| As the penetration rates of both electric vehicle loads and renewable energy in the power grid continue to rise,their disorder and uncertainty have increased the mismatched power fluctuations between power generation and load,and the contradiction between urban energy supply and demand has become increasingly prominent.In order to maintain the safe,reliable,and economic operation of the power grid,it is necessary to reasonably guide electric vehicles to participate in the demand response to the power grid.Therefore,this paper has taken large-scale decentralized electric vehicles as the research object,and has carried out research from three aspects of electric vehicle data mining,charging behavior cluster analysis,and orderly charging demand response scheme design.A sophisticated strategy based on data mining by Random Forest and orderly charging hybrid demand response of electric vehicles.The main research results of this paper are as follows.1)This paper proposes a charging load model construction method based on urban electric vehicle charging consumption records.The general steps of mining charging load information from charging consumption records are established.The processing rules of missing values of samples are based on the nature and proportion of missing values.The supplement to the construction of electric vehicle charging load model elements refers to the rated output power of the charging piles.The establishment of the abnormal sample identification process and the analysis of the cause of the abnormal sample are based on the actual charging scene of the electric vehicle.The example transforms the entire year of electric vehicle charging consumption records in Dundee,UK into corresponding charging behaviors and charging loads.It also analyzes changes in charging behavior,operating conditions of charging stations,and differences in charging loads on weekdays,weekends and holidays.2)A clustering technique for electric vehicle charging behavior based on Random Forest and Multi-dimensional Scaling Method is proposed.This paper analyzes the convergence of Random Forest and constructs unlabeled charging sample sets into labeled mixed sample sets to train decision trees for Random Forest.The sample dissimilarity matrix is generated from the voting results,and Multi-dimensional Scaling Method is used to obtain the clustering results from the dissimilarity matrix.In the example,clustering analysis is performed respectively on the charging behaviors on weekdays,weekends and holidays to verify the practicability of Random Forest in charging behavior clustering.The differences and distribution ranges of each type of charging behavior in the start time,end time,charging duration,charging rate and charging capacity are analyzed,as well as user characteristics and emerging scenarios.Finally,Mixed Gaussian Model is used to fit the characteristic parameter distribution.The research results verify the diversity of electric vehicle charging behavior models and provide a basis for designing electric vehicle participation elaborate demand response strategies.3)This paper proposes a hybrid targeted charging demand response strategy for electric vehicles based on the points mechanism and TOU power price.The charging points mechanism theory and implementation process are proposed.And according to the occurrence time and value of mismatched power,as well as the type and number of users participating in demand response,the charging reward and punishment points and the scope of the program are set,combined with TOU power price.Considering the uncertainty of the users’ response,the classical probability models of the users’ participation in demand response under the TOU power price,the points mechanism and the hybrid mechanism are established respectively.The benefits of both the power company and the users are considered comprehensively,and a multi-objective optimization model with reward and punishment coefficients and unit charging point value as variables is established.The example compares the changes in the load of electric vehicles under the three mechanisms and their impact on the parameters of grid operating.The results show that the hybrid targeted charging demand response strategy can better play a role in reducing the cost of power grid construction,optimizing power grid operation,protecting the company’s revenue,reducing user expenditure,and has a strong ability to adjust to emergencies. |
