Advanced wind turbine health monitoring from … wind turbine health monitoring from SCADA...
-
Upload
duongtuong -
Category
Documents
-
view
216 -
download
3
Transcript of Advanced wind turbine health monitoring from … wind turbine health monitoring from SCADA...
Advanced wind turbine health monitoring from SCADAmonitoring from SCADA
R li Wi d S G j i id ReliaWind - SuperGen joint side event
Dr. VIHAROS Zsolt JánosHead: Manufacturing and business processes groupg p g [email protected], www.sztaki.hu/~viharos
Engineering and Management Intelligence (EMI)Engineering and Management Intelligence (EMI)MTA SZTAKI
4/19/2010
AgendaAgenda
•Introduction
•The ReliaWind project•The ReliaWind project
WP3 Al ith L i l A hit t f •WP3: Algorithms-Logical Architecture of an Advanced WTG Health Monitoring System• Goals• Results by result groups
•Summary picturey p
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 2
Computer and Automation Res. Inst., HAS (SZTAKI)
EU Centre of Excellence i I f ti
Main Research TopicsKey figuresin Information
Technology, Computer Science and Control
Mathematics and CS• Combinatorial CS• Operations research
y g
Budgetp
• Modeling multi-agent systems• Stochastic systems
Information technology
15 M euros34% basic funding
• Basic and applied research in the field of mathematics, CS, IT Information technology
• Analogic & neural computing• Distributed systems • Cluster and GRID computing
Staff
295
mathematics, CS, IT and automation
• Contract based R&Dactivity mainly on p g
• Component and agent-based prog.• Embedded systems• Human-computer interactions
29567% scientific
complex systems, turn-key realizations
• Transferring up-to-date results and research
Automation• Systems & control theory• Geometric modeling & reverse eng.
Cost structure
personnel 38 %
results and researchtechnology to industry and universities
• Intelligent manufacturing• Digital Enterprises, prod. networks
operational 51 %investments 11 %
SZTAKI’s role in Hungary
• The only research institute on IT in the country
g y
in the country• Key role in national research
projects in CS, IT and controlP ti i ti i th d t d • Participation in the graduate and postgraduate education
• Contract-based projects with the l t i d t i l fi i largest industrial firms in Hungary (GE, Nuclear Power Plant Paks, MOL, Knorr Bremse)C t t ki g g th • Computer networking: e.g the largest regional center
• Supercomputing CentreANN model(s)
Search foroptimal solutions
Near tooptimal
solution(s)
Optimisation fordetermining optimal
solution(s)
• GRID Competence CentreANN model(s)replacing the
simulation
Plant
Simulation
ANN learning
Addition of thetested solution(s)
to the trainingset of the ANN
Performancecalculation(s)
Parameters tobe optimised
Performancecalculation(s)
Parameters tobe optimised
Simulation based testing
Training parameters
Inputs
Endsolution
International cooperation
• ERCIM, WWW ConsortiumCIRP IFAC IFIP t • CIRP, IFAC, IFIP, etc.,
• > 30 projects in the 5th Framework Programme
• > 30 projects in the 6th Framework Programme
• > 25 projects in the 7th 25 projects in the 7th Framework Programme
• NSF, ONR, ARO projectsC t t b d k• Contract-based work
• Virtual Institute with IPA-Fraunhofer, Germany
Engineering and Management Intelligence Laboratorywww.emi.sztaki.hu
EMI’s missionResearch and elaboration of techniques applicable Research and elaboration of techniques applicable
for handling complex production and businesssystems working in an uncertain, changing
environment, with special emphasis on artificial intelligence and machine learning approaches.
