摘要
电动汽车的快速发展导致充电负荷水平逐年升高,且具有强随机性、难预测的特点,因此关于充电站负荷预测的研究具有重要意义。首先,针对仅考虑负荷波动趋势的单因素模型预测精度不足问题,分析多重因素对充电站负荷预测精度的影响,建立考虑多重影响因素并基于CNN-LSTM(卷积神经网络-长短期记忆)混合网络结构的负荷预测模型;然后,考虑充电负荷的强随机性对模型的影响,提出基于RF(随机森林)算法的误差修正方法;最后,以真实充电站负荷数据为算例进行仿真验证。研究结果表明,经RF算法修正的CNN-LSTM模型的负荷预测结果能较为精准地覆盖真实值,相较于LSTM单模型和未经修正的CNN-LSTM模型,具有更高的预测精度和实用价值。
The rapid development of electric vehicles has led to a yearly increase in charging load levels,character⁃ized by strong randomness and unpredictability.Therefore,research on load forecasting for charging stations holds significant importance.Firstly,to address the inaccuracy of single-factor forecasting models that only consider load fluctuation trends,this paper analyzes the impact of multiple factors on the accuracy of charging station load fore⁃casting.A load forecasting model is established that takes into account multiple influencing factors and is based on CNN-LSTM(convolutional neural network,long short-term memory).Subsequently,given the impact of strong ran⁃domness of charging load on the model,an error correction method based on the random forest(RF)algorithm is proposed.Finally,the paper conducts simulation verification using real charging station load data as a case study.The research results indicate that the load prediction of the CNN-LSTM model,corrected by the RF algorithm,can accurately cover real values.Compared to the LSTM single model and the non-corrected CNN-LSTM model,it exhib⁃its higher forecasting accuracy and practical value.
作者
赵子鋆
彭清文
邓铭
李琳
邓亚芝
陈柏沅
吴东琳
ZHAO Zijun;PENG Qingwen;DENG Ming;LI Lin;DENG Yazhi;CHEN Boyuan;WU Donglin(State Grid Hunan Electric Power Co.,Ltd.Changsha Power Supply Branch,Changsha 410015,China)
出处
《浙江电力》
2024年第4期21-28,共8页
Zhejiang Electric Power
基金
国网湖南省电力有限公司科技项目(5216A522001Z)。
作者简介
赵子鋆(1985),男,硕士,高级工程师,主要从事电力系统方式计划管理相关工作。;通信作者:彭清文(1993),男,硕士,工程师,主要从事电网调度运行相关工作。
