摘要
针对数值天气预报模型输出的气象参数存在系统误差而导致风电场功率预测精度受到制约的问题,提出了一种基于卡尔曼滤波修正的风电场短期功率预测模型.使用卡尔曼滤波算法对数值天气预报输出的风速数据进行动态修正,并结合其他气象数据形成新的用于风电功率预测的修正气象数据集合;根据原始气象数据和修正气象数据这2个训练集分别建立了风电场功率输出的原始神经网络、修正神经网络的预测模型.经同一时间区间内的实测数据与模型分析数据的对比分析表明:通过卡尔曼滤波修正的风速数据能够很好地跟踪实际风速数据的变化趋势,平均误差与绝对平均误差比较小;所提模型能够显著降低预测结果的均方根误差,使其从未修正前的17.73%降低至11.32%,证明预测精度得到了明显提高.
A Kalman filter based correction model for short-term wind power prediction was proposed to solve the problem of wind energy prediction accuracy constraint induced by the systematic errors in meteorological parameters from the numerical weather prediction （NWP） model. The wind speed data from NWP were corrected dynamically by using the Kalman filter algorithm and the improved NWP set used for wind power prediction was formed by combining the corrected wind speed data with other meteorological data. The original neural network prediction model and the corrected neural network prediction model were trained by using the raw NWP set and the improved NWP set, respectively. The analysis on the comparison between the simulation data and the measured data in a same time interval shows that, the corrected wind speed series by the Kalman filter are very close to observed wind speed; the mean error and the mean absolute error are smaller; the root mean square error decreases from 17. 73% to 11.32%. It seems that the wind power prediction model proposed has a clearly higher accuracy.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2011年第5期47-51,共5页
Journal of Xi'an Jiaotong University
关键词
卡尔曼滤波
神经网络
功率预测
风力发电
Kalman filter
neural network
power prediction
wind power generation
