Truck type mobile charging station online collaborative optimization scheduling method

Abstract

The present application relates to the technical field of charging facility optimization, and particularly relates to a truck type mobile charging station online collaborative optimization scheduling method, a two-stage optimization scheduling model framework is constructed, the optimization scheduling model framework comprises: an offline training stage, a TMCS multi-period optimization decision model is established, and then a look-ahead rolling value function approximation algorithm (LRH-VFA) is established to iteratively learn from EV charging historical data, so that the influence of current period decision on future profit of the operator is considered; an online scheduling stage, based on the approximate value function obtained through offline training and short-time prediction and real-time information, TMCS online scheduling decision is updated rolling. The present application can fully consider the influence of EV charging demand uncertainty on TMCS scheduling results, effectively utilize the demand change adjustment decision updated dynamically, guarantee the quality of EV charging service, and improve the operator's income by coordinating TMCS to participate in power grid energy arbitrage.

Description

TECHNICAL FIELD

[0001] The application relates to a truck-type mobile charging station online collaborative optimization scheduling method and belongs to the technical field of charging facility optimization. BACKGROUND

[0002] According to the latest report Global EV Outlook 2024 released by the International Energy Agency (IEA), the EV sales are expected to reach 17 million in 2024, accounting for more than one fifth of the global automobile sales. However, the number of public charging facilities is still insufficient, and it is expected to increase by six times by 2035 [1] .

[0003] In fact, the traditional fixed charging station (FCS) still has some development obstacles that are difficult to overcome at present, including a long construction period, high expansion cost, insufficient flexibility and the like. Therefore, in recent years, the emerging truck-type mobile charging station (TMCS) is expected to provide a new way to cope with the above challenges. The TMCS integrates a certain number of charging piles and energy storage batteries in a container loaded by a truck. Since the TMCS is independent of the power grid and is easy to move, it is easier to expand than the FCS, and can provide on-demand charging services for EVs in any area [2][3] . Literature [4] schedules the TMCS to charge at a low electricity price period, and arrives at the FCS to provide electric vehicle charging services during the queuing peak period. Literature [5] establishes a mixed integer linear programming (MILP) model for TMCS scheduling and operator profit optimization, and solves it by using an improved genetic algorithm. Literature [6] optimizes the service location of the TMCS by using the flow refueling location model. Literature [7] reduces the peak load rate of the FCS by scheduling the TMCS to the charging peak area. In addition, some studies also focus on the application of the TMCS in the EV parking lot and its social fair access[8], as well as the energy transaction strategy based on auction between EVs and TMCS[9]. Literature

[10] proposes a day-ahead collaborative optimization framework to improve the utilization rate and economy of the TMCS during the non-charging period.

[0004] However, the above studies do not effectively consider the charging uncertainty problem faced by the TMCS during operation, which leads to the fact that the expected optimization effect cannot be achieved in actual operation.

[0005] REFERENCES

[0006] [1] IEA. Global EV outlook 2022 [EB / OL]. [2022-06-29]. https: / / www.iea.org / reports / global-ev-outlook-2022.

[0007] [2] AFSHAR S, MACEDO P, MOHAMED F, et al. Mobile charging stations for electric vehicles—A review[J]. Renewable and Sustainable Energy Reviews, 2021, 152: 111654.

[0008] [3] LIU L, QI X, XI Z, et al. Charging-Expense Minimization Through Assignment Rescheduling of Movable Charging Stations in Electric Vehicle Networks[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(10): 17212-17223.

[0009] [4] ZHENG Y, LI F, DONG J, et al. Spatiotemporal flexibility optimization and regulation of mobile charging vehicles considering resource elastic sharing

[0010] strategy[J]. Power System Automation, 2023, 47(14): 33-42.

[0011] [5] H. Li, D. Son, B. Jeong. Electric vehicle charging scheduling with mobile charging stations[J]. Journal of Cleaner Production, 2024, 434: 140162.

[0012] [6] WANG F, CHEN R, MIAO L, et al. Location Optimization of Electric Vehicle Mobile Charging Stations Considering Multi-Period Stochastic User Equilibrium[J]. Sustainability, 2019, 11(20): 5841.

[0013] [7] MOGHADDAM V, AHMAD I, HABIBI D, et al. Dispatch management of portable charging stations in electric vehicle networks[J]. eTransportation, 2021, 8: 100112.

[0014] [8] NAZARI-HERIS M, LONI A, ASADI S, et al. Toward social equity access and mobile charging stations for electric vehicles: A case study in Los Angeles[J]. Applied Energy, 2022, 311: 118704.

