory as this has been
done lately. Accordingly, a staged introduction of
automated control based on the intelligent principles
will improve the reliability of the entire power
system operation. Furthermore, it will increase the
quality of electricity supply to consumers.
An effective way to support these city goals is by
using technology to more intelligently monitor,
optimize and control key systems and infrastructure.
In other words, to operate as a ‘smart city’.
The evolution of the traditional electrical system
in the direction of the intelligent grid implies a
growing automation of the power grid management
Voropai, N., Kurbatsky, V., Tomin, N., Efimov, D. and Kolosok, I.
Intelligent Control and Protection of Power Systems in the Russian Cities.
DOI: 10.5220/0007656700190029
In Proceedings of the 8th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS 2019), pages 19-29
ISBN: 978-989-758-373-5
Copyright
c
2019 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
19
in order to increase operating efficiency, increase
reliability, expand the use of network assets, reduce
emergencies, etc. To achieve these objectives in
2014, President of the Russian Federation Putin V.V.
was announced one of the nine road maps of the
National Technology Initiative “EnergyNet”, which
is a high-level long-term program for the
development of technologies, standards and
communities in the field of building the electric
power industry for the new technological structure.
Integration of power systems, liberalization, and
modernization of the electric power industry
increase the changeability and unpredictability of
electric power systems operation and generate the
need to improve and develop principles as well as
systems of operation and emergency control.
Artificial intelligence application is an advanced
way to carry out smart emergency control in power
systems (Efimov, 2011; Voropai, 2011; Voropai,
2018).
2 SMART GRID CLUSTERS IN
RUSSIA
The process of intelligent power system formation
under the EnergyNet platform suggests pilot projects
implementation and territorial smart grid clusters
creation. These clusters are supposed to use
information, technological, and control systems
providing adaptive control of network parameters,
remote control of switching devices, and real-time
estimation of the network technical state under
normal, pre-emergency, and post-emergency
conditions (Efimov, 2011; Efimov, 2012).
Currently, along with implementation of the pilot
projects and the creation of smart grid clusters, new
equipment is being installed at the energy facilities
of the unified national electric grid of Russia as a
part of a modernization program. In Russia the
projects on formation of individual smart grid
clusters and implementation of pilot projects aimed
at creating the smart grid were launched in 2011
(Efimov, 2011; Voropai, 2011; Budargin, 2010).
Below we briefly indicate several most important
pilot projects, which are implementing in two (of the
seven) interconnected power systems belonging to
the unified energy system of Russia.
The following pilot projects of smart grid
clusters will be implemented in the north-western
region of Russia during the period through 2020
(Fortov, 2012; Efimov, 2011; Efimov, 2012;
Budargin, 2011): Karelskaya power system, power
systems of Komi Republic and Arkhangelsk city,
and “Big Ring” and “Small Ring” of electric
networks in St. Petersburg (Table 1).
Table 1: Pilot projects on creation of territorial clusters of
the smart grid in the Northwest interconnected power
system.
Smart grid cluster
Project goal
1. KOLA
(Karelskaya power
system)
To provide reliable power supply
and power quality under
conditions of parallel operation of
330 kV transit overhead
transmission lines
2. KOMI
(power system of
Komi Republic
and Arkhangelsk
power system)
To provide high reliability level of
power supply at the required
power quality
3. BIG RING
(St. Petersburg)
To ensure required reliability of
power supply to urban consumers
4. SMALL RING
(St. Petersburg)
To decrease current loading and
redundancy of existing lines
Table 2: Pilot projects on creation of territorial clusters of
the smart grid in the East interconnected power system.
Smart grid
cluster
Project goal(s)
1. ELGAUGOL
To provide power quality and
redundancy of tunneling and
traction power supply (a two-circuit
220 kV transit transmission line)
To ensure emergency and operation
control, considering development of
small-scale generation
2. NIZHNY
KURANAKH-
MAIYA
(Yakutian
power system)
To provide a high level of power
supply reliability and power quality
3. VANINO
To increase the reliability of the
power supply to traction substations of
the electrified railway in the
Khabarovsk Territory
4. PRIMORYE
TERRITORY
To supply electricity to the southern
part of the Primorye Territory
To increase transfer capabilities of
500 kV transit transmission lines by
350–400 MW
5. RUSSKY &
POPOV
ISLANDS
with
distributed
generation
To integrate wind generation and
mini-CPP into the grid
To provide power quality and
redundancy
To ensure emergency and operation
control, considering small-scale
generation expansion and
involvement of storage devices
SMARTGREENS 2019 - 8th International Conference on Smart Cities and Green ICT Systems
20
The pilot projects are intended to furnish
Northwest interconnected power system with
innovation technologies of smart grids that will
effectively solve the regional problems (limitations
on power output from power plants, insufficient
reliability level of power supply to consumers, etc.).
