Statistical Analysis And Re-generation Techniques For The Time Series Of Photovoltaic Power Output

There is an explosive development and deployment of PV power worldwide since 2009. Regardless of already-installed total capacity, generation, policy or market potential, most nations show the vigorous development force and high-spirited trend of PV power. Under this circumstance, how to accommodate solar power safely and stably is urgent to be researched.PV power is featured with random fluctuations and its output cannot be forecasted precisely. Meanwhile, due to the internal complexity, the stochastic characteristics of the PV power are hard to be described by simple probability density functions. For the analysis, optimal operation, and planning of the power systems with a large scale of PV station integrated, the time sequence of the PV output is often used. The focus area of this dissertation is to draw sufficient information on the statistic character from existing measurements and produce similar PV output sequences with the same concerned statistic characters.The component decomposition method, statistic characteristic analysis and re-generation strategy is proposed in the paper. Considering the fact that the PV power output is partially stochastic and partially deterministic, the PV power time series is decomposed into three parts: the normalized ideal output curve representing the regular changes of solar radiation, amplitude parameter corresponding to attenuations caused by atmosphere, and random component associated with local cloud movements. Stochastic features are analyzed respectively for these three major parts. Then stochastic time series are generated based on Vector Auto-Regressive(VAR) model considering the spatial and temporal correlations. The proposed method is verified based on real PV power measurements obtained from 6 PV power stations in Gansu and PV generation of 5 regions in Germany. The results show that the generated series keep all stochastic characteristics nearly the same with the original series. Moreover, temporal correlation characteristic of single PV power station and spatial correlation characteristic between multiple PV power stations are both preserved.