Posters

Abstract

Wind farms play a pivotal role in the transition to renewable energy, yet the underperformance of individual turbines can impact the overall efficiency of a wind farm. Wind farm owners and operators often face challenges in identifying and resolving these low-performance issues, particularly with limited access to full SCADA data. This paper presents a novel tool designed to enhance wind farm operation and maintenance by proactively identifying underperforming turbines and conducting automated root cause analysis. The tool leverages historical SCADA data from individual turbines within a wind farm to detect anomalous behavior and pinpoint potential sources of reduced energy production. The tool employs a multi-faceted approach to identify underperforming turbines. Firstly, it establishes baseline performance metrics for each turbine based on the available historical data, considering environmental characteristics. The operational data is clustered based on well-known factors such as seasonality, day and night operation and wind direction. A first comparison is made across the different turbines in the wind farm identifying operational factors that may induce better or worse performance. Subsequently, real-time SCADA data is continuously monitored and compared against these established baselines. Significant deviations from expected performance, such as lower power output or lower annual energy production may trigger an alert, flagging the turbine as potentially underperforming. Some additional criteria are used to flag turbines that may not be operating as expected, even if they are not properly underperforming. Those criteria involve increased downtime or abnormal component behavior. Upon detection of an underperforming turbine, the tool initiates an automated root cause analysis. This process involves: Data Aggregation: Relevant historical SCADA data, including power output, wind speed, rotor speed, pitch angle, yaw position, and fault codes, is collected and aggregated. Data filtering: A first classification and filtering is performed based on commonly known trends. Pattern Recognition: Advanced machine learning algorithms for model creation, clustering, and anomaly detection, are employed to identify patterns and anomalies in the collected data. These algorithms can detect subtle variations in turbine behavior, such as gradual performance degradation or intermittent faults. Fault Diagnosis: The tool integrates a knowledge base of known fault modes and their associated SCADA signatures. By comparing the identified anomalies with the knowledge base, the tool can generate a ranked list of potential root causes, such as gearbox issues, generator problems, siting challenges or control system malfunctions. Report Generation: A comprehensive report is automatically generated, summarizing the identified underperforming turbine(s), the detected anomalies, the ranked list of potential root causes, and visualizations of relevant data to aid in further investigation. This automated tool offers several key benefits: Proactive Maintenance: Early identification of underperforming turbines allows for timely maintenance interventions, minimizing downtime and maximizing energy production. Reduced Maintenance Costs: By pinpointing the most likely root causes, the tool can guide technicians directly to the source of the problem, reducing diagnostic time and minimizing unnecessary repairs. Improved Decision Making: The tool provides valuable insights into turbine performance, enabling data-driven decisions regarding maintenance schedules, spare part inventory, and overall wind farm optimization.