System Protection Schemes:

Limitations, Risks, and Management

Final Project Report

Power Systems Engineering Research Center

Empowering Minds to Engineer
the Future Electric Energy System

System Protection Schemes:

Limitations, Risks, and Management
Final Project Report

Project Team
James McCalley, Project Leader
Olatujoye Oluwaseyi, Venkat Krishnan, Renchang Dai
Iowa State University
Chanan Singh, Kai Jiang
Texas A&M University

PSERC Publication 10-19

December 2010

Information about this Project

For information about this project contact:
James D. McCalley
Professor
Iowa State University
Department of Electrical and Computer Engineering
Ames, Iowa 50011
Phone: 515-294-4844
Fax: 515-294-4263
Email: [email protected]

Power Systems Engineering Research Center

The Power Systems Engineering Research Center (PSERC) is a multi-university Center
conducting research on challenges facing the electric power industry and educating the
next generation of power engineers. More information about PSERC can be found at the
Centers website: http://www.pserc.org.

For additional information, contact:

Power Systems Engineering Research Center
Arizona State University
577 Engineering Research Center
Tempe, AZ 85287-8606
Phone: 480-965-1643
Fax: 480-965-0745

Notice Concerning Copyright Material

PSERC members are given permission to copy without fee all or part of this publication
for internal use if appropriate attribution is given to this document as the source material.
This report is available for downloading from the PSERC website.
2010 Iowa State University. All rights reserved.

Acknowledgements
This is the final report for the Power Systems Engineering Research Center (PSERC)
research project titled Special Protection Schemes: Limitations, Risks, and
Management (PSERC project S-35). We express our appreciation for the support
provided by PSERCs industrial members and by the National Science Foundations
Industry/University Cooperative Research Center program.
We very gratefully acknowledge the inputs received from the industry experts,
including Ali Chowdhury (CAISO), Jim Gronquist (BPA), Wenyuan Li (BC Hydro),
Tony Johnson, and Dede Subakti (MISO), for their insightful comments, which were
very helpful in shaping this work.

Executive Summary
Special protection schemes (SPS), known also as remedial action schemes, are
designed to detect abnormal system conditions and take predetermined, corrective action
(other than the isolation of faulted elements) to preserve system integrity and provide
acceptable system performance. Today, in many parts of the world, SPS represents a
viable planning alternative to extending transmission system capability. Although SPS
deployment usually represents a less costly alternative than building new infrastructure, it
carries with it unique operational elements among which are: (1) risks of failure on
demand and of inadvertent activation; (2) risk of interacting with other SPS in unintended
ways; (3) increased management, maintenance, coordination requirements, and analysis
complexity. Additionally, there is a dearth of simulation and assessment tools for
performing reliability studies of SPS and enable planners to evaluate the operational
complexity that SPS brings into the system along with its various economic and
operational advantages.
So the objective of the proposed work is to provide a structured framework for
identifying limitations of SPS deployment within a system, and assessing SPS risks and
develop proper standards and practices for maintaining SPS reliability over its lifetime.
One of the highlights of the report is the unique assessment framework proposed
based on both a process view and a systems view to identify risks and associated
consequences for SPS. The process view based framework will view SPS in terms of a
process which considers all the building blocks starting with the actuating signals, the
equipment and logic used to operate on those signals, communication equipments and so
on till the final action. The system view framework will view SPS in terms of its position
in and impact on the power system in which it exists. This framework proposed is one of
the significant contributions of this report, as even though a particular SPS design may
appear quite reliable from the process view, it may not be so from system view due to the
prospect of failures from interactions among many SPS in the system. So the system view
risk assessment framework addresses such critical issues, which otherwise could cause
cascading and catastrophic system consequences.
Some of the contributions from this project are:
1. A document with concentrated information pertaining to:

Background: Summary of relevant literatures on SPS, their operational and

maintenance strategies from IEEE transaction papers, CIGRE documents,
NERC standards, NERC sub-regions, and individual RTOs, ISOs, and
companies.

Technology or knowledge transfer: An overview of standards and methods

from industries such as process control, nuclear and power system, in order to
leverage interesting ideas from these mature industries that could be
applicable to reliability and maintenance studies related to SPS. Safety
instrument systems (SIS) of process control industry lend themselves very
well in this regard.

ii

2. SPS failure mode identification: The project provides SPS failure mode taxonomy
from both process and system point of view, and proposes approaches to identify
and evaluate such SPS failures.
3. Software design for system view risk assessment: Design of a simulation
capability to test various SPS logics, with flexibility to vary the SPS logic and
intelligence to vary operating conditions and events over a wide range.
Several reliability models and architectures for SPS and PMU-aided SPS have
been developed to facilitate system view reliability studies, which would enable
capturing impacts of SPS on system level phenomena.
4. Operational complexity metric: The operational and maintenance complexity due
to the proliferation of SPS in power system is quantitatively captured by a
proposed metric. System planning studies must incorporate such operational
complexity metric in their overall formulation to estimate the limit of SPS growth
for economical and reliable system operation.
The report identifies the importance of including such metric in power system
planning framework, and therefore illustrates a transmission expansion planning
study for SPS aided power systems.
In further investigations, the conceptual designs developed in this report will be
applied on real time utility systems.

iii

Table of Contents
1
2

Introduction................................................................................................................. 9
1.1 Types of SPS and typical components ............................................................... 10
1.2 Overview of report ............................................................................................. 14
Standards, practices and advancements on SPS ....................................................... 16
2.1 Current industry standards ................................................................................. 16
2.1.1 NERC standards and its relationship to other international standards ... 16
2.2 Existing industry practices ................................................................................. 19
2.2.1 Design of SPS ........................................................................................ 19
2.2.2 Documentation for reliability assessments ............................................ 21
2.2.3 SPS implementation and coordination in industries .............................. 23
2.3 Advancements in SPS ........................................................................................ 29
2.3.1 Relationship between SPS and synchrophasor technology ................... 29
2.3.2 Role of SPS in wide area monitoring, protection and control ............... 30
Relationship to other industries & applications ........................................................ 33
3.1 Process control industry ..................................................................................... 33
3.1.1 ISA S84.01 ............................................................................................. 33
3.1.2 ISA dTR84.02 ........................................................................................ 34
3.1.3 IEC work ................................................................................................ 34
3.1.4 Methods in Process control industry applicable to SPS......................... 35
3.2 Nuclear industry ................................................................................................. 39
3.2.1 IEC standards ......................................................................................... 40
3.2.2 IAEA standards ...................................................................................... 44
3.2.3 IEEE standards ....................................................................................... 46
3.2.4 Challenges to the U.S. nuclear standards ............................................... 48
3.2.5 Hints from nuclear industry ................................................................... 48
3.3 System operations for electric power ................................................................. 49
3.3.1 Operational security rules from decision trees....................................... 50
3.3.2 SPS logic design using decision trees .................................................... 51
3.3.3 Illustration .............................................................................................. 53
Risk assessment, process view.................................................................................. 55
4.1 Process view failure modes ................................................................................ 55
4.2 Process view risk assessment ............................................................................. 58
4.2.1 Failure mode and effect analysis............................................................ 58
4.2.2 Fault tree analysis .................................................................................. 59
4.2.3 Network modeling ................................................................................. 59
4.2.4 Markov modeling ................................................................................... 59
4.3 Formulation of SPS risk expression ................................................................... 60
4.3.1 Risk expression with GRS ..................................................................... 60
4.3.2 SPS reliability evaluation....................................................................... 62
4.3.3 State probabilities................................................................................... 64
4.4 Risk assessment of generation rejection scheme ............................................... 66
4.4.1 Illustration .............................................................................................. 67
4.5 Consideration for future SPS developments ...................................................... 71
4.5.1 Possible SPS architecture ....................................................................... 72
iv

7
8

4.5.2 Illustration of network reduction method .............................................. 73

4.5.3 Illustration of Markov modeling method to calculate MTTFF .............. 77
Risk assessment, systems view ................................................................................. 82
5.1 Systems view SPS failure- a case study ............................................................. 82
5.1.1 Background of SPS in Nordic grid ........................................................ 82
5.1.2 Event report ............................................................................................ 83
5.2 Systems view failure modes and their identification process ............................ 85
5.2.1 SPS logic design practices ..................................................................... 85
5.2.2 SPS logic validation and failure modes ................................................. 87
5.2.3 Simulation based system view SPS failure mode identification ............ 87
5.3 Systems view risk assessment ............................................................................ 91
5.4 Reliability analysis of power systems with impact of SPS operation ................ 92
5.4.1 Modeling SPS operation at the component level ................................... 92
5.4.2 Modeling SPS operation at the system level.......................................... 96
Consideration of SPS in planning ........................................................................... 102
6.1 Electric system long-term planning with SPS ................................................. 102
6.1.1 Electric system long-term planning ..................................................... 102
6.1.2 SPS planning related to wind ............................................................... 103
6.1.3 Accounting for SPS in planning .......................................................... 103
6.2 SPS interaction ................................................................................................. 103
6.2.1 Complexity for high penetration SPS .................................................. 104
6.2.2 Illustration ............................................................................................ 105
6.3 Incorporating SPS in generation and transmission expansion planning .......... 108
6.3.1 Case study ............................................................................................ 108
6.3.2 Transmission expansion options and specifications ............................ 110
6.3.3 Transmission expansion tree ................................................................ 110
6.3.4 Operational complexity ........................................................................ 115
6.3.5 Production costs ................................................................................... 116
6.3.6 Economic factors ................................................................................. 116
6.3.7 Study assumptions ............................................................................... 117
6.3.8 Optimization model ............................................................................. 120
6.3.9 Optimization methods .......................................................................... 120
6.3.10 Analytical Hierarchical Process ........................................................... 122
6.4 Optimization and AHP results.......................................................................... 124
6.4.1 Penalty factor method .......................................................................... 124
6.4.2 Non-linear penalty factor method ........................................................ 124
6.4.3 Ranking non-dominated solutions using AHP ..................................... 125
Summary and conclusions ...................................................................................... 127
7.1 Summary .......................................................................................................... 127
7.2 Conclusions ...................................................................................................... 128
References ............................................................................................................... 129

List of Figures
Figure 1: Buffalo ride area: MW requested vs. transmission capacity ............................. 10
Figure 2: General structure of a system protection scheme .............................................. 11
Figure 3: Process view and system view failure modes ................................................... 14
Figure 4: IEC 61508 life-cycle model .............................................................................. 18
Figure 5: CAISO RAS maintenance program .................................................................. 24
Figure 6: TSA display showing real-time generation shedding arming pattern ............... 24
Figure 7: BCTC central arming RAS system ................................................................... 25
Figure 8: BCTC - EMS/SCADA/protective relay integration .......................................... 25
Figure 9: California and BPA remedial action schemes controllers ................................. 26
Figure 10: Motivation for C-RAS ..................................................................................... 27
Figure 11: RAS range of functionalities ........................................................................... 28
Figure 12: Synchrophasor based vector processor from SEL ........................................... 29
Figure 13: Smart RAS from SCE...................................................................................... 30
Figure 14: Typical PMU ................................................................................................... 30
Figure 15: Wide-area monitoring and protection design .................................................. 31
Figure 16: Wide area control system ................................................................................ 32
Figure 17: Wide area control system using RAS .............................................................. 32
Figure 18: Economic analysis flowchart........................................................................... 37
Figure 19: Economic analysis of SPS aided transmission upgrade .................................. 38
Figure 20: Monte Carlo simulation based power system planning................................... 50
Figure 21: Efficient database generation approach ........................................................... 51
Figure 22: SPS logic design .............................................................................................. 52
Figure 23: Decision tree: shed load or not? ...................................................................... 53
Figure 24: Logic circuit for a GRS ................................................................................... 66
Figure 25: Markov model for failure mode 1 ................................................................... 68
Figure 26: Markov model for failure mode 2 ................................................................... 69
Figure 27: A conceptual all-digital SPS architecture ........................................................ 72
Figure 28: SPS reliability block diagram .......................................................................... 73
Figure 29: Reduction of SPS reliability block diagram .................................................... 75
Figure 30: Simplified SPS reliability block diagram ........................................................ 76
Figure 31: System logical chain of the SPS ...................................................................... 78
Figure 32: Algorithm to obtain off-diagonal elements of the matrix R ............................ 80
Figure 33: Extraction of submatrix R11 from matrix R ..................................................... 81
Figure 34: Nordic SPS case study ..................................................................................... 83
Figure 35: Confusion matrix- SPS logic testing ............................................................... 87
Figure 36: Simulation study- regular SPS failure mode identification ............................. 88
Figure 37: Process view SPS reliability assessment operating conditions sampling..... 90
Figure 38: System view SPS risk assessment ................................................................... 91
Figure 39: Modeling component i without SPS operation................................................ 92
Figure 40: Modeling component i with desirable SPS operation ..................................... 93
Figure 41: Modeling component i with desirable and undesirable SPS operation ........... 93
Figure 42: Modeling component i with all SPS operation effects .................................... 94
Figure 43: Modeling component j with SPS operation effects ......................................... 96
Figure 44: Modeling component k not involved in SPS operation ................................... 96
vi

Figure 45: Decoupling interactions caused by SPS failing to operate .............................. 97

Figure 46: Impact of SPS operation on modeling non-contingency system states ........... 98
Figure 47: Impact of SPS operation on modeling contingency system states .................. 99
Figure 48: Reliability modeling of power systems without SPS operation .................... 100
Figure 49: Reliability modeling of power systems with impact of SPS operation ......... 101
Figure 50: Complexity of SPS ........................................................................................ 105
Figure 51: Small system for illustrating concepts........................................................... 106
Figure 52: Small system with transmission upgrade ...................................................... 107
Figure 53: 5-bus system .................................................................................................. 109
Figure 54: Transmission tree expansion- numerical optimization .................................. 111
Figure 55: Transmission tree ......................................................................................... 112
Figure 56: Simple structure of analytical hierarchical process ....................................... 122
Figure 57: Structure of AHP for power system transmission expansion planning ......... 123

vii

List of Tables
Table 1: SPS survey studies ................................................................................................ 9
Table 2: IEC and ISA demand mode SIL in terms of availability .................................... 17
Table 3: FERC approved penalty matrix .......................................................................... 19
Table 4: Traditional vs. sychrophasor based RAS scheme ............................................... 29
Table 5: A sample risk matrix ........................................................................................... 35
Table 6: A prototype illustrating the relationship between SPS and risk matrix .............. 36
Table 7: Cost analysis ....................................................................................................... 39
Table 8: Ranking options .................................................................................................. 39
Table 9: Component failure and repair rate ...................................................................... 67
Table 10: Probability of each state.................................................................................... 71
Table 11: Probability of each state of simplified model ................................................... 71
Table 12: The initial state matrix ...................................................................................... 78
Table 13: The rearranged state matrix .............................................................................. 79
Table 14: Nordic SPS event description ........................................................................... 84
Table 15: Branch data for sample system ...................................................................... 106
Table 16: SPS designs for different wind plant growth stages ....................................... 107
Table 17: Time periods ................................................................................................... 108
Table 18: Forecasted load ............................................................................................... 109
Table 19: Branch data ..................................................................................................... 109
Table 20: Generator data ................................................................................................. 110
Table 21: Generation expansion plans ............................................................................ 110
Table 22: Possible solutions............................................................................................ 112
Table 23: Probabilistic load profiles .............................................................................. 116
Table 24: Different periods and associated value of load lost ........................................ 118
Table 25: Comparison of economic factors for planning options................................... 119
Table 26: Solution for the penalty factor method ........................................................... 124
Table 27: Solution for the non-linear penalty factor method .......................................... 124
Table 28: Non-dominated solutions ................................................................................ 125
Table 29: Comparison of economic factors and complexity .......................................... 125
Table 30: Judgment matrix for the criteria ..................................................................... 125
Table 31: Judgement matrix for operational complexity ................................................ 126
Table 32: Judgement matrix for reliability ..................................................................... 126
Table 33: Judgement matrix for total costs ..................................................................... 126
Table 34: Final weight and rank for non-dominated solutions ....................................... 126

viii

Introduction

System protection schemes (SPS) (also called remedial action schemes, RAS) are
designed to detect abnormal system conditions, typically contingency-related, and
initiate pre-planned, corrective action to mitigate the consequence of the abnormal
condition and provide acceptable system performance [1]. SPS actions include, among
others, changes in load, generation, or system configuration to maintain system
stability, acceptable voltages or power flows. SPS is also used as the acronym for
special protection scheme, with has the same meaning as system protection scheme.
However, it was recommended in [1] that word special be replaced by the word
system, since it can be argued that all protection is special in some fashion. IEEE uses
the System Integrity Protection System (SIPS), RAS is used by (BPA, WECC) others
use the term SPS [2].
Today, in many parts of the world, SPS represents a viable planning alternative to
extending transmission system capability. Although SPS deployment usually
represents a less costly alternative than building new infrastructure, it carries with it
unique operational elements among which are: (1) risks of failure on demand and of
inadvertent activation; (2) risk of interacting with other SPS in unintended ways; (3)
increased management, maintenance, coordination requirements, and analysis
complexity. The objectives of the work described in this report are to summarize the
state of the art in regards to SPS including closely related technologies in other
industries, provide a structured framework for assessing SPS risk, and examine SPS as
a viable planning alternative in which we consider how to identify limits of SPS
deployment within a system.
These objectives are motivated by the recognition that SPS has proliferated. For
example, SCE has 17 RAS on its transmission corridor and has planned to add another
57 [3]. Table 1 presents the results from three survey studies performed over the last
20 years, which indicates significant growth in the use of SPS.
Table 1: SPS survey studies
1989 Survey [4]

1996 Survey [5]

2009 Survey [6]

Respondents Schemes Respondents Schemes Respondents Schemes

18

93

49

111

110

958

According to Vinnakota et. al. [7], Due to the increased complexity of network
operation in the past 30 years due to several factors such as growth in load, changes
in market conditions and increased imports/exports, the network is more stressed in its
operation and RAS schemes have grown in numbers.
Furthermore, with the heavy growth in renewable resources, especially wind farms
that are being connected to the grid prior to the required upgrades, SPS or RAS has
become an increasingly critical application that enables quick and economic means of
interconnecting them while meeting the required RPS standards [8]. This is supported
9

by Figure 1 [9] where it is clear that the MW requested in Buffalo ridge area is way
ahead of the available transmission capacity, necessitating heavy SPS proliferation for
smooth and economical solution strategies. But, this growth in SPS has definitely
increased the operational complexity in managing the system with variable wind
resources, and poses an interesting challenge to the ISOs.

Figure 1: Buffalo ride area: MW requested vs. transmission capacity

So this work has been motivated by the above cited reasons, with a clear cut focus
on the topics such as current standards and industry practices related to SPS operation,
their limitations, and future advancements. Special focus has been provided on risk
assessment techniques, where the key concept on which our work is based is that SPSrisk may be examined from both a process view and a systems view. Viewing SPS in
terms of a process forces consideration of the reliability of each individual component
comprising the SPS. Viewing SPS in terms of the overall power system forces
consideration of the SPS and the system in which it operates.
In this chapter, we summarize types of SPS and the components comprising them,
we expand upon the process-view vs. systems view of SPS, and we will provide an
overview of the report.
1.1

Types of SPS and typical components

SPS generally refers to controllers having some or all of the following

characteristics [10]:

SPS can be armed or disarmed depending on the system conditions.

SPS are normally dormant systems; initiating events usually occur less than
once a year.

SPS usually employ discrete, feed-forward control laws.

The control action taken is predetermined in most cases.

10

Typically, some form of communication is involved in the control action.

Figure 2 shows the general structure of an SPS.

Disturbance

Power System

Electric variables
Direct detection

System Protection Scheme

ACTION

DECISION
PROCESS

INPUT

Figure 2: General structure of a system protection scheme

According to their control variables SPS can be classified as response-based or
event-based. Response-based SPS are based on measured electric variables (such as
voltage, frequency, etc.) and initiate their protective actions when the contingency has
caused the measured value to hit the trigger level. The two most common responsebased types of SPS are under-frequency load shedding and under-voltage load
shedding.
Event-based SPS are designed to operate upon the recognition of a particular
combination of events (such as the loss of several lines in a substation). Examples of
event-based SPS are below:

Generation rejection

Load rejection

System separation

Turbine valve control

Load and generator rejection

Out-of-step relaying

Discrete excitation controls

Dynamic braking

Generator runback

VAR compensation

Combination of schemes

It is useful to give some examples of equipment that are not SPS:

11

Transmission devices designed to provide dynamic control of electric system

quantities, which typically involve feedback control mechanisms using power
electronics to achieve the desired electric system dynamic response, during
normal operation conditions, are not considered as SPS but instead as
transmission control devices. Examples of such equipment and devices, which
are active during normal operational conditions, include: static var
compensator, power system stabilizer, active or reactive power flow control or
compensation.

Transmission line auto-reclosing is aimed to improve stability or supply

capability of the system, but the auto-reclosing is regarded as being a part of
the line protection. Hence transmission line protection with auto-reclosing is
regarded as a more advanced form of unit protection for the line than line
protection without auto-reclosing. Restoration is not protection.

Overload protection of equipment is not an SPS.

In general, SPS are distinguished from unit or equipment protection in that a

protection system is called an SPS when the focus of the protection is on the power
system supply capability rather than on specific equipment.
The composition of SPS generally includes measurement inputs, sensors, breakers,
communication equipment, logic solvers, and logic solver software [11, 12, 13], as
described in what follows:

Measurement inputs: These may include one or more of the following [14]:
o Power system voltage and/or currents, synchronized to local
measurements in the same substation, or they may be wide-area
synchronized.
o Power system frequency
o Polarity reversal
o Control signals including automatic voltage regulator, power system
stabilizer, governors, HVDC converters, and reactive power
compensators (e.g., HVDC converters and SVC)
o Status circuit breaker positions, tap changer positions, whether
generator field current limiter is activated or not.
o Last valid state data such as, for example, telemetry data during loss of
communication channel
o Arming levels and thresholds

Sensors: These are devices that measure the power system condition.
Generally, they include relays and breaker/switch status detectors. Relays may
be current, voltage, power, frequency, rate of change of each of these, out-ofstep, generator power output level, line loading power level, etc. Neither loss of
current nor loss of power can be used alone to determine that a line is open,
because they both go through zero as power flow reverses direction on the line.

12

Caution must be taken in determining settings required to distinguish between

local faults or system problems by using rate of change of current, voltage,
power, or frequency. Out-of-step relays may be used in some cases for
detecting pending instabilities. However, these are usually applied only where
it is acceptable to wait until the swing is coming out of the swing setting of
the relay before taking corrective action. Studies must be performed to
determine the proper setting to prevent out-of-step tripping on recoverable
swings.

Breakers: These are the tripping equipment.

Communication equipment: This is the equipment that communicates the

condition of a power system with the help of sensors to the logic solver, and the
logic solver output finally to the actuating elements. A robust communication
network offers low error-rate, low latency, high availability, high security and
is also deterministic. The Southern California Edison (SCE) uses IEC 61850
GOOSE (Generic Object-Oriented Substation Event) communication protocol
[15] to perform all the RAS functionalities, beginning from detecting the event
at remote location, transmitting the alarm over WAN for about 460 miles, then
processing it and deciding suitable control action, and finally implementing the
control with the help of RAS IEDs at the remote location. The usage of such a
digital communication technology ensures the whole process takes place within
50 milliseconds, which is very fast compared to traditional methods of
transmitting control messages.

Logic solver: The logic solver is that portion of an SPS that performs one or
more logic functions used to execute the SPS application logic and initiate
protective actions. Although it may be electrical or electronic, it is assumed in
what follows that it is a programmable electronic (PE) system such as a
microprocessor [16], micro-controller, programmable logic controller (PLC), or
application-specific integrated circuit (ASIC). If the logic solver is purchased
externally, the supplier should provide an integrated design including input
module(s),
output
module(s),
maintenance
interface
device(s),
communication(s), and utility software. The logic solver should have a
published mean time to failure (MTTF), unsafe failure mode listing, and
frequency of unsafe failure mode. It should have a method (internal and/or
external) to protect against covert faults (such as a watchdog timer). The
logic solver should be designed to ensure that the process will not restart
automatically when power is restored, unless it is required to do so. Detected
failure of the logic solver should not result in an unsafe system condition, if the
appropriate, documented, response action is undertaken.

Logic solver software: In developing software necessary for the logic solver,
good software development practices should be followed. For example, a
software requirements specification and a software design document should be
developed. These documents should specify the functionality of the design
using functional blocks so that the programmer does not need to make any
assumptions about the functionality of each software module. Software

13

architecture should be clearly identified including specification of the operating

system, databases, input/output subsystems, communication subsystems,
programming and diagnostic tools, and programming languages used. Coding
standards should specify good programming practices (e.g., readability,
traceability, checkability, and analyzability), proscribe unsafe language
features, and specify procedures for source code documentation. Testing plans
should be able to show that each individual software module, each software
subsystem, and the entire software system performs their intended functions
and does not perform any unintended functions.

1.2

Power supplies

Monitoring devices
Overview of report

The remainder of the report is organized as follows.

Chapter 2 presents the current industry standards for SPS design, maintenance and
operation. It also presents the way various prominent industries operate SPSs, and
what kind of technology they employ to achieve coordination among various SPSs.
Finally it presents some of the latest advancements in the way SPS are being used,
with primary focus on centralized SPS and its relationship with Sychrophasors.
Chapter 3 sheds focus on the standards and practices of three industries, namely
process control, nuclear and electric power industries. The aim is to extract relevant
information about processes and methods from these matured industries which may be
applicable for reliability studies concerning SPSs.
Chapter 4 and chapter 5 present risk assessment of SPS from two different views,
namely process view and system view. The process view framework will view
SPS in terms of a process which considers actuating signals, equipments, logic used to
operate on actuating signals, communication channels, and final action. The system
view framework will view SPS in terms of its position in and impact on the power
system in which it exists. System related failure modes are typically intended (by
design) but undesirable, reflecting faulty logic. Figure 3 presents an overview of these
two frameworks, and the kind of failure modes within each.

Figure 3: Process view and system view failure modes

14

Chapter 6 presents two planning studies, namely wind generation expansion study
and transmission expansion study. It proposes ways to incorporate SPS in such studies
and evaluate the reliability indices of such SPS-aided planning options.
Chapter 7 presents the conclusions.

15

Standards, practices and advancements on SPS

This chapter summarizes current industry standards and practices regarding SPS.
Section 2.1 focuses on standards, while Section 2.2 focuses on existing industry
practices related to SPS design, documentation, and technologies used for arming and
coordinating various SPS operations. Section 2.3 sheds light on some recent
advancement in SPS.
2.1

Current industry standards

The North American Electric Reliability Corporation (NERC) maintains the

following standards related to SPS, posted at the NERC website [17] under Standards Protection and Control:

PRC-004-WECC-1 Protection System and Remedial Action Scheme

Misoperation (WECC): This is a Regional Reliability Standard to ensure all
transmission and generation Protection System and Remedial Action Scheme
(RAS) Misoperations on Transmission Paths and RAS defined in section 4 are
analyzed and/or mitigated.

PRC-012-0 Special Protection System Review Procedure: To ensure that all

Special Protection Systems (SPS) are properly designed, meet performance
requirements, and are coordinated with other protection systems. To ensure that
maintenance and testing programs are developed and misoperations are
analyzed and corrected.

PRC-013-0 Special Protection System Database: To ensure that all Special

Protection Systems (SPSs) are properly designed, meet performance
requirements, and are coordinated with other protection systems.

PRC-014-0 Special Protection System Assessment: To ensure that all Special

Protection Systems (SPS) are properly designed, meet performance
requirements, and are coordinated with other protection systems. To ensure that
maintenance and testing programs are developed and misoperations are
analyzed and corrected.

PRC-015-0 Special Protection System Data and Documentation, PRC-016-0

Special Protection System Misoperations, and PRC-017-0 Special Protection
System Maintenance and Testing: To ensure that all Special Protection Systems
(SPS) are properly designed, meet performance requirements, and are
coordinated with other protection systems. To ensure that maintenance and
testing programs are developed and misoperations are analyzed and corrected.

2.1.1

NERC standards and its relationship to other international standards

This section gives more in-depth summary of current industry standards for special
protection schemes and how the standards can be improved by learning from more
established international standards such as the International Society of Automation
(ISA) and the International Electro-technical Commission (IEC).

16

2.1.1.1 PRC-004-WECC-1
According to PRC-004-WECC-1, mis-operations can be classified into two [18]:
1. Security-based misoperation: Security-based misoperation is a
misoperation caused by the incorrect operation of a protection system or
RAS. Security is a component of reliability and is the measure of a devices
certainty not to operate falsely.
2. Dependability-based misoperation: Dependability-based misoperation is
the absence of a protection system or RAS operation when intended.
Dependability is a component of reliability and is the measure of a devices
certainty to operate when required.
The ISA and IEC define Safety Integrity Level (SIL). Safety Integrity Level
(SIL) is a relative level of risk-reduction provided by a safety function, or it specifies a
target level of risk reduction. SIL may also be enforced for special protection
schemes to enhance reliability and availability based on Probability of Failure on
Demand (PFD). Table 2 describes SIL from IECs and ISAs view-point [19].
Table 2: IEC and ISA demand mode SIL in terms of availability
IEC SIL

ISA SIL

Availability required

PFD

1/PFD

N/A

>99.99%

1E-005 to 1E-004

100,000 to 10,000

99.90 99.99%

1E-004 to 1E-003

10,000 to 1,000

99.00 - 99.90%

1E-003 to 1E-002

1,000 to 100

90.00 - 99.00%

1E-002 to 1E-001

100 to 10

2.1.1.2 PRC-012-0, PRC-013-0, PRC-014-0, and PRC-015-0

PRC-012-0, PRC-013-0, PRC-014-0, and PRC-015-0 summarize the review of
SPS design, co-ordination with other protection schemes, establishing maintenance
programs, and proper and efficient documentations.
The IEC uses the safety life cycle approach as the framework for structuring safety
instrumentation system (SIS) requirements [20]. SIS performs specific functions to
achieve or maintain a safe state of the process, when unacceptable or dangerous
process conditions are detected. They are composed of similar elements as SPS, which
includes sensors, logic solvers, actuators and support systems. Therefore a similar
process can be developed for SPS as shown in Figure 4.

17

Figure 4: IEC 61508 life-cycle model

2.1.1.3 Penalty Matrix
NERCs sanction guidelines provide a matrix comprising violation risk factors and
violation severity levels, the intersection of which establishes base penalty amounts.
Table 3 presents the penalty matrix (per day rates) approved by FERC [21].