Table 2 presents pilot projects in the territory
covered by the East interconnected power system of
Russia (Amur region, Sakha Republic, Primorye,
and Khabarovsk Territories). The projects address to
electric power networks of megalopolises and
include several smart grid clusters.
Implementation of smart grid in the clusters
"ElgaUgol", "Vanino", and "Primorye Territory"
suggests construction of compact digital substations
furnished with innovative devices (new systems for
reactive power compensation and voltage
maintenance, active filters, equipment monitoring
and diagnosis systems, etc.)
The territorial smart grid cluster was formed at
Russky Island (Efimov, 2011; Voropai, 2011; R&D
Report, 2009). It involves the creation of a smart
automated control system, which aims to provide
centralized monitoring, dispatching, and process
control as well as solving the problems of creating
the smart grid of the region and optimal control of
electricity, heat, and gas supply facilities and
consumers. Alongside the network infrastructure, the
integrated smart grid includes gas-fired power plants
(mini-CHPPs), wind farms, boiler plants, electricity
and heat storage systems, “smart houses,” and a park
of electric vehicles with charging stations.
3 INTELLIGENT EMERGENCY
CONTROL AND
REQUIREMENTS TO LOAD
SHEDDING ALGORITHMS
Emergency control philosophy in the unified energy
system of Russia is a hierarchical approach and is
realized by the coordinated operation of many
control devices, which maintain power system
stability and interrupt the expansion of an
emergency situation in the case of stability violation
and a threat of undesirable cascade emergency
development. Coordinated emergency control is
realized by joint participation of generators,
networks, and consumers. Such approach can be also
realized for emergency control of large cities and
megalopolises based on smart grid technologies and
ideology of EnergyNet platform.
The components of emergency control system
are shown in Table 3.
Table 3: Emergency operation control services.
Conditions
Services
Aims
Pre-
emergency
Emergency
dispatching
Transfer
capability
control
To maintain transfer capability
margins of transmission lines
To increase transfer capabilities
of ties by voltage control
Emergency
Emergency
control
To provide EPS stability by
increasing voltages and
damping the oscillations
To provide EPS stability by
automatic stability control
To interrupt emergency
development by automatic
devices
Post-
emergency
Restoration
To provide fast EPS restoration
by observing the margins and
excluding restoration disruption
Table 4: Functions and types of emergency control
automations.
Functions of the
emergency
control
automations
Centralized hierarchical automation
Prevention of
stability violation
(this is the
function of
centralized
automatic stability
control system)
Standalone (local) automations
Elimination of
out-of-step
conditions
Limitation of
frequency
decrease or
increase
Limitation of
voltage decrease
or increase
Prevention of
inadmissible
overloads
Intelligent Control and Protection of Power Systems in the Russian Cities
21
According to Russian standards (National
Standard, 2012; National Standard, 2013), the
emergency control systems and devices are intended
for the detection of emergency conditions in energy
system, prevention of their development, and
elimination. The most important task of the
emergency control systems is prevention of system-
wide blackouts accompanied by an interruption in
electricity supply to consumers in a large territory.
Hence the main functions of emergency control and
corresponding types of the emergency control
automations may be classified as in Table 4.
Most of the difficulties in power system
operation are caused by electrical grid overloads.
Load shedding in the receiving subsystem is the
most efficient action among the remedies against
overload. Basing on those considerations the
principles and algorithms of distributed adaptive
load shedding against stability loss of the ties of
main network and overcurrents of controlled
subsystem lines are to be developed.
A new load shedding scheme was proposed in
(Voropai, 2018) as a combination of disconnections
either large or small consumers – it means
disaggregation and distribution of load shedding
automation (Fig.1).
From technical point of view, the easiest way of
emergency load shedding is a piecemeal
disconnection of large industrial consumers, which
are connected with the power system in high
voltages, see Fig.1(b). It is the usual present
practice. At the same time, the economic damages
caused by such kind of disconnections can be
considerable (for example, the loss of profit due to
technological cycle interruption). The compensation
of the potential damages (for example, with
preferential prices of electricity for consumers,
which agree to emergency limitation of electricity
supply) could be very unprofitable for power
system.
Coming from these considerations an alternate
load shedding scheme can be proposed as a
combination of disconnections either large or small
consumers – it means disaggregation and
distribution of load shedding automation, see
Fig.1(c). It is more difficult technically than
disconnection of large consumers, but present state
of automation and data transmission systems makes
it feasible. The advantages of small loads emergency
disconnections are much smaller total damage, and
ability of more accurate dosing the disconnections
(if the large consumers are disconnected then the
surplus load disconnection is practically imminent).