EMI’s present activities•Intelligent manufacturing processes and systems•Management of complexity, changes and disturbances in productionproduction•Production planning and scheduling A t b d t l t t •Agent-based control structures
•Production networks, supply chain information systemsTh di it l f t f t l i d l i ti •The digital factory, factory planning and logistics
management•Quality management modelling and optimisation of •Quality management, modelling and optimisation of process chains•Data mining (production related databases)g (p )
•Department of Manufacturing Science and Technology, Budapest University of Technology and EconomicsBudapest University of Technology and Economics
ReliaWind: Reliability focused research on ti i i Wi d E t d ioptimizing Wind Energy systems design,
operation and maintenance: Tools, proof of concepts guidelines & methodologies for aconcepts, guidelines & methodologies for a new generation• Identify Critical Failures and Components Understand • Identify Critical Failures and Components, Understand Failures and Their Mechanisms
•Define the Logical Architecture of an Advanced Define the Logical Architecture of an Advanced WTG Health Monitoring System – Leader: SZTAKI
• Failure -mitigation, -detection, -location, and -prognosis, i t ti d l i d i i tmaintenance generation and planning, decision support
19/04/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 8
ReliaWind partnersp
•Wind turbine and wind farm producers
•Component manufacturers
•Research institutions engineering consulting•Research institutions, engineering, consulting
•Wind farm owners and other companies: as members of the End user panel”members of the „End user panel”
ReliaWind stepsReliaWind steps
•Identify Critical Failures and Components•Understand Failures and Their Mechanisms•Define the Logical Architecture of an Advanced WTG H lth M it i S tWTG Health Monitoring System•Demonstrate the Principles of the Project FindingsT i t d th Wi d E t •Train partners and other Wind Energy sector
stakeholdersDisseminate the achieved new knowledge through •Disseminate the achieved new knowledge through
Conferences, Workshops, Web Site and Media
System architectureSystem architecture
Planned (to-do) maintenance activities
Scheduler Management aspects
Possible maintenance activitiesPreventive
maintenance
Failure
Structure, ReliabilityStructure, ReliabilityStructure, ReliabilityStructure, Reliability Detected
Failure prognosis
Failure & probability
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Mainte-nance
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Mainte-nanceMainte-nance
Detectedfailure and location
Failure (probability=1) Failure detection and prognosis
Failure effect mitigation
Maintenance/Event dataMaintenance/Event dataMaintenance/Event dataMaintenance/Event data
19/04/2010 Overview of the overall scope of Work Package 3 11
mitigation(control)
System architectureSystem architecture
Task 3.6: Planning for Maintenance Activities
Task 3.7: Support for Maintenance Decision Making
• developing maintenance decision making tools• algorithms will be tested in real conditions
Task 3.0: Overall Integration Assurance
• Guarantee overall consistency of monitoring architecture
Planned (to-do) maintenance activities
Scheduler Management aspectsTask 3 5: Generation of Maintenance Activities
g
• planning maintenance activities• algorithms will be tested in real conditions
• Ensure the harmonious and mutually beneficial development of tasks
Possible maintenance activitiesPreventive
maintenance
Failure Task 3.4: Agents for Fault Prognosis
• provide a prognosis for the incipient failure
Task 3.5: Generation of Maintenance Activities
• generating maintenance activities to effect improvements to the availability of the wind turbine• algorithms will be tested in real conditions
Structure, ReliabilityStructure, ReliabilityStructure, ReliabilityStructure, Reliability Detected
Failure prognosis
Failure & probability
Task 3.3: Agents for Failure Location
provide a prognosis for the incipient failure modes• the algorithms attempt to predict the residual life and reliability of the components• on the wind turbine and wind farm• algorithms will be tested in real conditions
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Mainte-nance
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Mainte-nanceMainte-nance
Detectedfailure and location
Failure (probability=1) Failure detection and prognosis
Task 3.2: Agents for Failure Detection
• detection of specific incipient failure modes• determination and identification which turbine system has progressed from a nominal to a malfunction condition• each turbine system will be adapted to each of the turbine system’s aims
• locate which turbine system has progressed from a nominal to a malfunction condition• integrated with the knowledge developed in T3.1 and T3.2.• on the wind turbine and wind farm• algorithms will be tested in real conditions
WP2
Failure effect mitigation
Maintenance/Event dataMaintenance/Event dataMaintenance/Event dataMaintenance/Event data Task 3.1: Mitigation of failures by control action
• drive designers towards more ‘intelligent’control algorithms• investigating the extent to which such advanced control algorithms might be extended to address reliability issues
b th l d l d i t l ill
the turbine system s aims• algorithms will be tested in real conditions
g
19/04/2010 Overview of the overall scope of Work Package 3 12
mitigation(control)
• both closed loop and supervisory control will be considered• algorithms will be tested in real conditions
3 0: Overall Integration Assurance3.0: Overall Integration Assuranceanalysis Business Workflows
The Workflow describes the diagrams processing informations of the Reliability Monitoring
«information»Manual records of
«information»O ti l R t
«information»
«information»Failure Forcast
Failure Forcast GenerationEvent1
Manual records of Maintenance
Operational Report Work Orders
«information»SCADA Data
«information»Failure Modes
Mitigation of Failure byControl
Mitigation of Failures byMaintenance
Failure DetectionMonitoring Signals
«information»Component Status
«information»«information»Pre enti e Maintenance Acti ities
«information»Sheduling Constraint Add-Hoc Maintenances
Maintenance Scheduling
Preventive Maintenance Activities Sheduling Constraint
«information»Maintenance PlanEvent2 Maintenance Plan
Decision support
13
Manager
Wind Turbine Logical NodesWind Turbine Logical Nodes
14
Critical components failuresCritical components, failures
•Critical components were selected in WP1(Id tif C iti l F il d C t ) b d (Identify Critical Failures and Components) based on field data
I ifi d b WP2 (U d d F il d Th i • It was verified by WP2 (Understand Failures and Their Mechanisms) using reliability models and calculations
Th 6 t iti l t l t d•The 6 most critical components were selected• The 5 most critical failures were selected for all of th iti l tthe critical components
6 t iti l t 5 t iti l f il 6 most critical components x 5 most critical failures
30 i i l f il id ifi d i id WP3= 30 critical failures were identified inside WP3
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 15
T3 1: Mitigation of failures by control actionT3.1: Mitigation of failures by control action
Specific objectives achieved:•Description of potential fault tolerance control algorithms.g•Assessment of controller mitigation actions for each failure mode when applicableeach failure mode, when applicable•Advanced control algorithms have beeninvestigated to address reliability issuesinvestigated to address reliability issues•Possible controller mitigating actions have b d ib d A l i f th t ti l been described. An analysis of the potential use of these mitigating actions on the most critical f il h b f dfailures has been performed.
