基于PCA-GBDT算法的电力数据用电量预测方法

刘晖

欢迎来到能源电力期刊网！

您当前的位置：

首页 >

文章列表页 >

基于PCA-GBDT算法的电力数据用电量预测方法

更新时间：2026-04-08

- 基于PCA-GBDT算法的电力数据用电量预测方法
- Power consumption forecasting method based on PCA GBDT algorithm
- 电力大数据 2026年第2期
- 作者机构：
  
  国网江西省电力有限公司宜春供电分公司
- 作者简介：
- 基金信息：
- DOI：
  中图分类号：
- 纸质出版：2026
- 稿件说明：
移动端阅览
刘晖. 基于PCA-GBDT算法的电力数据用电量预测方法[J]. 电力大数据, 2026,(2).

Liu Hui. Power consumption forecasting method based on PCA GBDT algorithm[J]. 2026, (2).
刘晖. 基于PCA-GBDT算法的电力数据用电量预测方法[J]. 电力大数据, 2026,(2). DOI：

Liu Hui. Power consumption forecasting method based on PCA GBDT algorithm[J]. 2026, (2). DOI：

摘要

在电力大数据结构复杂的背景下

研究针对传统梯度提升树算法在数据维度较高的情况下

其预测精度较低且效率低下等问题

提出一种结合主成分分析的梯度提升树算法模型

并对机器学习组件和算法参数进行了优化改进

旨在对电力数据的用电量进行精准预测

提高电力行业的工作效率。结果表明

迭代次数为5时

与支持向量机和未改进的算法相比

研究所提算法的误差分别降低了0.33%和0.63%。在居民用电中

该算法的平均绝对百分比误差与未改进的算法相比减少了18.48%。对该算法的参数进行优化后

发现研究所提优化方案的用电量误差最小

其均方误差和均方根误差分别为258.47万kwh和16.34万kwh。结果表明

研究所提算法能够对电力数据用电量进行精准预测

该算法对电力行业的发展规划和布局调整具有重要的研究意义。

Abstract

Under the background of complex power big data structure

aiming at the problems such as low prediction accuracy and low efficiency of traditional gradient lifting tree algorithm with high data dimensions

this paper proposes a gradient lifting tree algorithm model combined with principal component analysis

and optimizes and improves machine learning components and algorithm parameters

aiming at accurate prediction of power consumption of power data. Improve the efficiency of the power industry. The results show that when the number of iterations is 5

the error of the proposed algorithm is reduced by 0.33% and 0.63%

respectively

compared with support vector machine and unimproved algorithm. In residential electricity consumption

the average absolute percentage error of the algorithm is reduced by 18.48% compared with the unimproved algorithm. After optimizing the parameters of the algorithm

it is found that the power consumption error of the proposed optimization scheme is the smallest

and its mean square error and root mean square error are 2

584

700 kwh and 163

400 kwh

respectively. The results show that the proposed algorithm can accurately predict the power consumption of power data

and it has important research significance for the development planning and layout adjustment of power industry.

关键词

Keywords

references

陈伟伟,荆世博,边家瑜,等.基于长短期记忆神经网络的电力用电量预测[J].机械与电子.2024,42(05):18-23.

谭曾盛,王志兵.基于WPA-Prophet模型的区域用电量预测[J].现代信息科技.2024,8(06):132-135.

蔡翔,朱莉.改进深度神经网络的用户用电量预测方法[J].湖北工业大学学报.2023,38(01):15-20.

李丽,张振臣,刘国军.节能调度下短期预估分布式光伏发电功率模型研究[J].自动化仪表.2023,44(11):106-110.

魏梦飒,李强,许成娣,等.基于数据挖掘算法的用户用电量需求自动预测模型[J].自动化技术与应用.2023,42(12):15-17+94.

Hora SK, Poongodan R, De Prado RP, et al. (2021). Long short-term memory network-based metaheuristic for effective electric energy consumption prediction[J]. Applied Sciences. 2021, 11(23): 11263-11281.

成贵学,乔臻,滕予非,等.基于电力数据挖掘的涉污企业用电量预测方法研究[J].现代电子技术,2022,45(15):151-156.

李浩宇.基于SparkML+PCA-GBDT的行业数据挖掘模型构建[J].粘接,2024,51(04):193-196.

常俊晓,金之榆,卢姬,等.基于集成聚类和XGBoost的短期光伏发电功率预测[J].浙江电力.2021,40(10):102-107.

刘琼,张豹.基于GBDT算法的锂电池剩余使用寿命预测[J].电子测量与仪器学报.2022,36(10):166-172.

冮君泽,李海明.基于AGNN-GBDT的链上欺诈账户检测模型[J].国外电子测量技术.2023,42(08):102-110.

Yu L, Zhou R, Chen R et al. Missing data preprocessing in credit classification: One-hot encoding or imputation.Emerging Markets Finance and Trade[J].2022, 58(2): 472-482.

郭耀扬,张利,郝颖,等.基于分行业用电特性与多因素影响的区域级短期用电负荷曲线预测[J].电力系统保护与控制,2025,53(13):82-92.

Wu Y, Wu K, Li W, et al. Peak-load-regulation nuclear power unit fault diagnosis using thermal sensors combined with improved ICA-RF algorithm[J]. Sensors. 2021, 21(21): 6955-6982.

向俊伟,李玲娟.融合PCA降维和均值漂移聚类的协同过滤推荐算法[J].南京邮电大学学报(自然科学版).2023,43(03):90-95.

朱常鹏,刘元超,李刚.Spark实时大数据处理技术在线开放课程建设与实践[J].现代信息科技.2021,5(12):195-198.

李航.融合机器学习的用电检查与数据预测技术分析[J].集成电路应用.2024,41(07):220-221.

Antonesi G, Cioara T, Toderean L, et al. A machine learning pipeline to forecast the electricity and heat consumption in a city district[J]. Buildings. 2023, 13(6): 1407-1425.

黄子璇,夏壬焕,张雄涛.基于注意力机制和图卷积的电信客户流失预测[J].计算机工程与设计,2023,44(06):1685-1691.