摘要
电动汽车充电负荷数据的分散性、多样性,导致预测结果与实际观测结果偏差较大。为此,提出基于函数链神经网络(FLNN)的高质量配电网充电负荷需求预测方法。建立电动汽车的充电模型,在电网负荷等不确定性因素下得到电动汽车在充电时的初始荷电状态。基于初始荷电状态,提取出充电时间、充电功率等多个充电负荷特征,明确充电负荷特征值和充电负荷需求。在FLNN的作用下建立扩展函数,映射输入的充电负荷特征值。结合充电均衡指数,获取与实际值最接近的充电负荷预测值。试验结果表明,所提方法在实际应用中充电负荷需求预测值与实际值基本一致,预测的变异系数均值为0.98。该方法能准确预测电网的充电负荷需求。
Significant deviations between prediction results and actual observation results are caused by the dispersion and diversity of electric vehicle charging load data.To address this issue,a high-quality distribution network charging load demand forecasting method based on function link neural network(FLNN)is proposed.An electric vehicle charging model is established to determine the initial state of charge during electric vehicle charging under uncertainties such as grid load,etc.Based on the initial state of charge,multiple charging load features including charging duration and charging power,etc.,are extracted to define charging load characteristics and charging load demand.Under the influence of FLNN,the expanded function is established to map the input charging load feature values.Combined with the charging balance index,the charging load forecast closest to the actual value is obtained.Experimental results demonstrate that consistent charging load demand forecast values with actual values are achieved by the proposed method in practical applications,with a forecast average coefficient of variation of 0.98.The charging load demand of the power grid can be accurately predicted by this method.
作者
佟鹏飞
丁再贤
魏玉仁
屈斐
孙学成
TONG Pengfei;DING Zaixian;WEI Yuren;Qu Fei;SUN Xuecheng(State Grid Haidong Power Supply Company,Haidong 810600,China)
出处
《自动化仪表》
2026年第4期11-15,共5页
Process Automation Instrumentation
关键词
高质量配电网
函数链神经网络
充电负荷
需求预测
特征值
预测值
充电模型
High-quality distribution network
Function link neural network(FLNN)
Charging load
Demand forecasting
Characteristic value
Predicted value
Charging model