19/04/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 16
T3 2: Agents for Failure DetectionT3.2: Agents for Failure Detection
•Main achievements• An advanced unsupervised learning system has been set up to facilitate automatic extraction of data from SCADA d ill i f id ifi bl f l SCADA and an illustration of identifiable faults. • An unsupervised learning method for SCADA alarm clustering has been developed clustering has been developed. • supervised learning method has been developed to create SCADA signal models and generalise detection create SCADA signal models and generalise detection algorithms, giving examples on Generator & Gearbox.• Non-conform situation detection algorithms were • Non conform situation detection algorithms were developed and non-conform situations were identified
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 17
DatasourcesDatasources
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 18
Alarm clusteringAlarm clusteringNo Expression 1 Expression 2 Expression 3
1 TOP TOP
2 GROUND GROUND
3 CONNECTIONS UPS CONNECTION3 CONNECTIONS UPS CONNECTION
4 LIGHT LIGHTING
5 RELAY AND BASE RELAY BASE
6 BLADE REPAIR BLADE REPAIRING BLADES REPAIRING
7 ELEVATOR REPAIR ELEVATOR REPAIRING
8 VIBRATIONS SENSOR VIBRATIONS SENSORS
9 YAW SENSOR YAW SENSORS
COMMUNICATION FAILURE GROUND COMMUNICATION10 COMMUNICATION FAILURE GROUND
GROUND COMMUNICATION FAILURE
11 GEARBOX FILTER REPLACEMENT GEARBOX FILTERS REPLACEMENT
12 LOW BUS VOLTAGE LOW VOLTAGE IN BUS
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 19
12 LOW BUS VOLTAGE LOW VOLTAGE IN BUS
13 COOLING MOTOR FROZEN FROZEN COOLING MOTOR
Failure clustersFailure clusters
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 20
Maintenence driven failure detectionMaintenence driven failure detection
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 21
Non-conform situation detectionNon-conform situation detectionNon‐conform situation detection ‐ estimation of the gearbox bearing
temperature by a neural network modell (Model validity: ambient temperature between 4 and 10 C)
90
100
110
120
130
90
100
110
7%]
Values_for_Model_INPUT_2 Values_for_Model_INPUT_1
Gearbox bearing temperature_MODEL_ESTIMATES Gearbox bearing temperature_MEASURED
Ambient temperature (for model vaildity) Error_%
50
60
70
80
90
50
60
70
80
(%) [lim
it: +/‐17
peratures
0
10
20
30
40
20
30
40
50
stim
ation error (
Temp
‐30
‐20
‐10
0
10
Mod
el es
Time ‐ a year
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 22
T3 3: Agents for Failure LocationT3.3: Agents for Failure Location
•SCADA signals can be classified into operational signals and functional signals• SCADA operational signals, such as wind speed, shaft speed and power produced• SCADA functional signals, such as bearing and winding
d l b i i il temperatures and lubrication oil temperatures, pressure and pitch angle
•Location dependent•Location dependent
FMECA SCADA i l i d d •FMECA – SCADA signal mapping was needed to develop failure location algorithms
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 23
T3 4: Agents for Fault PrognosisT3.4: Agents for Fault Prognosis
•Reliability related estimation• Estimation of the residual life time of a component• …by considering cummulative SCADA data, too
• This ensures having concrete, field, turbine and component relevant estimation
Can be updated continuously• Can be updated continuously
•Statistical model building is possible based on past SCADA datap
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 24
T3 5: Generation of Maintenance Activities
•Preparation of a template
T3.5: Generation of Maintenance Activities
• for describing maintenance activities and• their circumstances
•This gives also the related database contentThis gives also the related database content
