A method and system for deploying electric bus charging piles considering double uncertainties

Abstract

The present application relates to the technical field of urban public transport, and discloses a charging pile deployment method and system for electric buses considering double uncertainty. The method constructs the charging facility planning as a two-stage stochastic programming model, determines the number, type and power grid expansion scheme of charging piles in the first stage, and optimizes the operation cost in combination with double uncertainty in the second stage. A multi-agent simulation system is constructed to restore the vehicle scheduling and charging process, and two algorithms of reinforcement learning and radial basis function agent optimization are used to solve the model, and the optimal deployment scheme is output. The present application can significantly reduce the total cost of the system, improve the operation stability under high uncertainty, and provide a quantitative basis for power grid expansion.

Description

Technical Field

[0001] This invention relates to the field of urban public transportation technology, specifically to a method and system for deploying electric bus charging stations that considers dual uncertainties. Background Technology

[0002] In the process of electrifying urban public transportation, electric buses have been widely used due to their zero emissions and low noise. However, due to limitations in battery capacity, the actual range of electric buses is significantly lower than the theoretical range, affected by factors such as temperature, humidity, battery degradation, and air conditioning load, requiring frequent charging at depots. Therefore, the scientific planning of electric bus charging facilities has become crucial to ensuring the efficient operation of the public transportation system.

[0003] Planning for electric bus charging infrastructure is a strategic decision that requires accurate cost estimation in conjunction with vehicle operation decisions. Two core uncertainties exist in actual bus operations: first, random fluctuations in travel time due to traffic congestion; and second, irreversible battery capacity degradation caused by charge-discharge cycles. These uncertainties directly alter the state of charge (SOC) and charging time of electric buses upon arrival at stations, significantly increasing the unpredictability of charging demand.

[0004] Existing charging facility planning methods generally do not simultaneously consider the uncertainty of travel time and battery degradation. They are mostly based on fixed timetables and ideal battery conditions for layout design, resulting in a disconnect between planning schemes and actual operation. This leads to problems such as wasted construction investment, high operating costs, lack of quantitative basis for grid expansion, and insufficient system stability, making it difficult to meet the long-term stable operation requirements of electric bus systems under high uncertainty environments. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for deploying electric bus charging stations that considers dual uncertainties, thereby solving the above-mentioned problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] This invention proposes a method for deploying electric bus charging stations considering dual uncertainties, comprising the following steps:

[0008] S1. The EB-CFP problem of electric bus charging facility planning is constructed as a two-stage stochastic programming model.

[0009] S2. Construct a multi-agent bus simulation system that includes intelligent agents for electric buses and intelligent agents for depots;

[0010] S3. Solve the two-stage stochastic programming model using reinforcement learning methods;

[0011] S4. Solve the two-stage stochastic programming model using the surrogate model optimization method to output the optimal charging pile deployment scheme.

[0012] Preferably, in step S1, the construction of the two-stage stochastic programming model involves the following specific steps:

[0013] S101. The first phase is to develop a deployment solution for charging facilities, including the number and type of charging piles at each bus origin / terminal station and depot, and to consider the possibility and cost of grid capacity expansion; the cost of the first phase includes the construction cost of charging piles and the cost of grid capacity expansion.

[0014] S102. The second phase, based on the facility deployment plan determined in the first phase, takes into account the uncertainties of driving time and battery degradation, and optimizes the daily operating costs of electric buses; the total cost includes the total cost of the first phase and the vehicle operating costs of the second phase.

[0015] Preferably, in step S2, the construction of the multi-agent bus simulation system involves the following specific steps:

[0016] S201. Define the intelligent agents in the bus simulation system: In the multi-agent bus simulation system, the intelligent agents include electric bus intelligent agents and depot intelligent agents; the electric bus intelligent agent performs timetable tasks, and the depot intelligent agent manages charging piles, parking spaces and charging waiting queues;

[0017] S202. Electric bus trip allocation: Bus trips are allocated according to the rules of sufficient power, location matching, minimum charging time, and highest SOC priority.

