Data Processing Method of Photovoltaic Power Based on Singular Spectrum Analysis

The fluctuation of the photovoltaic power(PV) and the noise interference in the communication measurement lead to the uneven quality of the measured data, this phenomenon brings great difficulty to the prediction based on statistical model. In this paper, based on the study of the chaotic characteristics of photovoltaic power fluctuations, an ultra short-term photovoltaic power prediction data preprocessing focused on the singular spectrum analysis is proposed. Firstly, a data preprocessing evaluation mechanism is established in terms of the noise reduction and prediction effect. Secondly, this paper discusses several parameter improvement methods, and puts forward an improved SSA method to solve the problem of subjective and unsystematic parameter selection. It is proved that chaotic time series preprocessing classical chaotic signals with Gaussian white noise of different signal noise ratios (SNR) is feasible. Finally, based on chaotic-radial basis function (Chaos-RBF) prediction model, it is further applied to the measured data of PV power to analyze its effectiveness and limitations for the preprocessing of the photovoltaic power time series(PPTS). The improved singular spectrum analysis based PV power data processing method proposed in this paper achieves good results in terms of sequence noise reduction and prediction accuracy improvement, and improves the efficiency of Chaos-RBF.