Predictive Analytics in Wind Power
Predictive Analytics in Wind Power Casey Walleck GE Intelligent Platforms
Avoid Surprises
2
Benefits of Predictive Analytics • Avoid surprise equipment failure • Reduce maintenance costs • Optimize resources • Improve availability and reliability 3
Improve Availability
4
Proficy SmartSignal Overview
5
The Proficy SmartSignal Solution Starts with Data from Your Equipment
Generator
Gearbox
Hub
Personalized Empirical Models Pulls every 5-10 minutes
Removes the Effects of Changing Conditions
Post s Emerging Problems t o Incident List
Proficy SmartSignal
6
The Proficy SmartSignal Solution
7
Groups of Sensors to Find Faults Cont rols Model Elect ro-Mechanical Model
Hub Model
8
Vocabulary Act ual - Real time value from data historian Est imat e - Predicted value from SmartSignal estimator engine Residual - Difference between Act ual and Est imat e Alert s (X) - Statistically significant differences between Act ual and Est imat e Incident s () - Represent when Alert thresholds set by rules have been exceeded
Incident s represent abnormal equipment behavior requiring attention 9
Vocabulary
10
Experience Currently monitoring more than 1000 turbines across current install base: • Invenergy • BP • SSE
Install base to increase by more than 70% • enXco • Iberdrola
11
Examples
12
Electro-Mechanical Model: Lube Oil Temperature Controller Issue Temperature controller fixed
13
Electro-Mechanical Model: Main Shaft Bearing Issue Main Shaft Bearing Fault
14
Electro-Mechanical Model: Generator Exhaust Ducting Issue Generator exhaust ducting fixed
15
Electro-Mechanical Model: Exhaust Fan Issue Exhaust fan issue
16
Hub Model: Blade Pitch Spread Issue Re-calibration of pitch controllers
17
Controls Model: Sensor Issue Sensor Failure
18
Conclusion
19
Faults Detected • Lube oil issues • Bearing issues • Generator exhaust ducting issues • Cooling fan issues • Blade pitch issues • Sensor issues 20
Fault Breakdown Performance 2% Equipment 14%
Instrumentation 84%
21
Summary - Key Points Proficy SmartSignal leverages existing SCADA historian data to provide predictive analytics in real time Many fault types can be detected using data from standard SCADA systems and existing instrumentation Optimal detection of impending gearbox problems requires additional vibration sensors 22
Improve Reliabilit y & Availabilit y
• Predictive Analytics • Proficy SmartSignal
Avoid Surprises
• Avoid surprise equipment failure • Reduce maintenance costs • Optimize resources • Improve availability and reliability 23
Avoid Surprises
2
Benefits of Predictive Analytics • Avoid surprise equipment failure • Reduce maintenance costs • Optimize resources • Improve availability and reliability 3
Improve Availability
4
Proficy SmartSignal Overview
5
The Proficy SmartSignal Solution Starts with Data from Your Equipment
Generator
Gearbox
Hub
Personalized Empirical Models Pulls every 5-10 minutes
Removes the Effects of Changing Conditions
Post s Emerging Problems t o Incident List
Proficy SmartSignal
6
The Proficy SmartSignal Solution
7
Groups of Sensors to Find Faults Cont rols Model Elect ro-Mechanical Model
Hub Model
8
Vocabulary Act ual - Real time value from data historian Est imat e - Predicted value from SmartSignal estimator engine Residual - Difference between Act ual and Est imat e Alert s (X) - Statistically significant differences between Act ual and Est imat e Incident s () - Represent when Alert thresholds set by rules have been exceeded
Incident s represent abnormal equipment behavior requiring attention 9
Vocabulary
10
Experience Currently monitoring more than 1000 turbines across current install base: • Invenergy • BP • SSE
Install base to increase by more than 70% • enXco • Iberdrola
11
Examples
12
Electro-Mechanical Model: Lube Oil Temperature Controller Issue Temperature controller fixed
13
Electro-Mechanical Model: Main Shaft Bearing Issue Main Shaft Bearing Fault
14
Electro-Mechanical Model: Generator Exhaust Ducting Issue Generator exhaust ducting fixed
15
Electro-Mechanical Model: Exhaust Fan Issue Exhaust fan issue
16
Hub Model: Blade Pitch Spread Issue Re-calibration of pitch controllers
17
Controls Model: Sensor Issue Sensor Failure
18
Conclusion
19
Faults Detected • Lube oil issues • Bearing issues • Generator exhaust ducting issues • Cooling fan issues • Blade pitch issues • Sensor issues 20
Fault Breakdown Performance 2% Equipment 14%
Instrumentation 84%
21
Summary - Key Points Proficy SmartSignal leverages existing SCADA historian data to provide predictive analytics in real time Many fault types can be detected using data from standard SCADA systems and existing instrumentation Optimal detection of impending gearbox problems requires additional vibration sensors 22
Improve Reliabilit y & Availabilit y
• Predictive Analytics • Proficy SmartSignal
Avoid Surprises
• Avoid surprise equipment failure • Reduce maintenance costs • Optimize resources • Improve availability and reliability 23