18

Table 3: FERC approved penalty matrix

Violation Security level
Lower
Moderate
High
Severe
Violation
Range Limits
Range Limits
Range Limits
Range Limits
Risk
Factor
Low
High
Low
High
Low
High
Low
High
Lower $1,000 $3,000 $2,000 $7,500
$3,000 $15,000 $5,000
$25,000
Medium $2,000 $30,000 $4,000 $100,000 $6,000 $200,000 $10,000 $335,000
High
$4,000 $125,000 $4,000 $300,000 $12,000 $625,000 $20,000 $1,000,000
The levels of risk factors in Table 3 are defined as:

Lower - administrative in nature and, which if violated, is not expected to

affect the state of the bulk power system or the ability to monitor, control, or
restore the system.

Medium - a requirement that, if violated, could directly affect the electrical

state or the capability of the bulk power system or the ability to effectively
monitor, control, or restore the system. But it does not result in instability,
separation, or cascading failures.

High - a requirement that, if violated, could directly cause or contribute to bulk

power system instability, separation, cascading failures, or hinder restoration
efforts.

2.2

Existing industry practices

RAS are fast acting automatic control devices, that utilizes protective relays and
fast telecommunication networks, to ensure acceptable (reliable and safe) power
system performance following critical outages on a power grid. Bonneville Power
Administration (BPA) uses SPS to maintain stability, reduce line overloads, maximize
transfer capabilities, and provide voltage support [ 22 ]. ERCOT employs SPS to
maintain system security and reliability in accordance with ERCOT and NERC
Reliability Standards, while facilitating the market [23]. BC Hydro heavily employs
SPS and RAS to maintain system integrity. It also shares some of these schemes with
the interconnected neighbors [ 24 ]. Southern California Edison (SCE) mitigates
transmission overload problems arising due to contingencies using RAS [ 25 ].
According to Alberta Electric System Operator (AESO) [26], RAS is used primarily
on a temporary basis when new market participants are being added to the system in
advance of the transmission systems capability to manage such loads or supplies.
Likewise, many utilities are using SPS especially in the event of increasing penetration
of intermittent renewable resources, and plenty of advancements are also being made
in such defense system designs to the extent that utilities have started designing PMU
based RAS [27].
2.2.1

Design of SPS

Design of SPS is very important due to the fact that SPS help keep the integrity of
a power system during extreme disturbance. Therefore SPS must be designed to be
highly reliable. One of the most important devices in SPS is the communication system
19

which enables data exchange between monitoring and controlling devices. Arming is
another important functionality in SPS design, wherein arming is usually automatically
or manually enabled. Redundancy is a useful technique used in SPS to help improve
reliability of SPS. Typically, SPS/RASs are comprised of three parts: monitoring,
event detection and mitigation. The quality of SPS is measured in terms of speed and
accuracy of operation, and redundancy in design.
The SPS design process comprises of the following five steps, namely [28]:
1. System study
System studies identify limitations under various contingencies. The limitations
could be thermal, voltage, or angular instability related system limits. Some of the
important aspects of system studies include understanding the requirements and
purpose of the application; identifying limits such as overload conditions, under
voltage, under frequency and so on; studying SPS requirements based on NERC
reliability and regional standards; evaluating many alternate solutions and so on.
2. Solution development
The solution based on system studies must be analyzed and specific
recommendations have to be made as a next step. The recommendations are about
the stability limits, conditions when SPS is to be armed, the amount of load to be
shed when needed, bus voltage limits, various other limits and so on.
3. Design and implementation
At the stage of implementation, typically many practical questions are to be
answered such as:

What technology is needed to meet the functional requirements of the SPS

in an economic fashion?

What kind of communication equipment is needed to meet the SPS

standards?

What is the most reliable way of building redundancy such as voting

schemes for the logic solver?

Reference [28] gives a detailed list of such practical issues to be addressed.

4. Commissioning & periodic testing
A comprehensive testing method must be in place for successfully
implementing the solution. The errors detected must be rectified and the system
specifications must be met. The test plan must include:

Lab testing: validation in a controlled environment

Field testing: test against undesirable system conditions likely in reality

The routine of tests are to be performed on a periodic basis and gradually

understand the working of SPS better in various conditions.

20

5. Training & documentation

Failure of SPS due to human intervention, faulty logic design and incorrect
settings has been reported to be about 20%, 12% and 10% respectively in past
surveys [5]. So proper training of operating and maintenance staff is necessary to
ensure reliable operation of SPS and its continuous improvement. A solid
documentation about SPS, its functionalities, past experiences, case studies will
definitely help in efficiently training the upcoming staff.
2.2.2

Documentation for reliability assessments

Electric system planners are facing unprecedented challenges in managing the risk
of using SPS in maximizing the usage of existing and planned transmission grid. The
following documenting procedure has been developed by Remedial Action Scheme
Reliability Subcommittee of WECC in order to assess the reliability of an SPS [29].
1. RAS scheme purpose and overview- The following information is included:
a. Name of the RAS
b. Purpose
c. Desired in-service date
d. Ownership
e. Person responsible for the operation and maintenance
f. SPS functionality
g. Single line drawings showing all sites involved, bus arrangement and other
protection systems such as breakers
h. Impact on the WECC power grid
2. RAS design
a. Design philosophy
b. Design criteria, including failure of which element or combination of
elements causes RAS failure mode
c. RAS Logic
d. RAS Logic Hardware
e. Redundancy
f. Arming method i.e. manual, automatic, or via SCADA
g. Define all inputs, including protective relay inputs such as angle, power,
current, voltage, frequency, rate of change of frequency power, current and
voltage.
h. List devices used to monitor inputs such as circuit breakers
i. Coordination with protection and control systems

21

j. Telecommunications with the diagram of telecom path, communications

system performance, supporting reliability information such as equipment
age, history, maintenance, communication architecture, bandwidth of
communication system and so on.
k. Transfer Trip Equipment
l. Remedial actions initiated
3. Monitoring
a. RAS monitoring equipment and time resolution
b. Station alarms
c. SCADA monitoring
d. Sequence of Events Recorders
e. Facilities monitored such as self diagnostics, telecommunications, transfer
trip equipment, RAS actions and so on.
4. RAS operating procedures - Procedure for abnormal system conditions like,
a. Incorrect operation (failure to operate or false operation)
b. Unavailability of redundant RAS system
c. Unscheduled or unplanned or uncoordinated loss of RAS
d. Partial or total loss of input for arming decisions
5. Commissioning, maintenance and testing
a. SPS commissioning
b. Overall functional test procedure
c. Preventative maintenance
d. Maintenance and test intervals
e. Maintenance and testing procedures
6. Performance and operational history- Based on system requirements provide
assurances on performance and operating time of the SPS. The following
information is provided:
a. How long has the RAS been in operation?
b. Failure to operate?
c. Incorrect operation?
d. Slow operation
e. Unnecessary operation?

22

7. RAS catalog information- Summary of the salient features of the newly

proposed or modified SPS
8. Revision History- Information about initial release and approval of the SPS by
regulatory bodies
2.2.3

SPS implementation and coordination in industries

SPS is designed to operate in stressed conditions. So properly arming SPS when

needed is very important, as failure in such a case could be very catastrophic.
Traditionally, RASs are localized, operated by fast algorithms installed at substations
or control centers. But with the increasing population of SPS, there have been many
disadvantages experienced by localized RASs especially in terms of maintenance and
coordination. Therefore, Centralized-RAS, a more technologically driven defense
scheme, has been gaining acceptance since it has lot of advantages. This section briefs
some of the utilities practices regarding SPS implementation, arming and coordination
under various conditions.
MISO
Midwest independent system operator (MISO) has about 15 RAS. MISO sends the
status of RAS to the control room personnel through ICCP (Inter Control Center
Protocol) points, who take the decision from real-time information displayed [30]. The
RTCA (Real-time Contingency Analysis) tool developed by MISO includes most of
the SPS, except a few very complicated ones. The RTCA tool [31] is either scheduled
to run at uniform intervals (say, every 5 minutes), or triggered based on a contingency
event or by an operator.
CAISO
In CAISO, RAS schemes are written using Data Base Language (DBL), which is
prevalently used in EMS related functionalities. The contingency analysis (CA)
module calls the RAS DBL code to initiate the simulation of RAS control actions
during contingency analysis [32]. Also, CAISO has developed a RAS maintenance
program, as shown in Figure 5. The more RAS in an electric grid, maintenance
scheduling of RAS and other electric power equipments becomes more and more
difficult. Also the electric grid experiences changes all the time, and RAS have to be
reviewed whenever there is a change in this system. So a proper documentation must
be done in-order to keep track of all the changes in the system.

23

Figure 5: CAISO RAS maintenance program

BC Hydro
In BC Hydro, RAS arming is done centrally at the control center [33, 34]. This has
proved to increase system limits as well as reliability. The arming of RAS is
automatically performed by TSA (transient stability assessment function) as shown in
Figure 6, which is within EMS. The arming is done either immediately after any
network configuration change or at systematic intervals of 4 minutes. At regular
intervals the system limits are updated automatically. In Figure 7, the patching matrix
of each RAS acts as the interface between inputs (multiple contingencies) and the
outputs, where outputs are suitable control actions such as generator shedding, line
tripping, shunt switching etc. to improve transient, voltage or thermal stability.

Figure 6: TSA display showing real-time generation shedding arming pattern

24

Figure 8 shows the way BC Hydro integrates EMS/SCADA and protective relays
in order to effectively handle contingencies [35].

Figure 7: BC Hydro central arming RAS system

Figure 8: BC Hydro - EMS/SCADA/protective relay integration

RTU: Remote terminal unit; SE: State estimation; TSAPM: Transient stability problem

25

ERCOT
To avoid unnecessary SPS operation, the SPS owner may provide a real-time
status indication to the owner of any generation resource controlled by the SPS to
show when the flow on one or more of the SPS monitored facilities exceeds 90% of the
flow necessary to arm the SPS. The cost necessary to provide such status indication
shall be allocated as agreed by the SPS owner and the generator owner [36]
BPA
BPA (Bonneville Power Administration) uses programmable logic controllers for
its SPS design. Most RAS in BPA are on their 500KV lines. The way RAS is operated
in BPA is that when there is a line loss detection, a transfer trip signal is sent to control
centers, from where control signals are sent to power plants and substations [37].
These RAS schemes are designed to be highly redundant using two out of three voting
schemes and also information is sent to two control rooms to improve redundancy. In
order to avoid frequency problems, there is a limit on generation tripping of 2700MW
[38]. BPA also co-ordinates with northern and central California, so that necessary
remedial actions can be taken in-order to keep up system integrity in Pacific NW and
California. Figure 9 presents the RAS controllers for the coordinated operation of
California and BPA RAS [39].

Figure 9: California and BPA remedial action schemes controllers

26

SCE
Southern California Edison (SCE) faces rapid growth of RAS in its footprint. This
is due to aggressive renewable generation expansion and load growth [8].
Consequently, having many isolated RAS leads to coordination problems, maintenance
issues with engineers having to traveling long distances for each RAS, and so on.
Figure 10 from the work [40] done by P. Arons gives a systematic outline of all the
associated problems due to proliferating RAS. So SCE is trying to introduce
centralized remedial action scheme (C-RAS), motivated by the fact that having many
localized RAS increases operational complexity.

Figure 10: Motivation for C-RAS

Some of the advantages and features of C-RAS developed by SCE are:
1. It will enhance transmission line capacity ratings
2. The various RAS will be to easier to maintain during different seasons and
different system operating conditions.
3. It will increase wide-area voltage stability
4. It possesses an additional study tool that can be used for transmission and
interconnection planning.

27

Figure 11 shows the system problems that can be addressed effectively and
economically using RAS, wherein RAS implements respective remedial action to
alleviate a stressed system which otherwise faces unstable conditions. So with the
advent of a Centralized-RAS approach, all the wide-area functionalities such as
monitoring, protection, and control could be fit in a single framework.

Figure 11: RAS range of functionalities

28

2.3
2.3.1

Advancements in SPS
Relationship between SPS and synchrophasor technology

Synchrophasor technology provides system status at a much faster rate, compared

to SCADA and other traditional state estimators. The publication [41] by Schweitzer
Engineering Laboratories (SEL) proposes a sychrophasor based vector processor
(SVP) shown in Figure 12, which collects synchronous phasor measurements from the
system, processes them and detects any unstable conditions within local as well as
large areas.

Figure 12: Synchrophasor based vector processor from SEL

This functionality of SVP when combined with RAS, provides a high improved
defense mechanism to enhance power system security. Table 4 [41] presents the
advantages of SVP based RAS compared to traditional RAS in terms of increased
speed, reduced equipments, and so on.
Table 4: Traditional vs. sychrophasor based RAS scheme

29

Southern California Edison (SCE) proposed Smart RAS [42], a Centralized RAS
technology that uses real power measurements of tie-line between two areas from
PMU to actuate suitable remedial action that prevents system out-of-step condition.
Figure 13 shows the smart RAS controller with its input and outputs.
The work in [27] presents a PMU based SPS developed and operated by Taipower
systems against transient instabilities caused by EHV line contingencies.

Figure 13: Smart RAS from SCE

2.3.2

Role of SPS in wide area monitoring, protection and control

After the advent of Phasor Measurement Units (PMUs), in 1988 Bonneville Power
Administration (BPA) first used it in WECC. Then on inspired by its many advantages
to record information useful for crucial system analysis, which were not earlier
possible using SCADA or IED data, many utilities have started deploying PMUs. Now
in WECC, various companies such as BPA, SCE, WAPA and PG&E take part in data
exchange program and benefit each other in having better reliability status of critical
transmission corridors [43]. Figure 14 shows a typical data exchange using PMU and
PDCs, which provide the ability to monitor system security and limits over a widearea, and take necessary control actions over a wide-area of the power network.

Figure 14: Typical PMU

30

Wide-area monitoring system (WAMS) based on PMUs is real time monitoring

application of electric power grid performing various tasks such as monitoring phase
angle, line loading, voltage stability, power oscillations, frequency stability, and event
archiving. WAMS has gained a lot of importance in the electric power grid
applications since there has been incessant load growth, aggressive renewable
generation expansion and, also frequent blackout threat. WAMS are also used in
implementing Wide-area protection schemes (WAPS), which are used to stop widespread blackout. According to [39], WAPS (Wide-area protection systems) schemes
are designed to detect abnormal system conditions and take pre-planned, corrective
actions intended to minimize the risk of wide-area disruptions and to increase system
power transfer capability. Figure 15 presents a basic design of WAMS with
protection module [44].

Figure 15: Wide-area monitoring and protection design

Figure 16 shows a response-based Wide-Area Control System (WACS) developed
by BPA using PMUs. WACS controller acts a centralized master control that receives
system data from PMUs and PDCs, processes them and issues relevant control signals
to WAPS controller.
Figure 17 shows the WACS implementation using RAS, where control signal from
the centralized master controller is sent to centralized RAS controller that performs the
WAPS functionalities. This wide-area system monitoring and control using PMU and
RAS respectively proves to be very efficient and faster, and helps in averting any
unnecessary RAS actions by proper coordination.

31

Figure 16: Wide area control system

Figure 17: Wide area control system using RAS

32

Relationship to other industries & applications

It is likely that individual companies have documented procedures for performing

SPS design, installation, and start-up. However, significant effort fails to identify
documentation of these procedures in the literature or in publicly available
documentation, available on the internet and elsewhere, with just a few exceptions,
including [11, 12, 13, 45, 46]. Yet, most of these are quite general and tend more
towards criteria rather than procedures. Indeed, the IEEE/CIGRE survey
conducted by Anderson and Le Reverend [5] found that most often utility criteria for
SPS contained at most general requirements for equipment redundancy.
It is found, however, that the other industries have confronted quite similar
problems. One of them in particular is the process control industry, summarized in
Section 3.1, and the other is the nuclear power industry, summarized in Section 3.2. In
addition, we find that, within the electric power industry, the approach to developing
operating rules is very similar to the approach for developing SPS logic, as described
in Section 3.3.
3.1

Process control industry

This industry is comprised of companies in the petroleum, pharmaceutical, power,

chemical, pulp and paper, and textile, and supporting areas. Often, the failure
consequence of the various processes implemented can be very high, and so a great
deal of attention is paid to standardizing procedures for designing, installing, and
maintaining safety instrumented systems (SIS). These are systems that are comprised
of sensors, relays, breakers, communication equipment, and logic solvers that take the
process to a safe state when predetermined conditions are violated [47]. The SIS
equipment and function are quite similar to the equipment and function of SPS in
power systems. The process control industry has done an extensive amount of work
investigating, modeling, and standardizing of SIS equipment and this work is well
documented. The documents in use are: ISA S84.01, ISA dTR84.02, IEC draft 61508
and S84/IEC 61511.
The Instrument Society of America (ISA) is the international society for
measurement and control. Members are from many industries with a large number
from the petroleum, chemical, and supporting industries. The International Electro
technical Commission (IEC) is the world organization that prepares and publishes
international standards for all electrical, electronic and related technologies. The
membership consists of more than 50 participating countries, including all the world's
major trading nations and a growing number of industrializing countries.
3.1.1

ISA S84.01

ISA S84.01 [48] is a standard titled Application of Safety Instrumented Systems

for the Process Industries that was approved on February, 1996 after many years of
development. This standard addresses the application of Safety Instrumented Systems
(SIS) for the process industries. ISA S84.01 addresses integrity levels for electrical,
electronic, and programmable electronic systems. These systems include
electromechanical relays, solid state logic, programmable electronic systems, motor33

driven timers, solid state relays and timers, hard-wired logic, and combinations of the
above. A key concept of this document is the definition and use of Safety Integrity
Levels (SIL).
The document provides standard criteria for function and integrity specifications,
conceptual design, detailed design, installation, commissioning, and prestart-up tests,
operation and maintenance procedures, periodic functional testing, management of
changes, and decommissioning of such systems. In particular, detailed design
requirements are specified for logic solvers and application logic, sensors and motor
starters, operator, communication, and maintenance interfaces, power sources, and
design of periodic testing capability. It also provides appendices which give SIL
assessment methods, design considerations, references, and an illustrative example.
3.1.2

ISA dTR84.02

ISA-dTR84.02 [ 49 ] is a supporting document for ISA S84.01 that provides

evaluation approaches for safety-instrumented system reliability. The focus of this
document is on modeling and calculation, and SIL is the basic performance measure. It
describes modeling of system failure elements, failure mode and effect analysis,
modeling of electrical, electronic, and programmable electronic systems, safety system
selection, common cause failure models, Markov model development and
quantification, uncertainty analysis, statistical sensitivity analysis, fault simulation test
procedures, and reliability analysis software. The objectives of ISA-dTR84.02 are to:
1) provide guidance on SIL analysis
2) present methods to implement SISs so that they achieve a specified SIL
3) identify failure rates and failure modes of SIS
4) address diagnostic, diagnostic coverage, covert faults, test intervals, common
cause, systematic failures, and redundancy of SIS
5) provide tools for verification of SIL and pre-startup safety review of SIS
6) present methods for determining the functional test interval
3.1.3

IEC work

The International Electro technical Commission (IEC) has prepared two

documents, IEC 61508 [50] and IEC-61511 [51], that define requirements common to
all industries (not just for process control) for electrical, electronic, and programmable
electronic systems. It is IEC's intent that, ultimately, there would be additional
standards developed to reflect specific requirements for the various industry sectors,
including nuclear, pharmaceutical, and aeronautical. Although the power industry is
never mentioned explicitly, it seems reasonably to include it as another sector. It is
therefore likely that anything the power industry wanted to do in the way of
standardization or guidelines would be of interest to IEC.
There are three basic ideas on which these ISA and IEC materials depend. One is
that the potential harm or danger can be measured by risk, which is the combination
(usually the product of) the probability of occurrence of the harm and the severity of

34

the harm. A second basic idea is that the safety instrumented functions, which mitigate
or prevent the harm and are therefore much like SPS, can be characterized by their
safety integrity. This is the probability of a safety instrumented function satisfactorily
performing the required functions under all the stated conditions within a stated period
of time. In the cited standards, safety integrity is quantified by a safety integrity level
(SIL). The SIL is a discrete number, 1, 2, 3, or 4, which specifies the requirements of
the safety instrumented functions to be allocated to the safety instrumented systems.
SIL 4 has the highest level of safety integrity, and SIL 1 has the lowest level. Each SIL
has associated target failure measures, according to whether the mode of operation is
low demand operation where frequency of demand for operation is not more than once
per year or high demand operation where this frequency is greater than once per year.
For low demand operation, the average probability of failure to perform the design
function on demand should lie in the range: 10-4 to 10-5 (SIL 4), 10-3 to 10-4 (SIL 3),
10-2 to 10-3 (SIL 2), and 10-1 to 10-2 (SIL 1). For high demand operation, the
probability of a dangerous failure per hour should lie in the range: 10-8 to 10-9 (SIL 4),
10-7 to 10-8 (SIL 3), 10-6 to 10-7 (SIL 2), and 10-5 to 10-6 (SIL 1). The third basic idea
embedded in these documents is that risk and SIL are keys in showing how the
establishment and maintenance of safety-instrumented system integrity involves many
activities over the lifetime of the equipment. This idea is captured via use of the term
safety life cycle, the necessary activities involved in the implementation of safety
instrumented function(s) occurring during a period of time that starts at the concept
phase of a project and finishes when all of the safety instrumented functions are no
longer available for use.
3.1.4 Methods in Process control industry applicable to SPS
3.1.4.1 Risk Matrix
One of the most common techniques used among refining, chemical and
petrochemical companies is the risk matrix. The Risk Matrix comprises of risk levels
based on probability and impact on its two dimensions as shown in Table 5 [52].

Risk frequency consequence

(3.1)

Table 5: A sample risk matrix

Consequence
Likelihood

Insignificant

Minor

Moderate

Major

Severe

Almost certain

Moderate risk

High risk

High risk

Extreme risk

Extreme risk

Likely

Moderate risk

Moderate risk

High risk

High risk

Extreme risk

Possible

Low risk

Moderate risk

Moderate risk

High risk

Extreme risk

Unlikely

Low risk

Moderate risk

Moderate risk

High risk

High risk

Rare

Low risk

Low risk

Moderate risk

High risk

High risk

35

The various risk levels are:

Extreme risk - detailed action/plan required

High risk - needs senior management attention

Moderate risk - specify management responsibility

Low risk - manage by routine procedures

The above concept of Risk Matrix can be used to assess SPS related risks. SPS
are programmed to engage in forced curtailments when an unwanted event is detected.
For instance, the fact that SPS increases operational transfer capability does not mean
that we should ignore the fact that frequent curtailments of load cause customer
dissatisfaction. Therefore the frequency of forced curtailments and the amount of load
interrupted should be regulated in order to minimize risk. Similarly, there are other
consequences for a desirable operation of SPS such as generation rejection, penalty
due to reduced export, etc. Also the consequences under an undesirable operation of
SPS could be equipment damage, system instability, etc. So a risk matrix can be used
to evaluate the risk of each of these consequences under different modes of SPS
operation. Table 6 shows a proposed matrix which can be used to minimize risk of SPS
actions.
Table 6: A prototype illustrating the relationship between SPS and risk matrix
Consequence
levels when
SPS is armed
to trip
Frequency of
SPS actions
Possibility of
repeated
events
Possibility of
isolated
events
Possibility of
occurring
sometimes

Insignificant
value of
consequence

Moderate
value of
consequence

Minor value
of
consequence

Significant
value of
consequence

Major value of
consequence

Significant risk

Significant risk

High risk

High risk

High risk

Moderate risk

Significant risk

Significant risk

High risk

High risk

Low risk

Moderate risk

Significant risk

High risk

High risk

Not likely to
occur

Low risk

Low risk

Moderate risk

Significant risk

High risk

Rare
occurrence

Low risk

Low risk

Moderate risk

Significant risk

Significant risk

3.1.4.2 Economic analysis of a Safety Instrumented System

Economic analysis of SIS is very important because sometimes installation of some
SIS may not be a wise idea. Any company that invests in a project wants to know
whether the benefit of the projects outweighs the costs. Economic models have been

36

developed to assess the benefit-cost ratio of installing a SIS [53], as shown in Figure
18.

Figure 18: Economic analysis flowchart

The above methodology of economic analysis can be applied to study SPS aided

37

transmission upgrade. In most cases SPS is a cheaper alternative to transmission

upgrade but we should not forget that SPS causes increase in operational costs such as
re-dispatch costs, EENS (expected energy not served) costs etc. and maintenance costs.
These when accumulated over the years it becomes very large. On the other hand
transmission lines are invested over long period of years and all these unnecessary
costs can be avoided over the years. A solution combining both SPS and transmission
line may strike the right balance. Therefore, it is necessary to do an economic sanity
check as shown in Figure 19 to know which transmission expansion plan aided by SPS
may be the better option over the long term.

Start

List all possible transmission upgrade that

can be corrected with SPS

Compute PCTU , PCSPS, CTL, CFC, CRED for all possible

transmission upgrade that can be corrected with SPS

Compute proposed benefit-cost index for

all possible options

Rank all options according to proposed

benefit-cost index

Make decisions based on the proposed

benefit-cost index

Start

Figure 19: Economic analysis of SPS aided transmission upgrade

A benefit-cost index is proposed and it can help to rank transmission upgrade based
on economic benefit.

38

B/C

( PCSPS PCTU )
CTU (CSPS CFC CSI )

(3.2)

where,
B/C = Benefit-cost index
PCTU = Production costs after transmission upgrade
PCSPS = Production costs if SPS is the preferred alternative
CTU = Cost for transmission upgrade
CSPS = Cost for implementing SPS
CFC = Expected cost due to forced curtailments (SPS)
CRED = Expected re-dispatch cost after generation tripping (SPS)
The example shown in Tables 7 and 8 assumes that the study is conducted for a 5year interval and all monetary values are in present value.
Table 7: Cost analysis
Option 1
Option 2
Option 3
Cost
$5,000,000 $8,000,000 $3,000,000
CTL
$250,000
$280,000
$186,000
CSPS
(PCSPS- PCTU) $1,700,000 $1,200,000 $900,000
$100,000
$100,000
$110,000
CFC
$120,000
$180,000
$90,000
CSI
Table 8: Ranking options
Options Benefit-cost index Rank
0.375
1
1
0.161
3
2
0.332
2
3
3.2

Nuclear industry

Safety is one of the most important issues in the nuclear power industry. The
consequence of safety-related failures in this discipline is always considered vital to
both people and the environment. Therefore, it is understandable that there are so many
safety-related instrumentation and control (I&C) systems in a nuclear power plant. In
addition, many organizations worldwide have developed various standards which give
guidance to nuclear I&C systems. These standards play an important role in guiding,
shaping, or even regulating the development of safety-related I&C systems in the
nuclear power industry.

39

3.2.1

IEC standards

The International Electro technical Commission (IEC) has been developing its
work on functional safety since 1985. As a result, the seven parts of a general industry
systems
standard
IEC
61508
titled
Functional
safety
of
electrical/electronic/programmable electronic safety-related systems were published
during the period of 1998-2000 [54, 55, 56, 57, 58, 59, 60]. In 2005 the part of IEC/TR
61508-0 was published [61]. After that, the revision to this standard has been under
progress and the new standard edition is expected to be published soon.
IEC 61508 adopts the overall safety lifecycle as the technical framework for the
strategy of achieving functional safety, from initial concept, through hazard analysis
and risk assessment, development of the safety requirements, specification, design and
implementation, operation and maintenance, and modification, to final
decommissioning and/or disposal. The functional safety requirements specification
consists of two elements:

Safety function requirements, and

Safety integrity requirements.

The safety function requirements are derived from the hazard analysis. It is the
safety function that determines what has to be done to achieve or maintain a safe state
for the equipment under control [62]. The safety integrity requirements are derived
from the risk assessment. It is the safety integrity that determines what degree of
certainty is necessary that the safety function will be carried out. In other words, the
safety integrity is the referred index of the safety performance.
IEC 61508 specifies four levels of safety performance for a safety function. These
are called safety integrity levels (SIL). Safety integrity level 1 (SIL1) is the lowest
level of safety integrity and safety integrity level 4 (SIL4) is the highest level. The
higher the level of safety integrity, the lower is the intended likelihood of a dangerous
failure. It should be noted that when determining the SIL, the mode of operation is an
important factor that influences the measures chosen. IEC 61508 classifies the mode of
operation into two categories:
(1) Low demand mode, where the frequency of demands for operation made on a
safety-related system is no greater than one per year and no greater than twice
the proof-test frequency.
(2) High demand or continuous mode, where the frequency of demands for
operation made on a safety-related system is greater than one per year or
greater than twice the proof-check frequency.
The target failure measures corresponding to the SIL are specified differently for a
safety function operating in different demand modes.
IEC 61508 is both a stand-alone standard and can also be used as the basis for
sector and product standards. As for nuclear power industry, IEC 61513 is the sectorapplication standard published in 2001 with the title Nuclear power plantsInstrumentation and control for systems important to safety - General requirements for
systems [63]. This standard has adopted a presentation format similar to basic safety
40

publication IEC 61508 with an overall safety life-cycle and a system life-cycle. The
standard also provides an interpretation of the general requirements of IEC 61508,
parts 1, 2 and 4, for the nuclear application sector. Compliance with this standard will
facilitate consistency with the requirements of IEC 61508 as they have been
interpreted for the nuclear industry.
The important parts of IEC 61513 are its normative clauses 5 to 8:
1) Clause 5 addresses the total architecture of the I&C systems important to
safety:
a) Defining requirements for the I&C functions, and associated systems and
equipment (I&C FSE) derived from the safety analysis of the NPP, the
categorization of I&C functions, and the plant lay-out and operation
context.
b) Structuring the totality of the I&C architecture, dividing it into a number of
systems and assigning the I&C functions to systems. Design criteria are
identified, including those to give defense in depth and to minimize
potential for common cause failure (CCF).
c) Planning the total architecture of I&C systems.
2) Clause 6 addresses the requirements for the individual I&C systems important
to safety, particularly the requirements for computer-based systems.
3) Clauses 7 and 8 address the overall integration, commissioning, operation and
maintenance of the I&C systems.
IEC 61513 also includes as its informative part some important annexes:
1) Annex A highlights the relations between IAEA and basic safety concepts that
are used throughout this standard;
2) Annex B provides information on the categorization/classification principles;
3) Annex C gives examples of I&C sensitivity to CCF;
4) Annex D provides guidance to support comparison of this standard with parts
1, 2 and 4 of IEC 61508. This annex surveys the main requirements of IEC
61508 to verify that the issues relevant to safety are adequately addressed,
considers the use of common terms and explains the reason for adopting
different or complementary techniques or terms.
In fact, IEC 61513 is the first level standard prepared by subcommittee 45A
(SC45A): Instrumentation and control of nuclear facilities, of IEC technical committee
45 (IEC/TC45): Nuclear instrumentation. SC45A core domain is instrumentation and
control (I&C) systems important to safety in nuclear energy generation facilities.
SC45A standards cover the entire lifecycle of these I&C systems, from conception,
through design, manufacture, test, installation, commissioning, operation,
maintenance, aging management, modernization and decommissioning. Some other
level SC45A standards that are relevant to IEC 61513 are:

41

IEC 61226: Nuclear power plants Instrumentation and control important to

safety Classification of instrumentation and control functions [64]
This standard classifies nuclear power plants instrumentation and control
systems according to their importance for safety. The Standard establishes a
method of classification of the information and command functions for nuclear
power plants, and the I&C systems and equipment that provide those functions,
into categories that designate the importance for safety of the function. The
resulting classification then determines relevant design criteria. The design
criteria are the measures of quality by which the adequacy of each function in
relation to its importance to plant safety is ensured.

