DETECTION OF MAJOR DISTURBANCES AND OPTIMIZATION OF TRANSMISSION LINE PROTECTIVE RELAYING OPERATIONS USING NEURAL NETWORKS

1
DETECTION OF MAJOR DISTURBANCES AND OPTIMIZATION
OF TRANSMISSION LINE PROTECTIVE RELAYING
OPERATIONS USING NEURAL NETWORKS

• CESAR RINCON
• LOUISIANA STATE UNIVERSITY

2
Outline

• Introduction
• Background
• Proposed Solutions
• Detailed Algorithm and Data Needed
• Next Step
• Discuss

3
Several major blackouts worldwide
4
Introduction

• Cascading outages
• Catastrophic economic and social impacts
• Lots of them occurred recently
• Increased research interests due to the lack of
  effective analysis tools
• Objectives
• To understand the phenomenon
• To develop and apply new techniques and tools

5
US Northeastern 1965
6
Cascading outage case study

• Aug 14, 2003 Northeastern Blackout Example
• Stage 1 slow steady state progress
• 12-1414pm, several lines and 1 gen outage
• 1505-1541pm, 3 FE 345KV lines outage
• 1539-1559pm, collapse of 138KV system
• Load shedding of 1000MW at 1541pm or 1500MW
  at 1605pm could have prevented the blackout
• 1605pm, trigger event outage of Sammis-Star
  line
• 1605-1609pm, 2 345KV 138KV lines outages
• Stage 2 fast transient progress
• 1609-1610pm, multiple power plants tripped
• 1610-1613pm, fully cascade in neighboring
  areas

7
Cascade Sequence
1) 406
2) 40857
3) 41037
4) 41038.6
8
Cascade Sequence (cont.)
5) 41039
6) 41044
7) 41045
8) 413
9
Interaction between system-wide and local levels

• Local disturbances to system security
• 12-1414pm, Reduced security although secure
• 1505-1541pm, 3 line outages, N-1 insecure
• System security to local disturbances
• 1539-1559pm, collapse of 138-KV system
• 1605pm, Sammis-Star outage due to overload and
  low voltage
• Local disturbances to system security
• Sammis-Star outage triggered cascading outage
• Possible good interactions
• Load shedding at 1541pm or 1605pm
• Backup relay not trip at power swing and
  overload, give time to other controllers and
  system operators

10
Outline

• Introduction
• Background
• Proposed Solutions
• Detailed Algorithm and Data Needed
• Next Step
• Discuss

11
Background
Typical Cascading Blackout
12
Causes

• Non-technical factors
• Change in operating procedures due to
  deregulation
• Aging infrastructure and lack of investment while
  the stress on the system is increased
• Inadequate personnel training for new operating
  conditions
• Conclusion
• More investment and better tools are needed

13
Causes (cont.)

• Technical factors
• Reduced operating margins
• Increased system complexity
• More difficult protection setting coordination
• Inadequate traditional security analysis
• Lack of understanding of the cascades and
  availability of effective support tools
• Conclusion
• Understanding and preventing cascades is a
  challenging problem

14
Relay Operations

• Problems
• Relay operation is a major contributing factor.
  So monitoring relay operation and extracting
  information are very important. Only aiming at
  relay behaviors under the situation when
  multi-events (frequency change, voltage change)
  happened simultaneously.
• Dependability
• Security

15
Background

• Impacts
• Local Level
• Relay operations can be assessed by real time
  monitoring tool
• Fault analysis and classification tool can also
  be implemented
• System Level
• Improve situational awareness of system operator
  under dynamic disturbance situation
• Provide reference of decision-making for system
  operator under emergent situation

16
Cascading Blackouts vs Time
t
Triggering Events
Slow Progression
Fast Progression
Final Blackout
Preconditions
Prevent blackout by acting as early as possible
The slow pace in the Initial stages allows
the time for remedial actions
An Example Tripped elements vs time (from Aug.
14, 2003 blackout final report)
17
Outline

• Introduction
• Background
• Proposed Solutions
• Detailed Algorithm and Data Needed
• Next Step
• Discuss

