摘要
由于风力发电所利用的近地风能具有波动性、间歇性、低能量密度等特点,对风电场的发电功率进行尽可能准确的预测是风电发展的关键.本文根据某风场的实测数据,采用了时间序列中的自回归移动平均模型(ARMA),对风电功率进行了实时预测;为进一步提高风电功率实时预测的精确性,本文提出了一种基于BP神经网络和ARMA组合模型的预测方法,并对上述实测数据采用该方法进行了实时预测.预测结果表明:组合模型的预测结果与单独的自回归移动平均模型相比,风电功率的实时预测的均方根误差和百分比误差分别减少了4.01%和3.25%,工程中可以采用该组合模型对风电功率进行预测.
Close wind of what wind power generation used features volatility, intermittence and low energy density etc. So it is key to the development of wind power that the accurately predicting the wind power generation power. According to measured data of a wind field, this paper uses time series auto-regression moving average（ARMA） model for wind power real-time prediction. To further enhance the accuracy of real- time prediction of wind power, this paper puts forward a predicting method based on the BP neural network and ARMA model portfolios; and with this method the measured data are used for the real-time prediction. The prediction results show that： compared the results of combination model with ARMA model, wind power real-time prediction of the RMS error and percentage error are reduced to 4.01% and 3.25% respectively; This combination predicting model of wind power can be used to predict the wind power.
出处
《三峡大学学报（自然科学版）》
CAS
2012年第5期48-51,共4页
Journal of China Three Gorges University：Natural Sciences
关键词
风电功率预测
ARMA模型
组合预测
wind power prediction
auto-regression moving average（ARMA） model
combined predicting
