关键词: 时空物负荷模型, 充电需求预测, 选址定容, 双层规划, 生活选择算法

Abstract: In the context of carbon peaking and carbon neutrality, automobiles are one of the main sources of greenhouse gas emissions such as carbon dioxide, so promoting new energy vehicles to replace fuel vehicles is a necessary condition for promoting energy transformation and achieving carbon peaking. As a key node and important foundation for promoting the use of electric vehicles and implementing green transportation, the planning and construction of charging and swapping infrastructure need to solve the problem of accurate prediction of electric vehicle charging and swapping demand. Electric vehicle load distribution is characterized by randomness and fluctuation in time and space, and is reasonably predicted and laid out according to the results. On the basis of the prediction results, we study the siting of electric vehicle charging stations, and the size and capacity setting of various types of equipment to meet the user demand, and consider the costs of construction, operation andmaintenance of charging stations to ensure a certain degree of economy while meeting the user demand.
This paper establishes a space-time-vehicle load model including private electric vehicle, electric taxi and electric bus, considering the characteristics of electric vehicle distribution, distinguishing behavioral characteristics of different vehicle types, multi-capacity charging piles and distinguishing different charging strategies for peak and idle periods, etc. The Monte Carlo method is used to determine the spatial and temporal distribution of electric vehicles to be charged, real-time speed updates, acceptable waiting time for electric vehicle users, remaining power and range. The selection of EV charging piles and charging strategies are also simulated. Based on the prediction results, a bi-level model is established for the location and capacity of electric vehicle charging-swapping-discharging-storage integrated stations. The upper-level planning considers the minimum sum of construction cost, energy storage cost and operation and maintenance cost to optimize the location. The lower-level planning takes the shortest electric vehicle driving distance as the goal to divide the service area, and returns the load to the upper level for capacity optimization. The upper and lower layer transfer to each other and influence each other. The bi-level planning balance the economy of investment and operation with the convenience of charging and driving for EV users.
The model is a NP-hard problem, and the exact algorithm can only solve relatively small-scale problems, so the intelligent optimization algorithm is chosen as the optimal solution to the problem. The life choice-based optimizer is an intelligent optimization algorithm that is simple to compute and efficient to solve on continuous optimization. This paper designs a hybrid coding method to apply to discrete practical problems, which not only retains the advantage of the algorithm's strong ability to find the best, but also allows the continuous optimization algorithm to be applied to solve practical problems in discrete domains. Considering that the algorithm still has the defect of easily returning to a local optimized result, the anchoring effect theory is introduced to design a new updating equation according to the characteristics of human decision-making behavior, improve the population quality and expand the algorithm to the search area, so as to obtain a novel life choice-based optimizer. This new algorithm improves convergence accuracy while balancing global exploitation capabilities with local exploration capabilities.
The new algorithm is combined with Dijkstra's algorithm to solve the new model in order to verify the effectiveness of the model. The experimental results show that the model is reasonable and necessary in distinguishing between multiple car models, different regional attributes, multi-capacity charging piles, and peak and idle charging strategies. Further, to verify the superiority of the algorithm, the new algorithm is compared with other five algorithms for experiments. The results prove that the new algorithm has the advantages of stability and fast convergence. The reliability of the new algorithm is verified by Wilcoxon rank sum test. How to take a more effective method to improve the life choice-based optimizer and apply it to solve the problems of site selection and volume of hydrogen refueling stations is the further research direction.

Key words: space-time-vehicle load model, charging demand forecasting, location and sizing, bi-level planning, life choice-based optimizer