18
Our Solution
Local Algorithm
Interaction between Local-level and System-level
System Algorithm
19
System Algorithm
System Monitoring and Control
Local Monitoring and Control
System Status
Security Analysis
Real-time Fault Analysis
Routine-based
Neural Network
Local Monitoring
Event-based
Synchronized Sampling
Disturbance Report
Security Control
Relay Operation Monitoring
Steady-State
Event Tree Analysis
Transient Stability
Measurements
Control
Control
Measurements
Power System and Protection System
20
Use of Local Information at the System Level

• Know exact local disturbance information in
  real-time
• Obtain local diagnostic support and predict
  future events (i.e., line overload, relay
  misoperation)
• Make better control decision based on correct
  local information
• Evaluate system security and take actions to
  preserve it
• Benefits Help operators have good situational
  awareness
• Provide operators with
  decision-making support

21
Use of System Information at the Local Level

• Identify threatening contingencies
• Identify vulnerable parts (lines and relays) and
  initiate local tool for careful monitoring
• Block relay misoperation during extreme
  conditions or make correction after system-wide
  analysis
• Find and store emergency control means ready for
  expected contingencies
• Find emergency control means for real time
  unexpected events
• Benefits Effective interaction between system
  and local
• actions and operator decision making
  support

22
Interactive scheme procedures

• Step 1 Routine security analysis performed by
  the system tool (a) decides security level and
  finds vulnerable elements, then sends monitoring
  command to the local tool (b) identifies
  critical contingencies, and starts associated
  control schemes to find the control means for
  those expected events.
• Step 2 Local monitoring performed by the local
  tool (a) starts analysis when disturbance
  occurs (b) if it finds relay misoperation, it
  makes correction or receives system control
  command for better control (c) reports
  disturbance information and analysis result to
  the system tool.
• Step 3 Event-based security analysis performed
  by the system tool (a) if it finds a match with
  expected event, activates the emergency control
  (b) if it does not find a match, analyzes if the
  system is secure or not (c) if it is not, finds
  new emergency control and activates it.
• Step 4 Update information and go to Step 1.

23
Graphic Demonstration
Substation Level
System Level
System Status Monitoring command
Local Monitoring Tool
Routine-based Security Analysis
Substation 1
Candidate control means
Selected control means
Local Monitoring Tool
Substation 2
Expected
Unexpected
Local Monitoring Tool
Event-based Security Analysis
Disturbance Report
Substation n
24
Outline

• Introduction
• Background
• Proposed Solutions
• Detailed Algorithm and Data Needed
• Next Step
• Discuss

25

• System Monitoring and Control

26
System-wide monitoring and control
27
Steady state control scheme
28
Steady state control scheme (cont.)

• Evaluation by vulnerability index (VI) and
  security margin index (MI)
• Identification of the vulnerable parts
• Prediction of successive events
• Steady state control by network contribution
  factor (NCF), generator distribution factor
  (GDF), load distribution factor (LDF), selected
  minimum load shedding (SMLS) and final control
  means
• Verified by AC load flow

29
Transient stability control scheme (cont.)

• Potential energy boundary surface (PEBS) method
• Analytical sensitivity of the transient energy
  margin
• Stability control classification
  admittance-based control (ABC) and
  generator-input-based control (GIBC)
• For each control, to calculate the energy margin
  variance, and find control means to make energy
  margin positive
• Verified by time-domain simulation method

30
Transient stability control scheme
31
Transient stability control scheme (cont.)

• Potential energy boundary surface (PEBS) method
• Analytical sensitivity of the transient energy
  margin
• Stability control classification
  admittance-based control (ABC) and
  generator-input-based control (GIBC)
• For each control, to calculate the energy margin
  variance, and find control means to make energy
  margin positive
• Verified by time-domain simulation method

32
Data Needed PSS/E

• Purpose Power Flow Analysis
• Power Flow Raw Data File (.raw)
• Slider Binary Data File (.sld)
• Purpose Fault Analysi
• Sequence Impedance Data File (.seq)