IEC 60709: Nuclear power plants Instrumentation and control systems

important to safety Separation [65]
For independence is required by IEC 61513, this standard defines the
assessments needed and the technical requirements to be met for
instrumentation and control (I&C) systems important to safety and their cables,
in order to achieve adequate physical separation between redundant sections of
a system and between a system and another system. This separation is needed
to prevent or minimize the impact on safety that could result from faults and
failures which could be propagated or affect several sections of a system or
several systems.

IEC 60987: Nuclear power plants Instrumentation and control important to

safety Hardware design requirements for computer-based systems [66]
This standard is directly referenced by IEC 61513 which addresses the
generic issue of hardware design of computerized systems. It is a second-level
SC45A standard applicable to computer-system hardware for systems of Class
1 and 2 (as defined by IEC 61513) in nuclear power plants. This standard
covers digital systems hardware, including multiprocessor distributed systems
and single processor systems. It also covers the assessment and use of predeveloped items, for example, commercial off-the-shelf items (COTS), and the
development of new hardware.

IEC 60880: Nuclear power plants Instrumentation and control systems

important to safety Software aspects for computer-based systems performing
category A functions [67]
This standard is directly referenced by IEC 61513 which deals with the
system aspects of high integrity computer-based I&C used in safety systems of
nuclear power plants together. It is a second level SC45A document tackling
the issue of software aspects for I&C systems performing category A functions,
as defined by IEC 61226. This standard provides requirements for the purpose
of achieving highly reliable software and addresses each stage of software
generation and documentation, including requirements specification, design,
implementation, verification, validation and operation. It also gives
recommendations on several key issues of using software in nuclear safety

42

systems, e.g. dealing with software aspects of defence against common cause
failures, use of software tools and pre-developed software.

IEC 62138: Nuclear power plants Instrumentation and control important for
safety Software aspects for computer-based systems performing category B
or C functions [68]
This standard provides requirements and recommendations for the software
aspects of computer-based I&C systems of safety classes 2 and 3, as defined by
IEC 61513. These I&C systems may be used for category B or C FSEs
(Functions, and associated Systems and Equipment), as defined by IEC 61226.
Its scope can be compared to the scope of revised IEC 60880, the difference
being that IEC 60880 addresses the software aspects of I&C systems of safety
class 1.
IEC 60880 and IEC 62138 together cover the domain of the software
aspects of computer-based systems used in Nuclear Power Plants to perform
functions important to safety, which correspond to IEC 61508, part 3 for the
nuclear application sector. Hence, they are consistent with, and complementary
to, IEC 61513.

IEC 62340: Nuclear power plants Instrumentation and control systems

important to safety Requirements for coping with common cause failure
(CCF) [69]
This standard provides requirements and recommendations for the total
architecture of instrumentation and control (I&C) systems. It provides
principles and requirements to overcome common cause failure (CCF) by
means which ensure independence. The implementation of these requirements
leads to various types of defense against initiating CCF events.

IEC 62342: Nuclear power plants Instrumentation and control systems

important to safety Management of ageing [70]
This standard provides strategies, technical requirements, and
recommendations for the management of ageing of nuclear power plant
instrumentation and control systems and associated equipment. It also includes
annexes on test methods, procedures, and technologies that may be used to
verify proper operation of such equipment and aim to prevent ageing
degradation from having any adverse impact on the plant safety, efficiency, or
reliability. This standard applies to all types of nuclear power plants and relates
primarily to safety.

IEC 60671: Nuclear power plants Instrumentation and control systems

important to safety Surveillance testing [71]
This standard lays down principles for testing I&C systems performing
functions important to safety, during normal power operation and shutdown, so
as to check the functional availability especially with regard to the detection of
faults that could prevent the proper operation of the important to safety
functions. It covers the possibility of testing at short intervals or continuous

43

surveillance, as well as periodic testing at longer intervals. It also establishes

basic rules for the design and application of the test equipment and its interface
with the systems important to safety. Further, the effect of any test equipment
failure on the reliability of I&C systems is considered.
3.2.2

IAEA standards

The International Atomic Energy Agency (IAEA) was set up as the worlds
Atoms for Peace organization in 1957 within the United Nations (UN) family. IAEA
works with its member states and multiple partners worldwide to promote safe, secure
and peaceful nuclear technologies. The IAEAs Statute authorizes IAEA to establish
safety standards to protect health and minimize danger to life and property. Many of
IAEAs member states have decided to adopt the IAEAs safety standards for use in
their national regulations.
In fact, the IEC/Technical Committee 45 (IEC/TC45) standards are closely related
to the IAEAs safety standards series. In order to avoid the potential for duplication
and even contradiction between IAEA documents and IEC technical standards, a
formal agreement of co-operation was reached in 1981 between IAEA and IEC/TC45.
The agreement states that IAEA is responsible for the development of safety principles
for instrumentation, control and electrical systems in nuclear power plants, while
IEC/TC45 is responsible for the design requirements that realize these safety principles
[72]. Therefore, the IEC/TC45 standards have been developed to be consistent with the
principles and basic safety aspects of the IAEA standards series, while the IAEA
documents apply to all IEC/TC45 instrumentation and control standards.
IAEA safety standards series covers nuclear safety, radiation safety, transport
safety and waste safety, and also general safety. The three categories within it are
Safety Fundamentals, Safety Requirements and Safety Guides. In particular, there are
four IAEA documents that are concerned with safety-related instrumentation and
control systems:

IAEA NS-R-1: Safety of Nuclear Power Plants: Design Safety Requirements

[73]
This Safety Requirements can be regarded as the overall guidance for
nuclear power industry. It establishes design requirements for safety functions
and associated structures, systems and components important to the safe
operation of nuclear power plants. It also establishes requirements for a
comprehensive safety assessment to identify the potential hazards that may
arise in the operation of plants. In relation to the design process, preventive and
mitigatory features for severe accidents, the management of safety, design
management, plant ageing and wearing out effects, computer based safety
systems, external and internal hazards, human factors, feedback of operating
experience, and safety assessment and verification are considered.

IAEA NS-G-1.1: Software for Computer Based Systems Important to Safety in

Nuclear Power Plants Safety Guide [74]

44

This Safety Guide provides guidance on the collection of evidence and

preparation of documentation to be used in the demonstration of safety and
reliability of the software for computer based systems important to safety in
nuclear power plants for all phases of the system life cycle. It recommends how
to meet the requirements established in IAEA NS-R-1.

IAEA NS-G-1.3: Instrumentation and Control Systems Important to Safety in

Nuclear Power Plants Safety Guide [75]
This Safety Guide recommends how the requirements established in IAEA
NS-R-1 should be met for instrumentation and control (I&C) systems important
to safety. It provides guidance on the design of I&C systems important to
safety in nuclear power plants, including all I&C components, from the sensors
allocated to the mechanical systems to the actuated equipment, operator
interfaces and auxiliary equipment.
The main content of this Safety Guide is its sections 2 to 7:
1) Section 2 discusses the identification of I&C functions and systems
within the scope of this Safety Guide, and their further classification
into safety and safety related functions and systems.
2) Section 3 describes the determination of the design basis for I&C
systems important to safety.
3) Section 4 provides design guidance for I&C systems important to
safety. It includes guidance that applies to all I&C systems important to
safety as well as guidance that applies only to safety systems.
Applicability of the guidance to these two classes is identified and
summarized.
4) Section 5 provides additional guidance that is specific to certain I&C
systems, namely protection systems, power supplies and digital
computer systems. The guidance for these systems comprises the
general guidance provided in Section 4 and the specific guidance
provided in Section 5.
5) Section 6 expands on the guidance given in Section4 in the area of
humanmachine interfaces.
6) Section 7 expands on the guidance given in Section 4 in the area of
design processes to ensure quality.

IAEA NS-G-1.8: Design of Emergency Power Systems for Nuclear Power

Plants Safety Guide [76]
This Safety Guide provides general recommendations and guidance for all
types of Emergency Power Systems (EPS) electrical and non-electrical
and specific guidance on the safety requirements for design and the features of
the electrical and non-electrical parts of the emergency power supplies. The
recommendations and guidance are focused on the power supplies necessary to
power loads important to safety.

45

3.2.3

IEEE standards

The Institute of Electrical and Electronics Engineers (IEEE) has also been
developing nuclear power plant standards through Nuclear Power Engineering
Committee (NPEC), one of the Technical Committees of the IEEE Power & Energy
Society (PES). NPECs established policy is to improve, clarify, update and provide
application guidance on the standards already produced and when appropriate, to
produce new standards.
Although the collections of IEEE and IEC standards have some overlap, but in
many cases cover significantly different topics. For example, IEEE standards go to
great depth on environmental qualification of many specific types of components,
while IEC covers the topic only at the general level. Conversely, certain IEC standards
deal with specific instrumentation and control (I&C) functions, a topic area where
IEEE standards are largely mute [77]. In fact, IEEE standards look upon the safetyrelated systems in general, rather than focus on the safety-related I&C systems, which
are one but perhaps the most significant part of the former. The following IEEE
standards are considered important to affect nuclear safety-related I&C systems:

IEEE Std 603: IEEE Standard Criteria for Safety Systems for Nuclear Power
Generating Stations [78]
This standard can be regarded as the general principles for nuclear safety
systems. It establishes minimum functional and design criteria for the power,
instrumentation, and control portions of nuclear power generating station safety
systems. The intent of these criteria is to provide a means for promoting
appropriate practices for design and evaluation of safety system performance
and reliability. However, adhering to these criteria will not necessarily fully
establish the adequacy of any safety systems functional performance and
reliability; nonetheless, omission of any of these criteria will, in most instances,
be an indication of safety system inadequacy.
The important parts of this standard are its clauses 5 to 8:
1) Clause 5 provides a large set of safety system criteria, including singlefailure criterion, completion of protective action, quality, equipment
qualification, system integrity, independence, capability for testing and
calibration, information displays, control of access, repair,
identification, auxiliary features, multi-unit stations, human factors
considerations, reliability, and common-cause failure criteria.
2) Clause 6 provides the functional and design requirements on sense and
command features, including automatic control, manual control,
interaction between the sense and command features and other systems,
derivation of system inputs, capability for testing and calibration,
operating bypasses, maintenance bypass, and setpoints.
3) Clause 7 provides the functional and design requirements on execute
features, including automatic control, manual control, completion of
protective action, operating bypasses, and maintenance bypass.

46

4) Clause 8 provides the requirements on power source, including

electrical power sources, non-electrical power sources, and maintenance
bypass.

IEEE Std 7-4.3.2: IEEE Standard Criteria for Digital Computers in Safety
Systems of Nuclear Power Generating Stations [79]
This standard addresses the use of computers as part of safety systems in
nuclear power generating stations. It specifies additional computer-specific
requirements (incorporating hardware, software, firmware, and interfaces) to
supplement the criteria and requirements of IEEE Std 603. This standard
should be used in conjunction with IEEE Std 603 to assure the completeness of
the safety system design when a computer is to be used as a component of a
safety system. This standard recognizes that development processes for
computer systems continue to evolve. As such, the information presented
should not be viewed as the only possible solution. However, this standard does
not provide requirements associated with the operation and maintenance of the
computer following installation (i.e., surveillance testing frequency).

IEEE Std 338: IEEE Standard Criteria for Periodic Surveillance Testing of
Nuclear Power Generating Station Safety Systems [80]
The standard provides criteria for the performance of periodic testing of
nuclear power generating station safety systems. The scope of periodic testing
consists of functional tests and checks, calibration verification, and time
response measurements, as required, to verify that the safety system performs
its defined safety function.

IEEE Std 336: IEEE Guide for Installation, Inspection, and Testing for Class
1E Power, Instrumentation, and Control Equipment at Nuclear Facilities [81]
This guide provides considerations for the pre-installation, installation,
inspection, and testing of Class 1E power, instrumentation, and control
equipment and systems of a nuclear facility. It is applicable to initial
construction, modification (backfit), and maintenance activities. However, this
guide does not apply to periodic testing.

ANSI/IEEE Std 352: IEEE Guide for General Principles of Reliability Analysis
of Nuclear Power Generating Station Safety Systems [82]
This Guide is also an American national standard endorsed by the
American National Standards Institute (ANSI). It provides the designers and
operators of nuclear power plant safety systems and the concerned regulatory
groups with the essential methods and procedures of reliability engineering that
are applicable to such systems. By applying the principles given, systems may
be analyzed, results may be compared with reliability objectives, and the basis
for decisions may be suitably documented.
The quantitative principles are applicable to the analysis of the effects of
component failures on safety system reliability. The principles are applicable
during any phase of the systems lifetime. They have their greatest value during

47

the design phase. During this phase, reliability engineering can make the
greatest contribution toward enhancing safety. These principles may also be
applied during the preoperational phase or at any time during the normal
lifetime of a system. When the principles are applied during either of these two
phases, they will aid in the evaluation of systems, in the preparation or revision
of operating or maintenance procedures, and in improving test programs.
Although not inherently limited, these principles are intended for application to
systems covered in the scope of IEEE Std 603.

IEEE Std 577: IEEE Standard Requirements for Reliability Analysis in the
Design and Operation of Safety Systems for Nuclear Facilities [83]
This standard sets forth minimum acceptable requisites for the performance
of reliability analyses for safety-related systems of nuclear facilities when used
to address the reliability requirements identified in regulations and other
standards. The requirement that a reliability analysis be performed does not
originate with this standard. However, when reliability analysis is used to
demonstrate compliance with reliability requirements, this standard describes
an acceptable response to the requirements.

3.2.4

Challenges to the U.S. nuclear standards

Some other organizations in the United States are also involved in developing
standards that significantly affect nuclear applications. These organizations include but
not limited to the American National Standards Institute (ANSI), the American
Nuclear Society (ANS), the American Society of Mechanical Engineers (ASME), the
American Society of Testing and Materials (ASTM), and the Instrumentation,
Systems, and Automation Society (ISA), etc. However, the existing standards are more
focused on specific types of nuclear components and corresponding safety-related
functions, rather than give guidance on the nuclear safety-related I&C systems.
The U.S. Department of Energy (DOE) has undertaken significant investment in
nuclear power as a source of non-greenhouse gas emitting energy. DOEs Office of
Nuclear Energy sponsored an initiative to identify codes and standards that will be
employed in the next generation of nuclear power plants to determine if any gaps exist
that could hinder their construction. According to the discussions among the U.S.
Nuclear Regulatory Commission (NRC), standards development organizations, and
industry, technological advances might provide unique challenges to construction
practices and construction quality. New standards may need to be developed to reflect
advances in technology, such as standards for digital instrumentation and controls,
cyber security, computers used in safety systems, and Probabilistic Risk Assessments
(PRA) used for setting surveillance intervals, etc [84].
3.2.5

Hints from nuclear industry

The nuclear safety-related I&C systems are so important that they are
indispensable to the safety of nuclear elements and nuclear plants. Although they are
different from the SPS in electric power industry, their safety functions to maintain the
aimed object in a safe state under predetermined conditions are similar to those of SPS.

48

Therefore, the contents of the nuclear standards mentioned previously can be useful to
the development of SPS. At least, we can have some beneficial hints from the nuclear
industry as follows.

3.3

In the design phase, reliability engineering can make the greatest contribution
toward enhancing reliability and safety. It is during this phase that the
quantitative reliability analysis, i.e. Probabilistic Risk Assessments (PRA) has
its greatest value.

The functional safety requirements specification consists of the safety function

requirements and the safety integrity requirements. The former is derived from
the hazard analysis. The latter is derived from the risk assessment, which can
be evaluated by the safety integrity level (SIL).

The initial design process has a comprehensive safety assessment to identify

the potential hazards that may arise in the operation. In addition, preventive and
mitigatory features for severe accidents should also be considered.

The system should be designed to have the capability against the single failure
of its components (N-1 criterion).

Although redundancy is a widely accepted principle to enhance the system

reliability, it is not enough. It is also important to guarantee adequate physical
separation between redundant sections of a system and between a system and
another system.

Common cause failures (CCF) should be paid attention to during design of the
total system architecture. Various means including independence against
initiating CCF events should be considered.

Digital computer systems have the trend to prevail in the application of

instrumentation and control systems. Therefore, reliability of both hardware
and software becomes significant to the system reliability.

Testing at short intervals or continuous surveillance, as well as periodic testing

at longer intervals are necessary to check the functional availability especially
with regard to the detection of faults that could prevent the proper operation.
System operations for electric power

Many companies develop so-called operating rules for use in guiding operators in
energy control centers during conditions for which contingencies may result in
violation of reliability criteria. For example, many companies in the Western US use
operating nomograms to do this, where secure operating regions are delineated from
insecure operating regions in the space of parameters such as flows, generation levels,
and load levels, so that proximity to a security boundary can be easily monitored, and
when encountered, the nomogram axes identify the parameters the operator must
control in order to move into a more secure operating condition. Next section describes
another approach to derive operating rules, which finds its application in relation to
SPS.

49

3.3.1

Operational security rules from decision trees

The French company RTE and others have developed a semi-automated way of
obtaining operating rules. Here Monte Carlo simulation of various operating
parameters is performed to form many basecases and a database of post-contingency
response is extracted, which will be used to derive significant planning and operational
information using data mining techniques as shown in Figure 20. Since, decision trees
are capable to provide explicit rules to system operators, French transmission operator
RTE has been using decision trees to define operational security rules [85, 86, 87, 88,
89, 90, 91, 92, 93, 94, 95, 96]. A similar approach has been applied to Entergy [97,
98].

Figure 20: Monte Carlo simulation based power system planning

Typically, the database generation approach using Monte Carlo simulation of
operating parameters plays a critical role in obtaining operational security rules with
good performance. Studies in the past have used a uniform or random sampling of
various parameters such as load level, exchanges at the boarders, etc. from their
probability distributions to form various basecases. But such an approach has two
problems:
(1) Computational requirements: Tedious and time consuming, as there could be
a tremendously large number of combinations of variables and topologies.
(2) Error Rates (Classification Accuracy): Some studies expend extra
computation after validating the operational rules to increase the unstable (rare)
situations in database to improve the accuracy. But, if the sampled unstable situations
are unlikely (namely a low probability of occurrence in reality), then it could make the
rules very conservative, i.e. either costly to respect or increase false alarms by
misclassifying acceptable scenarios as unacceptable.
Researchers at Iowa State University have developed an efficient sampling method
to generate likely influential operating conditions that captures high information
content for better classification and also reduces computing requirements. The method
developed is tested on the Brittany area of RTEs system for a voltage stability
50

assessment study [99], and decision rules are shown to have improved accuracy. A
database has high information content if there are many operating conditions for which
the post-contingency performance is close to the threshold or boundary region. So, the
influential operating conditions are generated directly from the boundary region using
Monte Carlo importance sampling by creating a probability re-orientation in the
operational parameter (input) state space towards the boundary region, which also
preserves the relative likelihood of the sampled conditions.
C. Singh et. al. [100] used a state space pruning method to identify the influential
region in the discrete operational parameter space under a single load level. We use
stratified sampling to quickly identify the boundary region in the operational parameter
state space in stage I, and then apply importance sampling to bias the sampling towards
the boundary region in stage II, as shown in Figure 21 depicting the developed
efficient sampling approach.

Database

Figure 21: Efficient database generation approach

3.3.2

SPS logic design using decision trees

The approach described in Figure 20 may also be utilized within the SPS logic
design step [101, 102, 103], recognizing that the only difference between operating
rules and SPS logic are:
1. The SPS logic is automated.
2. The SPS logic is not only limited to critical operating condition detection
with respect to some stability criteria, but also to automatic
preventive/corrective action to safeguard the system against impending
instability or increase the loadability of the system.
A typical SPS logic design procedure is illustrated in Figure 22. A study was
performed to demonstrate the SPS logic design using decision trees, with the following
study assumptions:
1. SPS location and functionality (i.e., Load Reject, Generator trip, undervoltage load shed (UVLS) etc.) already available
2. Critical contingency for which SPS installed known
3. Study performed to find either or both:

51

a. Arming condition for SPS, i.e., IF <condition> and

b. Actuating function of SPS. i.e., THEN <SPS Action>
As shown in Figure 22, a post-contingency response database is formed with the
simulation having the SPS modeled within it, i.e., in the event of any basecase
becoming post-contingency unstable, the SPS action is taken to save it. For instance, it
may be a single generator trip function or UVLS etc. Then two sets of basecases are
obtained:
a. Stable post-contingency cases with and w/o SPS action
b. Unstable post-contingency cases with SPS action
The cases that do not become stable even after SPS has been armed, in an effort to
save it, are thrown as they do not give any valuable information for SPS logic
derivation. The remaining cases are used to form the database for training the decision
tree to form the arming logic for SPS.
If the decision to be taken is binary (DT1), i.e., whether to arm SPS or not, then the
target (class) attribute contains information on status of SPS for every record in the
database. Finally decision tree DT1 would induce operating rules in terms of other
system attributes for deciding when to arm.

Sampling
Load, Unit commitment, Line
availability, wind speeds etc. etc.

Contingency Analysis

Database
Stable cases with and w/o SPS
action
(UVLS or generator trip)

Unstable cases inspite

of SPS action, thrown
away!

Decision Tree (SPS Logic)

DT1: To shed (Gen trip) or not?
DT2: How much to shed (trip)?

Figure 22: SPS logic design

If the decision to be taken is quantitative (DT2), i.e., how much load to shed or
how many generators to trip, then the target attribute could be the discretized version
of the actual amount shed load (UVLS) or the numbers of unit trips.

52

3.3.3

Illustration

This section presents a numerical example of the methodology explained in the

previous section to design SPS logic using Monte Carlo simulation of operating
conditions and decision tree based inductive learning.
The test system considered is the example case study provided as part of the
ASSESS software from French transmission operator RTE-France [104]. The base
case consists of 28 nodes and 37 lines forming a 90 kV network, serving 330 MW of
total load consumption. It is divided into four zones, namely A, B, C and D. The node
names follow a random numbering order from N1 S41 to N27 S41 except for a node,
called INTERS41, where there is the HVDC connection. There are 16 hydraulic
generators, 11 thermal generators and 5 wind production areas located in 2 zones
considered to be windy, zones A and B. Each wind generator area can produce up to
12 MW at 20 kV and is modeled by an asynchronous generator combined with a
negative load for static computations. Other generators (hydraulic and thermal) are
modeled by synchronous machines.
The study is performed to develop arming logic for a UVLS SPS, which is a
response based SPS. The decision to be taken is whether to shed a load or not? The
main aim of the study is to analyze the effects of five wind turbines in the test system
under various load levels, line and generator unavailabilities. Hence, following are the
operating parameters whose effects are captured by probabilistic models in ASSESS:
1. Wind speed in A and B zones
2. Load level
3. Unavailability of some lines and/or generators
For each basecase, an AC optimal power flow is performed using the software
TROPIC (integrated with ASSESS), minimizing the production cost under voltage,
current, flow constraints in N. The cases that encounter problems with convergence are
subjected to load shedding to achieve convergence. The database contains 500 stable
basecases, among which 67 involved SPS arming, i.e., load shedding. So a decision
tree was induced from the database to derive operational rules for arming the SPS, as
shown in Figure 23.

Figure 23: Decision tree: shed load or not?

53

Hence the design logic for SPS arming can be:

As mentioned in section 3.3.1, when the operating conditions are sampled near the
boundary region of the operating parameter state space (multivariate state space
formed by load levels and wind speeds under various combinations of discrete
parameters, i.e., line and generator unavailabilities), the resulting operating rules will
have better classification accuracy and result in reliable SPS operations.

54

Risk assessment, process view

This chapter will focus on risk assessment from the process point of view. Section 4.1 will
describe process-related failure modes, and section 4.2 will summarize methods of process-view
risk assessment. Section 4.3 presents formulation of risk expression for a Generation Rejection
Scheme (GRS), which was already presented in the publication [105]. Section 4.4 presents an
illustration of reliability evaluation of GRS. Section 4.5 presents various advancements in SPS
architectures and their reliability models.

4.1

Process view failure modes

We define an SPS event to be any one of the following:

desirable SPS operation,

undesirable SPS operation, or

SPS failure to operate when needed.

An SPS operation may be desirable or undesirable, depending on the consequence of the

operation relative to the consequence had the SPS not operated. If the consequence of the
operation is less severe than the consequence had the SPS not operated, the operation is
desirable. This is case, for example, when the action of a generation rejection scheme trips one
out of three units following a disturbance when otherwise, all three units would have lost
synchronism.
If the consequence of the operation is more severe than the consequence had the SPS not
operated, the operation is undesirable. Undesirable operation may either be unintended, due to a
hardware, software, or human error, or it can be intended (according to the design), but still
undesirable due to a fault in the design logic. A nuisance operation, when an SPS takes
unnecessary action when there is no disturbance in the system, is an example of an undesirable,
unintended operation. An example of an undesirable, intended operation is when a generator
rejection scheme operates and trips a unit following a disturbance for which it was designed to
operate, but had the SPS not operated, the plant would have been stable. This situation can occur
if the disturbance is single phase to ground fault and the design criteria is based only on three
phase faults.
An SPS failure to operate occurs when the SPS fails to respond as designed to conditions for
which the SPS is supposed to operate. An SPS may fail to operate as expected for several
reasons, among which are:

hardware failure,

faulty design logic,

software failure, or

human error.

Hardware failure occurs when some physical stress exceeds the capability of one or more
installed components. Faulty design logic may occur as a result of inappropriate or incomplete
study procedure during the design. Software failure results from errors in vendor written and user

55

written embedded, application, and utility software. The vendor software typically includes the
operating system, I/O routines, diagnostics, application oriented functions and programming
languages. User written software failure results from errors in the application program,
diagnostics, and user interface routines. Human errors can be classified according to whether
they are associated with construction, operation, or maintenance.
Other failure modes that may lead to an undesirable operation or a failure to operate include
failure to arm (any failure of a SPS to arm itself for system conditions that are intended to result
in the SPS being armed), unnecessary arming (any arming of a SPS that occurs without the
occurrence of the intended arming system condition(s)), and failure to reset (any failure of a SPS
to reset following a return of normal system conditions if that is the design intent).
When correctly operating, SPS significantly improve system response following a
contingency. However, the failure of SPS to accurately detect the defined conditions, or the
failure to carry out the required pre-planned remedial action, can lead to serious and costly
consequences. The survey by IEEE-CIGRE [5] suggests that the cost of SPS failure can be very
high as most of the respondents selected the highest cost category when asked to estimate the
cost of an operational failure of SPS.
Review of the U.S. NERC System Disturbance Reports from 1986-1995 [10,106] indicates
that of the 30 cases that involved the operation of SPS, 21 were reported as successful operation
of SPS, while 9 involved operational failures. The reasons for these failure cases include flaw in
logic design, software failure, hardware failure, incorrect setting, and inadvertent failure to arm.
The following are brief descriptions of these failure cases:
WSCC - Northeast/Southeast Separation Scheme - April 4, 1988:
Scheme:

System separation.

Reason:

Flaw in design (the scheme was susceptible to misoperation due to short

bursts of communications circuit noise).

Consequence:

1902 MW of generation was lost and 253 MW of load was interrupted.

Lessons learned: Faulty design logic.

NPCC - Hydro-Qubec - April 18-19, 1988:
Scheme:

Load rejection.

Reason:

Hardware failure.

Consequence:

Systemwide blackout.

Lessons learned: Hardware failure.

NPCC - Hydro-Qubec - November 15, 1988:
Scheme:

Load rejection.

Reason:

Hardware failure.

Consequence:

3950 MW of load was interrupted.

Lessons learned: Hardware failure.

WSCC-British Columbia Hydro/TransAlta Separation - January 7, 1990:

56

Scheme:

Controlled opening of lines.

Reason:

Not armed (inadvertently).

Consequence:

It caused 230 kV Cranbrook-Nelway circuit to trip on the subsequent

swing and resulted in separation (islanding) of the eastern part of the
BCHA/TAUC system from the Interconnection.

Lessons learned: Human error.

WSCC-Garrison - Taft 500 kV No.1 and 2 outages - January 8, 1990:
Scheme:

Var Compensation (trip two 500 kV bus reactors).

Reason:

Flaw in the logic design.

Consequence:

It caused the unnecessary dropping of generation at Hauser, Morony,

and Ryan (119 MW) as well as the loss of customer load (25 MW) in
Helena.

Lessons learned: Faulty design logic.

WSCC-SE Idaho/SW Wyoming Outage - September 12, 1991:
Scheme:

Generator rejection.

Reason:

Hardware failure (telemetry that automatically arms this scheme was

out of calibration).

Consequence:

It caused the loss of a second 345 kV line which led to further loss of
transmission by overload and out of step conditions.

Lessons learned: Hardware failure.

WSCC-Pacific AC Intertie Separation - November 17, 1991:
Scheme:

System separation.

Reason:

Software failure in PG&E SPS programmable logic controller caused

the delay in initiating remedial actions (also maybe hardware failure).

Consequence:

Fail to separate WSCC system into two islands, but did not produce any
severe problems (it was expected that there would be load lost and outof-step conditions).

Lessons learned: Software failure and/or hardware failure.

WSCC-Minnesota - Wisconsin Interface 69 kV conductor burn down - October 13, 1992:
Scheme:

Controlled opening of lines.

Reason:

Incorrect setting.

Consequence:

Two 69 kV lines in the Northern States Power and Dairyland Power

Cooperative service burned open causing the lines to fall to ground and
trip out.

Lessons learned: Human error.

MAPP & MAIN - Eastern MAPP-Western MAIN Interface Separation -November 6, 1997:

57

Scheme:

Controlled opening of lines.

Reason:

Flaw in design (opened the circuit at an ampere level below its setting,
possibly due to an unbalanced load.).

Consequence:

Resulted in low voltages in the south-western Wisconsin, eastern Iowa

and western Illinois (Cordova), heavy loading of parallel, lower voltage
transmission systems, and a large phase angle across the open tie at
Arpin.

Lessons learned: Faulty design logic.