OVERLOAD, MW
Substation
Substation
Small consumers
Large
consumer
HV /
MV
MV
/ LV
(a)
Small consumers
Substation
Large
consumer
Normal flow, MW
Substation
HV /
MV
MV
/ LV
(b)
Small consumers
Normal flow, MW
Substation
Substation
Large
consumer
HV /
MV
MV
/ LV
(c)
Figure 1: Possible control actions against overload of the
line. Emergency situation – overload of the main line (a);
Conventional control action – a large load partial shedding
(b); Alternate control action – distributed load shedding
(c).
SMARTGREENS 2019 - 8th International Conference on Smart Cities and Green ICT Systems
22
Load shedding scheme is addressed, depending
on appeared emergency situation, to provision of
stability of the ties of main network or thermal
stability, and stability of loads of controlled
subsystem. Coming from the tasks to be fulfilled, the
load shedding algorithms should be subdivided as
follows (Voropai, 2018):
1. Algorithms addressed to providing the stability
of the ties of main network. These algorithms are
to operate with the close interaction with the
centralized automatic stability control system.
The automation operation under those conditions
consists in the procedure of choice and
consequent disconnection of required amount of
the load. If the necessary amount of the load to
be shed could not be collected from small
consumers then load of large consumers be
partially shed. To minimize the shedding, the
more precise balancing is needed by means,
among others, further splitting the steps of large
consumers disconnection.
2. Algorithms preventing the overload with the
current of controlled subsystem lines. These
algorithms are to provide an optimal amount of
load shedding, i.e. to minimize the cost of
unloading the overloaded transmission line. The
automation is to operate as an intellectual
emergency control system providing the fast
disconnection of the consumer’s load in a
minimal necessary amount to prevent the
equipment overload.
Table 5: Summary of requirements to load shedding
algorithms.
Requirements
Comments
1. Keeping the
balance
between
complexity and
simplicity of
implemen-
tation
On the one hand, the algorithms
are to be complex enough to
provide an acceptable level of
control action precision.
On the other hand, they are to be
simple enough to provide an
acceptable speed of calculations.
2. Providing a
high level of
fault tolerance
The algorithms do not fail in the case
of uncritical loss of information.
Should the loss of information be
critical, the algorithms either
implement the excessive control
actions, or delegate control to power
system operator.
3. High speed
diagnostic self-
testing
The possibility of implementing an
effective self-testing procedure is
related directly to the requirement of
maximal simplification of the
algorithms.
Coming from above analysis the requirements to
load shedding algorithms can be summarized as in
Table 5 (Voropai, 2018).
4 INFORMATIONAL SUPPORT
OF INTELLIGENT CONTROL
PROBLEMS
In the EnergyNet strategy, great significance is
attached to the subsystem responsible for solving the
problems dealing with control of the current
operating conditions of intelligent power system.
This subsystem includes the technical means for the
acquisition, transfer and processing of data on the
state of the network components and state variables
(SCADA and WAMS), as well as the software for
the calculation of current operating conditions (state
estimation) of the EPS, forecasting and monitoring
of operating parameters on the basis of the obtained
information.
Simultaneously with the development of
information technologies and orientation to the total
digitalization of information exchange, the threats of
cyber-intrusion into the management systems of the
EPS are increasing. Cyberattack is deliberate
physical damage of measuring sensors and
transmission channels, malicious intrusion into local
networks of the power enterprise for the purpose of
entering of obviously false information, partial or
complete blocking of a traffic, distortion in work of
system of synchronization of time and so on.
The systems of data collection and processing
– SCADA and WAMS– belong to the subsystems of
the Smart Grid, which are most vulnerable to the
physical failures and information attacks dangerous
in terms of their consequences (Rihan, 2013; Ten,
2007). The PMU measurements, as well as SCADA
measurements, need to be validated (Glazunova,
2011). Especially the requirement for their
validation is important when cyber-attacks (CAs)
occur at the power industry facilities. Along with the
development of technical measures to detect cyber
interference in the control system of the power
facility, we offer the development of the state
estimation algorithms.
State estimation is a statistical method of data
processing which is used to filter the measurement
errors and to calculate unmeasured variables.
Detection of bad data and suppression of their
influence on the state variable estimates are one of
the most pressing issues when solving the state
estimation problem. The results of state estimation,
i.e. the steady state variables, represent the basis for
Intelligent Control and Protection of Power Systems in the Russian Cities
23
solving the problems of operating control of
intelligent power system (EMS-application),
including the problem of security calculation and
analysis. The use of PMU measurements to solve the
control problems requires measurement validation,
which can be carried out on the basis of the state
estimation methods. The method of test equations
was developed to detect bad data in SCADA
measurements, then adapted to check PMU
measurements and analyze cyber security of
SCADA and WAMS (Gamm, 2002; Kolosok, 2014).