[0018] S203. Formulate charging rules for electric buses: implement charging scheduling according to the minimum charging time, first-come-first-served, and SOC safety zone rules;

[0019] S204. Considering travel time uncertainty and battery degradation uncertainty: embedding the random distribution of travel time and the distribution of battery capacity decay to simulate the impact of dual uncertainties on operation.

[0020] Preferably, in step S3, the reinforcement learning method solves the two-stage stochastic programming model, and the specific steps are as follows:

[0021] S301. Taking the deployment solution of charging facilities in the first phase as the state, simulate the decision-making process in the second phase, and then calculate the total cost value function corresponding to the state.

[0022] S302, using the number vector of fast charging / slow charging piles as the state variable to satisfy the site capacity constraint;

[0023] S303. Construct an action set using ±1 / 0 perturbation decision vectors, and update the state using a random strategy and Monte Carlo sampling;

[0024] S304. Calculate the state value through multi-agent simulation and fit the value mapping of the high-dimensional state space with a neural network.

[0025] Preferably, in step S4, the two-stage stochastic programming model is solved using a surrogate model optimization method, and the specific steps are as follows:

[0026] S401. Radial basis function (RBF) is selected as the proxy model to fit the mapping between the deployment scheme and the total cost;

[0027] S402. A dual-strategy hybrid search combining local and global search strategies is used to generate candidate sample points.

[0028] S403. Candidate points are selected by weighting the dual scores of the agent target value and spatial distance, and the process is iterated until convergence to output the optimal solution.

[0029] Preferably, the formula for calculating the cost of the first stage is:

[0030] (2),

[0031] In the formula, This represents the total cost of the first phase. Indicates the bus stop or depot number. This indicates a meeting point at the bus terminal. This indicates the bus terminal and meeting point. Indicates the type number of the charging station. This represents a set of charging station types. This indicates the cost of the charging station. Indicates the first The first site The number of charging piles, Indicates the first Maximum charging power of similar charging piles Indicates the first The power grid capacity of each station, This represents the unit cost of expanding the power grid.

[0032] Preferably, in step S202, the trip allocation rules for electric buses are as follows:

[0033] (1) The current energy of each electric bus should be greater than the sum of the energy consumption of the trip and the minimum energy that the electric bus should maintain;

[0034] (2) The current location of the electric bus should be consistent with the starting / ending station of the connecting bus route;

[0035] (3) If the electric bus is being charged and the charging time exceeds the minimum required charging time. Then the electric bus can be used to perform other travel tasks;

[0036] (4) If in time There are always multiple eligible electric buses available for the trip. Then, the "highest SOC priority" strategy is applied to select the execution process. The electric buses; specifically, the electric buses with the highest SOC (State of Charge) status among all eligible electric buses will be selected to execute the trip. If no suitable electric bus is available, a new electric bus will be generated to complete the trip. .

[0037] Preferably, in step S203, the charging rules for electric buses are as follows:

[0038] (1) When an electric bus is assigned to a charging station for charging, the minimum charging time must be met. ;

[0039] (2) If all charging stations at the bus origin / terminal station and depot are occupied, the electric bus must wait in the queue until a charging station becomes available;

[0040] (3) If there are multiple electric buses waiting in line for charging, once the charging station is released, the first-come, first-served (FIFS) strategy will be used to select the electric bus to be charged.

[0041] (4) In order to maintain the good health of the electric bus battery, the SOC of the electric bus should be maintained at Within the range.

[0042] Preferably, the motion disturbance step size δ∈{−1,0,1}, and each time only the number of fast charging / slow charging piles is adjusted in the smallest unit.

[0043] Preferably, an electric bus charging facility planning system includes a memory and a processor, wherein the processor executes the above-described method to achieve coordinated optimization of the number and type of charging piles and grid expansion.

[0044] The beneficial effects of this invention are as follows:

[0045] (1) The method of the present invention constructs a two-stage stochastic programming model, which simultaneously covers the construction of charging piles, the expansion of the power grid and the optimization of daily operation, and solves the problem that traditional planning only focuses on the construction layer and ignores the uncertainty of the operation layer.

