3/2/20101SCADA Data Mining and IT Needs to Improve Plant Operation and DowntimeAWEA Wind Power Asset Management WorkshopGordon RandallGlobal Energy Concepts, LLC116 John StreetLowell, MA [email protected] forWind Power Management classhttp://www.icaen.uiowa.edu/~ie_155/by Andrew KusiakIntelligent Systems Laboratory2139 Seamans CenterThe University of Iowa Iowa City, Iowa 52242 – [email protected]: 319-335-5934 Fax: 319-335-5669http://www.icaen.uiowa.edu/~ankusiakSCADA Defined•Supervisory Control and Data Acquisition system• Actual definitions and descriptions can varysome skip the“supervisory control”vary –some skip the supervisory control part and just handle the data side• Projects need data analysis, not just acquisitionThree Levels of SCADA Systems• Systems focused on site operations– Primarily systems provided by turbine manufacturers• Systems focused on project-level analysisPrimarily smallscale thirdparty systems–Primarily small-scale, third-party systems• Systems designed for enterprise-level, fleet-wide analysis– Almost exclusively third-party systems, especially when handling multiple turbine typesWhat They’re Generally Best At (Although All Systems Vary)TaskManufacturer’s Project SCADAThird-party Project SCADAEnterprise SCADADay-to-day project operationsBest  Good  Fair Month-to-month project performance analysisFair  Best  Good Evaluation and comparison of entire wind portfolioPoor  Fair  Best 3/2/20102Why Use Third-Party Systems?• Provides common system and interface when mixing turbine manufacturers• Generally more customizable for reporting and analysis purposesypp• Provides independent measurements and analysis – the system calculating availability will not be designed by the people who have to pay for low availability• Frequently more functionality and data storageWhy Not Use Third-Party Systems?• Often redundant to some extent, if turbine manufacturer requires use of their SCADA for O&M/warranty purposes•Can be more difficult to get full access to•Can be more difficult to get full access to systems for data collection purposes• Value of analysis tools is limited by quality of data going in• Cost (perceived or actual)Hardware and Software Needs –On-Site System• Most manufacturer’s systems are turnkey installations– Controller/interface at each turbine or other it i i tmonitoring point– Fiber optic cabling or wireless communications across project– Centralized server at operations buildingHardware and Software Needs –Telecommunications• Security important for control: crucial that unauthorized users not control turbines!• High-speed, reliable Internet access required for efficient data transmittalrequired for efficient data transmittal– Typically, T1 speed and reliability necessary– DSL/cable problematic due to speed/availability issues– May not be cheap to get wiring to remote sites3/2/20103Telecommunications (Continued)• Project-specific needs:– Data transmittal to utility– Data transmittal to forecasting servicesIt l tti–Internal presentation– Public presentationHardware and Software Needs –Off-Site/Enterprise Systems• Lots of Storage– Data need to be readily available in order to be useful–Desktop-type database systems generally Desktoptype database systems generally inadequate for management of long-term data– Large project can generate many GBs of data/year – multiply by several projects and several years– Data backup system is (of course) importantWhat Can be Learned by Mining SCADA Data?• Verification of turbine and plant performance• Assessment/prediction of failures– Predictive maintenance of large components (including condition monitoring)(g g)– Evaluation of faults and minor components• Quantification of effects of problems and prioritization of efforts to solve problems• Warranty claim supportObjective: Optimize Operations to Maximize Profit• … not turbine availability, energy production, or project revenue, if at the expense of cost or effortexpense of cost or effort• On-site operations are frequently driven by reactions to short-term problems and may not reflect the best overall strategy3/2/20104Condition Monitoring• The more data you have, the easier it is to discover impending problems– Comparison of measurements across a turbine fleet– Comparison of measurements over time•Modern turbines have huge numbers of sensorsModern turbines have huge numbers of sensors for trend analysis • Interpretation can be tricky! Weighing indications of potential failure vs. replacement cost is tricky• Use of full condition monitoring systems (e.g., vibration analysis)Fault Analysis:Overheating/Power Regulation2%3%ity of Fault400500600700er (kW)Faults Power Curve0%1%0 2 4 6 8 101214161820Wind Speed (m/s)Probabili0100200300PoweFault Analysis:Pitch System ProblemsDowntime vs. Time of Day• Cost-effectiveness of nighttime and weekend fault response plans• Better assessment of lost revenue, considering time-of-day pricing3/2/20105Fault Recovery Time• Are 30-minute faults more realistically 60-minute faults?• Is this being accurately considered in availability calculations?• Would it be more cost-effective to “ride through” faults with lower power output?Environmental Considerations (Examples)• Blade soiling assessment– Cost of reduced power output vs. cost of blade washing•Site access restrictions•Site access restrictions– Cost of lost power vs. cost of snow removal• Heated control anemometry– Cost of downtime due to frozen sensors vs. cost of changes to sensor typesComponent Failure Rate Analysis• SCADA can provide supporting data for component failure predictions and/or serial defect analysis– Tie SCADA to site maintenance logs/parts inventory (e.g., CMMS = Computerized Maintenance Mgt System)(g p g y )– Comparative studies of subcomponents from different vendors• Long data history is (again) very important– Major components should have zero failures until late in project life– By the time there are enough failures to make predictions, it may be too lateWarranty Claims• Is availability being calculated accurately?– Turbines “paused” or otherwise incorrectly included?Turbine kWh AvailabilityT01 10908 100.0%T02 10584 99.8%– Wrong numerator or denominator in availability calculations?– If warranty is