[0015] [9] THI KIM O T, LE T H T, SHIN M J, et al. Distributed Auction-Based Incentive Mechanism for Energy Trading Between Electric Vehicles and Mobile Charging Stations[J]. IEEE Access, 2022, 10: 56331-56347.

[0016]

[10] JIA Hongjie, HE Kecheng, MU Yunfei, et al. A mobile charging station day-ahead coordinated optimization scheduling method: 202310938963.5[P]. 2024-03-19.

[0017]

[11] J. Liu, J, Wang, J, and Cardinal. Evolution and reform of UK electricity market[J]. Renewable and Sustainable Energy Reviews, 2022, 161, 112317.

[0018]

[12] J.Yang,X.Jiang,K.Zhao.Multiobiective optimization method of location and capacity determination problems of highway charging stations[J].Journal of Chongqing Normal University(Natural Science),2021,38(1),11-21. SUMMARY

[0019] The present application aims at the deficiencies in the prior art, and provides a truck-type mobile charging station online collaborative optimization scheduling method, which fully considers the influence of EV charging demand uncertainty on TMCS scheduling results, effectively utilizes the dynamically updated demand changes to adjust decisions, ensures the quality of charging service of the operator, and improves the operator's income by scheduling TMCS to participate in power grid energy arbitrage.

[0020] The technical scheme for solving the above technical problems is as follows: a truck-type mobile charging station online collaborative optimization scheduling method, wherein the truck-type mobile charging station online collaborative optimization scheduling method is:

[0021] S1, constructing a TMCS space-time operation model: dividing the set of TMCS operation positions into a charging service node set and an arbitrage node set, and establishing a TMCS multi-period optimization decision model;

[0022] S2, MDP reconstruction: re-describing the TMCS multi-period optimization decision model as a Markov decision process, and solving it by using dynamic programming, so as to decompose the original multi-period optimization problem into multiple continuous single-period optimization problems which can be iteratively solved;

[0023] S3, constructing an LRH-VFA two-stage scheduling model: combining a rolling optimization algorithm with an ADP algorithm to fully utilize the continuously updated real-time information, and defining a post-decision state value function to quantify the influence of the current decision on the future income of the operator.

[0024] Further, in step S1, the set M of TMCS operation positions is divided into M c and M a two disjoint subsets, wherein M c represents the charging service node set, M a represents the arbitrage node set; m and u are EV charging service nodes of the road network, n and v are energy arbitrage nodes of TMCS and the power grid, Z c , Z a , Z erespectively represent the transfer process of TMCS between nodes in charging, arbitrage and charging-arbitrage, and then the multi-period optimization decision model of TMCS is established:

[0025]

[0026] where ω is the number of TMCS; is a Boolean variable, if ω moves on path (m, u) at time t, carries out EV charging service at node m, or carries out energy arbitrage at node n, then or

[0027] otherwise, 0; D ω is the total travel distance in the scheduling period; v a is the average moving speed of TMCS; t e is the end service time of TMCS.

[0028] Further, TMCS needs to meet the following power and operation constraints:

[0029]

[0030]

[0031] where Ω is the set of TMCS, is the set of TMCS participating in EV charging service at node m; ρ c is the EV charging demand response ratio, reflecting the charging service quality requirement of CFO; is the charging power of TMCS ω at node n when carrying out energy arbitrage at time t, is the discharging power of TMCS ω at node n when carrying out energy arbitrage at time t, is the maximum chargeable power of TMCS ω, is the maximum dischargeable power of TMCS ω; is the maximum output power of TMCS ω when carrying out EV charging service; and is a binary variable, if TMCS ω is charging at time t, then if TMCS ω is discharging at time t, then if TMCS ω is not charging at time t, then if TMCS ω is not discharging at time t, then E ω is the capacity of TMCS; η ch,ω is the charging efficiency of TMCS; η dch,ω is the discharging efficiency of TMCS; SOC maxis the maximum state of charge (SOC) value of TMCS, SOC min is the minimum SOC value of TMCS, is the SOC value of TMCSω at time t, is the SOC value of TMCSω at time t+1.

[0032] Further, the objective function of the multi-period optimization decision model of TMCS is shown in equation (13):

[0033]

[0034] where x t is the decision variable; are the charging service fee, the charging price and the discharging price of energy arbitrage, respectively; is the electricity price when TMCS ends operation and charges; c tmc is the energy consumption per kilometer of TMCS; c la , c mt are the labor cost and maintenance cost converted to period t, respectively; c mdc is the marginal aging cost of the maximum life cycle benefit of TMCS; q t is the calendar aging parameter of TMCS battery pack; r0 is the discount rate; κ(t) is the year number corresponding to the time t when TMCS is put into use.