4.2

Process view risk assessment

There are several existing methods that can be used in SPS reliability evaluation. We
summarize four of these methods [107, 108] in what follows. Although these methods are most
commonly applied in assessing hardware reliability, we emphasize that their use in SPS
reliability assessment must also include assessment of human error [109] and logic integrity [48],
as these aspects of SPS reliability are often the weak links in the design.
4.2.1

Failure mode and effect analysis

A Failure Mode and Effect Analysis (FMEA) is a systematic technique that is designed to
identify failure modes. It is a bottom-up method that starts with a detailed list of all
components with the system. An entire system can be analyzed one component at a time.
Alternatively, the system can be hierarchically divided into subsystems and modules as required.
The basic steps in the process are
1) break the system down into subsystems
2) list all components
3) for each component, list all failure modes
4) for each failure mode
a. list its effect on the next higher subsystem or system, and its failure rate
b. list the severity of the effect
5) when the next higher subsystem is the highest system, stop; otherwise, consider the next
higher subsystem as a component, and return to 3)
The output of this process is a list including component name, failure mode, failure rate and
failure effect.
The FMEA technique is generally poor at identifying combinations of failures that cause
critical problems. Since each component is reviewed individually, failures due to combination of
components are not addressed. Common cause failures are rarely identified since they require
more than one component failure. FMEA can be used an initial step to identity failure modes for
Markov modeling.

58

4.2.2

Fault tree analysis

Fault tree analysis is a top-down approach to the identification of failure modes. It is very
complementary to the FMEA in that it requires a deductive approach to finding failure modes.
The method is good at finding combinations of failures that may cause problems. The fault tree
is developed using fault tree symbols. Fault tree analysis begins with the determination of the top
event. The fault tree is constructed by determining the failures that lead to the primary event
failures. After the fault tree structure is fully developed, failure rate data, which can be obtained
from field experience or from industry published data, is employed to quantify the fault tree.
The basic steps to build the fault tree are
1) identify a system or level L=0 fault (the top event)
2) identify all combinations of Level L+1 events that lead to level L failure. The
sequences of events are connected by AND, OR, or other logic gates.
3) if level L+1 constitutes a set of basic or trigger events, then stop; otherwise, L=L+1
and go to step (2)
4.2.3

Network modeling

Many systems used in industry can be modeled through the use of simple networks. Network
modeling (or reliability block diagrams) is used to perform a system integrity analysis through
representing the system as a number of functional boxes interconnected to show the effect of
each box on the overall system. The resulting networks show components in series, in parallel, or
in combination configurations. The key step in the process of reliability modeling is to convert
from a physical system into a network model. A good qualitative understanding of system
operation during both normal conditions and during failure conditions must exist. A reliability
network model is drawn with boxes that represent the modules or the components that comprise
the system. Lines are drawn between the boxes to indicate operational dependency. The network
model may connect very differently from the physical model. A reliability network may be
viewed as showing the success paths. If the viewer can find a path from left to right through
the reliability network, those components are sufficient to allow the system to operate. Given a
network, the rules of probability are used to evaluate success and failure probabilities.
4.2.4

Markov modeling

Markov modeling involves definition of all mutually exclusive success/failure states in a

system. These are represented by labeled circles. The system can transition from one state to
another whenever a failure or a repair occurs. Transitions between states are shown with arrows
and are labeled with appropriate failure or repair probability (often approximated using
failure/repair rates). With time modeled in discrete increments (for example, once per hour),
calculations can be made showing the probability of being in each state for each time interval.
Since some states represent system success, the probabilities of these states are added to obtain
either system reliability or system availability as a function of time.
Markov modeling is applicable for systems for which at any given time the subsequent
system state only depends on the state at the given time and is not affected by the state at any
preceding time. We can assume that an SPS has this characteristic.

59

Markov modeling is well suited for use in SPS reliability modeling because its flexibility
provides that it can account for the variety of features which are common in SPS. Specifically,
Markov modeling can incorporate independent and common cause failures, partial and full
repairs, maintenance, and diagnostic coverage. Most importantly, it provides that all of these
features can be modeled as a function of time. This is in contrast to probability methods which
provide steady state results and are accurate only for short repair times and low failure rates.
The necessary data for Markov modeling approach are the failure rate and repair rate for each
component of the SPS. There are primarily three sources for them: 1) actual data (field data or
test data); 2) published literature (databases, papers, handbooks, technical reports); and 3)
experts' opinions.
4.3

Formulation of SPS risk expression

There are many types of SPS in use today. Since the most commonly used SPS type in
industry is generator rejection scheme (GRS), we will narrow our focus to the GRS case and
develop a reliability evaluation approach for GRS that can also be adapted for other forms of
SPS. The typical power plant in which a GRS is installed features high generation capacity and
multiple generation units, and the plant depends on two or more transmission lines as its outlets.
Without GRS, outage of any one of these lines may cause an out of step condition at the plant.
This means that all generators at the plant will accelerate and trip on over-speed. We define any
circuit that initiates GRS action during a forced outage condition as a critical circuit. A properly
designed GRS, activated by outage of any critical circuit, will trip a limited amount of generation
at the plant in order to avoid out of step conditions for the remaining units.
4.3.1

Risk expression with GRS

Fi :

event that there is a fault on circuit i .

A:

fault type random variable. We define one phase to ground, two phase to ground, three
phase to ground and phase to phase fault, represented by 1, 2, 3, 4, respectively, as all
possible values of A.

Nc :

number of critical circuits.

Ei :

initiating event. The first N c initiating events correspond to N-1 outages,

i.e.,

Ei F1 F 2 F i 1 Fi F i 1 F N , i=1 to N, and the N c 1 initiating event is no fault,

i.e. E N c 1 F1 F2 FN c Initiating event Ei ,i N c 1 corresponds to simultaneous
outage of two or more circuits, but we have not considered multiple outages in this
report.
K:

transient instability event.

X:

pre-contingency operating point; it is a vector of critical pre-contingency controllable

parameters which significantly influence the post-contingency system performance. As
we consider generation level as the only critical pre-contingency parameter.

T:

GPS tripping event.

Risk ( ) :

risk of an event.

60

Im( ) :

impact of an event.

Pr( ) :

probability of an event

A GRS is designed to trip some pre-selected generating unit(s) at a plant in order to prevent
blackout of the entire plant. This action instantaneously reduces the electrical power input to the
transmission system following the occurrence of specified contingencies. Each operation of a
GRS is classified into one of the following categories:
1) The GRS trips when a contingency occurs ( T Ei ) , i 1,2 , , N c
2) The GRS does not trip when a contingency occurs ( T Ei ) , i 1,2, , N c
3) The GRS trips when there is no contingency (T E N c 1 )

4) The GRS does not trip and there is no contingency ( T E N c 1 )

According to these categories, the risk for a system with a GRS comes from three sources:
1) If a GRS takes action promptly and correctly as designed, system stability will be
maintained, but non-zero impact will occur via a controlled trip of a block of generation
capacity.
2) If a GRS fails to take corrective measures when armed and initiated, the plant may or may
not experience an out of step condition, depending on the pre-fault operating condition, and
fault type and location.
3) If a GRS takes an unnecessary action when there is no outage for a critical line, then nonzero impact occurs via a controlled trip of a block of generation capacity. This is called a
nuisance trip.
We assume most nuisance trips are caused by failure modes inherent to the design.
Therefore, variation in operating condition does not significantly influence this risk, and we
concentrate on effects on evaluating the risk associated with the first two sources.
The risk of an event E i ,i 1,2 , , which causes either GRS trip T or instability K, is
Risk (( K T ) / X ) . For simplicity, we drop the dependence on X, leaving the reader to be
cognizant of it in what follows. Thus, the risk is

Risk ( K T ) Risk ( K ) Risk (T )

Nc

Nc

i 1

i 1

Pr( K T Ei ) I m ( K T Ei ) Pr(T Ei ) I m (T Ei )

(4.1)

We discuss both the impact and probability terms in the following two subsections.
1. Impact

The impact associated with GRS failure to trip, T , possibly resulting in instability K, is
denoted as Im( K T Ei ) . This term reflects the same impacts of the event "instability"
which include energy replacement costs, repair costs, and startup costs.

61

The impact associated with GRS trip, T, is denoted by Im( T Ei ) . This impact, although it
does not include an instability event, is nonetheless not zero because a unit does in fact trip. This
impact also includes energy replacement costs, repair costs, and startup costs. However, whereas
instability causes loss of an entire plant, a controlled trip typically includes only 1 unit.
Therefore, the impact of a controlled trip is usually much less than the impact of instability.
2. Probability

The probability of the GRS failure to trip, T, resulting in instability K, is denoted

as Pr( K T Ei ) . Assuming that Ei may occur in any of 4 different ways, n=1,2,3,4,
corresponding to the four basic fault types, we may expand the probability term according to

Pr( K T Ei )

Pr( K T Ei ( A n))

n 1

Pr(T Ei ( A n)) Pr( K /(T Ei ( A n)))

n 1
4

(4.2)

Pr(T Ei ) Pr(( A n) /(T Ei )) Pr( K /(T Ei ( A n)))

n 1

The terms Pr(T Ei ) in (4.1) and Pr(T Ei ) in (4.2) are the probabilities of SPS success
and failure, respectively, and are addressed in the next section.
4.3.2

SPS reliability evaluation

In this section, we develop an approach for computing Pr( T Ei ) and Pr( T E i ) . This
approach integrates three techniques, i.e., the Failure Modes Effects Analysis (FMEA) technique,
the Markov modeling technique, and Markov model simplification techniques.
Let S {S0 , Sk ,, Sn } represent a state space of the SPS, where
and exhaustive states. We have

Sk

is a set of mutually exclusive

Pr( Ei T ) Pr(( Ei T ) ( S 0 S1k S n ))

Pr( Ei T S k )

k 0
n

(4.3)

Pr(T | ( Ei S k )) Pr( Ei S k )

k 0

Since event Ei is independent of

SPS state, then

Sk ,

that is, the occurrence of a fault is independent of the

Pr( Ei S k ) Pr( Ei ) Pr( S k )

(4.4)

Hence,

62

Pr( Ei T )

Pr(T | ( Ei S k )) Pr( Ei ) Pr(S ki )

(4.5)

k 0

and
n

Pr( Ei T ) Pr(T | ( Ei Sk )) Pr( Ei ) Pr( Sk )

k 0

(4.6)

Here we concentrate on the procedure for getting Pr( Sk ) .

Each SPS state can be in one and only one of the following four categories according to the
response of each system state to system input events,

C1--If the input is an active signal, then the SPS trips successfully; if the input is an
inactive signal; then the SPS has a nuisance trip.

C2--If the input is an active signal, then the SPS trips successfully; if the input is an
inactive signal; then the SPS does not trip, as expected.

C3--If the input is an active signal, then the SPS fails to trip; if the input is an inactive
signal; then the SPS has a nuisance trip.

C4--If the input is an active signal, then the SPS fails to trip; if the input is an inactive
signal, then the SPS does not trip, as expected.

These four categories comprise another state space of the SPS where the original states Sk
(k=0, 1 n) have been condensed to Cj (j=1, 2, 3, 4). Based on this state space, we have
4

Pr( Ei T )

Pr(T | ( Ei C j )) Pr( Ei ) Pr(C j )

Pr( Ei T )

Pr(T

(4.7)

j 1

and
4

j 1

| ( Ei C j )) Pr( Ei ) Pr(C j )

(4.8)

Each basic input event Ei belongs to a group either active (denoted as AC) or inactive
(denoted as AC ). The active input is the input that triggers SPS to trip, and the inactive input is
the input that does not activate tripping. Given basic input event Ei and Cj, the system output
event is completely determined. Therefore, the conditional probability term in (4.5) and (4.6) is 0
or 1 as expressed in (4.9).

1
Pr(T | ( Ei C j ))

0
Ei AC
Pr(T | ( E C )) 1
i
j

j 1,2
j 3,4
j 3,4
j 1,2

63

(4.9)

and

1
Pr(T | ( Ei C j ))

0
Ei AC
Pr(T | ( E C )) 1
i
j

j 1,3
j 2,4
j 2,4
j 1,3

(4.10)

The remaining question is how to calculate the value of Pr(Cj) (j=1,2,3,4); once this is done,
then each term on the right hand side of (4.5) and (4.6) can be calculated. Thus the values of
Pr( T Ei ) , Pr( T E i ) can be obtained.
4.3.3

State probabilities

There are six essential steps in evaluating the probability of each state, Pr(C j ) . We provide a
simple description for each step in what follows.
Step 1: Describe the system

There are two sub-steps. The first is to develop a logic block diagram of the understudy SPS.
The second is to identify the event to input mapping table, in which we list all possible system
input events together with the binary signal input to the SPS. In addition, we classify inputs as
active or inactive. An active event is one that should cause SPS activation; an inactive event
is one that should not.
Step 2: Identify the failure modes

We define that a component fails when the component cannot perform its predefined
functions. In this step, we use a procedure that is similar to a failure mode and effect analysis
(FMEA). We assume that there are two failure modes for each logic gate: output stuck on 1
(mode 1) or output stuck on 0 (mode 2). We do not consider the possibility of having a failure
mode such that the output is always the complement of that which it is supposed to be.
Step 3: Define the system states

System states are represented by the combinations of states of all system components. Given
that we have defined the modes, e.g.,
0--normal mode
1--failure mode 1
2--failure mode 2
Then we can define the state space of the system as the set of components where each
component may be in any of its modes. At this step, it may be possible to merge some states
based on physical observation of the system.
Step 4: Classify the states

The Markov model should be simplified as much as possible to retain simplicity. So we

classify states into categories, according to some criteria. For GRS, we can classify system states
into C1, C2, C3, and C4 according to the response of each system state to system input events.

64

Step 5: Reduce the states

We introduce two concepts

A transition state is a state that has non-zero entry transition probability from other
state(s) and non zero exit transition probability to other state(s).

An absorbing state is a state that has a 1.0 transition probability to itself.

The reduction steps are as follows

Merge absorbing state belonging to the same category. Entry transition probabilities are
added.

For each absorbing state, eliminate all preceding states that are in the same class Ci as
the absorbing state have only one exit transition probability. Add the entry probabilities
as the entry probabilities to the absorbing states.

Merge all transition states in the same class Cj that have identical transition
probabilities to common states. Entry probabilities are added. Exit probabilities remain
the same.

Step 6: Calculate state probabilities

We assume that the failure of the SPS components has approximately an exponential
distribution. Therefore the pdf of component failure is f (t ) e t , where is the failure rate
per unit time interval. Then the probability that the component fails before time t is
t

F (t ) e t dt 1 e t

(4.11)

When t is small, F(t) can be approximated by t, so

F (t ) t

(4.12)

With this model, we can write a n+1 by n+1 transition matrix B, where Bpq (p=0, 1, , n,
q=0, 1, n) indicates the probability that the system transfers from state Sp to Sq, and n=the
number of states.
Assume the probability list at initial time t=t0 is
Pr (0) (Pr( S 0 ' (t 0 )) Pr( S k ' (t 0 )) Pr( S n ' (t 0 ))) ,

(4.13)

After v time intervals, the probability list is

Pr (v ) Pr ( 0) B v (Pr( S 0 ' (t v )) Pr( S k ' (t v )) Pr( S n ' (t v ))) B v

(4.14)

The elements in the probability list Pr (v ) (Pr( S 0 ' (t v )

Pr( S k ' (t v ))
Pr( S n ' (t v )))
provide the probability that system is in state Sk' after v time intervals. Then we get

65

Pr(C1 )

Pr(C 2 )
Pr(C3 )
Pr(C 4 )

Pr( S k ' )
Pr( S k ' )
Pr( S k ' )
Pr( S k ' )

S k ' C1
S k ' C2
S k ' C3

(4.15)

S k ' C4

This concludes the description of the procedure used to obtain Pr(Cj). Once obtained, these
values are used in (4.5) and (4.6) to compute Pr( Ei T ) and Pr( E i T ) , respectively.
4.4

Risk assessment of generation rejection scheme

In this section, we will use Markov model to assess SPS reliability and operation risk. We
take generator rejection scheme as example to illustrate the approach. Generator rejection
scheme is one of the widely used special protection scheme by the industry. According to a
survey by industries it the most widely used special protection scheme employed by utilities
accounting for about 21.6% of all the special protection schemes used. Therefore proper
reliability evaluation of the generation rejection scheme is very important in reliability evaluation
of special protection schemes. The generator rejection scheme is designed to improve the
transient stability performance of a power system. Figure 24 shows a portion of the IEEE
Reliability Test System together with an illustration of the GRS logic. Line 12--13 and line 13-23 are critical lines. The way the generation rejection scheme works is that When the GRS
detects a line outage on either of these two lines, it trips promptly only one generator to keep the
other two generators in service. The way the GRS logic works is that when there is a fault on a
critical line, the breakers on this line open; an open signal (high level signal) from any breaker
energizes the output of the OR gate. The high level signal from the OR gate output, together with
the high level arming signal, sets the AND gate output in high level, which is input to the 2 out
of 3 voting scheme. When two or more of the voting scheme input signals are high, the voting
scheme output signal is high; otherwise, it is low. The high level signal from the voting scheme
will trip the selected generator [105].
Bus 23

Bus 12
Figure 24: Logic circuit for a GRS
66

4.4.1

Illustration

Markov modeling is one of the methods used to compute the reliability of safety
instrumented systems (SIS). Markov models can be used to compute probability of failure on
demand of an SIS and a detailed example is shown in [110]. In this section, this method has been
applied for a generation rejection scheme.
Notation

DD = Logic gate failure rate for detected failure

DU = Logic gate failure rate for undetected failure
= Repair rate for detected failure per year
T = Repair rate for undetected failure per year
1. Failure modes

1 - OR gate has failed

2- Two AND gates have failed
2. Component failure and repair rate

Table 9: Component failure and repair rate

Notation

DD
DU

Rate
0.02 (once in 50 years)

0.01 (once in 100 years)

1460(6 hours per year)
2(6 months time interval test)

3. Markov Model For Failure mode 1

For failure mode 1, there are 3 states as shown in Figure 25.

01: OR gate is working normally
11: OR gate has failed and failure is detected
21: OR gate has failed but the failure is not detected
where, state 21 is most severe

67

DD

DU

Figure 25: Markov model for failure mode 1

The steady-state equations are:

DD DU P01 P11 T P21 0

P11 DDP01 0
(4.16)

T P21 DU P01 0
P01 P11 P21 1
In (4.16), adding last equation to first equation we get,

1 DD DU P01 ( 1) P11 ( T 1) P21 1

(4.17)

Then, the probability of each state can be calculated by:

P01 0.97 1461 3
P 0.02 1460 0
11

P21 0.01
2
0

1
0.9950
0 1.3630 105

0
4.9750 104

(4.18)

As we can see, the probability of the most severe state 21, P21 is low .To reduce the
probability of the state 21, the redundancy of the OR gate can be employed.
4. Markov Model For Failure mode 2

In the Markov model shown in Figure 26, four states are identified:
02: All AND gates are working normally)
12: One AND gate has failed and failure is detected
22: One AND gate has failed but the failure is not detected
32: Sub -system is in the fail state and the condition is detected (2 AND gates has failed)
42: Sub -system is in the fail state and the condition is not detected (2 AND gates has
failed)
where, the state 42 is most severe.

68

2DD

2DU

3DU

3DD

2DD

2DU

Figure 26: Markov model for failure mode 2

The corresponding steady-state equations are:

3 DD 3 DU P02 P12 T P22 P32 T P42 0

3 DD P02 2 DD 2 DU P12 0
3 DU P02 2 D D 2 DU T P22 0
2 DD P12 2 DD P22 P32 0
2 DU P12 2 D U P22 T P42 0
P02 P12 P22 P32 P42 1

(4.19)

In the set of equations of (4.19), adding 4th equation to 1st equation, we get

(1 3DD 3DU ) P02 ( 1) P12 (T 1) P22 ( 1) P32 ( T 1) P42 1

(4.20)

Then, the probabilities of each state can be calculated by:

1461
3
1461 3
P02 0.91

P 0.06 1460.06
0
0
0
12
P22 0.03
0
0
0
2.06

P
0
0.04
0.02
1460
0

32

P42 0
0.02
0.04
0
3

69

1 0.9856
0 4.0503 105

0 1.4353 102

7
0 1.9773 10
0 1.9165 104

(4.21)

As we can see, the probability of the most severe state 42, P42 is low because the 2oo3 logic
improves the reliability by redundancy design. This result verified that redundancy is an effective
way to improve the reliability.
5. Combination of the failure modes

With the increase of number of components, the states increase exponentially that renders
system reliability assessment by Markov modeling time consuming. To simplify the
computation, the failure modes are combined. In the failure mode 1, we have 3 states and in the
failure mode 2, we have 5 states. Altogether we have 15 states as outlined below, with their
probabilities given in Table 10.
0102: All gates are working normally
0112: OR gates are working and 1 AND gate has failed and is detected (fail safe)
0122: OR gate is working and 1 AND gate has failed and is undetected (fail safe undetected)
0132: OR gates are working and 2 AND gates have failed and the condition is detected
(system fail detected)
0142: OR gate is working and 2 AND gates have failed and the condition is undetected
(system fail detected)
1102: OR gate has failed and failure is detected, and all AND gates are working (system fail
detected)
1112: OR gate has failed and failure is detected, and 1 AND gate has failed and is detected
(system fail detected)
1122: OR gates have failed and failure is detected, and 1 AND gate has failed and is
undetected (system fail detected)
1132: OR gates have failed and failure is detected, and 2 AND gates have failed and is
detected (system fail detected)
1142: OR gates have failed and failure is detected, and 2 AND gates have failed and is
detected (system fail detected)
2102: OR gate has failed and failure is undetected, and all AND gates are working (system
fail undetected)
2112: OR gate has failed and failure is undetected, and 1 AND gate has failed and is detected
(system fail undetected)
2122: OR gate has failed and failure is undetected, and 1 AND gate has failed and is
undetected (system fail undetected)
2132: OR gate has failed and failure is undetected, and 2 AND gates have failed and is
detected (system fail undetected)
2142: OR gate has failed and failure is undetected, and 2 AND gates have failed and is
detected (system fail undetected)

70

Table 10: Probability of each state

State
0102
0112
0122
0132
0142
1102
1112
1122
1132
1142
2102
2112
2122
2132
2142

Probability
0.980547605350883
0.000040294820980
0.014279819495401
0.000000392331899
0.000143201143164
1.343215897755340e-005
5.519838491000001e-010
1.956139656925000e-007
5.374409600000000e-012
1.961659495400000e-009
0.004902738026755
2.0146e-007
7.1394e-005
1.9615e-009
7.1595e-007

After the model is simplified by combining the states, the probabilities of each state are
shown in the Table 11.
Table 11: Probability of each state of simplified model
States
All gates are working
Fail safe
Fail safe undetected
Fail detected
Fail undetected

Probability
0.980547605350883
0.000040294820980
0.014279819495401
1.3228e-004
0.0050

According to the simplified Table 11, we are going to assume that P (fail undetected) is the
probability when the system is unsecure. From the previous results, we can see that the 2oo3
component increases the system reliability of the GRS, and the GRS itself has a high reliability
as all the gates are working normally during 98.05% of the time.
Next section presents some future developments in SPS, especially about the process and the
architecture getting strengthened by the use of Phasor Measurement Unit (PMU). This in turn
promises increase in process view SPS reliability.
4.5

Consideration for future SPS developments

As the era of the so-called "Smart Grid" emerges, the instrumentation, monitoring, control,
and protection systems in power industry are facing potentially significant changes due to the
penetration of information-age technologies. These technologies include but are not limited to
digital signal processing and digital communication, etc. With the development of powerful
microprocessors, the trend has been for digital systems to replace the analog ones and the
71

application of computer relays is such an example. Digital communication plays an important

role in the development of digital systems and it is possible to see in the future that bundles of
copper wires could be replaced by digital communication equipments such as Ethernet switches
and Ethernet media.
In addition, non-conventional instrument transformers have already been available which can
directly output digital signals of current and voltage measurements [111]. The synchronized
phasor measurement unit (PMU) can provide real-time information of phasors for the same time
stamp and will be the foundation of various kinds of wide-area protection and control schemes
[112].
It is also notable that a new concept of digital process bus has been raised in IEC 61850
[113]. With the development of intelligent electronic devices (IED), a one unique secondary
platform based on IEC 61850, called merging unit (MU), has been developed to interface all IED
such as protections, meters and control devices, etc [114]. It seems that the power industry is
getting ready for more aggressive steps by replacing switchyard hardwired schemes with plugand-play fiber-based virtual wiring solutions [115].
The SPS can be regarded as one kind of protection and control systems with special
protection purposes. Hence, its development is also likely to be exposed to the potential changes
mentioned above from the perspective of its constituent elements. For example, microprocessors
can prevail in the future for the choice of SPS logic solvers because their functions become more
and more powerful while the costs decrease rapidly. Although the sensor and the actuator are
different in functions, their data flow can be integrated in the same platform of merging unit with
the help of functional IED. If PMU is utilized as the measurement device, the synchrophasors of
voltage and current can be easily obtained and then sent to merging units for data process.
We can see that a possible SPS scheme in the future can be typically composed of digital
logic solvers, merging units (including auxiliary IED), PMU, Ethernet switches, and Ethernet
communication media. All the signals in them are digitally conveyed through a virtual process
bus. Since the all-digital SPS scheme has obviously more electronic components than a
conventional hardwired one, it should have some influence on its reliability. Yang Wang et. al.
have made considerable contributions towards quantitative reliability evaluation of PMU and
WAMS [116, 117, 118], two primary drivers for SPS advancements. The following section
presents a way to analyze the possible impacts on SPS considering these developments.
4.5.1

Possible SPS architecture

The all-digital SPS can have a variety of complex structures in the future. However, some
architecture of all-digital protection systems has already been proposed [119]. Based on the idea
of these configurations, we propose conceptual all-digital SPS architecture possible in the future
as shown in Figure 27.

Figure 27: A conceptual all-digital SPS architecture

72

In this architecture, the SPS consists of two redundant functional sets which are independent
of each other. Each set comprises a digital logic solver (LS), an Ethernet switch (SW), Ethernet
communication media (EM), merging units (MU), and phasor measurement units (PMU).
Specifically, components LS1, EM3, SW1, EM1, MU1, and PMU1 constitute one set of SPS
while components LS2, EM4, SW2, EM2, MU2, and PMU2 make up the other set. In addition,
the PMU of the two sets, i.e. PMU1 and PMU2 are shared with each other and can act as mutual
backup. For simplicity, we assume that one PMU/MU alone can perform the full function needed
for one set of SPS, instead of several units actually needed for processing different signals,
respectively. We also assume that the Ethernet interface is part of the host device (i.e. LS, SW,
MU, etc.) and its reliability is already included in the host device.
The main functional parts of SPS can be designed to be allocated in physically isolated
places against mutual interference and fire spreading. Also, every set of SPS can be designed to
be supplied by multiple power sources simultaneously, including AC, batteries, and UPS. Hence,
we assume its power supplies to be extremely reliable against common cause failure of
components due to power supply failures. Therefore, we assume for simplicity that components
of the all-digital SPS are independent of each other. In addition, the component state durations
are assumed to be exponentially distributed.
As we know, if there is any fault with the component recognized by either self-test routine or
manual test procedure, utilities would either fix or replace the problematic component so as to
keep the whole protection and control system up. Thus, we will analyze the proposed SPS
scheme with a repairable model. Since the estimated repair time for any failed equipment is
usually prescribed in power industry and the maintenance staff always conforms to this
guideline, we can assume constant repair rates for our repairable model.
Although we only illustrate the reliability analysis of the SPS scheme in Figure 27, we
emphasize the methodology to obtain important reliability indices such as state probability of
system failures, frequency of system failures, the mean time to failure (MTTF) and the mean
time to first failure (MTTFF). Therefore, similar analysis can be conducted for other possible
SPS configurations.
4.5.2

Illustration of network reduction method

According to the relationship of the functional components, the reliability block diagram of
the SPS architecture in Figure 27 can be drawn as shown in Figure 28. In general, it is not easy to
obtain the reliability indices directly for such a complex system. However, we can use the
network reduction method to analyze the system.