Test equations are steady-state equations which
contain only measured variables
y
. When the
values of measurements are written in such equation
a discrepancy appeared due to the presence of
random noise in those measurements. The TE
method based on comparing the value of the
discrepancy magnitude with some statistical
threshold allows one to judge on the presence of
gross errors in measurements. The TE method does
exclude gross errors and replace erroneous
measurements with pseudo measurements. The
constant presence of gross error in a Test Equation
during a long time period forces the operator to
analyze the reasons for the corruption of initial data
with subsequent generation of recommendations to
compensate for the systematic errors in
measurements or fix the fact of cyber-attack.
The TE technique is applied not only to bad data
detection but to the SE problem too. To this end a
set of measurements is divided into basic
measurements and redundant ones. (The basic
measurements are a minimum set of measurements
that provide observability of electric power system.)
First the estimates of basic measurements are
calculated, after that – the estimates of redundant
measurements and unmeasured variables are found
out. The main advantages of the TE method are the
opportunity to reduce the dimensionality of the
problem and to use the obtained test equations for a
priori detection of bad data in measurements.
Unforunatelly, if cyber-attacks are nontrivial or a
great amount of software and technical means
appear to be attacked simultaneously, the state
estimation method becomes ineffective.
In Energy Systems Institute SB RAS (ESI SB
RAS) (Glazunova, 2011), we suggested a technique
for a two-level distributed state estimation based on
singling out the areas in the scheme of electric
power systems, which are monitored using PMU.
The PMU measurements coming at a high frequency
make it possible to implement fast linear algorithms
of state estimation for such areas. Along with the
traditional algorithm of the linear SE solved through
the state vector, the proprietary algorithm of the
linear SE by the TE method is developed in ESI SB
RAS (Kolosok, 2014). Increase the redundancy of
measurements and their accuracy greatly affects the
results of data verification accuracy of estimates.
According to the above considerations we
suggest the following technique to identify
cyberattacks on SCADA and WAMS:
1. Local areas which are totally observable by
PMU measurements are singled out in the
scheme of an electric power system;
2. For such local areas the local state estimation
is performed using linear algorithms with
unconditional a priori verification of the
initial data;
3. A great number of bad data in measurements
or the measurements that go beyond the
technological limits, measurements
diagnosed as doubtful and unchecked
(belonging to one facility), should mean a
complete failure of PDC operation and
initiate finding the reasons for such a failure;
4. If there are formal signs of a WAMS failure,
an independent simultaneous ( by one
timestamp) bad data detection and state
estimation by SCADA measurements are
performed at this energy facility;
5. A sharp discrepancy between the results of
the independent procedures for the WAMS
and SCADA measurement verification are
used to make a conclusion if there is a
malicious attack on one or another system.
The efficiency of the suggested approach was
tested using a fragment of the scheme equipped with
SCADA and PMU measurements (Kolosok, 2018).
The calculation network was divided into areas that
have either a SCADA-server or a WAMS phasor data
concentrator (PDC) at the center of each of them
(Fig.2). The node 5 where both SCADA-server and
PDC were installed becomes to be boundary one.
Linear SE
Non-Linear SE
2
1
6
4
6
1
2
4
5
WAMS
SCADA
Figure 2: Dividing the area of node 5 on SCADA and
WAMS.
SMARTGREENS 2019 - 8th International Conference on Smart Cities and Green ICT Systems
24
Cybersecurity of the EPS SE
Measurements Equivalent Network (EN)
Algorithms
Desynchroniz
ation
Critical
meas&groups
Lost of
observability
Bad Data
Faults of
ENP
Splitting EN
Corruption of
database
Fault due to the
observability
lost
Fault due to
great amount of
baddata
Fault of the
Linear SE
CA 4 CA 1
Fault
of
PMU
Fault
of
PDC
Channel
fault
PMU
Functio
nal fault
CA 2 CA 1
Vector
bad
data
CA 1
desynch
ronizati
on
Bad
Data
Lack of
measure-
ments
Weath
er
TS
bad
data
CA 5
c3
c22
c1c1c5c1c10c1c21c6c8c3c6c1c4
c7
c3
c7 c9
CA 3
Critical
meas&
groups
Fault of
ENP
CA 1
CA 2
CA 3
CA 1
CA 2
CA 3
CA 4
CA 5
c21
c3
c4
c5
c1
c22
c21c22 c21
c22
c6
c7
c21
c22
c3
c5
Figure 3: A tree of the power system state estimation software faults.