[0035] Further, the MDP modeling in step S2 includes state variable S t , decision variable x t and exogenous variable W t , the state variables of two consecutive stages are linked through a transition function, and the state variable reflects the current state of the system, which is represented by equation (17):

[0036]

[0037] and are binary variables, if TMCSω at time t is located between nodes m and u at the previous moment, if TMCSω at time t is located between nodes n and v at the previous moment, if TMCSω at time t is located between nodes m and n at the previous moment, if TMCSω at time t is not located between nodes m and u at the previous moment, if TMCSω at time t is not located between nodes n and v at the previous moment, if TMCSω at time t is not located between nodes m and n at the previous moment, is the EV charging load of node m at time t, is the SOC value of TMCS ω at time t.

[0038] In the MDP framework, the decision variable x t is expressed as:

[0039]

[0040] Exogenous information is used to simulate the deviation between the predicted value and the actual value of uncertain factors. In the online scheduling decision of TMCS, the sources of uncertainty include EV charging demand, adjustment of charging price and fluctuation of energy arbitrage price. The exogenous information process W t can be defined as (19):

[0041]

[0042] From the time perspective, W t represents the random information obtained after the end of the last time period (t-Δt) and before the current decision x t is made, so the transformation sequence of the decision-making process is (W t , S t , x t , W t+Δt , S t+Δt , x t+Δt , …), where W t+Δt , S t+Δt , x t+Δt represent the exogenous variable, state variable and decision variable of the next time after time t, respectively.

[0043] The transformation function refers to the process of the system moving from the current state S t to the next state S t+Δt according to the decision x t and the exogenous information W t+Δt . The transformation function is defined to obtain the state information of the system in the next period:

[0044]

[0045] S t+Δt (2)=x t (4) (21)

[0046] S t+Δt (3)=x t (5) (22)

[0047] S t+Δt (4)=x t (6) (23)

[0048]

[0049] Where Δt is the time step; the numbers in the parentheses of formulas (20) to (24) correspond to S in formula (17). t Formula (18) in formula x t And (19) W t The order of these steps leads to the rewriting of the objective function (13) as follows:

[0050]

[0051] f t (S t ,x t )=R(S t ,x t )-C OM (S t ,x t )-C DEG (S t ,x t (26)

[0052] Where: E[·] represents the expectation operation; X t Assuming a feasible set of decisions, the TMCS online collaborative scheduling optimization model is modeled as an MDP. The optimal decision sequence is then obtained by solving the Bellman equation using dynamic programming, as shown below:

[0053]

[0054] Where: V t (S t () is a value function, representing the system in state S. t The cumulative immediate return for the next Δt period; f t is the objective function; γ is the discount factor that affects the importance of immediate and future rewards in the MDP, and γ is set between 0 and 1.

[0055] Furthermore, in step S3, the post-decision state value function is:

[0056]

[0057] in, For the state variables after the decision, The post-decision state-value function represents the system's state after making a decision but before receiving any random information. The Bellman equation is rewritten using a piecewise linear function with a monotonically increasing slope as follows:

[0058]

[0059] Where: N a For piecewise linear functions the number of segments, a is the segment index; SOC value of the post-decision TMCS, H is the optimization interval determined by the prediction model; is the slope corresponding to segment a; is the length of segment a mapped to the horizontal axis; the approximation function is evenly divided into N a segments, so:

[0060]

[0061]

[0062] wherein: and are the maximum and minimum storage capacities; t1 is the starting time; is the post-decision state of charge of the TMCS before execution; equation (31) indicates that the TMCS capacity is equal to the sum of the segment values, equation (32) indicates that each segment value does not exceed the upper and lower limits, equation (33) indicates that the slope monotonically increases with the segment number, i.e., to ensure that the PLF is a convex function, and the constraint equation (34) indicates the relationship between the TMCS capacity before and after decision-making; the approximate optimal decision is obtained by recursively solving the following equation:

[0063]

[0064] Further, the difference iteration method is used to train and update the slope of the PLF. After the (k-1)th iteration, the segment linear function is known, so equation (35) in the kth iteration can be rewritten as equation (36), and the slope sampling value of the segment a is equation (37), and the slope of the PLF is updated by equation (38);

[0065]

[0066] wherein, is the approximate optimal decision at the kth iteration; indicates the slope value of segment a in the PLF after the (k-1)th iteration; is the gradient value of the state variable at the kth iteration; θ k-1 is the slope update step. The slope of the PLF represents the influence of the unit energy change of the TMCS on the subsequent period CFO profit. To ensure that the updated slope still satisfies the monotonic increasing property, the leveling algorithm is used to update the slope value of each segment:

[0067]

[0068] wherein, is the slope value corresponding to the segment b after the kth iteration.

