Multi-scale analysis of wind power and load time series data

This paper presents novel analyses of high-resolution wind power and electric system load time series data. We use a discrete wavelet transform to resolve the data into independent time series of step changes (deltas) at different time scales, and present a variety of statistical metrics as a function of the time scale. We show that the probability distribution for wind power deltas is not Gaussian, has an exponential shape near the center and is well fit by a power-law in the tails. We provide a physical interpretation for the observed power-law behavior, and discuss the potential significance for modeling studies, prediction of extreme events, and the extrapolation of statistical characteristics to higher wind penetration levels. Several metrics are presented to quantify the degree of auto-correlation in the wind data, and of the correlations between wind and load. We show that the shape of the autocorrelation function is the same at different time scales, a property of self-similar statistical processes that is consistent with the observed power-law behavior.