Figure 28: SPS reliability block diagram

73

As we can see in Figure 28, components MU1, EM1, SW1, EM3, and LS1 comprise the
subsystem S1 of a series structure. If we regard S1 as a composite component using the concept
of equivalent transition rates [120], this composite component will have the same values of
failure and repair rates, state probabilities of success and failure, and frequencies to success and
failure with the original subsystem, respectively. Similarly, we can use another composite
component to represent the series subsystem S2 formed by components MU2, EM2, SW2, EM4,
and LS2. Then the reliability block diagram is reduced to a simpler one as shown in Figure 29.
Here we represent the failure and repair rates of a general component i by i and i ,
respectively. In addition, we use pi ,s , pi , f and f i ,s , f i , f for component i to represent the state
probabilities of its success and failure, and the frequencies to its state of success and failure,
respectively. Then the equivalent reliability parameters for subsystems S1 and S2 can be
calculated as following.
The state probabilities of success are
p S 1,s p MU 1,s p EM 1,s pSW 1,s p EM 3,s p LS 1,s
iS 1

i i

p S 2,s p MU 2 ,s p EM 2,s pSW 2,s p EM 4,s p LS 2,s

iS 2

S1 {MU 1, EM 1, SW 1, EM 3, LS1}

i i

S 2 {MU 2, EM 2, SW 2, EM 4, LS 2}

(4.22)
(4.23)

The state probabilities of failure are

p S 1, f 1 pS 1,s 1
iS 1

p S 2 , f 1 p S 2 ,s 1
iS 2

S1 {MU 1, EM 1, SW 1, EM 3, LS1}

(4.24)

S 2 {MU 2, EM 2, SW 2, EM 4, LS 2}

(4.25)

i i
i i

The equivalent failure rates are

S1 i ,

S1 {MU 1, EM 1, SW 1, EM 3, LS1}

(4.26)

S 2 {MU 2, EM 2, SW 2, EM 4, LS 2}

(4.27)

iS 1

S 2

iS 2

The frequencies to state of success and failure are

i ,
f S 1,s f S 1, f pS 1,s S 1

i iS 1
iS 1 i

i ,
f S 2,s f S 2, f pS 2,s S 2
iS 2 i i iS 2

S1 {MU 1, EM 1, SW 1, EM 3, LS1}
S 2 {MU 2, EM 2, SW 2, EM 4, LS 2}

(4.28)
(4.29)

The equivalent repair rates are

S1

f S 1,s
i

i

pS 1, f iS 1 i i iS 1

i
1
,
iS 1 i i

74

S1 {MU 1, EM 1, SW 1, EM 3, LS1}

(4.30)

S 2

f S 2 ,s
i

i

pS 2, f iS 2 i i iS 2

i
1
, S 2 {MU 2, EM 2, SW 2, EM 4, LS 2} (4.31)
iS 2 i i

Figure 29: Reduction of SPS reliability block diagram

Now let us see the reduced SPS reliability block diagram of Figure 29. There are two parallel
structures in this diagram, i.e. components PMU1 and PMU2 comprise one parallel subsystem P
while the composite components S1 and S2 form the other parallel subsystem S. Again, if we
regard these two subsystems as two composite components in a higher level, we can further
simplify the SPS reliability block diagram to a much concise one as shown in Figure 30. By the
concept of equivalent transition rates, the new higher-level composite components will have the
same values of transition rates, state probabilities, and corresponding frequencies with the
original subsystems P and S, respectively. The equivalent reliability parameters for higher-level
subsystems P and S can be calculated as follows.
The state probabilities of failure are
p P , f p PMU 1, f p PMU 2, f
iP

p S , f pS 1, f pS 2, f
iS

i i

i i

P {PMU 1, PMU 2}

(4.32)
(4.33)

S {S1, S 2}

The state probabilities of success are

p P ,s 1 p P , f 1
iP

p S ,s 1 p S , f 1
iS

P {PMU 1, PMU 2}

(4.34)

S 2 {S1, S 2}

(4.35)

i i
i i

The equivalent repair rates are

P i ,

P {PMU 1, PMU 2}

(4.36)

S i ,

S {S1, S 2}

(4.37)

iP

iS

The frequencies to state of failure and success are

i ,
f P , f f P ,s p P , f P
iP i i iP

P {PMU 1, PMU 2}

75

(4.38)

i ,
f S , f f S ,s p S , f S
iS i i iS

S {S1, S 2}

(4.39)

The equivalent failure rates are

P
S

i

iP i i iP

i
1
,
iP i i

i iS
iS i

i
1
, S {S1, S 2}
iS i i

f P, f
p P ,s
fS, f
p S ,s

P {PMU 1, PMU 2}

(4.40)
(4.41)

Figure 30: Simplified SPS reliability block diagram

The simplified SPS reliability block diagram of Figure 30 consists of only two composite
components in series. Thus, it is now easy to get the reliability indices of the whole system of
SPS using the parameters of composite components derived previously.
The system reliability of SPS (i.e. state probability of SPS success) is
p SPS ,s p P ,s p S ,s

P S
P P S S

(4.42)

The system unavailability of SPS (i.e. state probability of SPS failure) is

p SPS , f 1 p SPS ,s

P S P S S P
P P S S

(4.43)

The system failure rate of SPS is

SPS P S

(4.44)

The frequency of SPS system failure is

f SPS , f p SPS ,s SPS

P S P S
P P S S

(4.45)

The mean cycle time of SPS is

TSPS

1
f SPS , f

P P S S
P S P S

(4.46)

The mean down time (MDT) of SPS is

MDTSPS

pSPS , f
f SPS , f

P S P S S P
P S P S

(4.47)

76

The mean time to failure (MTTF), i.e. the mean up time (MUT) of SPS is

MTTFSPS MUTSPS TSPS MDTSPS

4.5.3

SPS

1
P S

(4.48)

Illustration of Markov modeling method to calculate MTTFF

The mean time to first failure (MTTFF) is also an important reliability index. It represents the
mean value of time from the moment the system starts operating until it fails for the first time.
MTTFF is actually the concept of the first passage time applied to the reliability engineering
field. However, it cannot be obtained from the previous network reduction method. In fact, the
calculation of MTTFF is more complex than that of MTTF. Based on the model of continuous
parameter Markov chains, we can derive the ultimate formula for calculating MTTFF using the
transition rate matrix of the system as follows [121].
MTTFF p (0)( R11 ) 1U k

(4.49)

R12
R
Here, R11 is the sub-matrix of the full system transition rate matrix R 11
and
R21 R22
represents the set of transition rates from system success to system success. p (0) is the
probability row vector of system success states for the initial state (all components up), while U k
is the unit column vector of dimension k which is equal to the number of states of system
success.

In practice it is not simple to utilize this formula for computing the MTTFF of the SPS. We
do not know how many success states this SPS would have just by looking at the system
structure. We only know that the total number of states of this SPS is 212 4096 since the
system consists of 12 components. In addition, it seems that we also cannot give the details of
vectors p (0) and U k unless we know the number of states of system success, or the dimension
of the matrix R11 . But the details of R11 are even more difficult to know. So, we must use a
systematic strategy to obtain the MTTFF value of our SPS. The key issue is that we can get R11
after we obtain the full system transition rate matrix R which is 212 212 in its size. The strategy
to achieve this is illustrated in the following steps.
Step I: Initializing the state matrix

In order to analyze the system states, we initially form a state matrix which can represent the
status of the system and all of its components. In this state matrix, each row represents a distinct
state of the system and each column represents a component state. For a system consisting of n
components, the size of this state matrix would be 2 n n . For our SPS to be analyzed, this state
matrix size is 212 12 . Now every element of this matrix represents the status of a component in
a specific system state. If we use the values 0 and 1 indicating the success and failure states of a
component, respectively, the complete system states can be represented by this state matrix
consisting of exhaustive combinations of 0s and 1s. We can arrange an initial state matrix as
shown in Table 12.

77

Table 12: The initial state matrix

Step II: Identifying system states of success and failure

Since we can draw the reliability block diagram as shown in Figure 28, it is natural to think
that we can use the minimal cut set method to distinguish system states of success and failure.
The method could be carried out in three steps: Firstly, we find all the minimal cut sets of the
system. Secondly, we use minimal cut sets to find all the system states of failure. Finally, the rest
of the states represent system success. However, this method is not smart and convenient for our
SPS architecture to be analyzed. For one reason, it is not easy to find out all the minimal cut sets
if the number of components of the system is relatively large. For another reason, there will be
some overlapping system states of failure based on different minimal cut sets. Unless we can
identify all the overlapping states, this method is likely to yield a wrong number of system states
of failure.
Here we propose a better way to distinguish states of success and failure for our SPS.
Although this method is also based on the concept of cut set, the main difference is that we do
not need to search all the minimal cut sets of the system. Since the reliability block diagram of
this SPS can be decomposed into combinations of simpler series and parallel structures, we can
get the logical chain of the system as shown in Figure 31.

(PMU1) fails

(P) fails AND

(PMU2) fails

SPS fails OR
(S1) fails ( MU1) OR (EM1) OR (SW1) OR (EM3) OR (LS1) fails

(S) fails AND

(S2) fails ( MU2) OR (EM2) OR (SW2) OR (EM4) OR (LS2) fails

Figure 31: System logical chain of the SPS

In the previous Step I of initializing the state matrix, we have already used the values 0 and 1
indicating the success and failure states of a component, respectively. Now in the logical chain of
Figure 31, let us replace each component by its state value (0 or 1) and treat the relationship
conditions "AND" and "OR" as the corresponding logical operation symbols. Thus, the logical
chain of Figure 31 is translated into a Boolean calculation. And the final result of the Boolean
calculation is just the indication of the system state, i.e. the value 0 of "SPS fails" indicates the

78

system success and the value 1 of "SPS fails" as system failure. If we scan each row of the initial
state matrix already setup in Step I and do the Boolean calculation, all the system states can be
distinguished as success or failure without omission or overlapping.
For our SPS architecture to be analyzed, the number of states of system success and failure
are counted to be 189 and 3907, respectively. After all the system states are identified, we can
reorder the initial state matrix in a better form as shown in Table 13. All the system states of
success are moved to the first 189 rows of the state matrix and all the system states of failure are
gathered in the latter part of 3907 rows. This rearrangement will be better for use in the
following steps.
Table 13: The rearranged state matrix

Step III: Forming the full system transition rate matrix R

Since we have identified all the states of system success and failure and rearranged the state
matrix, it is now possible to obtain the full system transition rate matrix R . However, the
diagonal and off-diagonal elements of the transition rate matrix R are very different. They
represent single-step transition rates from a given state to itself and to another state, respectively.
As a strategy, we need to get the off-diagonal elements of the transition rate matrix R first and
then obtain the diagonal elements from the off-diagonal ones.
It is recognized that there are two types of relationships between any two system states [122].
Suppose we choose two arbitrary system states i and j . If we need at least two components to
change their status for a transition between system states i and j , the interstate relationship is
not a single-step transition and thus the corresponding transition rates do not exist, i.e. the
elements (i, j ) and ( j, i ) of the matrix R are both zeroes. If, however, there is only one
component, say component k , that changes its status between system states i and j , then the
interstate relationship is indeed a single-step transition and the corresponding transition rates do
exist. Further in this case, if component k is working in system state i and fails in system state
j , then the transition rate from state i to j is the failure rate of component k , i.e. the element
(i, j ) of the matrix R is k . Accordingly, the transition rate from state j to i is the repair rate
of component k , i.e. the element ( j, i ) of the matrix R is k . After we scrutinize all the
interstate relationships of any two distinct system states, we can get all the off-diagonal elements
of the transition rate matrix R . Figure 32 is a brief flow chart of this algorithm.

79

Figure 32: Algorithm to obtain off-diagonal elements of the matrix R

Now it is easy to calculate the diagonal elements of the full transition rate matrix R because
the diagonal element of any row has a definite relationship with the off-diagonal elements of the
same row, i.e.

rii rij 0

(4.50)

j i

80

wherein rii and rij represent diagonal and off-diagonal elements, respectively. The subscripts i
and j represent the row and column indices of the matrix R , respectively.
Therefore, the diagonal elements can be obtained by the formula as below.

rii rij

(4.51)

j i

Hence, the full system transition rate matrix R is known after we obtain all of its elements.
Step IV: Extracting the submatrix R11 from the full system transition rate matrix R

If we form the full system transition rate matrix R based on the rearranged state matrix as
shown in Table 13 of Step II, we can see that it is quite easy to obtain its submatrix R11 , i.e. the
set of transition rates between states of system success. Because all the system states of success
are located in the first 189 rows of the rearranged state matrix, R11 is just the upper left square
submatrix (size of 189 189 ) of the matrix R (size of 4096 4096 ) obtained in Step III. In a
word, we can extract R11 directly from R as shown in Figure 33.

state i

state j

R12

(189 189 ) (189 3907 )

R12
R 22

(3907 189 ) (3907 3907 )

R11

R ( 4096 4096 )

Figure 33: Extraction of submatrix R11 from matrix R

Step V: Computing the MTTFF of the system

Since the submatrix R11 is obtained, we can give the details of vectors p (0) and U k as
follows.
The probability row vector of system success states for the initial state (all components up) is

p (0) (1 0 0 0)

(4.52)

188 0 's

The unit column vector of dimension k which is equal to the number of states of system
success is
U k (1 1 1) T

(4.53)

189 1's

As the final step of the strategy, we are now able to compute the MTTFF of our SPS using
the following formula.
MTTFFSPS p (0)( R11 ) 1U k

(4.54)

81

Risk assessment, systems view

In the previous chapter, SPS risk assessment from a process view was thoroughly
discussed. It basically involved studying the influence of each and every component in the SPS,
and ascertaining the risk associated with them using traditional methods such as fault-tree, failure
mode and effect analysis, Markov modeling etc. A particular SPS design may appear quite
reliable from the process view, i.e., from sensor to actuation. However, there remain questions
such as:
1. Are there system operating conditions (topology, loading, flows, dispatch, voltage
levels etc.) that may generate a failure mode for the SPS, e.g., an undesirable operation
or a failure to operate?
2. Are there two or more SPS that may interact to produce a failure mode?
Most companies answer these questions using engineering experience and judgment, coupled
with tedious trial and error testing involving manual computer simulations. However, it would be
beneficial to have a decision-support tool that has an automated simulation capability having
flexibility to vary the SPS logic, and having intelligence to vary operating conditions and events
over a wide range so that various SPS logic may be tested against various operating conditions
and events. Monte Carlo simulation techniques have been used prevalently in many fields over
the years to model and study complex factors influencing a phenomenon. So in this chapter the
focus is on developing a Monte Carlo simulation based procedure for systems view SPS risk
assessment.
Section 5.1 presents a report of an interesting case study of a typical system view SPS failure
happening in NRDIC grid. Section 5.2 focuses on identifying ways that SPS can fail from a
systems view, and proposes methods to identify such failures. Section 5.3 briefly proposes a
Monte Carlo simulation based system view SPS risk assessment. Section 5.4 presents various
reliability models of power system, considering the impact of SPS on the network.
5.1

Systems view SPS failure- a case study

The following section presents a report on mis-operation of special protection schemes in the
Nordic grid on December 1st 2005 [ 123 ]. This case study demonstrates the possibility of
interactions between SPS leading to cascading effect on SPSs in the network, which is
undesirable.
5.1.1

Background of SPS in Nordic grid

In Norway two SPS have been implemented to deal with problems concerning high generation
in the northwest region. These SPS relieve the burden in important transmission corridors by
shedding generation, and thus maintain system integrity. The SPS are actuated by monitoring
critical bottleneck from the focus area with respect to a predefined limit. A brief description of
both the SPSs functionalities are given below:
(i)

SPS Nordland

The shaded area in Figure 34 shows the northern Scandinavia region, which is rich in
generation with a very low demand. It contains about 6000 MW hydroelectric power, which is
about 15% of the installed capacity in the Nordel grid. In the event of any critical transmission

82

corridor contingency (i.e., either 420 kV through Northern Sweden or 300 kV through Middle
Norway), the rest of the transmission lines must be protected against overload. So the SPS
Nordlands functionality is to:
(1) Generation Shedding: up to 1200 MW
(2) Net Split: if there is surplus generation in the northernmost part of Norway, it is
disconnected from the main Nordel grid.

Nordland

stland

Figure 34: Nordic SPS case study

(ii) SPS stland

The eastern part around Oslo, shown by the yellow-shaded region in Figure 34, is the main
load center in Norway. The SPS stland is actuated on outage or overload of central lines in
the Oslo area, thereby shedding up to 1200MW of generation on the west coast of Norway,
shown by ellipsis in Figure 34. So in cases where there are high power transfers from west to
east, i.e., either into the Oslo area or from Norway to Sweden, this SPS will protect the
remaining lines in the Oslo area from overloading.
5.1.2

Event report

This section provides the description of events related to SPS mis-operation, their sequence
and the consequences. At 3:02 pm (CET) on December 1st 2005, the breaker failure protection
disconnected a 420kV busbar at Porjus power plant, due to an occurrence of fault on the 420kV
breaker at the power plant. Consequently, important transmission corridors out of northern
Scandinavia got overloaded, which as per design should have triggered the SPS at Nordland
instantaneously. But this SPS operation failed, leading to Nordel grid going out of limits and
ensuing a series of cascading events. The operation of the grid outside design limits led to
83

undesirable triggering of the second SPS at stland, which was supposed to trip about 1150 MW
of generation from the Nordic grid. Fortunately this SPS also failed, and an imminent system
breakdown was evaded. Table 14 summarizes the sequence of events, following the first SPS
mis-operation at Nordland.
Table 14: Nordic SPS event description
Events

Time
elapse

0s

Descriptionofevents

Mis
operation?

HighhydroelectricproductioninthenorthwestNordelregion
Breaker fault while switching out 420kV reactor at Porjus
powerstation,NorthernSweden
Fault cleared by tripping a line out of Northern Scandinavia
throughSweden,whichoverloadedremaininglines(2300MW
outofScandinavia)

Yes
TriggeredNordlandSPS,whichtripped600MWgenerationas
designed but failed to instantaneously trip 1030 MW
generationinNordelgrid
0.8s
1700 MW of surplus generation went south and caused
overloads, which finally lead to tripping of a 220kV line in
Sweden(0.8s)and2300kVlinesinMiddleNorway(0.85s)
2.2s
No,but
Nordland activated Grid split and tripped about 487 MW
generationinboththesplitislands(Middle&NorthNorway) Undesirable
cascading
tocounterrapidfrequencyrise
3.3s
Nordland Generator tripping function of step 2 above finally Yes,delayed
response
worked, tripping 300 MW generation in North Norway that
leadto128MWsheddingbyautomaticUFLS

Duetogridsplit,underfrequencyconditionprevailedinmain
Nordel grid leading to west coast generation pickup, causing
highwesttoeastMWtransferinSouthernNorway

No,but
Duetoheavywesteasttransfer,SPSstlandgottriggered
Undesirable
cascading

stland failed to actuate 1150 MW generation trip in west Yes,dueto

humanerror
coastofNorway
15mins MiddleNorwaywasjoinedbacktomainNordelgrid

10

24mins NorthNorwaywasjoinedbacktomainNordelgrid

2
3
4
5
6
7
8

From Table 14, event 2 (Nordland SPS mis-operation) has been the initiating SPS failure
which has caused the rest of the consequences in a cascading manner, starting with the event 4
when Nordland implemented the Grid split due to surplus generation. The reason for Nordland
mis-operation was found to be due to changes made in SCADA and communication systems that
slowed the response time of the SPS Nordland. Thereby, event 5 in Table 14 is the delayed
response of Nordland SPS, which is unnecessary at that instant. All these created a scenario that
demanded SPS stlands triggering as intended by design (event 7), though under current

84

stressed circumstances its successful activation would have worsened the situation. This is a
typical example of potential failures in power system due to interactions between SPS. So had
the event 8 in Table 14, i.e., SPS stland tripping 1150 MW generation, taken place, there might
have been one of the two post-event consequences:

Best scenario: Under-frequency load shedding of 2400 MW

Worst Scenario: Nordel grid breakdown

Fortunately it failed to operate, due to a technician forgetting to reclose switches following a

routine maintenance test of the SPS! On the whole, this sequence of disturbances due to SPS
failures incurred a total loss of 250MW load, and the two SPS were deemed unfit for operation
until they are rectified. The costs involved with this event for Nordic was about 4.2 million
Norwegian kroner.
Therefore, this event shows how critical it is to categorize and investigate SPS failure modes
from a system view-point also. Depending upon the system in which SPSs are functioning and
the associated operating scenarios, many SPS related failures may arise that may lead to
catastrophic system collapses. Hence identification of such failure modes and the associated
power system scenarios, will aid us to perform system view SPS risk assessment and SPS logic
re-design.
5.2

Systems view failure modes and their identification process

SPS operations can be classified into desirable, undesirable or failure to operate as explained
in section 4.1. The systems view failure mode identification method will focus on identifying
the possibility of SPS undesirable operations and failure to operate happening due to fault in
logic design (i.e., intended failures). The study also focuses on identifying scenarios that may
induce interactions among SPS, giving rise to undesirable cascading SPS operations even if they
were originally intended according to logic design. This type of failures due to interaction could
be a valid case for concern especially in todays SPS-rich power networks, where SPS are touted
as effective and economic means of strengthening transmission networks, increasing power
transfer capability and deferring investment decisions. Some examples of such undesirable
interactions could be:

A desirable (or undesirable) operation of SPS followed by inadvertent operation of

another SPS under certain operating conditions producing undesirable consequences, or

Operation of two separate SPS under rare events such as loss of two elements that initiate
those two SPS, or

A failure of one SPS leading to intended but undesirable triggering of another SPS, as
was the case with Nordic grid event.

In the following sub-sections, we shed some light on current industrial practices in designing
operational logic for SPS, and we propose ways of identifying system view failure modes due
to limitations in SPS logic design.
5.2.1

SPS logic design practices

The operating rules for SPS and other similar protection systems are conventionally derived
based on deterministic techniques that consider only the worst-case scenario, combined with
85

experts judgment. But due to the probabilistic nature of power system that comprise a wide
range of network topologies, operating conditions and possible events; such deterministic
techniques do not give forth optimal level of settings for generator rejection schemes or remote
load shedding schemes or any other SPS [102]. Many recent studies have utilized a risk-based
analysis to ensure reasonable coverage of operational scenarios and events, and design robust
rules for SPS operations. Van Cutsem et. al. [124] use such simulation methods to build a set of
training scenarios to find the minimal load shedding settings for the protection device. They find
the optimal triggering rule for the load shedding controllers using combinatorial optimization, in
terms of average voltage over several buses in a study area and reactive reserve of critical
generators. This kind of load shedding scheme has been programmed in French transmission
operator RTEs ASSESS software [125], which is being used by them for their planning studies.
The works in [126] and [127] perform many contingency simulations under various operating
conditions to design generation trip and load shedding defense plans against transient, small
signal and voltage stability problems in longitudinal Taiwan power system. Here the monitoring
locations for SPS arming is decided based on operational experience and the thresholds are found
out from simulation studies. The work in [128] performs simulation of several outage scenarios
such as critical generators and tie lines, for different values of anticipated system overloads and
study the loci of frequency-frequency rate plot to decide on the settings of under frequency load
shedding SPS for preserving the integrity of the system in the event of islanding. Inspite of so
many systematic methods to design SPS logic, it is still a challenging task to come up with
appropriate settings under wide range of system events and achieve coordination between other
SPS in the system.
BC Hydro [33] designed and implemented a centralized automatic RAS arming scheme that
ensures systems integrity under critical events based on the measurements from EMS. This
provides opportunity to achieve coordination among various RAS actions. The pre-outage
security limits and operating rules for the various RAS schemes such as generator rejection are
derived from Monte Carlo simulation based operational planning studies performed offline. Any
system growth, topological/network change, generator change or any major equipment change
triggers RAS logic redesign process.
The process of deriving operating rules has been made more effective with the introduction
of data mining techniques such as decision trees, association rule mining etc. They seem to
provide good information about most critical system attributes, which are otherwise not possible
from conventional analytical techniques or manual interpretation by experts. Further due to such
automatic learning techniques ability to process huge database and extract knowledge easily,
database of larger sizes with diverse set of situations and events can be used to identify the most
relevant attribute and their appropriate settings for SPS [102, 103,]. Hsiao et. al. [129] proposed a
risk based contingency selection method for SPS applications. Here the risk of a contingency is
computed as the product of its probability and severity under various loading conditions and
inter-zone transfers, where severity is quantified in terms of amount of load shedding and
generation tripping.
Inspite of all these efforts, system view type of failures may still occur due to inherent
limitations in logic design procedures or incomplete/inappropriate study procedures such as
limited types of faults considered in design phase etc. This is because it is practically impossible
to capture all the influencing factors and interesting events, and come up with efficient operating
rules for SPS logic that can accurately detect and operate under various operating conditions.

86

5.2.2

SPS logic validation and failure modes

Generally operating rule can be validated against an independent test database containing
post-contingency simulation results and SPS arming status for a wide range of operating
conditions and events. The testing results of the operating rule against the test database can be
expressed in the form of a confusion matrix, as shown in Figure 35.
Armed

Not armed

Armed

True
Positive

False
Positive

Not armed

False
Negative

True
Negative

Predicted
status

Actual status

Figure 35: Confusion matrix- SPS logic testing

The percentage of instances/scenarios that are tested false positive and false negative give the
expected failure rates of SPS. The false positives denote intended inadvertent operation such as
over tripping of generation rejection scheme etc., and the false negatives denote failure to
operate. Among these, the false negatives are expensive failures as in a stressed conditions
failure of SPS operation would have maximum damage to components, system and the economy
than inadvertent operation. Consequently, we can further analyze these situations that produce
false negative failures for improving the SPS logic. Furthermore, even though this study gives
false positive failure percentage (i.e., inadvertent operation of SPS), it doesnt give any
information on whether the consequence of inadvertent operation of SPS is desirable or
undesirable from economic or reliability point of view. Therefore, this further requires a
complete simulation study, which is explained in the next section.
Apart from these two types of regular SPS failures, there is possibility of another type of
failure from a system point of view, which is due to interactions among SPS, as mentioned
earlier using Nordic grid event. Identifying such failure modes require time domain simulation of
many interesting system conditions with as many system dynamics modeled, especially of all the
SPSs in the system, and other important equipments & protective devices.
The next subsection focuses on designing a simulation methodology to identify ways that
SPS can fail from a systems view, i.e., undesirable intended SPS operation or failure to operate,
and failures due to SPS interactions.
5.2.3

Simulation based system view SPS failure mode identification

5.2.3.1 Regular SPS failure mode identification

The typical failure modes due to undesirable operation and failure to operate of a single SPS
is referred here as regular failure modes. Figure 36 depicts the risk based simulation
methodology to identify such failures using a comparative study between two sets of simulation,
one with SPS and another without the particular SPS. This can be done for every single SPS in
the system to detect its regular failure modes.

87

Choose contingency
Model SPS and other
important devices in
power system

Efficient sampling operating

conditions for SPS logic testing

Testing database

Contingency analysis using

Transient simulation

SPS OFF

SPS ON

Store post contingency

database and performance

Store post contingency

database and performance

Comparative study

Identify regular SPS failure modes

1. Undesirable operation
2. Failure to operate

Figure 36: Simulation study- regular SPS failure mode identification

(1) Efficient Sampling of operating conditions: First the corresponding contingency for
which the SPS is planned is chosen. Then efficient sampling of operating conditions
based on importance sampling method, as explained in chapter 3, is performed. This is
because the traditional Monte Carlo sampling generates operating conditions according to
the probability distribution of the operational parameter state space. But in the case of
SPS, since they are especially designed to operate in stressed operating conditions,
operating conditions close to stability threshold should be given more importance in the
sampling. So importance sampling is utilized to bias the sampling and generate operating
conditions from the stability boundary region of the operational parameter state space.

88

Furthermore, the proposed sampling strategy based on importance sampling also reduces
the computational burden in validating SPS logic by focusing only on important
situations. Some conditions from the far tail of the parameter distribution (lower
probability) may also be sampled, as the likelihood of SPS mis-operation and their
consequences are high for these situations.
This proposed efficient sampling procedure can be realized starting with identifying
the key operating parameters that would have influence on the operation of the particular
SPS under consideration. The stability boundary region is then found in that parameter
state space, and finally important operating conditions from that region are sampled
according to their relative likelihood. Depending upon the nature of the operating
parameter probability distribution, i.e., if parametric, correlated etc., suitable parametric
[130] or non-parametric methods (Copulas [131], Latin Hypercube Sampling [132]) are
used to generate realistic operating conditions.
(2) Simulation and Comparative study: Then two sets of transient contingency simulation
are performed on these sampled scenarios, one with the SPS functionality modeled and
other without SPS. Then comparative studies are done to see the effectiveness of the SPS
and obtain factual information on the kind of regular failures happening, i.e., the false
positives and false negatives.

The software available for such simulations at Iowa state university are Siemens PTI
PSS/E and Eurostag embedded to ASSESS. PSS/E has the capability to model SPS
using the tripping functionality in multiple-contingency analysis module. Generation trip,
load shedding, branch tripping based on bus voltage, line loading and generator output
are some variety of special protections that can be designed in PSS/E. The risk based
analysis in PSS/E has to be automated using scripting language such as Python or IPLAN
or using batch files. Eurostag can also be used to model SPS such as UVLS, UFLS,
overload branch tripping and other such automata. The advantage in working with
Eurostag is the ability to interface Eurostag with ASSESS software, which facilitates
automation of such risk based statistical simulation studies and post-processing of the
simulation results.
5.2.3.2 Interaction based SPS failure mode identification

The failures resulting from interaction can also be detected using comparative study by
modeling all the system SPSs in the simulation, as depicted by Figure 37. The generation of
operating conditions is similar to what was explained before, i.e., sample from the stability
boundary region of a contingency for which the particular SPS is installed. Since the interest is
for capturing interactions, we could sample operating conditions from the union of pairwise
intersections of stability boundary regions of all the SPS in the study. To reduce the complexity
of the study (i.e., the number of combinations of possible SPS interactions), firstly the SPS can
be grouped. The grouping of SPS may be done with respect to some of the following criteria:
1. Contingency locations corresponding to various SPS,
2. SPS locations,
3. Intersection among SPS solution strategies (ex. common generators tripped?)
4. Intersection among stability boundaries

89

5. Probability of any two contingencies happening in a cascading manner

6. Sensitivity of one SPSs action on anothers actuating criteria
Figure 37 depicts the process of sampling operating conditions in an efficient manner for
process failure mode identification, where SPS A and SPS B are considered to be grouped
together. The sampling state space is considered to be made of two operating parameters (1 & 2).
Parameter 2

Bias!

Comparative
Study for
SPS A

SPS A ON

Contingency A
Boundary
Contingency B
Boundary
SPS OFF

1. Failuretoop.
2. Undesirable

SPS A and B OFF

SPS B ON

Comparative
Study for
SPS B

Parameter 1
Parameter 2

Parameter 2

Contingency A
Boundary

Contingency B
Boundary
SPS A ON

SPS B ON SPS
A
ON

SPS
B
ON
Parameter 1

Parameter 1
Comparative
Study for
Interaction

Figure 37: Process view SPS reliability assessment operating conditions sampling
Figure 37 shows that in the case of identifying regular failure modes related to SPS A
operation, the operating conditions sampling process is biased towards the stability boundary
region of contingency A when both SPS are in OFF state. Similarly, in the case of identifying
regular failure modes related to SPS B operation, the sampling process is biased towards the
stability boundary region of contingency B with both SPS in OFF state. When it comes to
identifying interactions between SPS A and SPS B, two sets of operating conditions are sampled.
One set is to test the SPS B operation on operating conditions that are biased towards the
stability boundary region of contingency B, where the state space is made of post-contingency A
operating conditions subject to SPS A operation. The other set is to test the SPS A operation on
90

operating conditions that are biased towards the stability boundary region of contingency A,
where the state space is made of post-contingency B operating conditions subject to SPS B
operation. Then the two sets of simulation results are analyzed to identify any failure or
undesirable scenarios due to interactions among the two SPS. This can be extended to identifying
interactions among several SPSs. While this will surely require extra computational requirements,
the process of efficient sampling that biases the sampling procedure only to the important
conditions will reduce the burden of computation. Furthermore, linear sensitivity measures may
be used to further decrease the computational requirements to a greater degree.
5.3

Systems view risk assessment

The effects of various system view failures (identified in previous section) on the system
reliability can be estimated using a risk index, which is the product of probability of the failure
and its severity. The severity value of a particular failure mode in terms of performance measure
is computed from Monte Carlo simulation study (similar to failure identification study) as shown
in Figure 38. Various initiating SPS failures are embed in the simulation, and the system-level
SPS failures and their severities are computed. The severities can be quantified in terms of
reliability indices such as expected cost of electricity (EC), tripped generator cost (TGC),
expected energy not served (EENS) due to SPS forced curtailments and so on.

Figure 38: System view SPS risk assessment

If a particular failure mode imposes big risk, then the appropriate SPS logic can be re-defined
by using decision tree based method as explained in section 3.3, wherein the operating conditions
that lead to this failure mode in the simulation study are included in the database for training the
decision tree for rule re-design. Consequently, any tightening of SPS logic and its consequent
effect on solution strategy, say for example, the rule is re-designed to contain some expensive
generator trip strategy, will enable evaluating the additional expense necessary to maintain a
system with a large number of SPS.

