改进PSO-BP算法的短期电力负荷预测方法
Short-term power load prediction method using improved PSO-BP algorithm
杨亚东 1耿丽清 2杨耿煌 2郝夏毅 1陈庆斌1
作者信息
- 1. 天津职业技术师范大学自动化与电气工程学院,天津 300222
- 2. 天津职业技术师范大学自动化与电气工程学院,天津 300222;天津职业技术师范大学天津市信息传感与智能控制重点实验室,天津 300222
- 折叠
摘要
针对电力负荷的周期性、随机波动性等复杂特点易造成预测精度低等问题,提出一种基于相似日分析、混沌映射优化粒子群算法(particle swarm optimization,PSO)和BP神经网络相结合的短期电力负荷预测方法.采用乘积法量化气象因素与时间因素间的综合相似度,选出综合相似度高的若干历史日作为相似日集;采用相似日集与非相似日集分别训练PSO-BP模型,相似日集的平均绝对百分比误差(mean absolute percentage error,MAPE)降低 3.9%;利用Sine映射对PSO中粒子的速度和位置进行优化,增强PSO算法的全局搜索能力和寻优精度,采用 2 个集合分别训练SPSO-BP模型,相似日集的MAPE降低 19.4%.结果表明,基于相似日分析和SPSO-BP模型的短期电力负荷预测方法可有效提高电力负荷的预测精度.
Abstract
This paper proposes a short-term power load prediction method based on similar day analysis and chaotic map-ping particle swarm optimization(PSO)algorithm and BP neural network to address the complex characteristics of pow-er load,such as periodicity and random fluctuations,which can lead to low prediction accuracy.The product method is used to quantify the comprehensive similarity between meteorological and temporal factors,enabling the selection of several his-torical days with high comprehensive similarity as similar day set.The PSO-BP models are trained separately using the simi-lar day set and the non-similar day set,resulting in a reduction of the mean absolute percentage error(MAPE)for the simi-lar day set by 3.9%.Sine mapping is implemented to optimize the velocities and positions of particles in the PSO and en-hance the global search ability and optimization accuracy of the PSO algorithm.The training of the SPSO-BP models for both datasets leads to a further reduction in the MAPE of the similar day set by 19.4%.The results show that the short-term power load prediction method based on similar daily analysis and the SPSO-BP model can effectively improve the accuracy of pow-er load prediction.
关键词
短期电力负荷预测/相似日/粒子群算法/BP神经网络/混沌映射Key words
short-term power load prediction/similar day/particle swarm optimization(PSO)algorithm/back propagation(BP)neural network/chaotic mapping引用本文复制引用
基金项目
天津市科技计划项目(23YDTPJC00320)
天津市高等学校科技发展基金资助项目(2022ZD010)
出版年
2024
NETL
NSTL
万方数据