[0069] The beneficial effects of the present application are:

[0070] (1) The present application proposes a TMCS online collaborative optimization scheduling framework to coordinate its real-time collaborative operation scheduling between EV charging services and energy arbitrage;

[0071] (2) The proposed lookahead rolling value function approximation (LRH-VFA) two-stage scheduling model can iteratively learn from EV charging historical data to consider the impact of current period decisions on future profits of operators;

[0072] (3) The proposed optimization strategy explores the impact of EV charging demand uncertainty on TMCS scheduling results, and only uses short-term prediction and real-time information during online operation to roll out TMCS online collaborative optimization scheduling results, which improves the utilization of TMCS and the profits of operators while ensuring the quality of charging services for operators. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 is an architectural diagram of a real-time optimization scheduling method for a truck-type mobile charging station;

[0074] Figure 2 is an LRH-VFA algorithm;

[0075] Figure 3 is a flowchart of a real-time optimization scheduling method for a truck-type mobile charging station;

[0076] Figure 4 is a mesh highway network and the distribution of related node locations;

[0077] Figure 5 is the distribution of EV departure times;

[0078] Figure 6 is an EV origin-destination (OD) matrix;

[0079] Figure 7 is the TMCS scheduling result of site 6 under the test scenario;

[0080] Figure 8 is the scheduling result of TMCS 3 under the test scenario. DETAILED DESCRIPTION

[0081] The specific embodiments of the present application will be described in detail below. The present application can be implemented in many different ways than described herein, and the skilled in the art can make similar improvements without departing from the spirit of the present application, therefore the present application is not limited to the specific embodiments disclosed.

[0082] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used only for the purpose of describing specific embodiments is not intended to be limiting of the present application.

[0083] The optimization model framework designed by the present application is as shown in Figure 1 The CFO operation has a certain number of FCS and TMCS. According to the traffic flow prediction data and the configuration information of FCS and TMCS, the EV model is used to generate the space-time distribution of EV charging demand of FCS and TMCS. According to the day-ahead operation plan of the power grid, the power limit and electricity price information of each node of the distribution network are obtained. Further, the day-ahead optimization scheduling strategy of TMCS is formulated to maximize the profit of CFO.

[0084] I. TMCS space-time operation model

[0085] In addition to assisting FCS to provide EV charging services, TMCS can also be dispatched to some distribution network nodes to participate in energy arbitrage during non-charging period to earn profits. In this paper, the set M of TMCS operation location is divided into M c and M a Two disjoint subsets, where M c represents the set of charging service nodes, and M a represents the set of arbitrage nodes. m, u are the EV charging service nodes of the road network, n, v are the energy arbitrage nodes of TMCS interacting with the power grid, Z c , Z a , Z e respectively represent the transfer process of TMCS between charging, arbitrage and charging-arbitrage nodes. Further, the multi-period optimization decision model of TMCS is established:

[0086]

[0087] Where ω is the number of TMCS; is a Boolean variable, if ω at time t moves on path (m, u), performs EV charging service at node m, or performs energy arbitrage at node n, respectively, then or Otherwise, it is 0 respectively; D ω is the total travel distance in the scheduling period; v a is the average moving speed of TMCS; t e is the end of service time of TMCS.

[0088] Equation (1) ensures that the TMCS is in either running or moving state, and Equations (2) and (3) represent the transition relationship between the running and moving states. In addition, considering the large demand of FCS and TMCS due to the rapidly growing EV charging load, the TMCS needs to meet the following power and operation constraints:

[0089]

[0090]

[0091] where Ω is the set of TMCSs, is the set of TMCSs participating in EV charging service at node m; p c is the EV charging demand response ratio, reflecting the charging service quality requirement of CFO; is the charging power of TMCS ω at node n when it performs energy arbitrage at time t, is the discharging power of TMCS ω at node n when it performs energy arbitrage at time t, is the maximum chargeable power of TMCS ω, is the maximum dischargeable power of TMCS ω; is the maximum output power of TMCS ω when it performs EV charging service; and is a binary variable, if TMCS ω is charging at time t, if TMCS ω is discharging at time t, if TMCS ω is not charging at time t, if TMCS ω is not discharging at time t, E ω is the capacity of TMCS; η ch,ω is the charging efficiency of TMCS; η dch,ω is the discharging efficiency of TMCS; SOC max is the maximum state of charge (SOC) value of TMCS, SOC min is the minimum SOC value of TMCS, is the SOC value of TMCS ω at time t, is the SOC value of TMCS ω at time t+1.