91

5.4

Reliability analysis of power systems with impact of SPS operation

It is obvious that SPS operation would have an impact on the power system reliability as SPS
can operate inadvertently or simply fail to operate when needed. However, it is not simple to
evaluate the power system reliability including the impact of SPS operation. One reason is that
there are various types of SPS such as generation rejection, load rejection, system separation, etc.
and these different SPS have varying design and operational action. Another reason is that SPS
failures can cause such complicated interactions between current-carrying components that both
component and system states experience intricate changes.
Thus, we first need to unify the effects of different SPS operations in order to make
reliability analysis of power systems feasible. In steady state point of view, effects of all SPS
operations can be classified ultimately into two categories: 1) tripping current-carrying
components; 2) changing bus power injection where load is regarded as a negative injection. If
we further divide the power injection at each bus into two parts, unchanging part and changing
part due to SPS operation, we can think of these two parts as two imaginary generators
connected to the bus independently. Therefore, all SPS operations can be uniformly represented
by the tripping effect for steady state reliability analysis.
All SPS operations including SPS failures along with their ultimate tripping effects can be
summarized into three cases as following. Firstly, for a desirable SPS operation, some
components of the power system trip as designed. These components are called intended
components hereafter. Secondly, for an undesirable SPS operation, i.e. the SPS operating
inadvertently, the intended components are also tripped down but undesirably. Thirdly, for the
case if the SPS fails to operate, the intended components do not trip initially. However, the
failure of SPS operation can only result in more severe and wider impacts on the power system
than if SPS operates as designed. As the ultimate tripping effect after possible transient stability
process, we can assume that more components of the power system including intended
components will trip eventually.
5.4.1

Modeling SPS operation at the component level

It is important to consider the reliability modeling of current-carrying components including

the impact of SPS operation. Including SPS operation at system level may be difficult or even
impossible if we do not incorporate this effect into the component modeling.
5.4.1.1 Modeling intended components

Suppose component i is an intended component of a SPS design. We realize that even if not
involved in the SPS design, component i can trip due to possible faults on it. Considering repair
of the component, we use a two-state Markov model to represent this effect as shown in Figure
39. The up and down states of the component are illustrated by i and i , respectively. The
parameters i and i in the figure are failure and repair rates of the component, respectively.
i

Figure 39: Modeling component i without SPS operation

92

When involved in SPS operation, we assume that there is no simultaneous fault on

component i when it is called upon for a SPS operation. Hence, for a SPS operation component
i is tripped to a state distinct from state i .
For a desirable SPS operation, we use i to represent the down state of component i after the
SPS operation. When the power system is returned to a normal operation condition later,
component i can be restored to its up state again. Thus, we add another two-state process
representing the desirable SPS operation effect to the previous one and the Markov model of
component i becomes the one as shown in Figure 40. The parameters i and i in the figure are
failure and repair rates of the component for desirable SPS operation, respectively.
i

Figure 40: Modeling component i with desirable SPS operation

For an undesirable SPS operation, we use i to represent the down state of component i after
the SPS operation. However, since the SPS operation is inadvertent and the power system is
actually in a normal operation condition, component i will be corrected back to its up state again
after some checking processes later. Therefore, we need a third two-state process representing
the undesirable SPS operation effect integrated into the previous modeling. The Markov model
of component i will evolve to the one as shown in Figure 41. The parameters i and i in the
figure are failure and repair rates of the component for undesirable SPS operation, respectively.
i

Figure 41: Modeling component i with desirable and undesirable SPS operation
As for the case when the SPS fails to operate, we have assumed that component i will be
~
ultimately tripped from the steady state viewpoint. Here we use i to represent this down state of
component i. However, because the power system has suffered more severe impacts and more
components have been tripped, we consider the system restoration in two steps. In the first step,
the extra components tripped due to SPS failing to operate other than intended components are
switched back to their up states. The second step is the same as if SPS did operate as designed,
i.e. the power system returns to a normal operation condition and intended components are
restored to their up states again.

93

Now the operation effect of SPS failing to operate can be illustrated by the Markov process
~
~
loop connecting states i, i , and i as shown in Figure 42. In this figure, the parameter i is the
~
failure rate of component i towards its eventual down state i caused by SPS failing to operate.
The parameter ~ is the switching rate of the first restorative step after the ultimate tripping
i

effect of SPS failing to operate. The parameter i is the switching rate of the second restorative
step, which is the same as the repair rate of component i for its desirable SPS operation as
defined in Figure 40 and Figure 41. In fact, Figure 42 is based on and further developed from
Figure 41.
i

i
i

~i

~
i

Figure 42: Modeling component i with all SPS operation effects

The Figure 42 depicts the overall reliability model for the intended component i including
all SPS operation effects. In summary, component i has five states: the up state i , the down state
i due to faults on component i , the down state i due to desirable SPS operations, the down
~
state i due to undesirable SPS operations, and the down states i due to SPS failing to operate,
~
respectively. If we use Pi , Pi , Pi , Pi , and Pi to represent the probabilities of states i , i , i , i ,
~
and i , respectively, we can figure out these probabilities using the frequency balance approach
as following.
For state i , we have
~
Pi i Pi i Pii Pi (i i i i )

(5.1)

For state i , we have

Pi i Pi i

For state i , we have

94

(5.2)

~
Pi i Pi ~i Pi i

(5.3)

For state i , we have

Pi i Pii

(5.4)

~ ~
Pi i Pi ~i

(5.5)

~
Pi Pi Pi Pi Pi 1

(5.6)

~
For state i , we have

We also have

Using any four of the five equations (5.1)-(5.5) together with equation (5.6), we can solve
and obtain the state probabilities as below.
Pi 1 K i

(5.7)

Pi i ( K i i )

(5.8)

~
Pi (i i ) ( K i i )

(5.9)

Pi i ( K i i)

(5.10)

~ ~
Pi i ( K i ~i )

(5.11)

wherein
~

i i i
Ki 1 i i
~
i
i
i i

(5.12)

5.4.1.2 Modeling extra components tripped due to SPS failing to operate other than
intended components

In the case of SPS failing to operate, the power system suffers more severe impacts and
some components other than intended components will be ultimately tripped from the steady
state viewpoint. Here we suppose component j represents the extra component. Since we have
assumed the system restoration in two steps, component j is switched back directly to its up
state in the first step. Thus, the process can be represented by a two-state Markov model, which
is illustrated in Figure 43.
Figure 43 shows the overall reliability situation for the extra component j including SPS
operation effects. Component j has three states: the up state j , the down state j due to faults
~
on component j , and the down states j due to SPS failing to operate, respectively. In this
~
~
figure, the parameter j is the failure rate of component j towards its eventual down state j
caused by SPS failing to operate. The parameter ~ is the switching rate of the first restorative
j

step after the ultimate tripping effect of SPS failing to operate. The parameters j and j in the
figure are failure and repair rates of the component due to faults on itself, respectively.

95

~ j
~

j
~
j
Figure 43: Modeling component j with SPS operation effects
~
~
If we use Pj , Pj , and Pj to represent the probabilities of states j , j and j , respectively, it is
also easy to calculate these probabilities using the frequency balance approach. The results are
obtained as follows.

Pj 1 K j

(5.13)

Pj j ( K j j )

(5.14)

~ ~
Pj j ( K j ~ j )

(5.15)

wherein
~

K j 1 j ~j
j j

(5.16)

5.4.1.3 Modeling components not involved in SPS operation

All other components not analyzed above are those not involved in SPS operation. Suppose
component k is such a component. It is obvious that component k is only influenced by its own
faults. Thus, we can model it the same way as if SPS does not exist, which is shown in Figure 44.
In Figure 44, the up and down states of component k are illustrated by k and k ,
respectively. The parameters k and k in the figure are failure and repair rates of the
component, respectively.
k

Figure 44: Modeling component k not involved in SPS operation

5.4.2

Modeling SPS operation at the system level

After we have incorporated SPS operation effects into the component modeling, it becomes
feasible to analyze reliability at the power system level including SPS operation. However, the
interactions between current-carrying components caused by SPS operations may raise the
96

complexity and discourage modeling at the system level. Hence, we need to decouple these
component interactions first. We still use i , j , and k to represent the intended components, the
extra components, and the components not involved in SPS operation, respectively.
5.4.2.1 Decoupling component interactions by SPS operation

We assume that the faults on a component are independent of those on other components.
Thus, the failure mode of a component due to faults on itself has influence only on it without
interaction with other components. It is also obvious that in case of desirable SPS operation or
undesirable SPS operation, only intended components are tripped without interaction with other
components. Hence, the component interactions exist only in the case of SPS failing to operate
when called upon.
We have analyzed and modeled the extra components tripped due to SPS failing to operate
other than intended components in the previous section. However, this tripping effect is actually
not independent. It is always accompanied by the ultimate tripping effect of the intended
components from the steady state viewpoint. Considering these two types of components
together, we can see that their ultimate tripping effect is actually a kind of common cause failure.
For clear illustration, we extract out the related parts in Figure 42 of component i and in Figure
43 of component j, and then put them together with little modification as shown in Figure 45.
The common cause failure process can be explained as following.

Component
UP

i j

~
i

Component
DOWN

~
j

~i ~ j

Figure 45: Decoupling interactions caused by SPS failing to operate

In the first transition of this process, component i transfers from the up state i to the down
~
~
state i with the failure rate i , while component j transfers from the up state j to the down
~
~
~
state j with the failure rate j , which equals to i in value due to common cause failure of the
two components. The second transition of this process is actually the first restorative step after

97

~
SPS failing to operate as mentioned previously. Component i transfers from state i to another
~
down state i with the switching rate ~i , while component j returns from state j to its up state
j with the repair rate ~ j , which is the same as ~i due to the switching operation. The third and
last transition of the process is actually the second restorative step after SPS failing to operate.
Component i transfers from state i to its up state i with the switching rate i , which is the
same as the repair rate of component i for its desirable SPS operation. Therefore, if we are only
concerned with transitions between up and down states of component j , this kind of common
cause failure makes the extra component j experience a two-state process with its transition
rates derived directly from the intended component i .
5.4.2.2 Impact of SPS operation on modeling non-contingency system states

SPS operations do have influence on modeling system states. However, for system states in
which the intended components are already down due to their own faults, SPS cannot be put into
operation. Hence, there will be no change for modeling these states as if without SPS. The
system state (i j k ) in Figure 46 is such a case, which represents the intended
component i already being down due to faults on itself, the extra component j being up, and
component k not involved in SPS operation also being up.
i j k

i j k

i
i
i j k
Figure 46: Impact of SPS operation on modeling non-contingency system states
If the intended components are in their up states, SPS can be alarmed and put into operation.
Nevertheless, the impact of SPS operation on modeling system states is still dependent upon
whether the states are in contingency or not.
For non-contingency system states, SPS is not designed to operate. The case of desirable SPS
operation will not occur and so does the case of SPS failing to operate. But SPS could operate
mistakenly, i.e. undesirable SPS operation could exist. Figure 46 is the illustration of the impact
of SPS operation on modeling non-contingency system states. In this figure, the system state
(i j k ) is such a non-contingency state that components i , j , and k are all in up states.
We already know that undesirable SPS operations only influence intended components without
interaction with other components. Thus, a new system state (i j k ) with connection to
this non-contingency state is added, in which i represents the intended component i being
tripped down due to undesirable SPS operation.

98

5.4.2.3 Impact of SPS operation on modeling contingency system states

For contingency system states, SPS is designed to operate. Hence, inadvertent SPS operation,
i.e. undesirable SPS operation will not occur. So, there are two types of SPS operation effects
relating to this situation, i.e. desirable SPS operation and the case of SPS failing to operate.
Figure 47 is the illustration of the impact of SPS operation on modeling contingency system
states. In this figure, the system state (i j k ) is a contingency state with components i
and j being up but component k being down.
k

i j k

i j k

i j k

i
i
i j k

~i

i
~ ~
i j k

Figure 47: Impact of SPS operation on modeling contingency system states

Although desirable SPS operations only influence intended components without interaction
with other components, it is always practical to restore the power system to a normal or noncontingency state rather than the original contingency state after desirable SPS operations.
Therefore, after the system transfers from the contingency state (i j k ) to a new state
(i j k ) representing the desirable SPS operation, it will not go back to state
(i j k ) again as component i being restored to its up state. Instead, we assume the
system will be restored to a non-contingency system state which is in the nearest connection but
has more components up than the original contingency state (i j k ) . Suppose the state
(i j k ) is one of such non-contingency system states. The three system states
(i j k ) , (i j k ) , and (i j k ) form a Markov loop as shown in Figure
47.

For the case of SPS failing to operate, we previously analyzed the existing common cause
failures and decoupled the interactions between current-carrying components. Based on the
information, we know that the system will transfer from the contingency state (i j k ) to
~ ~
a new state ( i j k ) representing the ultimate tripping effect of SPS failing to operate
from the steady state viewpoint as shown in Figure 47. Then as the first restorative step, the
system will transfer to the state (i j k ) representing the desirable SPS operation. The
second restorative step is all the same as the restorative process of the desirable SPS operation,
i.e. the system is restored to the non-contingency state (i j k ) . The four system states
~ ~
(i j k ) , ( i j k ) , (i j k ) , and (i j k ) also form a Markov
loop as shown in Figure 47.

99

5.4.2.4 Reliability modeling of power systems with impact of SPS operation

We have assumed the failure mode of a component due to faults on itself is independent.
Therefore, the reliability modeling of power systems without SPS operation is a one-layer
Markov chain as illustrated in Figure 48. In this figure, each block represents a system state and
all system states are independent of each other. In addition, all transitions between any two states
are single-step transitions.

i j k

i j k

i j k

Figure 48: Reliability modeling of power systems without SPS operation

However, if the impact of SPS operation needs to be considered, the reliability modeling of
power systems will become a two-layer Markov chain as shown in Figure 49. The first layer is
all the same as that in Figure 48. We name it the primary layer, which consists of the
independent system states with single-step transitions. The second layer is a dependent one
added to the primary layer to reflect the impact of SPS operation. As we have explained in
previous content, different system states correspond to different SPS operation effects.
For a system state in which the intended components are already down due to their own
faults, SPS cannot be put into operation. Hence, there will be no dependent second-layer state
attached to this system state, e.g. (i j k ) .

100

For a non-contingency system state with the intended components up, the tripping effect of
undesirable SPS operation exists. Thus, there will be a dangling second-layer state attached to
this system state, e.g. (i j k ) .

i j k

i j k

i j k

Figure 49: Reliability modeling of power systems with impact of SPS operation
For a contingency system state with the intended components up, there exist two types of
SPS operation effects, i.e. desirable SPS operation and that of SPS failing to operate. Therefore,
there will be a group of two second-layer states attached to this system state, e.g. (i j k ) .
In addition, this second-layer group also attaches to a non-contingency system state, e.g.
(i j k ) , which is in the nearest connection with but has more components being up than
the original contingency state, i.e. (i j k ) .
The methodology of power system reliability modeling with impact of SPS operation can be
summarized as following.
1) Set up the primary layer Markov chain.
2) Classify system states of the primary layer.
3) Attach the second-layer states to non-contingency states of the primary layer.
4) Attach the second-layer states to contingency states of the primary layer.
5) Use analytical method or Monte Carlo Simulation for reliability calculation.

101

Consideration of SPS in planning

This chapter considers SPS within the context of long-term planning. With the background of
fast growing wind energy in power system, SPS applications motivated by wind energy becomes
increasingly interesting in long-term planning. This topic is discussed in section 6.1.
High penetration of SPS increases the complexity of system planning and operation. The
complexity increases the possibility of undesirable and unintended interactions among SPS,
potentially degrading to system reliability. In this chapter, section 6.2 conceptually discusses the
increase in operational complexity with higher levels of SPS penetration.
It is typical that reliability problems encountered within planning studies may be corrected
either by building new transmission or by installing an SPS. SPS are almost always much less
expensive ways to correct reliability problems, since they generally require only relays,
communication equipment, computing and associated actuation logic, and tripping devices.
Section 6.3 presents a framework to address the problem of SPS-aided transmission expansion
planning in power system, and section 6.4 discusses the results.
6.1
6.1.1

Electric system long-term planning with SPS

Electric system long-term planning

The electric system planning process is the systematic assembly and analysis of information
about electric energy supply, transport, and demand, and the presentation of this information to
decision-makers who must choose an appropriate course of action. The composite power system
expansion planning usually is developed by reliability justification or economic justification.
Traditional system expansion planning is reliability based.
Reliability evaluation has been segregated into hierarchical levels HL-I (generation only),
HL-II (generation and transmission), and HL-III (generation, transmission, and distribution),
where the last is normally addressed by assuming the generation and transmission sides are
perfectly reliable. The generation and transmission adequacy of composite system at hierarchical
level HL-II is generally evaluated with the loss of load probability (LOLP) and the loss of load
expectation (LOLE) [133].
In generation and transmission adequacy assessment, power system uncertainties, including
load uncertainty, generation outage and transmission system contingency are modeled to
calculate the reliability indices. Practically, there are two ways to perform the reliability
evaluation: enumeration and Monte-Carlo simulation. Monte Carlo simulation, in theory, can
include effects of all possible events on system. The required number of samples is independent
of the size of the system to maintain a given accuracy level. Monte Carlo simulation is more
flexible when complex operating conditions are incorporated for large-scale system. Reference
[134] gives a thorough description of reliability evaluation for generation and transmission
planning procedures.
Wind energy is growing fast and becoming a major portion of energy portfolio in many
states. The federal government committed to supply 20% or more of the nations electric energy
needs from wind energy resources by 2030 [ 135 ]. Identifying what type and location of
transmission is required to effectively integrate wind power is an important planning issue. It is a
long term decision which involves decision criteria in terms of security, economy and

102

environment. Generation and transmission expansion for wind power integration rely on adding
generation and transmission capacity to support transmission and reserve margin requirements
when confronted with increasing wind power. To perform the planning procedure for wind
power integration, the wind power uncertainty is typically characterized by a probability
distribution. Then by using Monte-Carlo simulation, the reliability indices are calculated and
generation and transmission capacity requirements can be chosen.
6.1.2

SPS planning related to wind

In the US, generation is growing 4 times faster than transmission, and the transmission
capacity growth rate has been ~1/3 of peak load. Transmission upgrades are considered to be
expensive and is constrained by time, land and environment. In Texas, which has about 25% of
U.S. wind power, significant growth in 2008 pushed generation past transmission capacity by
65% by the end of that year [136]. In the near future, utilities will need devices such as SPS to
trip non-priority generation since if transmission does not grow at the required pace. Thus, SPS,
mainly in the form of wind generation rejection, is used to facilitate meeting reliability
requirements while interconnecting wind power to the grid.
6.1.3

Accounting for SPS in planning

Special protection scheme, different from conventional local protection relay to isolate the
faulted elements, is a wide area protection scheme. SPS uses wide area measurements to observe
system behaviors [137], and takes wide area countermeasures to avoid unstable and unusual
stresses on power systems. SPS actions include, among others, changes in load, generation, or
system configuration to maintain system stability, acceptable voltages or power flows. Most of
SPS are implemented by feed forward control. With feed forward control, the effect of
disturbances or contingencies must be predicted accurately, and there must not be any
unmeasured disturbances. Otherwise, the unmeasured disturbances will result in unintended
action by SPS. Once a control signal has been sent by feed forward control in SPS, it cannot be
further adjusted; any corrective adjustment must be by way of a new control signal. The
reliability of SPS on precise control thus highly depends on the original control logic design.
The benefit of SPS is that it is a relatively inexpensive way to expand secure operation
conditions. The cost of SPS is that its implementation results in exposure to additional failure
modes. In generation and transmission expansion planning, potential unintended events
corresponding to these additional failure modes can be predicted, assessed, and considered
within a planning decision framework.
6.2

SPS interaction

The topic of SPS interaction and the proposed ways to identify such fault modes and their
reliability indices using Monte Carlo simulation were presented in chapter 5. In this section,
further insights into it in the light of planning is presented, especially weighing the issues related
to high SPS penetration with respect to high cost transmission upgrade solution. The undesirable
interactions among SPS degrade the reliability of system operation. Two major sources trigger
undesirable and unintended interactions among SPS. One is failure of SPS. Another is faulty
design logic of SPS.
Typically, SPS involves input (measurements), decision making system and action. Each
component has failure rate though usually the failure rate is very low. The component failure of
103

SPS may result in interactions with other SPS. For example, a mis-operation of a generation
rejection scheme which mis-trips a generation unit may result in a under frequency load
shedding. Thus, a reliable SPS requires an appropriate level of redundancy. To enhance the
reliability of SPS itself, the following NERC standards apply to SPS design [17]:

SPS shall be designed so that cascading transmission outages or system instability do not
occur for failure of a single component of an SPS, which would result in failure of the SPS to
operate when required.

Mis-operation, incorrect operation, or unintended operation of an SPS when considered by

itself and not in combination with any other system contingency shall meet the system
performance requirements as defined under NERC Planning Standards on Transmission
Systems.

All SPS installations shall be coordinated with other system protection and control schemes.

All SPS operations shall be analyzed for correctness and documented.

Fault in SPS design logic is another source for undesirable interaction. Taking generation
rejection scheme as an example, when the system is stressed with heavy loading, after a
generation rejection scheme takes action to depress a stress of overload, the generation
reschedule changes the power flow may result in overload on another transmission line and trip
another generator by SPS, and result in cascading trip of generators if the SPS are not designed
properly. It is difficult to design a backup scheme to avoid the incorrect logic for high
penetration SPS. In contrary, transmission network upgrade can release the stress of operation
thoroughly.
In Chapter 4, the reliability model of SPS was presented. To analyze the impacts of SPS
interaction to system reliability, the SPS reliability model should be modeled in the simulation.
By Monte Carlo simulation, the impacts can be calculated and associated with reliability indices.
6.2.1

Complexity for high penetration SPS

The interactions among SPS discussed above increase the complexity of power system
planning, particularly for high penetration of SPS. Comparing with transmission planning, SPS
planning is relatively new in power industry. Analysis on complexity of high penetration SPS is
meaningful for power system planning.
According to Cambridge Advanced Learner's Dictionary, the definition of the word
complexity is when something has many parts and may be difficult to understand or find an
answer to. In the definitions of complexity in science and technology, it often used to describe a
system with numerous elements which have intricate relationships among them. Warren Weaver
categorized complexity as disorganized complexity and organized complexity [ 138 ] which
influenced contemporary thinking about complexity. For the interactions under high penetration
of SPS, it is a problem in which the number of elements is large, and the behavior of interactions
triggered by the chain of unpredictable events associated with a contingency is difficult to
manage or perhaps totally unknown which is characteristic of disorganized complexity.
Conceptually, the relation between the number of SPS and system operational risk is shown
in Figure 50.

104

Operational Risk

1. No Action
2. Install SPS
3.Transmission Upgrade

100GW

4. Combine 2&3
90GW

80GW
Acceptable Risk

70GW

B
A

60GW
50GW

Number of SPS
# of SPS Limit

Miles of new
transmission line

Figure 50: Complexity of SPS

Illustrated in Figure 50, when the system load increases, for example, from 50GW to 100GW
gradually shown by the dotted line, the system becomes increasingly stressed. Without any
action, as shown by the green line, the operational risk is increasing and becomes unacceptable
when the load demand is increased beyond 80GW as shown at point A in Figure 50. With SPS,
the operational risk can be reduced. But as the number of SPS increases, the operational risk
associated with the undesirable interactions among SPS increases. Without transmission upgrade,
utilizing more SPS reaches the limit of acceptable operational risk at a load of 100 GW, as
shown at point B. On the other hand, use of transmission network upgrades enable operation
within the acceptable risk limit for loading levels well beyond 100 GW, as shown by the dark
blue line in the figure. But use of transmission upgrades only is an expensive approach. A
practical way is to combine the two approaches, SPS installation and transmission network
upgrade, to achieve secure operation of the system. In Section 6.2.2 below, a small system case
study is used to illustrate the limit of SPS, the number of events increasing exponentially and
more lost load occurring with the increasing number of SPS.
6.2.2

Illustration

Small system for overload-motivated wind plant rejection

Assume three 100 MW wind plants are built in sequence A, B, and C, as shown in Figure 51.
Branch data for the system on a 100 MVA base are given in Table 15.

105

100
C
5: B-C
-j5
(60)

100

B
1: O-B
-j4.33
(120)

3; A-B
-j6.67
(75)
2: O-A
-j8
(190)

4: A-C
-j2.86
(60)

100
A

Figure 51: Small system for illustrating concepts

Table 15: Branch data for sample system
Line
O-A
O-B
A-C
B-C
A-B

Reactance
0.125
0.231
0.35
0.2
0.15

Susceptance
8
4.33
2.86
5
6.67

Matlab code was written to perform DC power flow analysis on this system, assuming Bus O
as the reference bus. Assessment was done for various combinations of wind plants on-line, line
outages, and wind plants tripped based on SPS. Results are summarized in Table 16. We can
notice from the simple illustration that the chances of interactions among SPS increases under
certain contingencies. Such Monte Carlo simulation based on the approach presented in chapter 5
can be done for many loading conditions for a system with many SPS, to identify the undesirable
consequences and plan to mitigate them.
All post-contingency overload problems may also be corrected without SPS by building one
transmission line with impedance 0.111 and capacity 182 MW from node C to node O, as shown
in Figure 52.

106

Table 16: SPS designs for different wind plant growth stages
Wind
plants
operating

A
A
A
A,B
A,B
A,B
A,B
A,B
A,B,C
A,B,C
A,B,C
A,B,C
A,B,C
A,B,C
A,B,C
A,B,C
A,B,C

Actuating
line
outages

none
O-B
O-A
none
O-B
O-B
O-A
O-A
none
O-B
O-B
O-A
O-A
A-C
A-C
B-C
B-C

Gen
trip

Post-contingency line flows

(percent of normal rating)

none
none
none
none
none
B
none
A
none
none
B,C
none
A,C
none
C
none
B,C

Events resulting in lost load

Lost
load,
SPS
works

1:
O-B
.22
.65
-

2:
O-A
.39
.53

3:
A-B
-.28
0

4:
A-C
-.09
0

5:
B-C
0.09
0

.64
1.05
.53

.23
1.05
0
0

.08
.36
0
0

1.0
2.5
.83
1.27
.75
1.39
.33

.95
1.58

.46
1.71
0.53
-1.4
0
1.10
.41
.65
-.16

.76
1.19
0
.12
0
1.67
0

.96
.69
1.31
.46

Lost
load,
SPS
fails

Inadvertent SPS operations

following first trip
Event
Event

-.08
-.36
0
0

NA
NA
NA
NA
NA
100
NA

NA
NA
NA
NA
NA
200
NA

What
trips
NA
NA
NA
NA
NA
A
NA

.90
.48
0
1.55
0
1.67
0
-

NA
NA
200
NA
200
NA
100
NA
200

NA
NA
300
NA
300
NA
300
NA
300

NA
NA
A
NA
B
NA
A
NA
A

Lost
load
NA
NA
NA
NA
NA
100
NA
NA
NA
100
NA
100
NA
100
NA
100

What
trips
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
B
NA
NA

Lost
load
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
100
NA
NA

100
5: B-C
-j5
(60)

100

B
1: O-B
-j4.33
(120)

C
4: A-C
-j2.86
3; A-B (60)
-j6.67
(75)

2: O-A
-j8
(190)

6: O-C
-j9
(182)
100
A

Figure 52: Small system with transmission upgrade

Although construction of the line would be expensive, there is significant benefit in two main
ways:

107

6.3

Reduced level of complexity: The overall system is simpler to operate and maintain.
Without SPS,
o there is no need to perform periodic maintenance, including self-checks for
operability, to ensure integrity of sensing, communication, and actuation;
o there is no need to monitor arming of schemes;
o there is no anxiety felt by individuals responsible for ensuring the integrity of the
SPS and of the power system being operate
More reliable system: Without SPS,
o there is no possibility of lost load due to an inadvertent trip during normal
conditions;
o there is no tripping for the actuating line outages (this refers to the lost load when
SPS works in Table 16)
o there is no possibility of lost load due to SPS failure
o there is no possibility of lost load due to inadvertent SPS operations following a
first-trip
Incorporating SPS in generation and transmission expansion planning

The electric grid has been undergoing rapid growth in load. Also more and more generation
especially renewable sources are being connected to the grid. However, limited transmission
poses a problem to this growth, which if not taken care of will lead to the electric grid not being
able to meet demand and accommodate cheaper generating resources. SPS are economically
cheaper than transmission expansion and could help alleviate the problem of ever increasing load
and aggressive renewable generation installation.
6.3.1

Case study

The system used is a 5 bus system with 4 generators as shown in Figure 53. This system is
used to illustrate and solve a 15 year planning problem that incorporates SPS and transmission
expansion plans into achieving demand and generation expansion plans from 2015-2029. The
goal is to minimize production costs and investments costs while also minimizing operational
complexity that comes with increasing number of SPS. The loads at Bus 2, Bus 3, and Bus 4 are
assumed to be growing at 3%, 4%, and 3.5% respectively as shown in Table 18. Table 19
presents the test system branch data. Table 20 shows the generator cost data, and Table 21
presents the generation expansion plans through the various planning period.
Table 17: Time periods
Periods Time Periods
1
2015 - 2019
2
2020 - 2024
3
2025 - 2029

108

Figure 53: 5-bus system

Table 18: Forecasted load

Bus\Year 2010 2015
2020
2025
Bus 2
500 579.65 671.95
779
Bus 3
400 486.68 592.08 720.36
Bus 4
350 415.70 493.71 586.36
Table 19: Branch data
Line Reactance Capacity
1-2
0.2
90MW
1-3
0.25
90MW
2-3
0.1
800MW
2-5
0.1
800MW
3-4
0.1
800MW
3-5
0.1
800MW
4-5
0.1
800MW

109

Generator cost curves are assumed to be in a quadratic form,

z a bP cP 2

(6.1)
Table 20: Generator data

Generator c co-efficient b co-efficient Capacity at 2015

G1
0.09
20
80MW
G3
0.25
20
800MW
G4
0.20
40
800MW
G5
0.01
40
800MW
Table 21: Generation expansion plans
Generator (G1)
G1
G1
G1
6.3.2

Year
Expansion plans
2015 80MW to 200MW
2020 200MW to 300MW
2025 300MW to 400MW

Transmission expansion options and specifications

Options

Build line from bus 1 to bus 2

Build line from bus 1 to bus 3

Build double lines from bus 1 to bus 2, and bus 1 and bus 3

Specifications

Line capacity should vary between 50MW to 200MW

SPS cannot be used on a new line

Operational constraints

6.3.3

Only one generating unit should be tripped at a given time

Not more than two lines can be tripped at the same time

Transmission expansion tree

Transmission expansion tree is a tree that shows possible actions (i.e., transmission only or
SPS-aided transmission expansion options) that can be taken at different periods to accommodate
more generation expansion plans and load growth, as shown in Figures 54 and 55. We found 27
possible solutions that include transmissions only solution as well as many SPS-aided
transmission expansion options, as shown in Table 22 that is continued over 3 pages. In Table 22,
for instance build 1-23 means building third line from bus 1 to bus 2, where the original line is
names as 1-21.
110

Input planning year, load and

system parameters

Starting from the line building

option bus 1-bus 2, compute the
impedance necessary for 50MW
Increase rating by 10MW and
compute new impedanece

Compute lodf and

optimal power flow

Apply contingency

Does new line survive all N-1

contingency and does system obey
all system operating conditions?

