Wind Turbine Diagnostics Based on Power Curve Using Particle Swarm Optimization

Abstract

In wind energy industry, power curve, the plot of the generated power versus the ambient wind speed, is an important indicator of the performance and health of wind turbines. The nominal power curves differ by manufacturers and types. The actual power curve will deviate from the nominal one because of the turbulence in the incoming wind, turbine health, etc. Power curve is widely used for visual inspection and performance evaluation, but there is no et a quantified approach to use it for diagnostic purpose. We propose an inverse transformation based change detector, called Inverse Diagnostic Curve Detector (IDCD), to track the variation of power curve over time for diagnostics. IDCD is adaptable to different wind turbine types. We use two example wind turbine types to illustrate the adaptation procedure. We select the Gaussian CDF (cumulative density function) in the inverse data transformation method for its fitting accuracy and one-to-one mapping property in its inversion. The dynamic fitting is optimized by particle swarm optimization (PSO) algorithm. IDCD simplifies abnormality detection with a scaler decision threshold. Some failures are predictable such as some major component failure, which causes degradation; other failures are not predictable from turbine information alone such as lightning strike, which happens suddenly and quickly. Early detection of either degradation or sudden faults is beneficial. After a deviation pattern is discovered by comparing it with historical data, the pattern can be used for prognostics to help predict the remaining useful life of a turbine and create an optimal schedule for maintenance and repair tasks.