[0092] Constraints (6)-(7) establish the feasible set of charging / discharging power of TMCS ω in arbitrage state. Constraint (8) defines the charging / discharging constraints related to TMCS arbitrage operation mode. Constraints (9)-(10) establish the feasible set of discharging power of TMCS ω in EV charging service mode. (11) and (12) are SOC constraints, and (12) determines the SOC of TMCS ω at the end of time t.

[0093] The objective function of the TMCS multi-period optimization decision model is shown in equation (13).

[0094]

[0095] where x t is the decision variable; are the charging service fee, the charging price of energy arbitrage and the discharging price, respectively; is the price when the TMCS ends operation and charges; c tmc is the energy consumption per kilometer of the TMCS; c la , c mt are the labor cost and maintenance cost converted to period t, respectively; c mdc is the marginal aging cost of the TMCS life cycle benefit maximum; q t is the calendar aging parameter of the TMCS battery pack; r0 is the discount rate; and K(t) is the year number corresponding to the time t when the TMCS is put into use.

[0096] II. MDP reconstruction

[0097] The above TMCS multi-period optimization decision model is a mixed integer linear programming (MILP) problem. Although the optimal solution of the optimization problem can be obtained, it needs to obtain the full-period accurate state information of the EV charging demand at the time of optimization decision. Due to the exogenous uncertainty of the EV charging demand, its accurate prediction result cannot be obtained in the real-time scheduling stage, so that the MILP model is not suitable for the real-time scheduling scenario. Scenario method and robust optimization are traditional effective methods for dealing with uncertainty and enhancing model robustness, the former relies on modeling uncertainty through probability distribution, and the latter focuses on ensuring the feasibility of the solution in the worst case. However, since these two methods are static, they cannot effectively utilize the changes in demand to dynamically adjust the decision.

[0098] In fact, if the online scheduling process of the TMCS is analyzed, since the scheduling period (usually one day) is divided into T periods, it is necessary to determine the global optimal decision from the current period to the final period at each stage. In addition, the current location and battery capacity already contain the influence of previous actions on the current state, which indicates that the scheduling optimization problem exhibits Markov property. Therefore, the above multi-period optimization model is re-expressed as a Markov decision process (MDP), and then solved by dynamic programming (DP). This method decomposes the original multi-period optimization problem into multiple continuous single-period optimization problems that can be solved iteratively, where the transition function reflects the transition process between adjacent periods.

[0099] MDP modeling mainly includes state variable S t , decision variable x tand exogenous variable W t The state variables of two consecutive stages are linked by a transition function. The state variable reflects the current state of the system, which is represented by equation (17).

[0100]

[0101] In view of the influence of the position information of the previous period on the scheduling range of the TMCS of the next period, the state variable considered by the present application includes a binary variable and If the TMCS ω is located between nodes m, u at the previous moment of time t, then If the TMCS ω is located between nodes n, v at the previous moment of time t, then If the TMCS ω is located between nodes m, n at the previous moment of time t, then If the TMCS ω is not located between nodes m, u at the previous moment of time t, then If the TMCS ω is not located between nodes n, v at the previous moment of time t, then If the TMCS ω is not located between nodes m, n at the previous moment of time t, then is the EV charging load of node m at time t, is the SOC value of the TMCS ω at time t. The TMCS scheduling process needs to determine the coordinated optimization decision according to the current state and the prediction information. In the MDP framework, the decision variable x t is represented as:

[0102]

[0103] The exogenous information is used to simulate the deviation between the predicted value and the actual value of the uncertain factor. In the online scheduling decision of the TMCS, the main sources of uncertainty include the EV charging demand, the adjustment of the charging price, and the fluctuation of the energy arbitrage price. The exogenous information process W t can be defined as (19):

[0104]

[0105] From the time perspective, W t represents the random information obtained before the current decision x t is made, therefore, the conversion sequence of the decision process is (W t , S t , x t , W t+Δt , S t+Δt , x t+Δt , …), wherein W t+Δt , S t+Δtx t+Δt Let represent the exogenous variable, state variable, and decision variable at the time following time t, respectively.