It can be used to feed the scheduler with the •It can be used to feed the scheduler with the maintenance assignments
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 25
T3 6: Planning for Maintenance ActivitiesT3.6: Planning for Maintenance Activities
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 26
The scheduler and a resultThe scheduler and a result
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 27
T3 7: Support for Maintenance Decision MakingT3.7: Support for Maintenance Decision Making
•A set of decision support reports
4/19/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 28
Advanced wind turbine health monitoring from SCADA
Planned (to-do) maintenance activities
Scheduler Management aspects
Possible maintenance activitiesPreventive
maintenance
Failure No Expression 1 Expression 2 Expression 3
1 TOP TOP
2 GROUND GROUND
No Expression 1 Expression 2 Expression 3
1 TOP TOP
2 GROUND GROUND
Structure, ReliabilityStructure, ReliabilityStructure, ReliabilityStructure, Reliability Detected
Failure prognosis
Failure & probability
3 CONNECTIONS UPS CONNECTION
4 LIGHT LIGHTING
5 RELAY AND BASE RELAY BASE
6 BLADE REPAIR BLADE REPAIRING BLADES REPAIRING
7 ELEVATOR REPAIR ELEVATOR REPAIRING
8 VIBRATIONS SENSOR VIBRATIONS SENSORS
9 YAW SENSOR YAW SENSORS
10 COMMUNICATION FAILURE GROUND
GROUND COMMUNICATION FAILURE
11 GEARBOX FILTER REPLACEMENT GEARBOX FILTERS REPLACEMENT
12 LOW BUS VOLTAGE LOW VOLTAGE IN BUS
13 COOLING MOTOR FROZEN FROZEN COOLING MOTOR
3 CONNECTIONS UPS CONNECTION
4 LIGHT LIGHTING
5 RELAY AND BASE RELAY BASE
6 BLADE REPAIR BLADE REPAIRING BLADES REPAIRING
7 ELEVATOR REPAIR ELEVATOR REPAIRING
8 VIBRATIONS SENSOR VIBRATIONS SENSORS
9 YAW SENSOR YAW SENSORS
10 COMMUNICATION FAILURE GROUND
GROUND COMMUNICATION FAILURE
11 GEARBOX FILTER REPLACEMENT GEARBOX FILTERS REPLACEMENT
12 LOW BUS VOLTAGE LOW VOLTAGE IN BUS
13 COOLING MOTOR FROZEN FROZEN COOLING MOTOR
Non‐conform situation detection ‐ estimation of the gearbox bearing temperatureby a neural networkmodell
Non‐conform situation detection ‐ estimation of the gearbox bearing temperatureby a neural networkmodell
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Mainte-nance
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Statistical Analysis
Maintenance/
Reliability
ReliabilityModel
Mainte-nanceMainte-nance
Detectedfailure and location
Failure (probability=1) Failure detection and prognosis
‐30
‐20
‐10
0
10
20
30
40
50
60
70
80
90
100
110
120
130
0
10
20
30
40
50
60
70
80
90
100
110
Mod
el estim
ation error (%) [lim
it: +/‐17
%]
Temperatures
Time ‐ a year
temperature by a neural network modell (Model validity: ambient temperature between 4 and 10 C)
Values_for_Model_INPUT_2 Values_for_Model_INPUT_1
Gearbox bearing temperature_MODEL_ESTIMATES Gearbox bearing temperature_MEASURED
Ambient temperature (for model vaildity) Error_%
‐30
‐20
‐10
0
10
20
30
40
50
60
70
80
90
100
110
120
130
0
10
20
30
40
50
60
70
80
90
100
110
Mod
el estim
ation error (%) [lim
it: +/‐17
%]
Temperatures
Time ‐ a year
temperature by a neural network modell (Model validity: ambient temperature between 4 and 10 C)
Values_for_Model_INPUT_2 Values_for_Model_INPUT_1
Gearbox bearing temperature_MODEL_ESTIMATES Gearbox bearing temperature_MEASURED
Ambient temperature (for model vaildity) Error_%
Failure effect mitigation
Maintenance/Event dataMaintenance/Event dataMaintenance/Event dataMaintenance/Event data
19/04/2010 Overview of the overall scope of Work Package 3 29
mitigation(control)
ContactContact
Dr. VIHAROS Zsolt János
Head: Manufacturing and business processes Head: Manufacturing and business processes group
[email protected], www.sztaki.hu/~viharos
Engineering and Management Intelligence (EMI)g g g g ( )MTA SZTAKI
19/04/2010 Dr. Zsolt János VIHAROS: Advanced wind turbine health monitoring from SCADA 30