基于聚类分析的风电场短期功率预测方法研究

The Study of Short-Term Wind Power Generation Prediction Methods Based on Clustering Analysis in Wind Farm

摘要: 为了充分利用历史风速数据所蕴含的信息，本文根据风速和风电功率的日相似性提出基于聚类分析的短期功率预测方法，通过对原样本数据进行预处理，选取与预测日NWP特征参数相似的历史日数据，以此作为建立模型的训练样本，将气象部门提供预测日的NWP信息作为预测日的特征参数，计算特征参数间的欧式距离，以此作为相似性度量的依据，最后利用聚类后的相似样本建立预测模型，以NWP数据为输入参数，实际风电功率为目标值，经过训练后得到聚类风电功率短期预测模型。经实际风电场测试，预测精度明显提高。

Abstract: In order to make full use of information contained from historical wind data, this paper proposed a short-term power prediction method based on clustering analysis according to the daily similarity property of wind speed and wind power. By the preprocessing of the original sample data, when calculating the Euclidean distance among the characteristic parameters, the history data which are similar to the NWP characteristic parameter of the prediction day are used as the training sample; the NWP information provided by the meteorological department is used as the characteristic parameter of the prediction day. This Euclidean distance is used to be the basis of the similarity measure. Finally, this paper uses the similar samples after clustering to establish the short-term prediction model which is using the NWP data as the input parameter, using the actual wind power generation as the target data. Testing by the actual wind farms, the prediction precision is improved obviously.