[0106] The transformation function refers to the function of the system based on decision x. t and exogenous information W t+Δt From the current state S t Transition to the next state S t+Δt The process. Therefore, a transition function is defined to obtain the system's state information in the next time period:

[0107]

[0108] S t+Δt (2) = x t (4) (21)

[0109] S t+Δt (3) = x t (5) (22)

[0110] S t+Δt (4) = x t (6) (23)

[0111]

[0112] Where Δt is the time step; the number in parentheses corresponds to S in (17). t (18) x t And (19) W t The order of these steps. Therefore, the objective function (13) can be rewritten as follows:

[0113]

[0114] f t (S t ,x t )=R(S t ,x t )-C OM (S t ,x t )-C DEG (S t ,x t (26)

[0115] Where: E{·} represents the expectation operation; X t This represents the set of feasible decisions. After modeling the TMCS online collaborative scheduling optimization model as an MDP, the optimal decision sequence for this problem can be obtained by solving the Bellman equation using dynamic programming (DP), as shown below:

[0116]

[0117] where: V t (S t ) is the value function, representing the cumulative immediate reward of the system in state S t at time period Δt; f t is the objective function; γ is the discount factor that affects the importance of immediate reward and future reward in MDP, usually set between 0 and 1.

[0118] III. LRH-VFA two-stage scheduling model

[0119] However, due to the curse of dimensionality, it is difficult to enumerate the entire solution space for DP algorithm in computation. Moreover, in stochastic environment, the system in a certain state at the current stage can transfer to any different state at the next stage, resulting in the problem of "dimensional explosion", which is difficult to be used for real-time scheduling optimization of TMCS.

[0120] Therefore, ADP algorithm can be used to approximate the optimal solution of the original problem while ensuring the feasibility of online decision-making. ADP method mainly includes policy function approximation and value function approximation. For the latter, linear function, piecewise linear function (PLF), table function and non-parametric model can be used to approximate the value function. It should be noted that when making online scheduling decisions for TMCS, the energy arbitrage price is usually known

[11] . The impact of EV charging demand uncertainty on CFO profit will ultimately be reflected as the change of SOC of TMCS. Studies have shown that for optimization problems involving energy storage, using PLF with convex / concave properties to approximate the true value function can achieve excellent approximation results, which has been verified in related literature. In addition, since the transfer and charging / discharging operations of TMCS may be difficult to complete within a time period, we combine the rolling optimization algorithm with the ADP algorithm to make full use of real-time information updated continuously. The post-decision state value function is defined to quantify the impact of the current decision on the future revenue of the operator:

[0121]

[0122] where, is the post-decision state variable, is the post-decision state value function, which is the state of the system before any random information is received after the system executes the decision. The evolution process of pre-decision state S t , decision x t , random information W t and post-decision state is shown in Figure 2 . Further, the Bellman equation is rewritten as:

[0123]

[0124] Where: N a For piecewise linear functions The number of segments, where 'a' is the segment index; denoted as SOC value of TMCS after decision; H is the optimization interval determined by the prediction model. Let be the slope corresponding to segment 'a'; for The segment 'a' is mapped to the length of the horizontal axis. This article will use the approximate function. Divided into N a Therefore:

[0125]

[0126] in: and t1 represents the maximum and minimum stored capacity; t1 is the starting time. For decision making The state of charge of TMCS before execution. Formula (31) indicates that the charge of TMCS is equal to the sum of the values ​​of each segment, formula (32) indicates that the values ​​of each segment do not exceed the upper and lower limits, formula (33) indicates that the slope increases monotonically with the segment number, that is, to ensure that PLF is a convex function, and constraint formula (34) indicates the relationship between the change of TMCS charge before and after the decision; approximate optimal decision The following equation is obtained by recursively solving it:

[0127]

[0128] The optimality of the above model depends on how close the approximate function is to the true function, which in turn depends to a large extent on the slope of each segment in the PLF. To ensure optimal decision-making, a differential iterative method is used to train and update the slope of the PLF. Note the piecewise linear functions corresponding to each time segment after the (k-1)th iteration. Since it is known, (35) in the k-th iteration can be rewritten as (36), defined The slope sampling value is (37), and then the slope of PLF is updated by equation (38).

[0129]

[0130] in, This represents the approximate optimal decision at the k-th iteration. This represents the slope value of segment a in the PLF after the (k-1)th iteration; The gradient value of the state variable at the k-th iteration; θ k-1 The slope is updated step size. The slope of the PLF characterizes the impact of a unit energy change in the TMCS on the CFO profit in subsequent periods. To ensure that the updated slope still satisfies the monotonically increasing property, the leveling algorithm is used to update the slope values ​​of each segment:

[0131]

[0132] in, Let be the slope value corresponding to segment b after the k-th iteration. A flowchart of a real-time optimization scheduling method for truck-type mobile charging stations is shown below. Figure 3 As shown.

[0133] IV. Examples

[0134] The embodiments of the present invention are verified using the mesh highway network in reference

[12] (e.g. Figure 4 (As shown). This road network consists of 10 nodes and 18 connections, with CFO operating 20 FCSs and 6 TMCSs. See details below. Figure 5 , Figure 6 See Tables 1 and 2. The load electricity price adopts the typical domestic time-of-use pricing. Other parameters are shown in Table 3. The scheduling results of a test scenario are as follows: Figure 7 , Figure 8 As shown in Table 4, the average results based on 200 test scenarios are presented.

