用户侧电力碳排放量预测方法  

Prediction Method for Carbon Emissions of Power on User Side

作　　者：张禄 严嘉慧 王立永 李香龙 ZHANG Lu;YAN Jia-hui;WANG Li-yong;LI Xiang-long(Electric Power Research Institute,State Grid Beijing Electric Power Company,Beijing 100075,China)

机构地区：[1]国网北京市电力公司电力科学研究院,北京100075

出　　处：《计算机仿真》2024年第9期494-499,共6页Computer Simulation

基　　金：国网北京市电力公司科技项目(20223220005)。

摘　　要：为了帮助电力行业分摊碳排放责任,优化碳减排策略,研究了用户侧电力相关碳排放预测方法。按照先预测用电量,再根据电碳转换换算的方法实现用户侧电力相关碳排放的预测。其中,用户的用电量预测是核心。设计一种多模态嵌入的循环神经网络,考虑用电量的多重影响因素,建模用电序列的短期依赖关系;提出一种历史注意力机制,考虑用户用电习惯的周期性特点,捕获用电序列中的周期性因素。实验结果表明,上述方法的用电量预测结果性能明显地优于一些常用的用电量预测方法,有助于用户侧电力相关碳排放的准确预测。In order to help the power industry apportion the responsibility of carbon emissions and optimize the carbon emission reduction strategy,the prediction of carbon emissions related to power on the user side was studied.According to the method of predicting the power consumption first and then the electric carbon conversion,the carbon emission related to the user side can be predicted.The power consumption prediction of users is the core.A multi-modal embedded recurrent neural network was designed to model the short-term dependence of power consumption sequence considering the multiple influencing factors of power consumption;A historical attention mechanism was proposed to capture the periodicity factors in the power consumption sequence by considering the periodicity of users' power consumption habits.The experimental results show that the performance of the proposed method is significantly better than that of some commonly used power consumption prediction methods,which is helpful for the accurate prediction of power-related carbon emissions on the user side.

关 键 词：用户侧 碳排放 用电量预测 循环神经网络 注意力机制 周期性 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]