The results of bad data detection using two
independent systems (SCADA and WAMS) and
combination of a-priori and a-posteriori bad data
detection make it possible to conclude whether or
not there is a malicious cyberattack The efficiency
of the suggested approach was tested using a
fragment of the scheme equipped with SCADA and
PMU measurements (Kolosok, 2018). The
calculation network is divided into areas that have
either a SCADA server or a WAMS concentrator at
the center of each of them (Fig.2). The node 5 where
PDC is installed becomes to be boundary one.
A problem of resistance of the state estimation
procedure to cyberattacks on the system for gathering
and processing of PMU measurements was
investigated, and the SE procedure itself using an
analytical approach to the assessment of SE software
operability based on the fault tree technology. The
Fault Tree of the SE cyber security consists of three
main components: Measurements, Equivalent
Network and Algorithms. The failure of any of these
components may lead to a failure of SE procedure
itself. Every component has elements, for example,
{desynchronization of measurements; critical
measurement and critical groups; loss of
observability; bad data} in the first component,
which are vulnerable to different potential
cyberattacks (
i
CA
). The fault tree technology helps
one make visual solution for constructing
countermeasures
i
CA
c
.
At present the SE procedure is getting
increasingly more important in the conditions of
adjustment of the automated systems for acquisition
of phasor measurements as well as in the conditions
of various cyber threats. The SE mathematical tools
allow us to straighten out a tangle of true and
erroneous measurements and make certain
conclusions on the operability of the devices for
collection and primary processing of measurements.
An algorithm for a two-level SE on the basis of
SCADA and WAMS measurements is effective in
terms of detection of malicious attacks on energy
system.
5 VOLTAGE/VAR CONTROL AND
OPTIMIZATION USING AI
APPROACH
In the nearest future the development of the
networks in megalopolises and large industrial
centers in Russia will result in the formation of
systems with a complex multi-loop structure. In
these conditions we should expect large-scale
system emergencies which will occur according to
the scenario in which the electrical current and
voltage constraints become decisive in case of
emergency operating conditions unfolding. The first
system blackout of such kind happened in the UPS
of Russia in Moscow power system in May 2005.
5.1 Multi-agent Approach
Traditional voltage and VAr control (VVC) and
optimization (VVO) techniques has a number of
downsides: low robustness to erroneous inputs;
computational complexity, erroneous identification
of states, etc (Tomin, 2018). In addition, the use of
classical optimization techniques, such as linear
programming or decoupled Newton-based optimal
Intelligent Control and Protection of Power Systems in the Russian Cities
25
power flow and mixed integer programming, usually
provides limited the useful results as show in
(AlRashidi, 2010). A relatively new approach to
design the problem is the application of AI
techniques, such as machine learning, multi-agent
systems, fuzzy logic control, etc.
The approaches implemented on the basis of
MASs as a whole have a single methodology of
decentralized or partial decentralized (hybrid)
control of electric power system (Sidorov, 2018).
The principle of distributed intelligence is
implemented at permanent or periodic exchange of
messages among the agents to implement specified
protocols. Any of the elements of a power grid
(generators, loads, lines etc) could be modified to be
an agent to provide specified protocols - tap
changers block, load shedding, increasing of reactive
power production etc. The examples of the
completed developments in this area can be
exemplified by a multi-agent control system
(MACS) of voltage and reactive power control
developed by Center of systems studies and
development at JSC “STC FGC UES” Russia
(Arkhipov, 2014) and ESI SB RAS, Russia
(Sidorov, 2018; Kurbatsky, 2016).
In ESI SB RAS, we developed In ESI SB RAS,
we developed MACS, which is provided a
decentralized automatic Volt/VAr control associated
with determination of the time of critical overload
and switching to the load shedding procedure, rather
than ensuring the best efficiency of secondary
control. The agents are integrated by means of a
common information environment in which they can
exchange messages. The knowledge of an agent
about subsystem is formed as a basis of sensitivity
coefficients (elements of Jacobi matrix of steady-
state equations).
A decentralized MACS was applied to make the
generator agents (GAs) of a power system interact
effectively to prevent voltage collapse. The GA
receives the following local information: stator and
rotor currents, bus voltages, the local stop-off
signals, and the numbers of tap changers of the
generator transformers. If the value of the stator or
rotor current exceeds the maximum permissible
value (approaching the limit value), the GA tries to
exclude the possibility of switching off the generator
due to overload. The condition for generator
overload is the increase in the current generation of
reactive power, Q
g
above the maximum value, Q
max
:
If the excitation current of the GA goes beyond
its normal range, the GA tries to decrease it to
exclude the possibility of generator tripping. The
overloaded GA determines the rate of reactive power
increase according to sensitivity coefficients. It is
important to note that we do not need to know the
exact value of the coefficients for the current
conditions. If any of the requested GA stops
increasing Q
g
it informs the overloaded GA. If all
the requested GA stop increasing Q
g
but the
overload is not eliminated, the overloaded GA starts
the load shedding procedure.