No

Yes

Search among all possible planning options that meet the

load demand, with minimum production costs

Then go to the next possible line connection (bus 1bus 3) and double line building option (bus 1-bus 2
and bus 1-bus 3)
Figure 54: Transmission tree expansion- numerical optimization

111

Figure 55: Transmission tree

Table 22: Possible solutions
Year
X1

X2

X3

X4

X5

X6

2015

2020

Build 1-22 100MW

X =0.18
Build 1-32 50MW
X= 0.45; NO SPS
Build 1-22 100MW
X =0.18
Build 1-32 50MW
X= 0.45
NO SPS
Build 1-22 100MW
X =0.18
Build 1-32 50MW
X= 0.45
NO SPS
Build 1-22 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-22 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-22 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

2025

Build 1-23 60MW

X = 0.30
Build 80MW 1-33
X= 0.2813; NO SPS
Build 1-23 70MW
X=0.2571
Put SPS on 1-31
Put SPS on 1-32
2 SPS
Build 90MW 1-33
X= 0.25
Put SPS on 1-21 Put
SPS on 1-22
2 SPS
Build 100 MW 1-23
X=0.1636
Put SPS on 1-31
1 SPS

Build 1-24 50MW

X=0.36
Build 1-34 90MW
X= 0.25; NO SPS
Build 1-24 90MW
X= 0.36
Put SPS on 1-31
Put SPS on 1-32
2 SPS
Build 90MW 1-23
X=0.20
Build 80MW 1-34
X= 0.2813
NO SPS
Build 100 MW 1-24
X=0.1636
Put SPS on 1-31
1 SPS

Build 100 MW 1-23

X=0.1636
Put SPS on 1-31
1 SPS

Build 150MW 1-24

X=0.12
Build 200MW 1-32
X=0.1125
NO SPS
Build 90MW 1-23
X=0.20
Put SPS on 1-31
Put SPS on 1-32
2 SPS

Build 130MW 1-32

X=0.1731
Put SPS on 1-21
Put SPS on 1-22
Put SPS on 1-31
3 SPS

112

Year

2015

2020

X7

Build 1-22 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

Build 130MW 1-32

X=0.1731
Put SPS on 1-21
Put SPS on 1-22
Put SPS on 1-31
3 SPS

X8

Build 1-22 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

X9

Build 1-22 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-22 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-22 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

Build 130MW 1-32

X=0.1731
Put SPS on 1-21
Put SPS on 1-22
Put SPS on 1-31
3 SPS
Build 120MW 1-32
X=0.1875
Build 70MW 1-23
X=0.2571
NO SPS
Build 120MW 1-32
X=0.1875
Build 70MW 1-23
X=0.2571
NO SPS
Build 120MW 1-32
X=0.1875
Build 70MW 1-23
X=0.2571
NO SPS

X12

Build 1-22 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

Build 120MW 1-32

X=0.1875
Build 50MW 1-23
X=0.36
1 SPS

X13

Build 1-22 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

Build 120MW 1-32

X=0.1875
Build 50MW 1-23
X=0.36
1 SPS

X14

Build 1-32 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

X15

Build 1-32 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

Build 160MW 1-22

X=0.1125
Put SPS on 1-21
Put SPS on 1-31
Put SPS on 1-32
3 SPS
Build 160MW 1-22
X=0.1125
Put SPS on 1-21
Put SPS on 1-31
Put SPS on 1-32
3 SPS

X10

X11

113

2025
Build 140MW 1-33
X=0.1607
Put SPS on 1-21
Put SPS on 1-22
Put SPS on 1-31
Put SPS on 1-32
4 SPS
Build 80MW 1-23
X=0.225
Build 90MW 1-33
X=0.25
NO SPS
Build 50MW 1-24
X=0.36
Put SPS on 1-31
Put SPS on 1-32
2 SPS
BUILD 100MW 133
X=0.225
NO SPS
BUILD 100MW 133
X=0.225
BUILD 50MW 1-24
X=0.36
NO SPS
Build 70MW 1-24
X=0.2571
Put SPS on 1-31
Put SPS on 1-32
2 SPS
BUILD 100MW 133
X=0.225
BUILD 50MW 1-24
X=0.36
NO SPS
Build 90MW 1-23
X=0.2
Put SPS on 1-31
Put SPS on 1-32
2 SPS
Build 120MW 1-33
X=0.1875
Put SPS on 1-21
Put SPS on 1-22
2 SPS

Year

2015

2020

X16

Build 1-32 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

X17

Build 1-32 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-32 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-32 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-32 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-32 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-32 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-32 120MW
X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

X18

X19

X20

X21

X22

X23

Build 160MW 1-22

X=0.1125
Put SPS on 1-21
Put SPS on 1-31
Put SPS on 1-32
3 SPS
Build 170MW 1-22
X=0.1059
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 170MW 1-22
X=0.1059
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 170MW 1-22
X=0.1059
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 190MW 1-22
X=0.0947
Put SPS on 1-21
1 SPS
Build 190MW 1-22
X=0.0947
Put SPS on 1-21
1 SPS
Build 190MW 1-22
X=0.0947
Put SPS on 1-21
1 SPS

2025
Build 100MW 1-33
X=0.225
Build 50MW 1-23
X=0.36
NO SPS
Build 90MW 1-23
X=0.2
Put SPS on 1-31
Put SPS on 1-32
2 SPS
Build 120MW 1-33
X=0.15
Put SPS on 1-21
Put SPS on 1-22
2 SPS
Build 100MW1-33
X=0.225
Build 50MW 1-23
X=0.36
NO SPS
Build 90MW 1-23
X=0.2
Put SPS on 1-31
Put SPS on 1-32
2 SPS
Build 120MW 1-33
X=0.15
Put SPS on 1-21
1 SPS

Build 110MW 1-33

X=0.2045
Put SPS on 1-21
1 SPS

Build 100MW1-33
X=0.225
Build 50MW 1-23
X=0.36
NO SPS
Build 190MW 1-22
X=0.0947
Put SPS on 1-21
1 SPS

X24

Build 1-32 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

Build 110MW 1-33

X=0.2045
Put SPS on 1-21
1 SPS

Build 120MW 1-34

X=0.1875
Put SPS on 1-21
1 SPS

X25

Build 1-32 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

Build 110MW 1-33

X=0.2045
Put SPS on 1-21
1 SPS

Build 50MW 1-34

X=0.45
Build 180MW 1-22
X=0.1
No SPS

114

Year

6.3.4

2015

2020

2025

X26

Build 1-32 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

Build 150MW 1-22

X=0.12
Build 50MW 1-33
X=0.45
NO SPS

Build 70MW 1-34

X= 0.3214
Put SPS on 1-21
Put SPS on 1-22

X27

Build 1-32 120MW

X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS

Build 150MW 1-22

X=0.12
Build 50MW 1-33
X=0.45
NO SPS

Build 1-23 50MW

X=0.36
Build 1-34 50MW
X=0.45
NO SPS

Operational complexity

SPS are economically cheaper but their increase can actually increase operational complexity
of the electric grid. In regards to power system maintenance, SPS increases operation complexity
because with the increased transmission system utilization that comes with the application of
SPS, planned transmission outages may become more difficult to schedule. In the area of
protection co-ordination, coordinating multiple SPS with other protection systems causes coordination complexity. With regards to mis-operation, mis-operation of one SPS could bring
about serious cascading consequences. Therefore, it is necessary to come up with a metric that
can measure how much complexity we bring into the system by using SPS, and use that metric as
one of the decision criteria at the planning stage.
An operational complexity metric proposed in this report is the total number of states of SPS
failure modes that will be encountered in reliability studies. The states will increase as the
number of SPS in the system increases, which will indicate the possibility of increasing
operational complexity. For instance, one SPS has at least 3 states:
1. No failure mode
2. Will fail to operate
3. Will inadvertently operate
Likewise,

2 SPS will have 8 states

3 SPS will have 27 states

4 SPS will have 64 states

20 SPS will have 8000 states

50 SPS will have 125000 states

For example, in option X16 of Table 22,

For 2015 - Build 120MW 1-32 and put SPS on 1-21 and 1-31 (2 SPS)

For 2020 - Build 160MW 1-22 and put SPS on 1-21, 1-31, 1-32 (3 SPS)

For 2025- Build 100MW 1-33 and 50MW 1-23 (no SPS)

115

Based on our definition, we have a total complexity of 2^3 + 3^3 + 0^3 = 35 states.
6.3.5

Production costs

Production cost refers to the operational costs accompanied with producing electric energy.
The optimized production cost for a planning option is computed as,

Min Pr oCij * Pi *8760

j

(6.2)

where,
ProCij is the production cost for ith loading scenario of jth year,
Pi is the probability of ith loading scenario of jth year.
The optimization is subject to generator capacity constraint, transmission capacity constraint,
power balance, and network flow constraints. The system loading scenarios used for computing
production cost of various planning options are represented in terms of percentage of peak load.
The probabilities of load profile shown in Table 23 were estimated from MISO load duration
curve [139].
Table 23: Probabilistic load profiles

6.3.6

Loading scenarios

Probability

40%
45%
50%
55%
60%
65%
70%
75%
80%
85%
90%
95%
100%

0.0297
0.0191
0.0894
0.1084
0.1998
0.2317
0.145
0.097
0.0365
0.0228
0.0114
0.008
0.0034

Economic factors

6.3.6.1 EENS costs

EENS is the expected energy not served due to forced curtailment actions by SPS. The cost
of a particular occurrence is computed by multiplying the associated EENS and a coefficient
called the value of lost load (VOLL).

EENS

jLC

kj

D j fi

(6.3)

where,

116

Lkj is load curtailed at bus k due to contingency j,

Dj is duration (hours) of load curtailment due contingency j, and

fi is frequency of occurrence of outage j.

EENS cos t EENSi * VOLLj

(6.4)

where,
k is number of years,
EENSCOST is Expected Energy not served due to forced curtailments of SPS actions,
EENSi is Expected Energy not served due to forced curtailments of SPS actions ith option, and
VOLLj is value of load lost at period j.
6.3.6.2 System interruption costs

This is the difference between post-contingency production costs and pre-contingency

production costs due to SPS action of tripping a cheaper generator and employing a more
expensive generator.

SysInt cos tij ( PostC Pr odij Pr eC Pr odij ) * DTG *(1.05) k

j

(6.5)

where,

k is 0, 1, 2, 3, 4 (number of years in time period)

j is (2015, 2020, 2025)
SysIntcostij = System interruption cost associated with ith option at period j.
PreCProdij = Pre-contingency production cost associated with ith option at period j
PostCProdij = Post-contingency production cost associated with ith option at period j
DTG = Down time of tripped generator
6.3.6.3 Investment costs

There is cost associated with the investment in SPS and transmission expansion projects,
with their inflation rate assumed to be about 5%.
6.3.6.4 Congestion rent

Congestion rent is the difference between the price of electricity at the point of delivery and
at the point of generation. Congestion rent is also called the re-dispatch cost because it is the
extra cost expended for dispatching more expensive generators that would be needed if the
transmission system had enough capacity and did not constrain power transfer.
6.3.7

Study assumptions

The cost of transmission upgrade is assumed to be $600,000/MW. It is also assumed

that the cost of materials for building transmission lines has an inflation rate of 2%.

117

The cost of SPS is assumed to be $1,000,000. It is also assumed to have an inflation

rate of 2%. All the costs are expressed in present value.

The congestion rent is the difference between constrained and unconstrained

production costs. The production costs are computed for different loading scenarios
for 15 years which includes three 5-year planning intervals.

Similarly, EENS cost is computed for 15 year planning period. For options with SPS,
EENS is computed from the forced curtailments. In this study, the probability of each
contingency that enables SPS to operate is assumed to be 0.1. The forced load
curtailment is assumed to be the generator output of G1, due to load-generation
balance. The duration for lost load is assumed to be 3hrs. The VOLL (Value of load
lost) for the 3 planning intervals are given in Table 24.
Table 24: Different periods and associated value of load lost
Period
VOLL
$300/MWh
1
$500/MWh
2
$800/MWh
3

Table 25 presents a comparison of different economic factors for various possible planning
options comprising of transmission only and SPS aided transmission expansion options.

118

Table 25: Comparison of economic factors for planning options

Opti
ons

Transmission
costs

SPS
costs

Congestion
rent

Interruption
Costs

EENS
(MWh)

EENS
Costs

Com
plexity

X1

$285,140,400

$0

$0

$0

$0

X2

$202,198,200

$1,804,138

$3,274,100

7800

$4,809,0000

16

X3

$273,959,400

$0

$666,050

1800

$90,000

X4

$211,386,000

$14,347,00
0

$3,543,100

7800

$5,160,000

14

X5

$394,236,000

$0

$1,590,900

4050

$1,845,000

11

X6

$223,945,800

$0

$2,947,100

7050

$4,785,000

41

X7

$260,515,800

$0

$1,353,100

3450

$1,905,000

99

X8

$282,457,800

$0

$821,710

2250

$945,000

35

X9

$234,437,400

$0

$1,839,000

4500

$3,150,000

16

X10

$271,007,400

$0

$244,900

900

$54,000

X11

$307,577,400

$0

$244,900

900

$270,000

X12

$235,816,200

$0

$2,031,200

4950

$3,375,000

19

X13

$294,328,200

$0

$437,170

1350

$495,000

11

X14

$243,819,600

$0

$2,415,800

5850

$3,825,000

43

X15

$265,761,600

3450

$1,905,000

43

X16

$287,703,600

X17

$250,444,200

X18

$272,386,200

X19

$294,328,200

X20

$263,693,400

X21

$285,635,400

X22

$307,577,400

X23

$283,836,600

X24

$232,638,600

X25

$313,092,600

X26

$255,690,000

X27

$277,632,000

$2,208,
200
$2,208,
200
$2,000,
000
$2,000,
000
$4,323,
100
$4,323,
100
$3,104,
100
$4,438,
000
$2,000,
000
$2,000,
000
$3,219,
000
$2,000,
000
$3,104,
100
$4,323,
100
$3,104,
100
$3,219,
000
$3,219,
000
$2,000,
000
$4,323,
100
$2,000,
000
$2,000,
000
$2,000,
000
$2,000,
000
$3,219,
000
$4,323,
100
$2,000,
000

$0

$1,353,100

$0

$821,710

2250

$945,000

35

$0

$2,223,500

5400

$3,600,000

24

$0

$1,160,800

3000

$1,680,000

24

$0

$629,440

1800

$720,000

16

$0

$2,031,200

4960

$3,375,000

19

$0

$702,840

1950

$975,000

14

$0

$437,170

1350

$495,000

11

$0

$1,279,700

3300

$1,650,000

14

$52,911

$2,342,400

5700

$3,570,000

14

$0

$1,014,000

2700

$1,170,000

11

$0

$776,250

2700

$1,230,000

16

$0

$244,900

900

$270,000

119

Budget
$285,140,
400
$204,406,
400
$276,167,
600
$213,386,
000
$396,236,
000
$228,268,
900
$264,838,
900
$285,561,
900
$238,875,
400
$273,007,
400
$309,577,
400
$239,035,
200
$296,328,
200
$246,923,
700
$270,084,
700,
$290,807,
700
$253,663,
200
$275,605,
200
$296,328,
200
$268,257,
100
$287,635,
400
$309,577,
400
$285,836,
600
$234,638,
600
$316,311,
600
$260,013,
100
$279,632,
000

Total
costs
$285,140,
400
$214,293,
638
$276,923,
650
$236,436,
100
$399,671,
900
$236,001,
000
$268,097,
000
$287,328,
610
$243,864,
400
$273,306,
300
$310,092,
300
$244,441,
400
$297,260,
370
$253,164,
500
$273,342,
800
$292,574,
410
$259,486,
700
$278,446,
000
$297,677,
640
$273,663,
300
$289,313,
240
$310,509,
570
$288,766,
300
$240,603,
911
$318,495,
600
$262,019,
350
$280,146,
900

6.3.8

Optimization model

Multi-objective optimization (or programming), also known as multi-criteria or multiattribute optimization, is the process of simultaneously optimizing two or more conflicting
objectives subject to certain constraints.

Min X i INVTCij X i INVSPSij X i EENSFCij X i Con Re ntij X i SysInt cos tij

Min X i OPCij

for all i, j
subject to
Budget
EENS
X i Binary
where,

(6.6)

Xi = ith possible planning option,

INVTCij = Transmission investment cost associated with ith option at period j,
INVSPSij = SPS investment costs associated with ith option at period j,
EENSFCij = EENS costs due to forced curtailments associated with ith option at period j,
ConRentij = Congestion rent associated with ith option at period j,
OPCij = Operational complexity associated with ith option at period j.
The conflicting part here is the cost vs. operational complexity, as even though SPS is
cheaper option, its increase in the system will increase the operational complexity as discussed
before. In this formulation, it is to be noted that the influence of transmission investment
decision on the operational complexity metric has not been modeled. In reality, a transmission
investment decision will certainly have its effect on operational complexity, which could even be
a negative effect, i.e., of reducing the maintenance requirements, and making system operation
and maintenance a lot easier. So it is very important to model it in this problem. But in this case,
we have made an assumption that any transmission investment decision, irrespective of its high
investment cost, will not contribute to increase in the system operational complexity. So the
second objective of this problem always supports transmission investment decisions.
6.3.9

Optimization methods

6.3.9.1 Penalty factor method

This approach is based on weighting the objectives, similar to the approached in [140]. The
bi-level optimization problem can be converted into a single objective optimization by
introducing a cost-complexity penalty factor as follows,

Cd C *
Dc D *

(6.7)

where,
120

C* is the optimal value of complexity objective function of the optimal solution,

D* is the optimal value of cost objective function of the optimal solution,
Cd is the cost value of the optimal solution for complexity objective function, and
Dc is the complexity value of the optimal solution for cost objective function.
The problem now becomes
Min T w * f1 ( x) (1 w) * p * f 2 ( x)
subject to
Budget
EENS

(6.8)

X i Binary
where,
f1(x) is the cost function,
f2(x) is the operational complexity metric,
p is the cost-complexity penalty function, and
w is the weighting factor such that 0<=w<=1.
When w = 0, the optimization is minimizing complexity, and when w = 1 it is minimizing
cost.
6.3.9.2 Non-linear penalty factor method

This technique is analogous to the weighted method but different weights are assigned to
objective values of different options, making the cost-complexity penalty factor independent and
non-linear. The nonlinear relationship between complexities of various planning options is
assumed be an exponential ratio.
k

C C C *
wi i d

Cd Dc D *

(6.9)

where, Ci is the complexity value for the ith option.

The problem now becomes

Min T f1 ( x) wi * f 2 ( x)
subject to
Budget

(6.10)

EENS
X i , Binary
where,
wi is a co-efficient computed for each option.

121

6.3.10 Analytical Hierarchical Process

Analytical Hierarchical Process (AHP) was developed by Thomas Saaty, and is being
prevalently used for decision making in various fields. AHP is a structured method for ranking a
list of objectives. AHP doesnt prescribe a correct decision, but helps in finding the best
alternative according to the users needs and problem comprehension.
The following are the steps for implementing the AHP algorithm [141]
Step 1: A structural model of the hierarchy as shown in Figure 56 is set up.
Step 2: A judgment matrix is formed.

Depending upon the users knowledge on the relative importance of every pair of criteria,
each element in the judgment (or comparison) matrix is attributed a value. The measurement
scale used by AHP consists of the following elements, {1, 2, 3, 4, 5, 6, 7, 8, 9}, and their
reciprocal. In the pair-wise comparison matrix, the number in the ith row and jth column gives
the relative importance of the criterion Ci as compared to criterion Cj. For instance,
aij = 1 indicates the criteria are equal in importance
aij = 3 indicates Ci is weakly more important than Cj
aij = 5 indicates Ci is strongly more important than Cj
aij = 7 indicates Ci is very strongly more important than Cj
aij = 9 indicates Ci is absolutely more important than Cj

The values 2, 4, 6 and 8 can also be used as intermediate values. These values form the upper
triangle of the comparison matrix. The lower triangle is filled with reciprocal of the value in the
upper triangle, indicating the relative importance of the criteria in a reciprocal manner. The
diagonal elements are always 1.

Goal

Criteria 1

Alternative 1

Criteria 2

Criteria n-1

Alternative n-1

Criteria n

Alternative n

Figure 56: Simple structure of analytical hierarchical process

122

Step 3: The maximal eigenvalue of the judgment matrix and its corresponding eigenvector is
computed. This provides the value of the weighting coefficients of all criteria.
Step 4: Hierarchical rank and consistency of results are checked.

The elements values in the eigenvector gives the relative importance of all the criteria. So
the hierarchical ranking is performed according to these values of eigenvector elements. The
hierarchy rankings consistency index is checked using the below formula,

CI

max n
n 1

(6.11)

where,
max - the maximal eigenvalue of judgment matrix, and
n - the dimension of the judgment matrix.
Generally, a value less than 0.1 is a very acceptable consistency index.
Application of AHP to SPS aided transmission planning

Most electric power planning decisions are usually based on reliability and economic factors.
When incorporating SPS into transmission and generation expansion, it is necessary to take into
account the operational complexity that SPS adds to the electric power grid operations. Because
without SPS, planned outages will be easier to schedule and maintenance becomes easier.
Analytical hierarchical process (AHP) is used to evaluate the non-dominated solutions based on
the conflicting objectives of total costs and proposed operational complexity index. Three
criteria, i.e., the total cost, reliability (EENS - forced curtailments due to SPS) and operational
complexity are used in this problem to rank the alternatives. Figure 57 presents the structure of
the hierarchy for power system SPS aided transmission expansion planning. Since AHP is a
preference based ranking, the power system planner has the facility to weigh the various criteria
based on the utilitys needs and their understanding of the problem.
Goal

Total costs

X1

Operational
Complexity

Reliability

X2

X4

X10

Figure 57: Structure of AHP for power system transmission expansion planning

123

6.4

Optimization and AHP results

The optimization methods are solved and the corresponding results are presented in the next
section. The budget constraint is set to be $285,140,400.
6.4.1

Penalty factor method

The computed cost-complexity penalty factor is:

p

285,140, 400 214, 293, 638

4, 427,922.63$ / complexity
16 0

The solutions as we change the weights are presented in Table 26.

Table 26: Solution for the penalty factor method
Weight Optimal solution
X1
0
X1
0.1
X1
0.2
X1
0.3
X1
0.4
X2
0.5
X2
0.6
X2
0.7
X2
0.8
X2
0.9
X2
1
6.4.2

Non-linear penalty factor method

The solutions as we assign different exponential powers (k) for the non-linear method are
presented in Table 27.
Table 27: Solution for the non-linear penalty factor method
k Optimal solution
X2
0
X2
1
X2
1.5
X10
2
X10
2.5
Infeasible
3
X4
3.5
X4
4

Therefore, X1, X2, X4 and X10 are the four non-dominated Pareto optimal solutions. Table
22 and Table 25 presented earlier contain the description of each of these solution options, and
`their corresponding economic factors and complexity measure. Table 28 and Table 29 present
the extracted contents of Tables 22 and 25 for the four non-dominated solutions.
124

Table 28: Non-dominated solutions

Options
X1

X10

X4

X2

2015
Build 1-22, 100MW, X =0.18
Build 1-32, 50MW, X= 0.45
NO SPS
Build 1-22, 120MW, X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-22, 120MW, X=0.15
Put SPS on 1-21
Put SPS on 1-31
2 SPS
Build 1-22, 100MW, X=0.18
Build 1-32, 50MW, X= 0.45
NO SPS

2020
Build 1-23, 60MW, X = 0.30
Build 1-33, 80MW, X= 0.2813
NO SPS
Build 1-32, 120MW, X=0.1875
Build 1-23, 70MW, X=0.2571
NO SPS

2025
Build 1-24, 50MW, X=0.36
Build 1-34, 90MW, X= 0.25
NO SPS
Build 1-33, 100MW, X= 0.225
NO SPS

Build 1-23, 100MW, X=0.1636

Put SPS on 1-31
1SPS

Build 1-24, 100MW, X=0.1636

Put SPS on 1-31
1SPS

Build 1-23, 70MW, X = 0.2571

Put SPS on 1-31
Put SPS on 1-32
2 SPS

Build 1-24, 90MW, X=0.36

Put SPS on 1-31
Put SPS on 1-32
2 SPS

Table 29: Comparison of economic factors and complexity

X1
X10
X4
X2

6.4.3

Transmission
costs
$285,140,
400
$271,007,400
$211,386,
000
$202,198,
200

SPS
costs
$0

Cong.
Rent
$0

Interr.
cost
$0

(MWh)

EENS

$2,000,0
00
$2,000,0
00
$2,208,2
00

$0

$244,900

900

$14,347,
000
$1,804,1
38

$3,543,
100
$3,274,
100

7800

7800

EENS
Cost
$0

Comp
lexity
0

$54,00
0
$5,160,
000
$4,809,
000

8
14
16

Budget
$285,14
0,400
$273,00
7,400
$213,38
6,000
$204,40
6,400

Total
costs
$285,14
0,400
$273,30
6,300
$236,43
6,100
$214,29
3,638

Ranking non-dominated solutions using AHP

Table 30 presents the judgment matrix that shows the relative importance of various criteria
in selecting the best planning option. Tables 31, 32 and 33 present the judgment matrices that
show the relative importance of every planning option with respect to the various decision
criteria. The consistency index of every judgment matrix is also computed, which are found to be
less than 0.1 in all the cases.
Table 30: Judgment matrix for the criteria
Total costs
Reliability
Operational complexity

Total costs Reliability Operational complexity Avg score

1
4
4
0.666
1/4
1
1
0.1667
1/4
1
1
0.1667

Consistency index= 0

125

Table 31: Judgement matrix for operational complexity

X1 X2 X4 X10 Avg score
1
9
8
4
0.639
X1
1/9
1
1/2
1/4
0.0545
X2
1
1/4
0.0822
X4 1/8 2
4
4
1
0.2241
X10

Consistency index=0.0341
Table 32: Judgement matrix for reliability
X1 X2 X4 X10 Avg score
1
8
8
2
0.5303
X1
1 1/7
0.0581
X2 1/8 1
1 1/7
0.0581
X4 1/8 1
7
7
1
0.3535
X10

Consistency index=0.0133
Table 33: Judgement matrix for total costs
X1
1
X1
7
X2
5
X4
X10 2

X2 X4 X10 Avg score

1/7 1/5 1/2
0.0596
1
4
6
0.5908
1/4 1
4
0.2562
1/6 1/4
1
0.0933

Consistency index=0.0549
Table 34 provides the final ranking of the non-dominated planning options using AHP, which
is obtained by weighing various options according to the relative importance attributed by the
user for various decision criteria, namely the total cost, reliability (EENS) and operational
complexity. For the given relative importance, planning option X2 is ranked high.
Table 34: Final weight and rank for non-dominated solutions
Total costs Reliability Operational complexity Weight Rank
(0.666)
(0.167)
(0.167)
0.0596
0.5303
0.639
0.2346
2
X1
0.5908
0.0581
0.0545
0.4122
1
X2
0.2562
0.0581
0.0822
0.1940
3
X4
0.0933
0.3535
0.2241
0.1584
4
X10

126

7
7.1

Summary and conclusions

Summary

Special Protection Schemes (SPS) have been proved to be a quick and economic way of
ensuring power system reliability, especially in the wake of drastically increasing renewable
generation resources and an invariably stagnant transmission up-gradation policy. SPS postpones
transmission upgrades while maximizing the usage of transmission capacity by enabling system
operation closer to stability limits, and at times even beyond. Therefore this technology very
much bolsters the current markets paradigm of optimizing the network resources, especially
transmission usage, while supplying uninterrupted and economic power.
While the usage of SPS is encouraging, it has many downsides, especially in the eve of its
tremendous proliferation in the system due to the increase of intermittent generation facilities.
Many utilities have started to implement SPS for tripping wind farms to unburden the system
during transmission overloads. So this increase in SPS has raised several reliability issues, one of
which is the serious and undesirable consequences of inadvertent interactions among SPS. But
there is a dearth of simulation and assessment tools that could capture such phenomenon during
reliability studies and enable planners to come up with reliable planning option at the system
planning stage.
This report has shed a considerable focus on these and other related range of topics, such as:
1. Introduction to SPS and its components
2. Current industry standards, practices and advancements in SPS A survey of various
technologies used by a cross section of industries to achieve SPS centralization and
coordination have been presented.
3. Standards and methods of related industries such as process control, nuclear and power
system planning The attempt is to leverage interesting ideas from these mature
industries that could be applicable to reliability and maintenance studies related to SPS.
4. Risk assessment of SPS based on two frameworks process view and system view.
Process view framework is the traditional way of computing risk associated with a
system taking into account its various individual components and processes that serve as
the building block of the entire system. System view framework is a new idea proposed
in this report where the nature of the system and the operating conditions faced in reality
are considered in the process of estimating risk associated with SPSs operation.
5. A design of Monte Carlo simulation based reliability assessment and SPS failure mode
identification has been presented.
6. The report identifies the importance of including SPS in the power system planning
framework, and so it illustrates system planning studies for SPS aided power systems.
Two illustrations have been presented:
a. Accommodating more wind generation using SPS
b. Incorporating SPS in generation and transmission expansion planning

127

7.2

Conclusions

1. Special Protection Schemes (SPS), also known as Remedial Actions Schemes (RAS)
have been a major technological advancement that aids in economical usage of
transmission resources and smooth interconnection of renewable generation.
2. SPS has proven to be greatly economical and easy to implement compared to
transmission lines, and many utilities are favoring SPS to meet their generation and
transmission expansion goals.
3. Maintenance standards and documentation have been developed by industries deploying
SPS to ensure meeting NERC reliability standards. One of the prominent features of all
standards has been the emphasis in embedding redundancy into SPS architectures, to
ensure SPS operations are immune to failures and uncertainties.
4. The advent of synchrophasors (PMUs) has given a major boost to SPSs operational
performance and has increased the range of SPS applications. SPS along with PMUs and
PDCs (Phasor Data Concentrators) have been instrumental in advancing the Wide Area
Monitoring, Protection and Control Systems (WAMPACS).
5. Power industry has seen a drastic proliferation in SPS, which is proving to offset the
advantage these individual SPS brings in by causing coordination and maintenance
issues. This has served as great motivation for industries to move from a localized-RAS
to Centralized-RAS technology with the help of EMS and PMUs.
6. As the dependence on SPS is growing, there is a greater need to build our knowledge
base and expertise in understanding SPS better and maintaining them. Interestingly, this
can be accomplished by extracting relevant standards and practices from existing
industries. Safety instrument systems (SIS) of process control industry in one such
example. The process of building operational rule for power system operators using
Monte Carlo and machine learning techniques is another example that could contribute in
SPS logic derivation and evaluation.
7. In modern SPS-rich systems, inadvertent interactions among SPS may prove to be
catastrophic. Therefore the report emphasizes the need to perform studies that identify
SPS failures and quantify risk associated with SPS from system point of view. The
report also proposes a study design based on operational planning framework using
Monte Carlo simulation to identify system view SPS failures, estimate their
reliability/risk indices and re-design SPS logic. The proposed system study and decision
support tool harnesses the advancement in computing power.
8. Several reliability models and architectures for SPS and PMU-aided SPS have been
proposed in this report to facilitate system level reliability studies that account SPS.
9. The report has introduced the concept of operational complexity due to the proliferation
in SPS and has provided a quantitative definition. System planning studies must
incorporate such operational complexity metric in their overall formulation to estimate
the limit of SPS growth for economical and reliable system operation, and also find the
best SPS-aided transmission expansion plan.