[0046] (2) The method of the present invention uses a multi-agent simulation system to realistically simulate the entire process of vehicle scheduling, charging queuing, and trip execution, accurately quantifies the impact of trip time and battery degradation on operating costs, and improves the matching degree between planning schemes and actual operation.

[0047] (3) The method of the present invention uses two algorithms, reinforcement learning and surrogate model optimization, to solve the problem. It takes into account both the efficiency of solving large-scale road networks and the optimality of the solution, and can quickly output a stable and reliable deployment scheme.

[0048] (4) The method of the present invention can provide a quantitative decision-making basis for power grid expansion, minimize the total construction and operation costs under the premise of meeting operational needs, and significantly improve the economy and stability of electric bus system under high uncertainty environment. Attached Figure Description

[0049] Figure 1 Overall flowchart of the method of this invention;

[0050] Figure 2 A framework diagram of the reinforcement learning method of this invention;

[0051] Figure 3 Flowchart of the proxy model optimization SBO method of the present invention. Detailed Implementation

[0052] The present invention will be further described below with reference to the embodiments. It should be noted that these are merely examples and descriptions of the inventive concept. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the inventive concept or exceed the scope defined in the claims, they should all be considered to fall within the protection scope of the present invention.

[0053] Example 1:

[0054] like Figure 1 , Figure 2 , Figure 3 As shown, the present invention proposes a method for deploying electric bus charging stations considering dual uncertainties, comprising the following steps:

[0055] S1. The EB-CFP problem of electric bus charging facility planning is constructed as a two-stage stochastic programming model.

[0056] First, let's describe the problem in detail. Suppose that a city's bus company operates a fleet of electric buses (denoted as set) with non-homogeneous battery capacities. This system serves multiple bus routes operating on publicly available timetables. Each route consists of several stops, including a starting station and a terminal station (for loop routes, the starting and ending stations are the same). The daily timetable for each route specifies the departure time of each bus from the starting station and the arrival time at the terminal station. After the end of each day's operations, the buses are parked at the depot and depart from the depot the following day to perform new tasks at the starting station. The basic assumptions are as follows:

[0057] Fully charged departure: Electric buses depart from the depot daily with a full charge, perform their duties according to the published timetable, and return to the depot after completing their tasks;

[0058] Flexible charging: After completing a single or multiple trips, buses can be charged at the depot or using two types of charging piles (set K = {fast charging, slow charging}).

[0059] The specific steps for constructing the two-stage stochastic programming model are as follows:

[0060] S101. The first phase is to develop a deployment solution for charging facilities, including the number and type of charging piles at each bus origin / terminal station and depot, and to consider the possibility and cost of grid capacity expansion; the cost of the first phase includes the construction cost of charging piles and the cost of grid capacity expansion.

[0061] The objective function for total cost can be written as:

[0062] (1),

[0063] In the formula, the first term represents the total cost of the first stage, and the second term represents the vehicle operating cost of the second stage. Considering the uncertainty of the operating stage, it is written in the expected form. The charging facility deployment plan determined for the first phase of the problem is considered as known parameters in the second phase. This represents the expectation operator based on the empirical distribution Q, where y, z, and T are the second-stage decision variables and random variables. This represents the uncertainty of travel time and battery capacity degradation during operation. It is usually assumed that the distribution of travel time and the distribution of battery capacity degradation are independent of each other.

[0064] First, the first phase involves developing a deployment solution for charging infrastructure, including the number and type of charging stations at each bus origin / terminal and station, as well as the expansion of the power grid capacity. The cost calculation formula for the first phase is as follows:

[0065] (2),

[0066] In the formula, the first term represents the construction cost of the charging pile, and the second term represents the cost of expanding the grid capacity. This represents the total cost of the first phase. Indicates the bus stop or depot number. This indicates a meeting point at the bus terminal. This indicates the bus terminal and meeting point. Indicates the type number of the charging station. This represents a set of charging station types. This indicates the cost of the charging station. Indicates the first The first site The number of charging piles, Indicates the first Maximum charging power of similar charging piles Indicates the first The power grid capacity of each station, This represents the unit cost of expanding the power grid.