5.2 Machine Learning Approach
Significant support in the development of system
operator adviser intelligent systems involved the
approaches based on the machine learning
algorithms, namely artificial neural networks
(ANNs) (Methods, 2010; Diao, 2009), decision trees
(Kurbatsky, 2016; Zhukov, 2018), which have high
approximating abilities. This made it possible to
effectively train such models to successfully solve a
whole complex of real-time problems within the
framework of applications of automated
management systems. The majority of solutions in
this direction are connected with the transformation
of the classical optimization problem into the
regression /classification task, which allows to
significantly reduce the computing time while
maintaining acceptable accuracy. The most
successful developments were obtained in the
application of various decision trees algorithms for
VVC/VVO.
In ESI SB RAS, we developed an online
VVC/VVO technique based on the model of online
decision trees (Proximity Driven Streaming Random
Forest (PDSRF) (Zhukov, 2018) and deep learning
models (DeepCS). The combination of original
properties of machine learning and capabilities of
voltage stability L-index indicator as a target vector
makes it possible to reformulate typical VVO
problem as a machine learning problem. Thus, we
presented the classic optimization problem as a
multi-output regression problem, which aims to
simultaneously predict multiple real valued
output/target variables. As a result, the obtained
values of injections ∆Q from PDSRF or DeepCS
were used for online VAR compensation by using
reactive power sources, which decreased L
sum
.
In (Zhukov, 2018) was clearly showed that the
power system will work in the optimal operating
condition when the sum of local L-indices, L
sum
, is
minimal. This AI update enables us to apply the
classical methodology in real time.
SMARTGREENS 2019 - 8th International Conference on Smart Cities and Green ICT Systems
26
5.3 Case Study
The efficiency of the proposed intelligent system
were tested on IEEE 6 (Fig. 4) and IEEE 118 - Bus
test systems. The proposed decentralised MACS are
implemented in MATLAB/PSAT. PDSRF models is
implemented in C++. DeepCS is implemented in
Python using TensorFlow library. A set of the
obtained system states was used to offline calculate
the values of global L-index, and on the basis of
L
sum
, the reactive power injections were found for
each load node.
At each step of the load increase in the IEEE
systems, we simulated the following three cases of
VVC:
1. Case 1 – local devices: overexcitation limiter
(OEL) in AVR, there are no VAR
compensators,
2. Case 2 – all generators have MACS agents
for preventing generator overload, there are
no VAR compensators
3. Case 3 – MACS + DeepCS or PDRSF (there
are VAR compensators).
Figure 4 shows the location of all AI-elements in
the IEEE 6 system: generators with AVR that are
agents and loads with 25 Mvar VAR compensators
that are controlled by machine learning models
(PDRSF or DeepCS).
Gen 2
Gen 1
Gen 3
Load 3
Load 2
Load 1
RPS 1
RPS 3
RPS 2
Machine learning
model (PDRSF,
DeepCS)
Agent 2
AVR
AVR
Agent 1
Agent 3
AVR
Figure 4: IEEE 6-Bus system with AI-grid elements.
Figure 5 demonstrates the simulation results. The
use of MACS in Case 2 allows maintaining voltage
stability at the critical step of overload and
preventing voltage collapse. At load factor N=19,
MACS starts the load shedding procedure, because
all the requested agents stopped increasing Qg but
the overload was not eliminated. The joint use of
MACS and machine learning algorithms (PDRSG or
DeepCS) in Case 3 allows us not only to optimize
the voltage/VAR profiles before critical overload
(secondary control stage), but also improve voltage
profiles in emergency control stage of the IEEE 6-
Bus system. Moreover, in this case, the load
shedding procedure was started later (N=22) than in
Case 2.
0 10 20 30 40 50 60 70
0
0.5
1
1.5
2
2.5
3
Load Factor, N
The sum of local indices, L
sum
Case 1 (local AVR)
Cace 2 (MACS)
Case 3 (MACS+machine learning)
Load shedding
(Case 2, N=19)
Load shedding
(Case 3, N=22)
Voltage collapse
Figure 5: The curves L
sum
for IEEE 6-Bus system.
As shown in Table 6, the DeepCS gives better
results than PDRSF.
Table 6: Comparative test results for different multi-output
regression models (IEEE-6 test scheme).
Models
MAE x 10
-3
, p.u.
RPS1
RPS2
RPS3
PDRSF
0.385
0.456
1.126
DeepCS
0.087
0.034
0.540
0 5 10 15 20 25 30 35 40 45 50
2
3
4
5
6
7
8
9
10
11
Load Factor, N
The sum of local indices, L
sum
Case 1 (local AVR)
Case 2 (MACS)
Case 3 (MACS+machine learning)
Load shedding
(Case 3, N=21)
Load shedding
(Case 2, N=15)
Voltage collapse
Figure 6: The curves L
sum
for IEEE 118-Bus system.
