基于贝叶斯随机波动模型的短期风速预测  

SHORT-TERM WIND SPEED PREDICTION BASED ON BAYESIAN STOCHASTIC VOLATILITY MODEL

作　　者：汤鑫[1] 黄仙[1] 

机构地区：[1]华北电力大学控制与计算机工程学院,北京102206

出　　处：《太阳能学报》2014年第11期2236-2241,共6页Acta Energiae Solaris Sinica

基　　金：国家自然科学基金(60974051;61273144);北京市自然科学基金(4122071)

摘　　要：提出一种基于随机波动(SV)模型的短期风速预测方法。该方法引入贝叶斯推理以解决SV参数的估计问题,并以我国西北某风电场实采风速数据作为样本,建立标准SV模型和某一参数服从t分布的SV-t模型。对所建模型与标准广义自回归条件异方差(GARCH)模型的预测能力做了综合比较分析。算例表明:SV模型不但参数辨识简便易行,而且更适合于描述风速序列的波动性。As classical time series not suitable for analyzing the volatility of wind speed sequence, a short-term wind speed prediction approach based on stochastic volatility （SV） model was proposed. Firstly, Bayesian inference was adopted to estimate model parameters; then, field data collected at a wind farm in the northwest of China were applied to develop standard SV model and SV-t model （SV-t was extended from SV in order to describe the diversity of practical wind sequence fluctuation）. The developed models were compared and analyzed comprehensively with corresponding standard Generalized Autoregressive Conditional Heteroscedasticity （GARCH） model. The results showed that SV model is not only more convenient in parameter estimation but also more suitable for describing wind speed volatility.

关 键 词：风速预测 贝叶斯估计 随机波动模型 MCMC 

分 类 号：TM74[电气工程—电力系统及自动化]