Smart Asset Management Stephen McArthur s.mcarthur@eee.strath.ac.uk
Drivers Environment Solar radiation Wind speed and direction Atmospheric pressure Relative humidity Precipitation Conservator tank Key requirements: State and health of assets Real-time rating Prognostics Main tank (3 phases) Temperature Vibration Acoustic Internal UHF probe Tap changer Temperature Vibration Cooling circuit Dissolved gas Temperature (external and internal) Moisture Fans Load current Oil pump motor Temperature Load current Vibration Cooling radiators Condition monitoring is increasing: In terms of new sensors and sensor technology In terms of more condition monitoring systems In terms of deployment, both on-line & offline Improved engineering support is necessary: In terms of managing and interpreting data In terms of corroborating evidence from different sensors and monitoring systems Provision of decision support
Smart Asset Management New sensor technology, artificial intelligence and advanced software techniques to embed intelligence within plant and equipment, integrated with: Knowledge/models of physical behaviour Knowledge/ models of degradation Materials knowledge Statistical models of asset performance Self-learning monitoring and diagnostic systems: Adapt to new plant and equipment Can diagnose defects in the absence of detailed experience of applying the monitoring technologies
ASSET MANAGEMENT PROGNOSTIC FRAMEWORK Across multiple plant items Reusable, generic, framework ASSET MANAGEMENT DECISION METHODS Optimal Condition Monitoring Policies Strategic Management of System Assets Optimal Outage Planning Asset Management Decision Making Under Extreme Uncertainty ASSET PROGNOSTICS Unified Model of Plant Combine: Online monitoring data Codified expertise / knowledge Statistics-based health & degradation models Physics based degradation models Using: AI based methods Statistical techniques Continuous Learning of Asset Behaviour and Degradation Physical Models of Asset Degradation Statistical Models of Asset Degradation On-line Condition Monitoring
Unlocking the true value of CM Combine condition monitoring with real time network control decisions Local intelligence Local data management Link condition monitoring with utility asset management systems combine business and technical information Substation D Local intelligence Local data management Local intelligence Local data management Local intelligence Local data management Substation A Substation B Substation C
EPSRC AMPerES Demonstrator - Two sister transformers - Manufacturer: GEC Witton - 275/132kV, 180MVA - One fine, one in poorer health - Transfix on-line dissolved gas monitoring
Environment Solar radiation Wind speed and direction Atmospheric pressure Relative humidity Precipitation Conservator tank Tap changer Temperature Vibration Main tank (3 phases) Temperature Vibration Acoustic Internal UHF probe Cooling circuit Dissolved gas Temperature (external and internal) Moisture Fans Load current Oil pump motor Temperature Load current Vibration Cooling radiators
SGT1 sensors Oil cooling circuit Main tank Pumps (2) Gases (Transfix & Hydran), top oil temp., bottom oil temp., bottom oil humidity External temp. (6), vibration (4), acoustic emission, oil pressure Temperature, vibration, load current Fans (4) Load current Environment Weather station, solar radiation
SGT2 sensors Oil cooling circuit Main tank Pumps (2) Gases (Transfix & Hydran), top oil temp., bottom oil temp., bottom oil humidity External temp. (6), vibration (4), acoustic emission, oil pressure Temperature, vibration, load current Fans (4) Load current Environment Weather station, solar radiation
User requirements Generic handling of data sources Learn per-item normal behaviour Periodic re-learning Conventional data interpretation Retain all data User interface is important
How do we deliver a Smart Grid which employs innovative products and services together with intelligent monitoring, control, communication, and self healing technologies? Distribute intelligence and control: Provide localised autonomy within the power system Break down the complexity Manage and interpret data locally Arbitrate and co-operate globally Implement automated data interpretation techniques Automatically aggregate interpreted data into meaningful information Provide plug and play architectures flexible and extensible Deliver tailored information to support various engineering functions Control centres Asset managers Field support
Extensible & Dynamic Architecture
Anomaly Detection -A class of machine learning techniques -Potential for false alarms -Therefore, Conditional Anomaly Detection
Conditional Anomaly Detection -Aims to reduce false alarms -Classifies an anomaly if context doesn t explain outliers Context is environmental weather parameters Only looks for anomalies in transformer data Uses statistical models of environment and transformer parameters
The need for wireless sensing A generic diagnostic condition monitoring architecture has been created through AMPerES A number of novel sensors have also been developed through AMPerES However, deployment of new sensors is challenging.. Widescale deployments can be underpinned by Wireless Sensor Networks (WSNs) with in-built data processing and diagnostics
Partial discharge diagnostics: The conventional approach Integrate into a wireless CM sensor
Sensor architecture overview
+ + Wireless sensor technology Agent-based architecture Knowledge-based Diagnostics
How do we deliver a Smart Grid which employs innovative products and services together with intelligent monitoring, control, communication, and self healing technologies? Distribute intelligence and control: Provide localised autonomy within the power system Break down the complexity Manage and interpret data locally Arbitrate and co-operate globally Implement automated data interpretation techniques Automatically aggregate interpreted data into meaningful information Provide plug and play architectures flexible and extensible Deliver tailored information to support various engineering functions Control centres Asset managers Field support