基于支持向量机的新能源台区相户关系辨识方法

Identification method of phase household relationship in new energy stations based on support vector machine

提出一种基于随机模糊模拟理论的改进T型关联度和支持向量机的用户相户辨识方法。针对风、光不确定性,采用随机模糊模拟理论构建风电与光电的随机模型,确定电动汽车离并网时刻、日行程数的概率密度函数,利用拉丁超立方抽样技术对概率密度函数进行抽样,采用频谱聚类统一管理的方法获得聚类相似的电车用户,对聚类相似的电车用户采用不同的充放电策略,得到电动汽车充放电功率时刻分布;对不同风光场景下和不同电动汽车充放电策略下的用户电压时间序列特征数据进行重构;利用改进T型关联度和支持向量机的两阶段相户辨识方法对单一风、光场景和多风、光场景下的电动汽车用户进行相户辨识。通过实例仿真分析证实所述方法具有较强的辨识精度,普适性强且易实现,可适应智慧城市的发展。

This paper presents a user identification method based on stochastic fuzzy simulation theory with improved T-type correlation degree and support vector machine(SVM).Aiming at the uncertainty of wind and light,the stochastic model of wind power and photoelectricity is constructed by using the stochastic fuzzy theory,and the probability density function of electric vehicle off-grid time and daily travel number is determined.The Latin hypercube sampling technology is used to sample the probability density function,and the EV users with similar clusters are obtained by using the method of spectrum clustering unified management.Different charging and discharging strategies are adopted for the EV users with similar clusters to obtain the time distribution of electric vehicle charging and discharging power.The characteristic data of user voltage time series under different wind and light scenes and different EV charge and discharge strategies are reconstructed.A two-stage household identification method based on improved T-type correlation degree and support vector machine is used to identify the household of electric vehicle users in single wind and light scene and multiple wind and light scenes.Through the example simulation analysis,it is proved that the method described in this paper has a strong identification accuracy,is universal and easy to implement,and can adapt to the development of smart city.