[0067] S102. The second phase, based on the facility deployment plan determined in the first phase, takes into account the uncertainties of driving time and battery degradation, and optimizes the daily operating costs of electric buses; the total cost includes the total cost of the first phase and the vehicle operating costs of the second phase.

[0068] Having determined the deployment solution for charging infrastructure, the second phase will seek the optimal public transport schedule by minimizing total operating costs while considering travel time and battery degradation uncertainties. The overall objective function for the second phase is:

[0069] (3),

[0070] (4),

[0071] (5),

[0072] (6),

[0073] (7),

[0074] (8),

[0075] (9),

[0076] (10)

[0077] (11),

[0078] (12)

[0079] (13)

[0080] (14)

[0081] (15)

[0082] (16)

[0083] (17)

[0084] For details on the specific meaning of each parameter in the above formulas (3)-(17) and other formulas in the steps of this invention, please refer to Appendix Table 1.

[0085] The objective function formula (3) aims to minimize operating costs, including fixed costs and charging costs. This represents the unit fixed cost of operating an electric bus (EB). The unit charging cost is represented; constraint formulas (4)-(7) are the basic operational requirements and flow conservation conditions; constraint formula (8) ensures that the minimum terminal station dwell time between two consecutive trips is Where M is a sufficiently large positive number; constraint formulas (9)-(12) describe the energy state conservation, that is, the change in the electric bus's charge between two consecutive trips should be equal to the amount of charge; constraint formula (13) specifies the relationship between charging time, charging rate and amount of charge. yes The charging rate of the type of charger; the constraint formula (14) shows that when hour, The constraint formula (15) describes the electric bus's completion of the journey. The subsequent energy change should equal the energy consumption required for the journey. The constraint formulas (16)-(17) describe the relationship between charging time and charging time.

[0086] S2. Construct a multi-agent bus simulation system that includes intelligent agents for electric buses and intelligent agents for depots.

[0087] The specific steps for constructing a multi-agent bus simulation system are as follows:

[0088] S201. Define the intelligent agent in the bus simulation system:

[0089] A multi-agent simulation system consists of agents that interact within the system; in a multi-agent bus simulation system, the main agents are the electric bus agent and the depot agents, including the bus starting station and the terminal station.

[0090] (1) Each electric bus agent simulates the behavior of an electric bus driver. The goal of these agents is to complete all the travel tasks required by the published bus timetable based on the current location and battery capacity (SOC) status, i.e., the electric bus agents execute the timetable tasks;

[0091] (2) The station agent at each bus starting and ending station manages a series of bus charging piles and parking spaces, and controls which buses can use the charging piles during each time period. The agent is responsible for allocating charging piles and parking spaces to buses. If all charging piles are occupied by buses, the station agent also needs to manage the waiting queue. That is, the station agent manages charging piles, parking spaces and charging waiting queues.

[0092] S202, Electric Bus Trip Allocation:

[0093] Electric bus trip allocation aims to assign suitable electric buses to designated timetables, considering factors such as the bus's current location, current state of energy (SOC), and charging station availability. It primarily allocates bus trips based on factors such as sufficient battery power, location matching, minimum charging time requirement, and maximum SOC priority. The specific rules for electric bus trip allocation are as follows:

[0094] (1) The current energy of each electric bus should be greater than the sum of the energy consumption of the trip and the minimum energy that the electric bus should maintain;

[0095] (2) The current location of the electric bus should be consistent with the starting / ending station of the connecting bus route;

[0096] (3) If the electric bus is being charged and the charging time exceeds the minimum required charging time. Then the electric bus can be used to perform other travel tasks;

[0097] (4) If in time There are always multiple eligible electric buses available for the trip. Then, the "highest SOC priority" strategy is applied to select the execution process. The electric buses; specifically, the electric buses with the highest SOC (State of Charge) status among all eligible electric buses will be selected to execute the trip. If no suitable electric bus is available, a new electric bus will be generated to complete the trip. .