33
Data Needed PSS/E

• Purpose Contingencies Analysis
• Subsystem Description Data File (.sub)
• Monitored Element Data File (.mon)
• Contingency Description Data File (.con)
• Load Throwover Data Files (.thr)
• Purpose System Dynamic Analysis
• Dynamics Data File (.dyr)
• Machine Impedance Data File (rwm)

34
Fault Calculation and Analysis
35
Dynamic Stability Analysis
36

• Local Monitoring and Control

37
Local monitoring and control
38
SSFD and NNFDC
39
Synchronized Sampling based Fault Diagnosis
40
Fault Detection Feature

• Short Line Model
• Long Line Model

When no internal fault, those features equal to
zero When there is an internal fault, those
features are related to fault current
41
Synchronized Sampling based Fault Diagnosis
What we need Synchronized Sampling based fault
diagnosis provides a very high accuracy in fault
detection and location. Not depend on any
assumptions about system operating conditions,
fault resistance, fault waveforms, etc. We need
synchronized raw samples of voltage and current
from both ends of transmission line under
specified sampling rate. Sources of
data Relays, PMU, DFR, synchronized with GPS
42
Neural Network Based Fault Diagnosis and
Classification
Overall Scheme
43
Training and Testing Process
Input Pattern (using normalized raw samples)
Pattern Space
(2-D demo)
Testing (Fuzzy K-nearest neighbor classifier)
Training (Self-Organized Clustering Technique)
44
Neural Network Based Fault Diagnosis and
Classification
What we need Neural Network based approach
provides a more accurate fault detection and
classification by using the same data inputs as
distance relay. We need ample enough raw samples
of current and voltage under different situations
to finish and verify the training and testing
process. Once the neural network is well trained,
it is capable for online use. Sources of
data Can be share with synchronized sampling
based approach.
45
Event Tree Analysis
t
46
Example Event Tree for No-fault Condition
Node Scenarios Reference Action
1 No fault in preset zones Keep monitoring
2 Relay does not detect a fault Stand by
3 Relay detects a fault and initiates a trip signal Check the defects in relay algorithm and settings
4 Trip signal blocked by the other device in the system
5 Trip signal failed to be blocked Check communication channel Send blocking Signal if necessary
6 Circuit breaker opened by a trip signal
7 Circuit breaker fails to open Check the breaker circuit.
8 Autoreclosing succeeds to restore the line
9 Autoreclosing fails to restore the line Send reclosing signal to the breaker
10 Breaker failure protection trips all the breakers at the substation
11 No Breaker failure protection or it doesnt work Check the circuit of the breaker failure protection.
47
Event Tree Analysis
What we need Event Tree Analysis provides an
efficient way for real time observation of relay
operations and an effective local disturbance
diagnostic support. According to the
characteristics of generic design for event tree,
the number of event trees is finite. However, it
need much work to set up the system. It is
feasible. Sources of data Relay trip signal,
circuit breaker contact signal.
48
Outline

• Introduction
• Background
• Proposed Solutions
• Detailed Algorithm and Data Needed
• Next Step
• Discuss

49
Implementation

• Obtain the raw synchronized data from Relays,
  PMUs and DFRs
• Starts local fault analysis when disturbance
  occurs using NNFDC and SSFL Tools
• Monitoring and analyzing relay operations
• Reports disturbance information and analysis to
  the Central.

50
Algorithm Description
System Modeling
Relay Testing
Relay Monitoring
Fault Detection
Fault Classification
51
Scenario Demonstration
Data Library
Relay Setting
Relays
Relay Operation
DFRs
Data of Collection
Synchronized sampling data
PMUs
Monitoring Software
Measurement System
52
Monitoring Procedure
Relay Operations
Monitoring Software
ETA
Fault Analysis
Relay Operations Decision
Internal Fault?
Potential Fault?
N
Y

• Relay Operation
• Misoperation
• Un-intended Operation
• Failure Operation

SSFL
Y
SSFDC
NNFDC
53
Outline

• Introduction
• Background
• Proposed Solutions
• Detailed Algorithm and Data Needed
• Next Step
• Discuss

54
Thank You!!!
???