Research On Optimal Scheduling Of Electric Vehicles Based On Virtual Partition

Development of electric vehicles is the solution to environmental degradation andresource depletion and other issues, and effective means of achieving sustainablesocio-economic development.For the distribution network, pure electric vehicles (ElectricVehicles; EVs) is both a power and a load. Disorderly conduct charge and discharge oflarge-scale distribution network EVs will increase the peak load curve poor; while adjustablecharacteristics of EVs charging and discharging behavior, provides a new method for loadshifting for the distribution network load curve. For this reason, analyzing the impact oflarge-scale EVs disorderly conduct charge and discharge to the load curve of the distributionnetwork and scheduling policies for EVs charging and discharging behavior to achieve loadcurve of load shifting distribution network have very important significance.In order to study the charging and discharging characteristics of EVs in distributionnetwork, a EVs battery model has established to reflect the relationship between the opencircuit voltage、battery state of charge and the battery current. A battery inverter model isalso established to describe the relationship between power、 non-power that the batteryinjected into distribution network and current and voltage of the battery.After that, according to EVs day mileage, last returned moments which are suitable forlognormal distribution and normal distribution characteristics, the Monte Carlo samplingmethod is used to analyze the impact of EVs charging load, discharge capacity, integratedfeatures on daily load curve in the distribution network. The examples show that a massiveEVs disorderly conduct charge and discharge will increase the difference between peak andvalley distribution network.EVs that access to the distribution network have characteristics of large-scale number,single station with a small-capacity, fragmented accessing position, which is not conducive tobe unified. In order to make EVs that have similar attributes to be scheduled as a class unit, avirtual partition is needed to the electric vehicles, which use Scale space hierarchicalclustering algorithm (SSHC) and the partition index of dispatch capacity and access locationand achieve the similar properties in the same area of EVs points. The examples show thescale space-based clustering algorithm can effectively achieve EVs virtual partitions.Furthermore, in order to maximize the reduction of peak load curve distribution network,a philosophy of scheduling the EVs by stratification and partition scheduling philosophy isintroduced, which can build a double-optimal scheduling model based on virtual partitions of EVs. The upper level of the model has the goal of minimizing the total system load variance;the lower has the goal of minimizing the results of the actual scheduling model results and theupper scheduling deviation. Top model belongs to non-linear programming problem, whichuses genetic algorithm based on improved ELM; the lower belongs to large-scale model ofinteger programming problem, which uses branch and bound method. The example showsthat the improved genetic algorithms based ELM has fast convergence rate and can quicklysolve the upper model. Double optimal scheduling model based on virtual partitions of EVscan effectively stabilize the peak load.