[0098] S203. Establish charging rules for electric buses:

[0099] When the State of Charge (SOC) of an electric bus falls below a certain threshold, the bus should be charged at the bus origin / terminal station or depot to maintain a good SOC. The charging process should consider minimum charging time and the availability of charging stations. Charging scheduling is primarily based on minimum charging time, first-come, first-served, and SOC safety zone rules. The specific charging rules for electric buses are as follows:

[0100] (1) When an electric bus is assigned to a charging station for charging, the minimum charging time must be met. ;

[0101] (2) If all charging stations at the bus origin / terminal station and depot are occupied, the electric bus must wait in the queue until a charging station becomes available;

[0102] (3) If there are multiple electric buses waiting in line for charging, once the charging station is released, the first-come, first-served (FIFS) strategy will be used to select the electric bus to be charged.

[0103] (4) In order to maintain the good health of the electric bus battery, the SOC of the electric bus should be maintained at Within the range.

[0104] S204. Considering the uncertainty of travel time and battery degradation:

[0105] Trip time uncertainty and battery degradation uncertainty will affect the trip schedule and further influence the energy consumption of each trip. Considering the distribution of travel time, a timetable incorporating these uncertainties can be generated in advance. Battery degradation uncertainty primarily affects the battery capacity of the electric buses. Considering the distribution of battery degradation, the battery capacity of each electric bus can be determined when generating new electric buses to perform trips. In other words, by embedding the random distribution of trip time and the distribution of battery capacity degradation, the impact of these two uncertainties on operation can be simulated.

[0106] S3. Solve the two-stage stochastic programming model using reinforcement learning. The specific steps are as follows:

[0107] S301. Taking the deployment solution of charging facilities in the first stage as the state, simulate the decision-making process in the second stage, and then calculate the total cost value function corresponding to the state.

[0108] Several key variables exist in reinforcement learning methods: state, policy, action, and reward. In a two-stage stochastic programming model, the charging facility deployment scheme involved in the first stage problem can be considered as a state. For a specific state (i.e., the charging facility layout scheme), the multi-agent electric bus simulation system will simulate the second-stage decision-making process (such as vehicle scheduling and charging arrangements), and then calculate the value function corresponding to that state (i.e., the total cost of the two stages).

[0109] S302, using the number vector of fast charging / slow charging piles as the state variable, satisfies the site capacity constraint.

[0110] The charging facility deployment scheme involved in the first stage problem can be represented by the number and type of charging piles in each bus starting / ending station and depot. Since only two types of charging piles are considered, namely fast charging piles and slow charging piles, the state variables can be represented by the first stage decision vector. This indicates that it should satisfy the following constraints:

[0111] (18)

[0112] That is, the number of charging piles in each bus starting / ending station and depot should not exceed the maximum capacity of the bus starting / ending station and depot.

[0113] S303. Construct an action set using ±1 / 0 perturbation decision vectors, and update the state using a random strategy and Monte Carlo sampling.

[0114] In each visit to the Monte Carlo (MC) method, a stochastic policy is used to implement state transitions. The action set A in the policy is perturbed by the decision vector of the first stage. Build. Define. For time step The action, and all possible actions, are defined by the following equation:

[0115] (19)

[0116] In the formula, Indicates the first The number of fast charging stations at each site Representing the The number of slow charging stations at each site and These represent their respective perturbation step sizes.

[0117] The decision variable is increased or decreased by a step size. Adjustments are needed; this value depends on the problem size. Smaller... It can improve algorithm accuracy, but it will significantly increase the computational burden; increase This can accelerate the solution process. It should be noted that... This is an integer value. To ensure algorithm accuracy, the method of this invention assumes... The range of values ​​is limited to That is, each step only allows the decision variable (number of fast charging / slow charging piles) to be increased, decreased or kept unchanged in the smallest unit.

[0118] S304. The state value is calculated through multi-agent simulation, and the value mapping of the high-dimensional state space is fitted by a neural network to achieve efficient solution under large-scale road network.

[0119] (1) Value Function: Since the goal of the Monte Carlo method is to maximize the value of the reward function, while the two-stage stochastic model in this invention seeks to minimize the total cost, the value function is set as the state. The negative total cost, i.e., the negative objective function:

[0120] (20)

[0121] In equation (20), It corresponds to the solution vector in the first stage. The state; This represents the empirical distribution, which is generated through simulation by a multi-agent simulation system.

