基于卷积神经网络的电力市场短期售电量预测方法  

A Method for Predicting Short Term Electricity Sales in the Electricity Market Based on Convolutional Neural Networks

作　　者：王蕾[1] 李斌[1] 李泠聪 张振明 姜涛[1] Wang Lei;Li Bin;Li Lingcong;Zhang Zhenming;Jiang Tao(Northeast Electric Power University,Jilin,China)

机构地区：[1]东北电力大学,吉林吉林

出　　处：《科学技术创新》2024年第1期85-88,共4页Scientific and Technological Innovation

基　　金：吉林省科技发展计划项目《基于新一代人工智能的电力市场售电智慧服务云平台》(20200401097GX)。

摘　　要：电力市场短期售电量预测的精度对优化用电结构以及提高供电可靠性具有重要意义,传统短期售电量预测方法没有考虑偏差电量考核影响、用电行为差异导致电预测精度低,提出基于卷积神经网络的电力市场短期售电量预测方法,首先根据用户的用电负荷率进行分类,获取不同行业的用电特征和需求模式,然后考虑正负偏差电量的影响,设计基于CNN-ResNet的短期售电量预测方法,通过实验分析表明,该方法能够有效提高多因素影响下售电量预测的准确率。The accuracy of short-term electricity sales forecasting in the power market is of great significance for optimizing the electricity consumption structure and improving power supply reliability.Traditional short-term electricity sales forecasting methods do not consider the impact of deviation in electricity consumption assessment and differences in electricity consumption behavior,resulting in low electricity prediction accuracy.A convolutional neural network-based short-term electricity sales forecasting method in the power market is proposed,which first classifies users based on their electricity load rates,Obtain the electricity consumption characteristics and demand patterns of different industries,and then consider the impact of positive and negative deviations in electricity consumption.Design a short-term electricity sales prediction method based on CNN ResNet.Experimental analysis shows that this method can effectively improve the accuracy of electricity sales prediction under multiple factors.

关 键 词：售电量预测 偏差电量 K-means++ CNN-ResNet 

分 类 号：TM73[电气工程—电力系统及自动化] TP18[自动化与计算机技术—控制理论与控制工程]