Potential Evaluation of Large-Scale Electric Vehicle Demand Response Resources Based on K-means+LSTM Network

The potential evaluation of large-scale electric vehicle demand response resources is an important prerequisite for Regional Integrated Energy Systems to carry out supply-demand interaction and optimal scheduling. However, most of the existing work does not use real charging data to analyze the demand response potential of electric vehicles. Thus, this paper proposes a large-scale electric vehicle demand response resource potential evaluation method based on the K-means+LSTM network. First, use the real data of a large-scale charging pile in an area to analyze the typical electricity consumption patterns of large-scale electric vehicle users through a cluster analysis; secondly, analyze the time period during which the complete charging events of each electric vehicle user are concentrated through a second clustering. Determine the aggregate capacity of electric vehicles that can participate in demand response every hour, finally establish an LSTM neural network model, train and test the proposed prediction model with the clustered data set, and use an area as an example to predict the scale of different time periods in the future the total potential value of electric vehicles that can participate in demand response. It is verified by examples that the model can be used as one of the reference -basis for determining the load reduction of each user, and it also provides key technical support for the subsequent formulation of an orderly charging and discharging optimization control strategy for electric vehicles and optimizing the coordinated operation of the Regional Integrated Energy System.