[0122] (2) Value function update:

[0123] ① Put all states of Set it to 0, and create a list for each state. To record the value of each access function;

[0124] ②In each episode, from the action set Randomly select an action Execute the selected action , transition to the new state ;

[0125] ③ Simulate the state through a multi-agent simulation system Calculate the corresponding total cost value during the following operation process. ;

[0126] ④ Save to list And update the mean of the list to ;

[0127] ⑤ Repeat steps ②-④ until the maximum episode is reached.

[0128] (3) A neural network is used to establish the mapping relationship between the state and the corresponding value. The value function value can be directly output by inputting the state parameter vector, so as to achieve efficient solution of high-dimensional state space.

[0129] In electric bus networks, the state space becomes extremely large when there are numerous depots, stations, or parking spaces. To address this issue, this invention employs a neural network to establish a mapping between states and their corresponding values. The core idea is to assume the existence of an unknown underlying function that stably maps charging facility deployment schemes (states) to total cost values ​​(values). By constructing a dataset consisting of inputs (state vectors) and outputs (total cost), supervised learning can be used to train a neural network to fit this function. Thus, the value is no longer stored in tabular form but rather as an input parameter vector. The neural network is used for characterization.

[0130] S4. Solve the two-stage stochastic programming model using the surrogate model optimization method to output the optimal charging pile deployment method.

[0131] Solving two-stage models using surrogate function-based optimization methods: The SBO method has been widely used to solve black-box problems or computationally expensive objective function problems. Its basic idea is to use a surrogate function to characterize the relationship between the input and simulated output of the decision variables. In the two-stage stochastic model of the method in this invention, the input is the deployment of charging facilities in the first stage, and the output is the total cost of the system. The specific steps are as follows:

[0132] S401. Radial basis function (RBF) is selected as the proxy model to fit the mapping between the deployment scheme and the total cost;

[0133] A surrogate function is selected, and an iterative process is designed to enhance the "closeness" between the surrogate function and the original objective function until some convergence criteria are met. In this invention, the radial basis function (RBF) model is used as the surrogate function model, as shown below:

[0134] (twenty one),

[0135] In the formula, For proxy functions, This is the RBF function. RBF assumes that the correlation between any two sample points depends only on the distance (given by the norm). definition).

[0136] The method of this invention uses a cubic RBF, namely . ,in All these parameters , and Both can be calculated by solving the following linear equations:

[0137] (twenty two),

[0138] in, , Represents the zero matrix.

[0139] .

[0140] S402. A dual-strategy hybrid search combining local and global search strategies is used to generate candidate sample points.

[0141] The core of the dual-strategy hybrid search method, which combines local and global search strategies, lies in generating and selecting new sample points based on the current proxy model. The specific sample point generation mechanism is as follows:

[0142] make To find the solution with the minimum objective function value among all sample points in the previous iteration, the algorithm uses a dual-strategy hybrid search to generate candidate sample points:

[0143] (1) Local search strategy: Based on the baseline, apply random perturbations (addition and subtraction operations) of different magnitudes (such as small, medium, and large) to generate neighboring candidate sample points;

[0144] (2) Global search strategy: randomly generate new candidate sample points in the full sample space.

[0145] S403. Candidate points are selected by weighting the dual scores of the agent target value and spatial distance, and the process is iterated until convergence to output the optimal solution.

[0146] (1) Target value criterion for the surrogate model:

[0147] Set candidate points For any candidate point , and These represent the maximum and minimum values ​​of the objective function in the candidate point set, respectively. Candidate points The score is calculated as follows:

[0148] (twenty three).

[0149] (2) Based on the newly generated candidate points and the sampled point set The spatial distance between them.

[0150] definition For the l-th newly generated candidate point and set The minimum Euclidean distance among all sampled points. Further definition. and This represents the maximum and minimum distances between all new candidate points and X. The score for candidate point l can be calculated as:

[0151] (twenty four).

[0152] Then, a weighted sum function is used to evaluate the performance of the newly generated candidate sample points, as shown below:

[0153] (25).

[0154] In the formula, and The weights represent the substitution function criterion and the distance criterion. .

