Sensitivity analysis of PCA method for wind ramp event detection

Wind is one of the fastest growing sources of green energy in the last few decades. A special attention has been recently focused on the wind ramps, sudden changes in wind power production caused by surges of wind speed (both increases and decreases). In this paper, sensitivity analysis of Principal Component Analysis (PCA) method is preformed on wind speed dataset with the 5 minute time resolution. Principal Component Analysis results can be used to evaluate conditional probability of forth-coming wind ramp event. Advantage of PCA method compared to conventional approaches is that numerical weather prediction model producing wind forecasts is not required. Two important input parameters are analyzed: the wind ramp duration and the wind ramp magnitude, respectively. Results show rapid increase of the first principal component value, when ramp duration is prolonging, nevertheless other two components seems to be insensitive in this case. Boxplots are illustrated for simple comparability of resultant distributions. The number of down ramp events is always lower than the number of up ramps, and with increasing percentage threshold these numbers are decreasing exponentially. Moreover trends of average PCA values for down ramp events are not monotone with significant jumps and drops.