Risk Assessment Of Shared Electric Vehicles Participating In Reserve Service

Under the background of high proportion of renewable energy penetration in the future,the power system will become an important form to promote the absorption of renewable energy through aggregation of flexible demand side resources.Due to the strong uncertainty of renewable energy output,it has become a hot research point to develop flexible resources other than conventional power sources to achieve the balance of power supply and demand under the scenario of high proportion of renewable energy.This paper evaluates the reserve service risk of shared electric vehicle(EV),a demand side resource belonging to a single economic entity with low deployment cost,participating in power grid reserve service,to provide reference for coping with the challenges of safe and reliable economic operation of high proportion of renewable energy power system and solving the problems of energy transformation and development.The main research contents of this paper are as follows:(1)In view of the lack of public data to support the forecasting of shared EV schedulable capacity,this paper proposes a method to construct a spatio-temporal data set of shared EV schedulable capacity based on the 3-minute time-sharing rental service data provided by a foreign car sharing operator.Firstly,the historical track data of shared cars are extracted to analyze their travel rules under different date types.Moreover,the functional types of community grid belonging to shared car parking spots are identified,discriminated,and aggregated to divide functional areas.Finally,the sample data set of shared EV schedulable capacity in each functional area is constructed by mileage estimation of vehicle historical travel trajectory,which provides data for training and verification of shared EV schedulable capacity forecasting model.(2)Aiming at the scheduling capacity prediction problem of shared EV,this paper establishes the scheduling capacity forecasting model based on the scheduling capacity data set,a model-agnostic meta-learning algorithm is introduced to train multi-tasks to obtain good model initialization parameters,and features are extracted from network inputs through convolution neural network(CNN).Then,combined with long short-term memory(LSTM)and attention mechanism,the spatio-temporal forecasting of shared EV schedulable capacity is realized.Finally,an example is used to verify that the proposed forecasting model has high prediction accuracy and strong generalization ability for multitask prediction,and the forecasting results provide high-precision data support for subsequent risk assessment of shared EV participating in power grid reserve auxiliary services as a demand side response.(3)Aiming at the risk assessment problem of shared EV participating in power grid reserve service,this paper constructed a risk assessment model of shared EV as system reserve considering the risk caused by demand response(DR)uncertainty,and used CVa R to measure the system risk loss caused by response uncertainty.The simulation results show that shared EV can replace part of the rotating reserve provided by conventional generator units,reduce the system load cutting risk while reducing the reserve cost of system configuration,and provide a new choice for high proportion of renewable energy power system reserve decision.