The Case 1 modeling leads the IEEE 118 system
to voltage collapse (Fig.6). Using MACS in the Case
2 allows to maintain voltage stability at the critical
step of overloading and prevent the voltage collapse.
Again, joint using MACS and machine learning
algorithms in the Case 3 allow to optimize
voltage/VAR profiles not only before critical
overloading secondary and emergency control
stages, as well as to postpone the start of load
Intelligent Control and Protection of Power Systems in the Russian Cities
27
shedding procedure for IEEE 118-Bus system. Thus,
the use of the proposed hybrid AI technique (Case 3)
provides intelligent secondary and emergency
control making the large test system a more stability
under disturbances.
The operability of the MACS in comparison with
traditional local automatic was also tested using the
IEEE 118-Bus (Fig. 7). The tests demonstrated a
lower probability of cascading disconnection in the
Case 2 where the MACS is used since the reactive
power redistribution makes the loading of
generators more balanced compared to the situation
where there is no MACS at all (the Case 1). This
happens because MACS is trying to redistribute the
reactive power from overloaded to underloaded
generators. To this regard, when using MACS the
number of steps towards stability limit Nmax will
always be bigger (N
max
=255), compared with the
situations without any automation (N
max
=250) and
with only local overloading control (N
max
=221).
Figure 7: Results of quasi-dynamic modeling in the IEEE
118 test scheme. Comparison of equipment overload under
different approaches.
6 CONCLUSION
Intelligent automation of electric power grid in large
cities is a key contributor and a prerequisite to
building the smart grids of the future. ESI SB RAS
has been driving the development of advanced
protection, supervision, control and management
techniques and systems for the complete power
delivery process.
The smart solutions are built on AI products for
protection and control, monitoring, measurement,
and communication. The creation of intelligent
power system under EnergyNet platform should
provide a qualitatively new level of efficiency of
electric power industry development and
functioning, raise system security, and increases the
quality and reliability of the electricity supply to
consumers in large cities.
ACKNOWLEDGEMENTS
This work was supported by the Russian Scientific
Foundation (No.19-49-04108) and German Research
Foundation (No. RE 2930/24-1) under the joint
project "Development of Innovative Technologies
and Tools for Flexibility Assessment and
Enhancement of Future Power Systems".
REFERENCES
ABB Power and Automation: Solid Foundations for Smart
Cities Available from: https://new.abb.com/docs/
default-source/smart-grids-library/abb_smart_grids_
white_paper_2013.pdf
Voropai, N., Kurbatsky, V., Tomin, N., Panasetsky, D.,
2016. Improving power system monitoring and control
in Russian modern megalopolises," 2016 18th
Mediterranean Electrotechnical Conference
(MELECON), Lemesos.
Efimov, D.N., 2011. Some developments, prospective
ways and projects of smart grid technologies in
Russia—an overview, Proc. of IEEE PES ISGT
Europe 2011, Dec. 5–7, Manchester, UK
Voropai, N.I., Efimov, D.N., Etingov, P.V., Panasetsky,
D.A. 2011. Smart emergency control in electic power
system, 18
th
IFAC World Congress, Milano, Italy,
Aug. 28–Sept. 2 (2011), Pp. 1658-1664
Voropai, N.I. et al., 2011. Emergency control in electric
power systems based on smart grid concept,
APPEEC’2011 Int. Conf., Wuhan, China, March 25–
28.
Voropai N., Kurbatsky V. Tomin N. et al., 2018.
Intelligent control and protection in the Russian
electric power system / in: L. Lamont, A. Sayigh (eds)
Application of Smart Grid Technologies. Case Studies
in Saving Electricity in Different Parts of the World,
Pp. 61-140
Efimov, D.N., 2012. Smart grid in Russia: today's and
tomorrow's practices, Proc. of the 7th Conf. on
Sustainable Development of Energy, Water and
Environment Systems, Ohrid, Macedonia.
Budargin, O., 2010. Modernization via innovation
development—creation of smart grid, Presentation for
Commission on Modernization and Technological
Development of Russian Economy, 2010 October, 26
(2010) (in Russian). Available from: http://www.fsk-
ees.ru/media/File/press_centre/speeches/Presentation_
Budargin_26.10.10.pdf
R&D Report, 2009. JSC “DALENERGOSETPROJECT”
Main Directions for Development of Primorye Region
Energy till 2012 and for Perspective to 2020 (2009),
Vladivostok (in Russian)
Rihan, M., Ahmad, M., and Beg, M.S., 2013.