128

References

[1] System Protection Schemes in Power Networks,'' International Conference on Large High Voltage Electric
Systems (CIGRE), January, 2001.
[2] Workshop on SPECIAL PROTECTION SYSTEMS for Transmission operations and Emergencies, Almaty,
Kazakhstan, available online: http://www.usea.org/programs/EUPP/SPS_Workshop_Almaty_KZ_2-18to202009/SPS_Handbook_English.pdf
[3] Dana Callell and Shashi Pandey (SEC), Centralized Remedial Action Scheme (C-RAS) using Emerging
Telecommuniction (IEC61850), Protection Technologies, and OSIsoft Controller software, Colorado Tarnsmission
Planning Meeting, 2008 Available online: http://www.westconnect.com/filestorage/CRAS%20Presentation%20for%20CRT%20Mtg_9-17-08.pdf
[4] W. Winter and B. LeReverend, Operational performance of bulk electricity system control aids, Electra, N. 123,
March 1989
[5] P. M. Anderson and B. LeReverend,, Industry experience with special protection schemes, discussion, IEEE
Trans. Power Syst., vol. 11, pp. 11671179, Aug. 1996.
[6] V. Madani, M. Begovic, et al., Global Industry Experiences with System Integrity Protection Schemes, IEEE
PSRC Working Group C4, Oct., 2009
[7] R. Vinnakota, M. Yao, D. Atanackovic, BCTC, Modeling issues of system protection schemes in energy
management systems, IEEE Electrical Power & Energy Conf., 2008
[8] J. Wen, P. Arons, W. Liu, The role of remedial action schemes in renewable generation integrations, IEEE,
2010
[9] D. Manjure, M. McMullen, D. Subakti, D. Tewari, Managing Wind Energy: From Interconnection planning to
real time operations, an integrated approach to ensure energy and transmission capacity, PES Meeting, 2009
[10] J. McCalley and W. Fu, Reliability of Special Protection Systems, IEEE Trans. on Power Systems, Vol. 14, No.
4, pp.1400-1406, November, 1999.
[11] Western Systems Coordinating Council, Guide for Remedial Action Schemes, April, 1991.
[12] Northeast Power Coordinating Council, Special Protection System Guideline, February, 1992.
[13] Northeast Power Coordinating Council, Bulk Power System Protection Criteria, August, 1995.
[14] V. Madani and W. Higinbotham, Advantages in Modern Communications as applied to remedial action
schemes and control centers, IREP Symposium Bulk Power System Dynamics and Control VII, August 19-24,
Charleston, SC, USA.
[15] David Dolezilek (SEL), Case Study Examples of Interoperable Ethernet Communications Within Distribution,
Transmission, and Wide-Area Control Systems, IEEE International Conference on Communications Workshops
(ICC), May 2010
[16] P. C. K. Lau, M. Grover, and W. Tanaka, Reliability Assessment of Special Protection Systems, CIGRE paper
AA-11, presented at the CIGRE Symposium on Electric Power System Reliability, Montreal, September, 1991
[17] NERC reliability standards, Protection and control, website http://www.nerc.com/page.php?cid=2%7C20
[18] WECC Standard PRC-004-WECC-1 Protection System and Remedial Action Scheme Misoperation available
online: http://www.nerc.com/files/PRC-004-WECC-1.pdf
[19 ] Safety Instrumented Functions (SIF) - Safety Integrity Level (SIL) Evaluation Techniques, The Instrumentation
Systems and Automation Society 2002, Available online:

129

http://www.isa.org/Content/Microsites195/SP5_2,_Binary_Control_Logic_Diagrams_for_Process_Operations/Hom
e193/Committee_Archives126/TR_8402p1.pdf
[20]Peter B. Ladkin, The Concepts of IEC 61508: An Overview and Analysis, Available online:
http://www.rvs.uni-bielefeld.de/Bieleschweig/fifth/download/B5-Ladkin.pdf
[21] Dan Schoenecker, Determination of Financial Penalties, Midwest Reliability Organization, 2008, Available
online:http://www.midwestreliability.org/01_about_mro/board_of_directors/presentations/Penalty%20tool%20deter
minants_MRO%20BOD.pdf
[22]An operations view of special protection systems IEEE Trans. Power Syst., vol. 3, pp. 1078-1083, Aug. 1988
[23] Nodal Operating Guides Planning section, ERCOT, 2007, Available online:
http://www.ercot.com/content/mktrules/guides/noperating/2007/11/05/05-110107.doc
[24] Balancing Authority/Transmission Operator Reliability Readiness Audit Report: British Columbia
Transmission Corporation (BCTC), Burnaby, British Columbia, Canada, 2006, Available online:
http://www.nerc.com/docs/rap/audits/BCTC_BA-TOP_ReadinessAuditReport.pdf
[25]Patricia L. Arons, SCE Pilots the Next Level of Grid Protection, Southern California Edison, Transmission and
Distribution World, 2007, Available online: http://tdworld.com/overhead_transmission/power_sce_pilots_next/
[26] Transmission Constraints Management: RAS in the Planning Stage, AESO Discussion Paper, 2009 Available
online: http://www.aeso.ca/downloads/RAS_Guideline_Discussion_Paper_July_03_2009.pdf
[27] T.J.Wang, C.W.Liu, and Y.H.Liu, A PMU based Special Protection Scheme: a Case Study of Taiwan Power
System, Electrical Power and Energy Systems 27, pp.215-223, 2005
[28] Madani, V.; Adamiak, M.; Thakur, M., "Design and implementation of wide area special protection schemes,"
Protective Relay Engineers, 2004 57th Annual Conference for , vol., no., pp. 392-402, 30 Mar-1 Apr 2004
[29] Procedure to submit a RAS for Assessment 2005, WECC, Remedial Action Scheme Reliability Subcommitee,
Available online:http://www.wecc.biz/library/Library/Remedial%20Action%20Schemes/RAS_Approval_Procedure04-2005.pdf
[30]Reliability Coordinator, Reliability Readiness Audit Report, MISO, 2006, Available online:
http://www.nerc.com/docs/rap/audits/MISO_RC_ReadinessEvaluationReport.pdf
[31] Document online: http://www.ferc.gov/industries/electric/indus-act/reliability/blackout/ch5.pdf
[32] Varghese, M.; Licheng Jin; Ghosh, S.; Lin, G.; Bunthath Pek; The CAISO experience of implementing
automated Remedial Action Schemes in Energy Management Systems IEEE power & energy society meeting 2009
[33] BC HYDRO, Operations Support, 2009 Available online:
http://transmission.bchydro.com/nr/rdonlyres/f56489b9-f09a-452b-a0aa-6100a8f13aaf/0/7t18.pdf
[34]Pai S.C,Sun J, BCTCs experience towards a smarter grid - increasing limits and reliability with centralized
intelligence Remedial Action Schemes Electric Power Conference, 2008. EPEC 2008. IEEE Canada ,no.,1-7,pp:
6-7 Oct. 2008
[35] R. Vinnakota, M. Yao, D. Atanackovic, A. Steed, and Q. Zhu, Managing Contingencies in Real Time using
EMS Advanced Network Applications, IEEE PES General Meeting 2009, Slides available online:
http://www.ieee.org/organizations/pes/meetings/gm2009/slides/pesgm2009p-001028.pdf
[36] Nodal Operating Guide, ERCOT, available online:
http://www.ercot.com/content/meetings/owg/keydocs/2009/1021/029NOGRR01_Synchronization_of_OGRR224,_Special_Protection_S.doc
[37] C. W. Taylor, BPAs Wide-Area stability and voltage Control System (WACS), Imperial College, London, 2006,
available online: http://www3.imperial.ac.uk/pls/portallive/docs/1/4859962.PDF
[38] Talking points on RAS Arming and Ability to Schedule Generator Interconnection Projects, BPA, 2008 available
online: http://transmission.bpa.gov/business/generation_interconnection/documents/GI_talking_points_060508.pdf

130

[39] A. Guzman, D. Tziouvaras, E. O. Schweitzer, and K. Martin, Local and Wide-Area Network Protection
Systems improve Power System Reliability, available online:
http://www.naspi.org/resources/archive/prtt/waps_wprc04.pdf
[40] P. Arons, Piloting a Centralized Remedial Action Scheme (C-RAS) with merging Telecom/Protection
Technologies, Transmission & Interconnection Planning, SCE, VOTAGE2007, available online:
http://videostar.osisoft.com/uc2007/ppt/MO-02-04B_Arons_SCEd.pdf
[41] E. O. Schweitzer, and D. E. Whitehead, Real-Time Power System Control Using Synchrophasors, Schweitzer
Engineering Laboratories, Inc., 61st annual conference for Protective Relay Engineers,pp. 78-88, 2008
[42] S. Wang, and G. Rodriguez (SCE), Smart RAS (Remedial Action Scheme), Innovative Smart Grid
Technologies, Jan 2010.
[43] K.E. Martin (BPA), S4-1: Power systems operation control based on the phasor measurement technology,
Phasor Measurement Systems in Western North America, Relay Protection and Substation Automation of Modern
Power Systems, Sept 2007
[44] D. Karlsson, L. Messing, and M. Akke, Wide Area Protection and Emergency Control, 8th IEEE
International conference on developments in Power System Protection, Vol. 2, pp. 746-751, 2004
[45] Northeast Power Coordinating Council, Maintenance Criteria for Bulk Power System Protection, 1996.
[46] Northeast Power Coordinating Council, Procedure for Reporting and Reviewing Proposed Bulk Power System
Protection, September, 1996.
[47] S. Gillespie, Safety Intrumented Systems, available online: http://www.idconline.com/technical_references/pdfs/instrumentation/Safety_Instrumented_Systems.pdf
[48] ANSI/ISA Standard S84.01-1996, Application of Safety Instrumented Systems to the Process Industries,
International Society for Measurement & Control, Research Triangle Park, NC, 1996
[49] ISA-dTR84.02 (draft), The Instrument Society of America (EA), September, 1997
[50] IEC 61508 Parts 1-7: 1998, Functional Safety of Electrical/Electronic/Programmable Electronic Safety-Related
Systems, International Electrotechnical Commission, Geneva, Switzerland.
[51] S 84/IEC 61511 Standard for SIS, FAQ sheet Available online:
http://www.primatech.com/info/faq_s84_standard_for_safety_instrumented_systems.pdf
[52] Guidelines for Cabinet Submissions and New Policy Proposals, Department of the Prime Minister and Cabinet,
Australian Government, available online: http://www.dpmc.gov.au/implementation/policy.cfm
[53] Mike Scott and Bud Adler, How to select a Safety PLC, Available online:
http://www.isa.org/Content/Microsites838/Safety_Division/Home818/ISA_2004_Safety_Papers/How_to_Select_a_
Safety_PLC.pdf
[54] Functional safety of electrical/electronic/programmable electronic safety-related systems Part 1: General
requirements, International Electrotechnical Commission (IEC) Standard IEC 61508-1, 1st ed., Dec. 1998
[55] Functional safety of electrical/electronic/programmable electronic safety-related systems Part 2:
Requirements for electrical/electronic/programmable electronic safety-related systems, International
Electrotechnical Commission (IEC) Standard IEC 61508-2, 1st ed., May, 2000
[56] Functional safety of electrical/electronic/programmable electronic safety-related systems Part 3: Software
requirements, International Electrotechnical Commission (IEC) Standard IEC 61508-3, 1st ed., Dec. 1998
[57] Functional safety of electrical/electronic/programmable electronic safety-related systems Part 4: Definitions
and abbreviations, International Electrotechnical Commission (IEC) Standard IEC 61508-4, 1st ed., Dec. 1998
[58] Functional safety of electrical/electronic/programmable electronic safety-related systems Part 5: Examples of
methods for the determination of safety integrity levels, International Electrotechnical Commission (IEC) Standard
IEC 61508-5, 1st ed., Dec. 1998

131

[59] Functional safety of electrical/electronic/programmable electronic safety-related systems Part 6: Guidelines

on the application of IEC 61508-2 and IEC 61508-3, International Electrotechnical Commission (IEC) Standard IEC
61508-6, 1st ed., Apr., 2000
[60] Functional safety of electrical/electronic/programmable electronic safety-related systems Part 7: Overview of
techniques and measures, International Electrotechnical Commission (IEC) Standard IEC 61508-7, 1st ed., Mar.,
2000
[61] Functional safety of electrical/electronic/programmable electronic safety-related systems Part 0: Functional
safety and IEC 61508, International Electrotechnical Commission (IEC) Technical Report IEC/TR 61508-0, 1st ed.,
Jan., 2005
[62] R. Bell, Introduction to IEC 61508, in: T. Cant, editor, 10th Australian Workshop on Safety Critical Systems
and Software, Conferences in Research and Practice in Information Technology, vol. 55, pp. 3-12, 2005
[63] Nuclear power plants Instrumentation and control for systems important to safety General requirements for
systems, International Electrotechnical Commission (IEC) Standard IEC 61513, 1st ed., Mar., 2001
[64] Nuclear power plants Instrumentation and control important to safety Classification of instrumentation and
control functions, International Electrotechnical Commission (IEC) Standard IEC 61226, 3rd ed., Jul., 2009
[65] Nuclear power plants Instrumentation and control systems important to safety Separation, International
Electrotechnical Commission (IEC) Standard IEC 60709, 2nd ed., Nov., 2004
[66] Nuclear power plants Instrumentation and control important to safety Hardware design requirements for
computer-based systems, International Electrotechnical Commission (IEC) Standard IEC 60987, 2nd ed., Aug., 2007
[67] Nuclear power plants Instrumentation and control systems important to safety Software aspects for
computer-based systems performing category A functions, International Electrotechnical Commission (IEC)
Standard IEC 60880, 2nd ed., May, 2006
[68] Nuclear power plants Instrumentation and control important for safety Software aspects for computerbased systems performing category B or C functions, International Electrotechnical Commission (IEC) Standard
IEC 62138, 1st ed., Jan., 2004
[69] Nuclear power plants Instrumentation and control systems important to safety Requirements for coping
with common cause failure (CCF), International Electrotechnical Commission (IEC) Standard IEC 62340, 1st ed.,
Dec., 2007
[70] Nuclear power plants Instrumentation and control systems important to safety Management of ageing,
International Electrotechnical Commission (IEC) Standard IEC 62342, 1st ed., Aug., 2007
[71] Nuclear power plants Instrumentation and control systems important to safety Surveillance testing,
International Electrotechnical Commission (IEC) Standard IEC 60671, 2nd ed., May, 2007
[72] E. A. Corte, and J. P. Bouard, International Standardisation in Nuclear I&C Engineering, 2005, Available
online: http://www.yellowdocuments.com/6460428-international-standardisation
[73] Safety of Nuclear Power Plants: Design Safety Requirements, International Atomic Energy Agency (IAEA),
Safety Standards Series No. NS-R-1, 2000, Oct. 31, 2000
[74] Software for Computer Based Systems Important to Safety in Nuclear Power Plants Safety Guide,
International Atomic Energy Agency (IAEA) Safety Standards Series No. NS-G-1.1, 2000, Nov. 14, 2000
[75] Instrumentation and Control Systems Important to Safety in Nuclear Power Plants Safety Guide, International
Atomic Energy Agency (IAEA) Safety Standards Series No. NS-G-1.3, 2002, Apr. 17, 2002
[76] Design of Emergency Power Systems for Nuclear Power Plants Safety Guide, International Atomic Energy
Agency (IAEA) Safety Standards Series No. NS-G-1.8, 2004, Oct. 13, 2004
[77] G. Johnson, Comparison of IEC and IEEE Standards for Computer-Based Control Systems Important to
Safety, in Proc. 2001 IEEE Nuclear Science Symposium Conference Record, vol. 4, pp. 2474-2481

132

[78] IEEE Standard Criteria for Safety Systems for Nuclear Power Generating Stations, Institute of Electrical and
Electronics Engineers (IEEE) Standard IEEE Std 603-2009, Nov., 2009
[79] IEEE Standard Criteria for Digital Computers in Safety Systems of Nuclear Power Generating Stations,
Institute of Electrical and Electronics Engineers (IEEE) Standard IEEE Std 7-4.3.2-2003, Dec., 2003
[80] IEEE Standard Criteria for Periodic Surveillance Testing of Nuclear Power Generating Station Safety Systems,
Institute of Electrical and Electronics Engineers (IEEE) Standard IEEE Std 338-2006, Jun., 2007
[81] IEEE Guide for Installation, Inspection, and Testing for Class 1E Power, Instrumentation, and Control
Equipment at Nuclear Facilities, Institute of Electrical and Electronics Engineers (IEEE) Standard IEEE Std 3362005, Feb., 2006
[82] IEEE Guide for General Principles of Reliability Analysis of Nuclear Power Generating Station Safety Systems,
American National Standards Institute (ANSI) / Institute of Electrical and Electronics Engineers (IEEE) Standard
ANSI/IEEE Std 352-1987, Nov., 1985 / Jun. 1985
[83] IEEE Standard Requirements for Reliability Analysis in the Design and Operation of Safety Systems for
Nuclear Facilities, Institute of Electrical and Electronics Engineers (IEEE) Standard IEEE Std 577-2004, Aug.,
2004
[84] Construction Codes and Standards: Avoidance of New Nuclear Power Plant Construction Delays. Energetics
Incorporated, The United States Department of Energy (DOE), Office of Nuclear Energy, Tech. Rep., Nuclear
Energy Standards Coordination Collaborative (NESCC) Meeting Jun. 2009, Available online:
http://publicaa.ansi.org/sites/apdl/Documents/Meetings and Events/2009 NESCC/NESCC Meeting - June 1,
2009/NESCC 09-004 Mattson Report.pdf
[85] C. Lebrevelec, P. Cholley, J.F. Quenet, and L. Wehenkel, A statistical analysis of the impact on security of a
protection scheme on the French power system, International Conference on Power System Technology,
Proceedings, POWERCON '98, Volume 2, pp. 1102 - 1106, 18-21 Aug. 1998
[86] Y. Schlumberger, C. Lebrevelec, and M. De Pasquale, Power systems security analysis-new approaches used
at EDF, IEEE Power Engineering Society Summer Meeting, Volume 1, pp. 147 151, 18-22 July 1999
[87] J. Pierre, C. Lebrevelec, and L. Wehenkel, Automatic learning methods applied to dynamic security
assessment of power systems, International Conference on Electric Power Engineering. PowerTech Budapest,
pp.180, 29 Aug.-2 Sept. 1999
[88] H. Martigne, P. Cholley, D. King, and J. Christon, Statistical method to determine operating rules in the event
of generator dropout on EDF French Guyana Grid, IEEE Power Tech Proceedings, Porto, Vol. 1, pp. 5, 10-13
Sept. 2001
[89] J. Paul and K. Bell, A Flexible and Comprehensive Approach to the Assessment of Large-Scale Power System
Security Under Uncertainty, Proc. of the 7th International Conference on Probabilistic Methods Applied to Power
Systems, Naples Italy, September 2002
[90] S. Henry, J. Pompee, M. Bulot, and K. Bell, Applications of statistical assessment of power system security
under uncertainty, International Conference on Probabilistic Methods Applied to Power Systems, pp. 914-919, 1216 Sept. 2004
[91] S. Henry, J. Pompee, L. Devatine, M. Bulot, and K. Bell, New trends for the assessment of power system
security under uncertainty, IEEE PES Power Systems Conference and Exposition, vol.3, pp. 1380-1385, 10-13
Oct. 2004
[92] S. Henry, E. Brda-Sys, H. Lefebvre, V. Sermanson and M. Bna, Probabilistic study of the collapse modes
of an area of the French network, Proc. of the 9th International Conference on Probabilistic Methods Applied to
Power Systems, Stockholm, Sweden, June 2006
[93] P. Cholley, C. Lebrevelec, S. Vitet, and M. de Pasquale, Constructing operating rules to avoid voltage
collapse: a statistical approach, International Conference on Power System Technology, Proceedings, POWERCON
'98, Volume 2, pp. 1468-1472, 18-21 Aug. 1998

133

[94] C. Lebrevelc, Y. Schlumberger, and M. de Pasquale, An application of a risk based methodology for defining
security rules against voltage collapse, IEEE Power Engineering Society Summer Meeting, Volume 1, pp.185-190,
18-22 July 1999
[95] S. Henry, C. Lebrevelec, and Y. Schlumberger, Defining operating rules against voltage collapse using a
statistical approach: The EDF experience, International Conference on Electric Power Engineering, PowerTech
Budapest 99. pp. 30, 29 Aug.-2 Sept. 1999
[96] Y. Schlumberger, J. Pompee, and M. De Pasquale, Updating operating rules against voltage collapse using
new probabilistic techniques, IEEE/PES Transmission and Distribution Conference and Exhibition: Asia Pacific.,
Volume 2, pp. 1139-1144, 6-10 Oct. 2002
[97] K. Sun, S. Likhate, V. Vittal, V.S. Kolluri, and S. Mandal, An Online Dynamic Security Assessment Scheme
Using Phasor Measurements and Decision Trees, IEEE Trans. Power Systems, Volume: 22, Issue: 4, pp. 1935-1943,
Nov. 2007
[98] R. Diao, K. Sun, V. Vittal, R.J. O'Keefe, M.R. Richardson, N. Bhatt, D. Stradford, and S.K. Sarawgi, Decision
Tree-Based Online Voltage Security Assessment Using PMU Measurements, IEEE Trans. Power Systems, Vol. 24,
Issue 2, pp. 832-839, May 2009
[99] V. Krishnan, J. D. McCalley, S. Henry, and S. Issad, High Information Content Database Generation for Data
Mining based Power System Operational Planning Studies, IEEE PES General Meeting, Minneapolis, July 2010
[100] C. Singh, and J. Mitra, Composite system reliability evaluation using state space pruning, IEEE Trans.
Power Systems, Volume 12, Issue 1, pp. 471-479, Feb. 1997
[101] J. Huang, G. Vanier, A. Valette, S. Harrison, F. Lvesque, and L. Wehenkel, Operation rules determined by
risk analysis for special protection systems at Hydro-Qubec, CIGRE, 2004
[102] J. Huang, G. Vanier, A. Valette, S. Harrison, and L. Wehenkel, Application of data mining to optimize settings
for generator tripping and load shedding system in emergency control, The International Journal for Computation
and Mathematics in Electrical and Electronic Engineering Vol. 23 pp. 21-34 No. 1, 2004
[103] N. Senroy, G. T. Heydt, and V. Vittal, decision tree assisted controlled islanding, IEEE trans. power syst., vol.
21, no. 4, November 2006
[104] ASSESS, TROPIC, METRIX website: http://www.rte-france.com/htm/an/activites/assess.jsp
[105] V. Vittal, J. McCalley, V. Van Acker, W. Fu, and N. Abi-Samra, "Transient instability risk assessment", Proc.
Of the 1999 IEEE PES Summer Meeting , pp. 1999
[106] System disturbances, North Amer. Elect. Rel. Council, 19861995
[107] P. M. Anderson, Power System Protection, Vol 111, Review Copy, 1996
[108] W. M. Gable, Evaluating Control Systems Reliability- Techniques and Applications, The Instrument Society
of America, (ISA), N. C., 1992
[109] B. S. Dhillon, Mechanical Reliability: Theory, Models and Applications, American Institute of Aeronautics
and Astronautics, Inc., Washington, DC, 1988
[110] Peter Morgan, Markov Modelling Primer- Calculating System PFDs for the design of Safety Instrumented
Systems, available online: http://www.isa.org/filestore/Markov_Modelling.pdf
[111] Non-conventional instrument transformer solutions, AREVA T&D, Available online: http://www.arevatd.com/solutions/liblocal/docs/NCIT/NCIT_BRen_1720.pdf
[112] P. Zhang, J. Chen, and M. Shao, "Phasor measurement unit (PMU) implementation and applications," Electric
Power Research Institute (EPRI), Palo Alto, CA, Tech. Rep. 2007.1015511 Available online:
http://my.epri.com/portal/server.pt?Abstract_id=000000000001015511
[113] Communication networks and systems in substation Part 9-2: Specific communication service mapping
(SCSM) Sampled values over ISO/IEC 8802-3, International Electrotechnical Commission (IEC) Standard IEC

134

61850-9-2, 1st ed., Apr. 2004.

[114] D. Chatrefou, J. P. Dupraz, and G.F. Montillet, "Interoperability between non-conventional instrument
transformers (NCIT) and intelligent electronic devices (IED)," in Proc. 2005/2006 IEEE Power Engineering
Society (PES) Transmission and Distribution Conf. and Exhibition, pp. 1274-1279, May 2006
[115] B. Kasztenny, J. Whatley, E. A. Udren, J. Burger, D. Finney, M. Adamiak, "Unanswered questions about IEC
61850: What needs to happen to realize the vision?" 32nd Annual Western Protective Relay Conference, 2005
[116] Y. Wang, W. Li, and J. Lu, Reliability Analysis of Phasor Measurement Unit Using Hierarchical Markov
Modeling, Electric Power Components and Systems, 37, pp. 517-532, 2009
[117] Y. Wang, W. Li, J. Lu, and H. Liu, Evaluating multiple reliability indices of regional networks in wide area
measurements system, Electric Power Systems Research, 79, pp. 1353-1359, 2009
[118] Y. Wang, W. Li, and J. Lu, Reliability Analysis of Wide-Area Measurement System, IEEE Trans. Power
Delivery, Vol. 25, No. 3, July 2010
[119] P. Zhang, L. Portillo, and M. Kezunovic, "Reliability and component importance analysis of all-digital
protection systems," in Proc. 2006 IEEE Power Engineering Society (PES) Power Systems Conf. and Exposition, pp.
1380-1387, Oct. 2006
[120] C. Singh and R. Billinton, "Frequency and duration concepts in system reliability evaluation," IEEE Trans.
Reliability, vol. R-24, no. 1, pp. 31-36, Apr. 1975
[121] C. Singh and R. Billinton, System Reliability Modeling and Evaluation, London, UK: Hutchinson Educational
Publishers, 1977, pp. 48-59, Available online: http://www.ece.tamu.edu/People/bios/singh/sysreliability
[122] C. Singh, "A matrix approach to calculate the failure frequency and related indices," Microelectronics and
Reliability, vol. 19, pp. 395-398, 1979
[123] J. Walseth, J. Eskedal, and O. Breidablik, Analysis of misperations of protection schemes in the Nordic
Grid, Protection, Automation, and Control World, March 2010
[124] T. Van Cutsem, C. Moors, D. Lefebvre, Design of load shedding schemes against voltage instability using
combinatorial optimization, IEEE PES Winter Meeting, Vol.2 2002
[125] V. Sermanson, and a. Giard, Dynamic voltage security assessment in operational planning at RTE, a way to
improve the power system security, Securing Critical Infrastructures, Grenoble, October 2004
[126] T. Y. Hsiao, C. N. Lu, Y. H. Liu, "Defense plan design in a longitudinal power system," PSCC 2002, in Sevilla,
Session 25, Paper 5, 2002
[127] W. T. Tsai and C. R. Chen, A Special Protection System Implementation-Based on an Isolated and
Longitudinal Power System, International Conference on Power System Technology, POWERCON, 21-24
November 2004
[128] H. Seyedi, M. Sanaye-Pasand , Design of New Load Shedding Special Protection Schemes for a Double Area
Power System, American Journal of Applied Sciences 6 (2): 317-327, 2009
[129] T.Y. Hsiao, C.A. Hsieh, and C. N. Lu; A risk-based contingency selection method for SPS applications,
IEEE Trans. Power Sys., Vol. 21, issue. 2, pp. 1009 1010, 2006
[130] B. D. Ripley, Stochastic Simulation, Wiley: New York, 1987
[131] G. Papaefthymiou, and D. Kurowicka, Using Copulas for Modeling Stochastic Dependence in Power System
Uncertainty Analysis, IEEE Trans. Power Systems ,Vol. 24, Issue 1, pp. 40 49, 2009
[132] W.G. Wyss, K.H. Jorgensen, A Users Guide to LHS: Sandias Latin Hypercube Sampling Software, Sandia
National Laboratories Report SAND98-0210, Albuquerque, NM, 1998.
[133] R. Billinton, R.N. Allan, Reliability evaluation of power systems, Second Edition, New York: Plenum Press,
c1996

135

[134] R. Billinton, and W. Li, Reliability assessment of electric power system using Monte Carlo methods, Springer
[135] 20% Wind Energy by 2030, Increasing Wind Energy's Contribution to U.S Electricity Supply, DOE/GO102008-2567, July 2008, available online: http://www.energy.gov
[136] Paul Davidson, Wind energy confronts shortage of transmission lines, USA Today, 2008, available online:
http://www.usatoday.com/money/industries/energy/environment/2008-02-25-wind-power-transmission_N.htm
[137] Phadke, A.G.; Thorp, J.S., History and Application of Phasor Measurements, IEEE Power Systems
Conference and Exposition, 2006. PSCE '06. Oct. 29 2006-Nov. 1 2006 pp. 331 335
[138] Weaver, Warren (1948), "Science and Complexity", American Scientist 36: 536
[139] Highlights of the Midwest ISO 2003 State of the Market Report, Independent Market Monitor, Midwest ISO,
2004, available online: www.ferc.gov/EventCalendar/Files/20040915134723-A-3-MISO.pps
[140] N. Cetinkaya, Optimization Algorithm for Combined Economic and Emission Dispatch with Security
Constraints
[141] Zhu J.; Momoh J.A, Optimal VAr pricing and VAr placement using analytic hierarchical process, Electric
Power Systems Research, Volume 48, Number 1, 1 December 1998 , pp. 11-17

136