[0155] In the above-described method for deploying electric bus charging stations considering dual uncertainties, the specific meanings of each parameter are detailed in Appendix Table 1.

[0156] Table 1. List of Specific Meanings of Each Parameter

[0157] gather definition Bus stop collection Meet at the bus stop terminal Collection of charging pile types All bus routes All electric buses gathered Bus routes The corresponding departure and arrival times, Bus routes The corresponding departure and arrival stations, Bus routes Travel time, Bus routes Energy consumption Lower and upper bounds of the bus battery state. Unit grid expansion cost Power grid capacity of bus origin / terminal stations and depots electric buses Battery capacity electric buses Battery health capacity No. Deployment costs of charging piles No. Maximum charging power of similar charging piles Minimum charging time requirement Shortest station stay time No. Power grid capacity of each station No. The station number of each site No. Charging rate of charging piles No. The first site Number of charging piles A binary variable. Represents an electric bus. After completing the route Accept the route later A binary variable. Represents an electric bus. After completing the route Accept the route later A binary variable. Represents the route. For electric buses The first task to be completed A binary variable. Represents the route. For electric buses The last task that needs to be completed electric buses Leave the terminal station to execute the route Battery status at time electric buses Leave the terminal station to execute the route Battery status at time electric buses Execution route Battery status upon arrival at the final destination electric buses Execution route Battery status upon arrival at the final destination electric buses Execution route With route The amount of charge between electric buses Execution route With route Charging time between

[0158] This invention presents a method for deploying electric bus charging stations considering dual uncertainties. It studies the deployment problem of electric bus charging stations under uncertainties and models the problem as a two-stage stochastic programming problem. A multi-agent simulation system is designed to simulate the operation of electric buses, and two heuristic algorithms are proposed to solve the two-stage programming problem, providing a feasible solution for the site selection of electric bus charging stations.

[0159] Example 2:

[0160] The present invention proposes an electric bus charging facility planning system, including a memory and a processor. The processor executes the steps of an electric bus charging pile deployment method considering dual uncertainties disclosed in Embodiment 1, thereby achieving coordinated optimization of the number and type of charging piles and grid expansion.

[0161] The above is an exemplary description of the invention. Obviously, the specific implementation of the invention is not limited to the above-described manner. Any non-substantial improvement made using the inventive concept and technical solution of the invention, or the direct application of the inventive concept and technical solution to other situations without modification, is within the protection scope of the invention.

Claims

1. A method for deploying charging stations for electric buses considering dual uncertainties, characterized in that, Includes the following steps: S1. The EB-CFP problem of electric bus charging facility planning is constructed as a two-stage stochastic programming model. S2. Construct a multi-agent bus simulation system that includes intelligent agents for electric buses and intelligent agents for depots; S3. Solve the two-stage stochastic programming model using reinforcement learning methods; S4. Solve the two-stage stochastic programming model using the surrogate model optimization method to output the optimal charging pile deployment scheme.

2. The method for deploying electric bus charging stations considering dual uncertainties according to claim 1, characterized in that, In step S1, the construction of the two-stage stochastic programming model involves the following steps: S101. The first phase is to develop a deployment solution for charging facilities, including the number and type of charging piles at each bus origin / terminal station and depot, and to consider the possibility and cost of grid capacity expansion; the cost of the first phase includes the construction cost of charging piles and the cost of grid capacity expansion. S102. The second phase, based on the facility deployment plan determined in the first phase, takes into account the uncertainties of driving time and battery degradation, and optimizes the daily operating costs of electric buses; the total cost includes the total cost of the first phase and the vehicle operating costs of the second phase.

3. The method for deploying electric bus charging stations considering dual uncertainties according to claim 1, characterized in that, In step S2, the construction of the multi-agent bus simulation system involves the following specific steps: S201. Define the intelligent agents in the bus simulation system: In the multi-agent bus simulation system, the intelligent agents include electric bus intelligent agents and depot intelligent agents; the electric bus intelligent agent performs timetable tasks, and the depot intelligent agent manages charging piles, parking spaces and charging waiting queues; S202, Electric Bus Trip Allocation: Bus trips are allocated based on the following rules: sufficient battery power, location matching, minimum charging time met, and highest SOC priority. S203. Formulate charging rules for electric buses: implement charging scheduling according to the minimum charging time, first-come-first-served, and SOC safety zone rules; S204. Considering travel time uncertainty and battery degradation uncertainty: embedding the random distribution of travel time and the distribution of battery capacity decay to simulate the impact of dual uncertainties on operation.

