Fractional analysis and synthesis of the variability of irradiance and PV power time series
The prediction of the power output of the photovoltaic (PV) system o ers useful information for planning the operation and management strategy, which also helps to maintain or even to improve the power-supply reliability and quality. For this purpose, the power output of a PV system is sampled and documented for the Milagro PV plant at Navarra. The long memory indicator, the Hurst parameter H, is characterised for the power output time series of several consecutive days of October 2009. It shows that the PV power time series is a non-stationary process whose behaviour resembles a long memory process. Therefore, both the ARMA(p,q) model and the ARFIMA(p,d,q) model are built for the PV power output prediction. The orders, p and q, of the models are determined by the Akaike Information Criterion (AIC). The prediction performance of each model is quantified by the mean square errors (MSE). Comparison shows that the ARFIMA model exhibits much better prediction performance than the ARMA model at the same or even smaller orders of p and q. In addition, the original 1s-interval time series is re-sampled at 30s, 60s and 300s intervals, and the two types of models are adopted to the re-sampled time series again in order to investigate the relationship between the sampling rate and their performance.
ARMA ARFIMA PV load prediction sampling rate.
Xianming Ye Xiaohua Xia Jiangfeng Zhang YangQuan Chen
Department of Electrical, Electronic and Computer Engineering,University of Pretoria, Pretoria, 0002 Center for Self-Organizing & Intelligent Systems (CSOIS) Electrical & Computer Engineering Dept. Uta
国际会议
The Fifth Symposium on Fractional Differentiation and Its Applications(第五届国际自动控制联合会分数阶导数及其应用会议)
南京
英文
1-7
2012-05-14（万方平台首次上网日期，不代表论文的发表时间）