[0135] The results show that this invention can meet the real-time control requirements of large-scale TMCS and provide a near-globally optimal online scheduling scheme. The proposed method will better coordinate TMCS operations between electric vehicle charging and energy arbitrage, thereby improving charging service quality and CFO profitability. TMCS can be adjusted to areas with high charging demand as needed and can be transferred at any time according to real-time changes in demand, providing a flexible solution to meet the needs of dynamic electric vehicle charging services and the energy market.

[0136] Table 1 Parameters of the Mesh Expressway Network

[0137]

[0138] Table 2 Number of charging stations for each FCS

[0139]

[0140]

[0141] Table 3 Other simulation parameters

[0142]

[0143] Table 4. Solution results for the mesh-like expressway network

[0144]

[0145] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are exhaustively listed. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0146] For those skilled in the art, various modifications and improvements can be made without departing from the concept of the present invention, and these modifications and improvements are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the appended claims.

Claims

1. A method for online collaborative optimization scheduling of truck-type mobile charging stations, characterized in that, The online collaborative optimization scheduling method for truck-type mobile charging stations is as follows: S1. Constructing the TMCS spatiotemporal operation model: Divide the set of TMCS operation locations into a set of charging service nodes and a set of arbitrage nodes, and establish a multi-period optimization decision model for TMCS. S2, MDP Reconstruction: The TMCS multi-period optimization decision model is reformulated as a Markov decision process and solved using dynamic programming, decomposing the original multi-period optimization problem into multiple continuous single-period optimization problems that can be solved iteratively; S3. Construct a two-stage scheduling model of LRH-VFA: combine the rolling optimization algorithm with the ADP algorithm to make full use of continuously updated real-time information, and define the state value function after decision to quantify the impact of the current decision on the operator's future revenue. In step S1, the set M of TMCS running locations is divided into M... c and M a Two disjoint subsets, where M c M represents the set of charging service nodes. a Represents the set of arbitrage nodes; m and u are EV charging service nodes in the road network, and n and v are energy arbitrage nodes interacting with the power grid. The objective function of the TMCS multi-period optimization decision model is shown in equation (13): Where, x t For decision variables; These are charging service fees, charging electricity prices for energy arbitrage, and discharging electricity prices; The electricity price when TMCS ceases operation and begins charging; c tmc c represents the energy consumption per kilometer of TMCS. la c mt These are the labor cost and maintenance cost converted to time period t, respectively; c mdc The marginal aging cost that maximizes the lifetime benefits of TMCS; q t These are the calendar aging parameters of the TMCS battery pack; r0 is the discount rate; κ(t) is the year number corresponding to the time t when the TMCS is put into use. It is the charging power at node n when TMCSω performs energy arbitrage at time t. D is the discharge power at node n when TMCSω performs energy arbitrage at time t; ω The total travel distance within the scheduling period; Step S2, MDP modeling, includes state variables S t Decision variable x t and exogenous variable W t The state variables of the two consecutive stages are linked by a transformation function. The state variables reflect the current state of the system, as expressed by equation (17): and Both are binary variables. If TMCSω is located between nodes m and u in the time preceding time t, then If TMCSω is located between nodes n and v at the time preceding time t, then If TMCSω is located between nodes m and n in the time preceding time t, then If TMCSω is not located between nodes m and u in the time preceding time t, then If TMCSω is not located between nodes n and v in the time preceding time t, then If TMCSω is not located between nodes m and n in the time preceding time t, then Let t be the EV charging load at node m. It is the SOC value of TMCSω at time t; In the MDP framework, decision variable x t Represented as: ω is the TMCS number; Let ω be a Boolean variable. If at time t ω moves along path (m, u), providing EV charging service at node m, or engaging in energy arbitrage at node n, then... or Otherwise, the values ​​are 0. and Let be a binary variable. If TMCSω is charged at time t, then If TMCSω discharges at time t, then If TMCSω is not charged at time t, then If TMCSω does not discharge at time t, then Exogenous information is used to simulate the deviation between predicted and actual values ​​of uncertainties. In TMCS online scheduling decision-making, sources of uncertainty include EV charging demand, adjustments to charging prices, and fluctuations in energy arbitrage prices. The exogenous information process W... t It can be defined as equation (19): From a time perspective, W t This indicates the current decision x after the previous time interval (t-Δt) ends. t Given the random information obtained before making a decision, the transformation sequence of the decision-making process is (W t ,S t ,x t W t+Δt ,S t+Δt ,x t+Δt ,…), where W t+Δt S t+Δt x t+Δt Let represent the exogenous variable, state variable, and decision variable at the time following time t, respectively. The transformation function refers to the function of the system based on decision x. t and exogenous information W t+Δt From the current state S t Transition to the next state S t+Δt The process involves defining a transformation function to obtain the system's state information for the next time period: S t+Δt (2)=x t (4) (21) S t+Δt (3)=x t (5) (22) S t+Δt (4)=x t (6)(23) Among them, E ω η is the capacity of TMCS; Δt is the time step; η is the t-value. ch,ω For the charging efficiency of TMCS; η dch,ω The discharge efficiency of TMCS; the numbers in parentheses in equations (20) to (24) correspond to S in equation (17). t Formula (18) in formula x t And (19) W t The order of these steps leads to the rewriting of the objective function (13) as follows: f t (S t ,x t )=R(S t ,x t )-C OM (S t ,x t )-C DEG (S t ,x t ) (26) Where: E[·] represents the expectation operation; X t Assuming a feasible set of decisions, the TMCS online collaborative scheduling optimization model is modeled as an MDP. The optimal decision sequence is then obtained by solving the Bellman equation using dynamic programming, as shown below: Where: V t (S t () is a value function, representing the system in state S. t The cumulative immediate return for the next Δt period; f t is the objective function; γ is the discount factor that affects the importance of immediate and future rewards in the MDP, and γ is set between 0 and 1.