4. The method for deploying electric bus charging stations considering dual uncertainties according to claim 1, characterized in that, In step S3, the reinforcement learning method solves the two-stage stochastic programming model, and the specific steps are as follows: S301. Taking the deployment solution of charging facilities in the first phase as the state, simulate the decision-making process in the second phase, and then calculate the total cost value function corresponding to the state. S302, using the number vector of fast charging / slow charging piles as the state variable to satisfy the site capacity constraint; S303. Construct an action set using ±1 / 0 perturbation decision vectors, and update the state using a random strategy and Monte Carlo sampling; S304. Calculate the state value through multi-agent simulation and fit the value mapping of the high-dimensional state space with a neural network.

5. The method for deploying electric bus charging stations considering dual uncertainties according to claim 1, characterized in that, In step S4, the two-stage stochastic programming model is solved using a surrogate model optimization method, and the specific steps are as follows: S401. Radial basis function (RBF) is selected as the proxy model to fit the mapping between the deployment scheme and the total cost; S402. A dual-strategy hybrid search combining local and global search strategies is used to generate candidate sample points. S403. Candidate points are selected by weighting the dual scores of the agent target value and spatial distance, and the process is iterated until convergence to output the optimal solution.

6. The method for deploying electric bus charging stations considering dual uncertainties according to claim 2, characterized in that, The formula for calculating the cost of the first stage is: （2）， In the formula, This represents the total cost of the first phase. Indicates the bus stop or depot number. This indicates a meeting point at the bus terminal. This indicates the bus terminal and meeting point. Indicates the type number of the charging station. This represents a set of charging station types. This indicates the cost of the charging station. Indicates the first The first site The number of charging piles, Indicates the first Maximum charging power of similar charging piles Indicates the first The power grid capacity of each station, This represents the unit cost of expanding the power grid.

7. The method for deploying electric bus charging stations considering dual uncertainties according to claim 3, characterized in that, In step S202, the trip allocation rules for electric buses are as follows: (1) The current energy of each electric bus should be greater than the sum of the energy consumption of the trip and the minimum energy that the electric bus should maintain; (2) The current location of the electric bus should be consistent with the starting / ending station of the connecting bus route; (3) If the electric bus is being charged and the charging time exceeds the minimum required charging time. Then the electric bus can be used to perform other travel tasks; (4) If in time There are always multiple eligible electric buses available for the trip. Then, the "highest SOC priority" strategy is applied to select the execution process. The electric buses; specifically, the electric buses with the highest SOC (State of Charge) status among all eligible electric buses will be selected to execute the trip. If no suitable electric bus is available, a new electric bus will be generated to complete the trip. .

8. The method for deploying electric bus charging stations considering dual uncertainties according to claim 3, characterized in that, In S203, the charging rules for electric buses are as follows: (1) When an electric bus is assigned to a charging station for charging, the minimum charging time must be met. ; (2) If all charging stations at the bus origin / terminal station and depot are occupied, the electric bus must wait in the queue until a charging station becomes available; (3) If there are multiple electric buses waiting in line for charging, once the charging station is released, the first-come, first-served (FIFS) strategy will be used to select the electric bus to be charged. (4) In order to maintain the good health of the electric bus battery, the SOC of the electric bus should be maintained at Within the range.

9. A method for deploying electric bus charging stations considering dual uncertainties according to claim 4, characterized in that, The step size of the motion disturbance is δ∈{−1,0,1}, and each time only the number of fast charging / slow charging piles is adjusted in the smallest unit.

10. An electric bus charging facility planning system, characterized in that, The device includes a memory and a processor, wherein the processor executes the method described in any one of claims 1-9 to achieve coordinated optimization of the number and type of charging piles and grid expansion.

Images