摘要
针对大型冶金企业专用母线负荷种类多、分布不均、规律性弱等特点,利用自组织特征映射神经网络(self-organizing feature map,SOM)对模糊聚类法进行改进,以选择待预测日的相似日,通过db4小波对相似日负荷数据进行分解、去噪和重构处理后作为后期预测模型的训练样本;采用混沌粒子群算法(chaos particle swarm optimization,CPSO)对最小二乘支持向量机(least square support vector machine,LSSVM)算法的惩罚参数和核函数覆盖宽度进行优化,构造了基于CPSO和LSSVM的母线负荷预测模型。仿真结果表明:该负荷预测模型,将预测结果的相对误差降低到1.998%,预测精度达到了97%,提高了专用母线负荷预测准确性。
Aiming at the large metallurgical enterprise special bus has loads variety,uneven distribution,weak regularity characteristics,using the self-organizing feature map(SOM)improving fuzzy clustering method to select the similar day as the predicting day,then the similar day load data as training samples of late period forecasting model is decomposed,de-noising and reconstructed with db4wavelet,decomposing,denoising and reconstructing the similar day load data,uses the chaos particle swarm optimization(CPSO)algorithm optimize the penalties parameters and kernel function coverage of the least square support vector machine(LSSVM)algorithm,constructes the bus load forecasting model based on CPSO-LSSVM.The simulation result shows that the relative error of prediction result reduced to1.998%,and the prediction accuracy reached97%,can improve the prediction accuracy of the special bus load.
作者
杨波
YANG Bo(Sichuan Mingxing Electric Power Co., Ltd., Suining Sichuan 629000, China)
出处
《宁夏电力》
2017年第5期25-33,共9页
Ningxia Electric Power