2. The online collaborative optimization scheduling method for truck-type mobile charging stations according to claim 1, characterized in that, Z c Z a Z e These represent the transition processes of TMCS between charging, arbitrage, and charging-arbitrage time nodes, respectively, thus establishing a multi-time-period optimization decision model for TMCS: Among them, v a t represents the average moving speed of the TMCS. e This is the end time of TMCS service.

3. The online collaborative optimization scheduling method for truck-type mobile charging stations according to claim 2, characterized in that, TMCS must meet the following power and operational constraints: Where Ω is the set of TMCS, Let ρ be the set of TMCSs participating in EV charging services at node m; c The EV charging demand response ratio reflects the CFO's charging service quality requirements; This is the maximum rechargeable power of TMCSω. This is the maximum dischargeable power of TMCSω; This is the maximum output power of TMCSω when providing EV charging services; SOC max It is the maximum SOC value of TMCS, SOC min It is the minimum SOC value of TMCS. It is the SOC value of TMCSω at time t. It is the SOC value of TMCSω at time t+1.

4. The online collaborative optimization scheduling method for truck-type mobile charging stations according to claim 3, characterized in that, In step S3, the post-decision state value function is: in, For the state variables after the decision, The post-decision state-value function represents the system's state after making a decision but before receiving any random information. The Bellman equation is rewritten using a piecewise linear function with a monotonically increasing slope as follows: Where: N a For piecewise linear functions The number of segments, where 'a' is the segment index; Let SOC be the SOC value of TMCS after the decision, and H be the optimization interval determined by the prediction model; Let be the slope corresponding to segment 'a'; for Map segment a to the length of the horizontal axis; approximate the function. Divided into N a Therefore: in: and t1 represents the maximum and minimum energy storage capacity; t1 is the starting time. For decision making The state of charge of TMCS before execution; Formula (31) indicates that the charge of TMCS is equal to the sum of the values ​​of each segment, Formula (32) indicates that the values ​​of each segment do not exceed the upper and lower limits, Formula (33) indicates that the slope remains monotonically increasing as the segment number increases, that is, ensuring that PLF is a convex function, and the constraint formula (34) indicates the relationship between the change of TMCS charge before and after the decision; approximate optimal decision The following equation is obtained by recursively solving it:

5. The online collaborative optimization scheduling method for truck-type mobile charging stations according to claim 4, characterized in that, The slope of the PLF is trained and updated using the differential iterative method, and the piecewise linear function corresponding to each time period after the (k-1)th iteration is obtained. Since it is known, formula (35) in the k-th iteration can be rewritten as formula (36), defined as follows: The slope sampling value is given by formula (37), and then the slope of PLF is updated by formula (38); in, This represents the approximate optimal decision at the k-th iteration. This represents the slope value of segment a in the PLF after the (k-1)th iteration; The gradient value of the state variable at the k-th iteration; θ k-1 Update the step size for the slope; The slope of the PLF characterizes the impact of a unit energy change in the TMCS on the CFO profit in subsequent periods. To ensure that the updated slope still satisfies the monotonically increasing property, the leveling algorithm is used to update the slope values ​​of each segment: in, This represents the slope value corresponding to segment b after the k-th iteration.

Images