基于AP聚类算法的RBF神经网络风速预测方法的研究  

RESEARCH ON WIND SPEED PREDICTION METHOD USING RBF NEURAL NETWORK BASED ON AP CLUSTERING ALGORITHM

作　　者：李昊 张煜成 Li Hao;Zhang Yucheng(State Grid Jiangsu Electric Power Co.,Ltd.Nanjing Power Supply Branch,Nanjing 210013,China)

机构地区：[1]国网江苏省电力有限公司南京供电分公司,南京210013

出　　处：《太阳能》2025年第2期54-61,共8页Solar Energy

摘　　要：近年来,江苏地区在迎峰度夏期间出现了较大的电能供给缺口,电力系统频率失稳的风险增加,因此,在电力保供工作中,稳定的风电输出功率愈发重要。考虑到风能的随机性和间歇性,准确的风速预测可以降低风电入网时的附加成本,协助电力系统调度部门调整调度计划,提升电力系统的风电消纳与稳定运行能力。从提高超短期风速预测精度的角度出发,提出了1种基于近邻传播(AP)聚类算法的径向基函数(RBF)神经网络风速预测方法(即“AP-RBF方法”)。首先建立AP-RBF模型,然后以江苏地区某风电场实地采集的实际风速数据为例进行算例分析,对AP-RBF模型的预测效果进行了验证,并对各类预测方法的预测精度和预测效率进行了对比分析。研究结果表明:1)AP-RBF方法通过采用“先计算聚类结果,再计算权值矩阵”的预测模式,克服了传统聚类方法对初值敏感的缺点。2)与常规预测方法相比,AP-RBF方法在整体预测精度上表现最佳,且在保证训练数据质量的基础上具有较快的预测速度。AP-RBF方法的应用对提高风电消纳能力与电力系统频率稳定性具有重要意义。In recent years,there has been a large power supply gap during the period of peak summer electricity consumption to ensure the smooth operation of the power grid in Jiangsu Province,increases the risk of power system frequency instability.Therefore,stable wind power output power has become increasingly important in ensuring power supply work.Considering the randomness and intermittences of wind energy,accurate wind speed prediction can reduce the additional cost of grid-connection for wind power,help the dispatching department of the power system adjust the dispatching plan,improve the wind power consumption and stable operation capability of the power system.This paper proposes a wind speed prediction method using RBF neural network based on AP clustering algorithm(that is"AP-RBF method")from the perspective of improving the accuracy of ultra short term wind speed prediction.Firstly,an AP-RBF model is established,and then the actual wind speed data collected from a wind farm in Jiangsu Province is used as an example for numerical analysis,and the predictive performance of the AP-RBF model is verified.The prediction accuracy and prediction efficiency of various prediction methods are compared and analyzed.The research results show that:1)The AP-RBF method overcomes the sensitivity of traditional clustering methods to initial values by first calculating the clustering results and then calculating the weight matrix for the prediction mode.2)Compared with conventional prediction methods,the AP-RBF method performs the best in overall prediction accuracy and has a faster prediction speed while ensuring the quality of training data.The application of AP-RBF method is of great significance for improving the wind power consumption capacity and frequency stability of power system.

关 键 词：清洁能源 风速 风电 近邻传播聚类算法 径向基函数神经网络 风速预测 精度分析 

分 类 号：TM614[电气工程—电力系统及自动化]