电动汽车接入配电网不平衡负荷数据渐进学习方法

Progressive learning method for unbalanced load data of electric vehicles connected to distribution network

为了降低电动汽车接入配电网不平衡负荷数据处理过程中错误数据影响,以最短的时间达到平衡状态,提出了电动汽车接入配电网不平衡负荷数据渐进学习方法。计算电动汽车接入配电网的总充电功率,分析电动汽车接入配电网不平衡负荷变化规律。通过添加合成样例,平衡少数训练信息,合成不平衡负荷数。采用渐进学习的方式学习不平衡负荷数据,并利用非对称更新策略,获取错误预测有标签数据。通过更新线性分类器方差,获取去除错分合成的不平衡负荷数据,经过渐进学习后逐渐实现负荷平衡。由实验结果可知,该方法应用后的负荷变化情况与实际情况一致,且最短耗时为0.2ms,具有良好的学习效果。

In order to reduce the influence of wrong data in the process of unbalanced load data processing of electric vehicle connected to distribution network and reach the equilibrium state in the shortest time,a progressive learning method of unbalanced load data of electric vehicle connected to distribution network is proposed.Calculate the total charging power of electric vehicles connected to the distribution network,and analyze the variation law of unbalanced load of electric vehicles connected to the distribution network.By adding synthesis examples,balance a few training information and synthesize the number of unbalanced loads.The method of progressive learning is used to learn the unbalanced load data,and the asymmetric update strategy is used to obtain the labeled data of false prediction.By updating the variance of the linear classifier,the unbalanced load data synthesized by removing misclassification is obtained,and the load balance is gradually realized after progressive learning.The experimental results show that the load change after the application of this method is consistent with the actual situation,and the shortest time is 0.2 ms,which has a good learning effect.