基于Goldstein滤波算法的短期售电量精准预测方法  

Accurate short-term electricity sales prediction method based on Goldstein filtering algorithm

作　　者：李云 杨冰 马晶 石文娟 程思远 LI Yun;YANG Bing;MA Jing;SHI Wen-juan;CHENG Si-yuan(Beijing China-Power Information Technology Company Limited,Beijing 100085,China)

机构地区：[1]北京中电普华信息技术有限公司,北京100085

出　　处：《信息技术》2025年第1期180-185,共6页Information Technology

摘　　要：为提升售电量预测的精准性,提出基于Goldstein滤波算法的短期售电量精准预测方法。采集历史用电数据并绘制电力负荷曲线图,依托于Goldstein滤波算法处理图像,得到可以准确反映用户短期用电规律的数据。利用皮尔逊相关系数筛选出相似电力用户,并合理选取参考日,得到可用于短期售电量预测的数据集。运用曲线拟合分析原理,搭建售电量增长趋势模型,并了解售电量变化影响因素,基于此建立多元回归预测模型,通过粒子群求解得出精准的售电量预测结果。实验结果表明:所提方法的短期售电量预测结果,平均绝对百分比误差保持在5%以下。To improve the accuracy of electricity sales prediction,an accurate short-term electricity sales prediction method based on Goldstein filtering algorithm is proposed.The historical electricity consumption data is collected and a power load curve is drawn,and the Goldstein filtering algorithm is adopted to process images,obtaining data that can accurately reflect users’short-term electricity consumption patterns.Using Pearson correlation coefficient to screen similar electricity users and selecting reference days reasonably,a dataset that can be used for short-term electricity sales prediction is obtained.Using the principle of curve fitting analysis,a model for the growth trend of electricity sales is established,and the influencing factors of electricity sales changes are understood.Based on this,a multiple regression prediction model is set,and the accurate electricity sales prediction results are obtained through particle swarm optimization.The experiment results show that the average absolute percentage error of the short-term electricity sales prediction results of the proposed method remains below 5%.

关 键 词：Goldstein滤波 售电量 皮尔逊相关系数 电量特征 粒子群算法 

分 类 号：TM154[电气工程—电工理论与新技术]