SHORT-TERM WIND POWER FORECASTING USING SIMILAR DAY BASED NEURAL NETWORKS WITH BAYESIAN LEARNING ALGORITHMS

摘要：

This paper presents an adaptive short-term wind power forecasting method that combines similar day method with BP neural networks trained with adaptive Bayesian learning algorithms. Wind speeds and the corresponding wind power measured every 10 minutes are used to construct the models input-output sample data. Four forecasting scenarios of using different inputs are modeled and compared: (1) only using n previous wind speed observations, (2) only using n previously wind power observations, (3) using n previous wind speed and wind power observations, and (4) using n previous wind power observations and 1 similar days wind power observation. For each scenario, the models parameters are adaptively optimized by using Bayesian training algorithm to ensure that at any given time t, the model outputs approximate the real wind generation at time t+k, where k is the forecasting horizon. To further verify and compare the models performance, each scenario is tested under two horizons: 10 minutes ahead and 30 minutes ahead. The results indicate that the scenario of using wind generation observations as inputs slightly outperforms that using wind speeds for short-term wind power forecasting, whereas the scenario of using both wind speed and wind power as inputs generates the higher forecasting accuracy compared with the first and second scenarios. Besides, the results indicate that combining similar day method can also further improve the forecasting accuracy. It is also verified that the shorter the forecast horizon, the higher the forecasting accuracy. Moreover, the Bayesian training method also demonstrates its ability to forecast the possible wind generation in the format of confidence interval for a given probability.

关键词： short-term wind power forecasting Bayesian learning Neural network similar day method

作者: Gong Li Chuanhua Zhou Xiuli Qu Xun Yu Jing Shi

作者单位: Department of Industrial and Manufacturing Engineering, North Dakota State University, Fargo,ND 5810 School of Management Science and Engineering, Anhui University of Technology, Maanshan,China 243002 Department of Industrial and Systems Engineering, North Caroline A&T State University,Greensboro, NC Department of Industrial and Manufacturing Engineering, North Dakota State University, Fargo,ND 5810

会议类型: 国际会议

会议名称: The 7th International Green Energy Conference & The 1st DNL Conference on Clean Energy(第七届绿色能源国际会议暨第一届DNL洁净能源会议(IGEC)

会议地点: 大连

会议语种:英文

页码: 54

在线出版日期: 2012-05-28（万方平台首次上网日期，不代表论文的发表时间）