基于K-means和改进KNN算法的风电功率短期预测系统

Wind Power Short-term Forecasting System Based on K-means and Improved KNN Algorithm

摘要:

为提高风电功率短期预测的准确性，针对KNN（K-Nearest Neighbor algorithm）算法在风电功率预测中的不足，提出了基于K-means和改进KNN算法的风电功率短期预测方法。利用K-means聚类方法确定风电历史样本的类别，对KNN算法中搜索相似历史样本集的方式进行了改进和优化，构建了预测模型，并采用C/S架构实现了预测系统的设计。该系统具有自修正功能，能够随着预测次数的增加，不断修正预测模型，逐渐降低预测的误差率。以吉林省某风电场历史数据为样本进行了仿真分析，结果显示该算法与其它算法相比平均绝对误差和均方根误差最大下降1.08%和0.48%，运算时间提升了5.45%，在风电功率超短期多步预测中具有推广应用价值。

Abstract:

In order to improve the accuracy of short-term prediction of wind power, in view of the shortcomings of KNN (K-Nearest Neighbor algorithm) algorithm in wind power prediction, a short-term wind power forecasting method based on K-means and an improved KNN algorithm is proposed . The K-means clustering method is used to determine the types of historical wind power samples, the method of searching for similar historical sample sets in the KNN algorithm is improved and optimized, a prediction model is constructed, and the C/S architecture is used to realize the design of the prediction system. The system has a self-correction function, which can continuously correct the forecast model as the number of forecasts increases, and gradually reduce the error rate of the forecast. A simulation analysis with historical data of a wind farm in Jilin Province is carried out. The results show that compared with other algorithms, the algorithm has the largest decrease in average absolute error and root mean square error by 1.08% and 0.48%, and the calculation time has increased by 5.45%,ultra-short-term multi-step forecasting has the value of promotion and application.