Vulnerability Analysis of Wide Area Measurement
System in the Smart Grid. Smart Grid and Renewable
SMARTGREENS 2019 - 8th International Conference on Smart Cities and Green ICT Systems
28
Energy. Pp. 1-7. Available: http://www.scirp.org/
journal/sigre.
Ten, C., Liu C., Govindarasu, M., 2007. Vulnerability
Assessment of Cybersecurity for SCADA Systems, in
Proc. 24-28 June 2007 Power Engineering Society
General Meeting.
Gamm, A., Kolosok, I., 2002. Test Equations and Their
Use for State Estimation of Electrical Power System,
Power and Electrical Engineering: in Scientific Proc.
of Riga Tech. University. Riga: RTU, Pp. 99-105.
Kolosok, E. Korkina, E. Buchinsky, 2014. The Test
Equation Method for Linear State Estimation Based on
PMU Data, unpublished, presented at the 18th PSCC.
Wroclaw, Poland.
Glazunova, A., Kolosok, I., Korkina E., 2011. Monitoring
of EPS Operation by the State Estimation Methods, in
“Innovative Smart Grid Technologies“ (ISGT Europe-
2011), Great Britain, Manchester, 5-7 December 2011.
Kolosok I.N., Korkina E.S., 2018. Decomposition of
Power System State Estimation Problem as a Method
to Tackle Cyberattacks / The1st IEEE International
Conference on Industrial Cyber-Physical Systems
(ICPS-2018), Saint Petersburg, Russia, May 15-18,
2018, SF-004928
Tomin N., Kurbatsky V., Zhukov A., Panasetsky D.,
Sidorov D., 2018. Voltage/VAR Control and
Optimization: AI Approach, Proc. of 10th IFAC
Symposium on Control of Power and Energy Systems,
(CPES 2018), Meiji University, Tokyo.
AlRashidi, R., AlHajri, M.F., Al-Othman, A.K., El-
Naggar, K.M., 2010. Particle Swarm Optimization and
Its Applications in Power Systems, Pp. 295-324.
Springer Berlin Heidelberg, Berlin, Heidelberg.
Voropai, N.I., Tomin, N.V., Sidorov, D.N. et al., 2018. A
Suite of Intelligent Tools for Early Detection and
Prevention of Blackouts in Power Interconnections //
Autom Remote Control. Vol.79, Pp. 1741–1755
Arkhipov I.L., et al., 2014 Multiagent control system for
voltage and reactive power, in Proc. Of the 22nd
conference "Relay protection and automatics of power
systems", Moscow. Pp. 243-252
A set of intelligent tools for prevention of large-scale
emergencies in electric power systems / Voropai N.I.,
Kurvatsky V.G. et al. - Novosibirsk: Publishing
House “Nauka”, 2015 - 332 p. (in Russian)
Methods and models for power system reliability studies,
Syktyvkar: Komi Scientific Center of Ural Branch of
RAS, 2010, 292 p. (in Russian)
Diao R. at el., 2009Decision tree-based online voltage
security assessment using PMU measurements, IEEE
Trans. on Power Systems. Vol. 24, no. 2, Pp. 832-839.
Zhukov A. On-Line Power Systems Security Assessment
Using Data Stream Random Forest Algorithm
Modification / Zhukov A., Tomin N., Sidorov D., et al
In: Zelinka I., Vasant P., Duy V., Dao T. (eds)
Innovative Computing, Optimization and Its
Applications. Studies in Computational Intelligence,
vol 741. - 2018. Springer, Cham
Budargin, O., 2011. Smart Grid—First Outcomes,
Presentation for Round Table “Smart Grid—Projects
of Future”, 2011 June, 16–18, 2011 (in Russian).
Available from: http://www.fsk-ees.ru/upload/docs/
01_ayladwlg.pdf
Fortov, V.E., and Makarov, A.A., 2012. A concept of
Russia’s intelligent power system with active-adaptive
network. Moscow JSC “NTC FSC EES», 2012, 235 p.
(in Russian)
National Standard P 55105-2012, 2012. United power
system and isolated working systems. Operative-
dispatch management. Automatic emergency control
of modes of power systems. Emergency control of
power systems. Norms and requirements. (in Russian)
National Standard P 55438-2013, 2013. United power
system and isolated power systems. Operative-
dispatch management. Relay protection and
automation. Interaction of actors, consumers of
electrical energy in creating (modernization) and the
exploitation. General requirements. (in Russian)
Kolosok I., Korkina E., and Mahnitko A., "A fault tree
cyber security analysis of power system state
estimation software", in Proc. 2017 of the 9th
International Scientific Symposium on Electrical
Power Engineering, ELEKTROENERGETIKA, Pp.
349-353.
Intelligent Control and Protection of Power Systems in the Russian Cities
29