Research On Vehicle To Grid Considering The Health State Of Power Battery

As a flexible and adjustable load,the wide application of electric vehicle(EV)is one of the measures to achieve the goal of "double carbon".Large scale EV disorderly access will expand the load peak valley difference of the distribution network and bring severe challenges to the safe and stable operation of the distribution network.EV can be reasonably arranged for orderly charge and discharge based on vehicle to grid(V2G)technology.When EV participates in V2 G,it will cause additional loss of battery.The loss cost of battery is related to the state of health(SOH).At the same time,SOH determines the maximum power that EV battery can store at present,which will affect EV load distribution.Therefore,this paper studies the health state estimation of power battery,electric vehicle load forecasting,V2 G scene division and V2 G strategy.Firstly,aiming at the problems of difficult parameter identification and poor expansion ability in the traditional SOH estimation methods,a data-driven SOH estimation method for power battery is proposed in this paper.This method extracts the health features from the battery charge curve,selects the health features by using Pearson and Spearman correlation coefficients,and constructs a data-driven SOH estimation model.SOH estimation is carried out on Oxford battery data set,NASA battery data set and battery pack data set.The SOH estimation methods based on SVR,LSTM and one-dimensional CNN are compared.The results show that the SOH estimation methods based on one-dimensional CNN and LSTM have high accuracy and certain universality.Secondly,aiming at the problems of low prediction accuracy and high randomness of model-driven EV load forecasting method,a data-driven EV load forecasting method considering battery SOH is proposed in this paper.This method comprehensively considers the factors such as EV time characteristics,travel characteristics and power characteristics,and constructs the EV load calculation model considering the battery health state.Considering the mapping relationship between other characteristics of EV users and travel end time,residence time and driving mileage,an EV feature prediction model based on two-dimensional CNN is constructed.The predicted EV features are introduced into the EV load calculation model to realize the high-precision prediction of EV load.Finally,this paper proposes a vehicle network interaction strategy considering the battery health state.By integrating the clustering results of K-means clustering,k-medoids clustering,fuzzy c-means clustering and SOM clustering,different typical V2 G scenes in residential areas are divided.Based on the principle of engineering economics,the cost model of unit power loss of power battery considering SOH is constructed,and the EV groups of disorderly charging,orderly charging and orderly charging and discharging are divided according to the time when EV can participate in V2 G and SOH judgment conditions.Aiming at minimizing the total load peak valley difference of distribution network and the total charge and discharge cost of EV users,a V2 G strategy model is established.Under different V2 G scenarios in residential areas,four schemes are compared and analyzed,i.e.disorderly charging without considering battery SOH,disorderly charging with considering battery SOH,participation of battery SOH in V2 G and participation of battery SOH in V2 G.The results show that the V2 G strategy considering battery SOH can reduce the total cost of EV charging and discharging and realize "peak cutting and valley filling".