A method and system for regional dispatching of electric buses

Abstract

The invention discloses an electric bus regional scheduling method, comprising the following steps: establishing a time extension network model to explain the pure electric bus multi-depot scheduling problem; setting nodes and arcs of the time extension network model, and setting the arc costs; then establishing a dual-objective integer programming model considering time-of-use electricity prices and energy consumption control; using a lexicographic order optimization method to process the dual objectives of the dual-objective integer programming model, and converting the dual-objective problem into the first objective and the second objective of a single-objective model for solving; designing a branch pricing algorithm to solve the first objective of the converted single-objective model; using the optimal solution obtained by the first objective as the initial solution of the second objective, and solving the second objective of the converted single-objective model through a commercial solver, thereby completing the electric bus regional scheduling; the invention can not only reduce costs, but also reduce the peak load of a power grid and improve the safety of the power grid, and the algorithm has good characteristics and strong practical application prospects.

Description

Technical Field

[0001] The present invention relates to the research field of electric bus regional dispatching technology, and in particular to an electric bus regional dispatching method and system considering time-of-use electricity price and energy consumption control. Background Art

[0002] Electric buses have good environmental and social benefits, and offer great development potential for future public transportation systems. However, the large-scale adoption of electric buses will face huge obstacles, not only the limited operating range and long charging time, but also the characteristics of the power grid, namely time-of-use electricity prices and peak load risks. On the one hand, the operating cost depends largely on time-of-use electricity prices and vehicle scheduling. On the other hand, the imbalance in charging demand caused by vehicle scheduling will bring about peak load risks, posing a potential threat to the security of the power grid. With the popularization of electric buses, there is an increasing need for meticulous design and management of electric bus scheduling, which can not only reduce system costs but also ensure the security of the power grid.

[0003] In the pure electric regional electric bus dispatch problem, the electric bus departs from different depots, completes a series of trips and uses idle time to charge, and finally returns to the original depot. The vehicle execution operation plan includes the processes of departure, collection, execution of trips, waiting, idling, and charging.

[0004] In the traditional bus regional scheduling model, the minimum total operating cost or the minimum fleet size is usually used as the optimization goal, where the total operating cost includes vehicle use cost, distance cost, waiting time cost and fuel replenishment cost. However, in the pure electric bus regional scheduling problem, it not only involves the operation of the bus company, but also the load problem of the power grid. When the electric bus is connected to the power grid, the city's electricity consumption increases dramatically, and due to the uncertainty of the daytime charging time and space, if the bus is charged on a large scale, it will increase the burden on the power grid and affect the safety of the equipment. However, the power grid load problem and the bus company cost problem are conflicting at some levels. For the bus company to have the best cost, there will inevitably be many vehicles gathered in the period of low electricity price for charging, which will lead to an increase in the peak load of the power grid. The decrease in the peak load of the power grid will inevitably spread the charging demand of the vehicles to various time periods. Some vehicles may choose to charge during the period of high electricity price, which will bring additional operating costs to the bus company. Therefore, in the pure electric bus regional scheduling model, it is necessary not only to consider the operating cost of the bus company, but also to consider the problem of the minimum peak load of the power grid. Summary of the invention

[0005] The main purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art and provide an electric bus regional scheduling method that takes into account time-of-use electricity prices and energy consumption control. By considering electricity prices in different time periods and the peak load of the power grid during charging, the problem that in the pure electric bus regional scheduling model, not only the operating expenses of the bus company need to be considered, but also the minimum load peak of the power grid should be taken into account is solved.

[0006] The first object of the present invention is to provide an electric bus regional scheduling method;

[0007] A second object of the present invention is to provide an electric bus regional dispatching system.

[0008] The first object of the present invention is achieved by the following technical solutions:

[0009] A method for regional dispatching of electric buses, comprising the following steps:

[0010] Establish a time-extended network model and explain the multi-depot scheduling problem of pure electric buses through the time-extended network model;

[0011] Set the time to expand the nodes and arcs of the network model, and set the cost of the arcs;

[0012] According to the nodes and arcs of the time-expanded network model and the arc costs, a dual-objective integer programming model considering time-of-use electricity prices and energy consumption control is established;

[0013] A lexicographic optimization method is used to process the dual objectives of a dual objective integer programming model, wherein the dual objectives include a first objective problem and a second objective problem, and the dual objective problem is converted into the first objective and the second objective of a single objective model for solving;

[0014] Design a branch-and-price algorithm to solve the first objective of the transformed single-objective model;

[0015] The optimal solution obtained for the first objective is used as the initial solution for the second objective, and the second objective of the transformed single-objective model is solved by a commercial solver to complete the regional scheduling of electric buses.

[0016] Furthermore, the time-extended network model is established as follows:

[0017] The network is layered according to the parking lot to establish a time-expanded network model G = (V k ,A k ), where V is the set of nodes at the kth level, and A is the set of arcs at the kth level;

[0018] In the time extension network, different layers represent different parking lots, and a train trip can be specified to be completed by a certain parking lot; a node represents the execution of a certain plan, and the nodes include: the starting or ending point of the parking lot, the train trip node, and the charging plan node; an arc represents a feasible connection between two plans, and the arcs include: departure arc, arrival arc, train trip connection arc, and charging arc.

[0019] Furthermore, the cost of setting the arc is specifically:

[0020] The cost in the multi-depot scheduling problem is reflected in the arc cost, and the arc cost is set as:

[0021]

[0022] in, For the starting arc, For train connection arcs, For the charging arc, To close the arc, Indicates the end position el from node i i To the starting position sl of node j j The empty driving distance, c d represents the unit distance cost, c w Indicates the unit waiting time cost, where the waiting time ID ij =st j -et i -DH ij , st j represents the starting time of node j, et i Indicates the end time of node i, DH ij Indicates the end position el from node i i To the starting position sl of node j j The idle time, c v represents the fixed cost per vehicle, Indicates the charging cost;

[0023] Charging costs The calculation formula is:

[0024]

[0025] Where P is the charging power of the electric bus, W(t) is the electricity price at time t, and st t is the starting time of the charging plan t, et t is the end time of charging plan t, c f Fixed cost for charging.

[0026] Furthermore, the nodes and arcs of the time-expanded network model and the arc costs are used to establish a dual-objective integer programming model that considers time-of-use electricity prices and energy consumption control, specifically:

[0027] The dual-objective integer programming model needs to consider the operating costs of the bus company and the minimum load peak of the power grid; then:

[0028]

[0029] Min L (3)

[0030] st

[0031]

[0032]

[0033]

[0034]

[0035]

[0036]

[0037]

[0038]

[0039]

[0040]

[0041]

[0042] in, is a Boolean variable, indicating whether the arc from node i to node j in the kth layer in the time extension network model is selected; is a continuous variable, which represents the cumulative mileage from the k-th layer vehicle to the node i in the time extension network model. This variable is used to track whether the cumulative mileage of the vehicle exceeds the upper limit of the cruising range; represents the cost from node i to node j in the kth layer; represents the distance from node i to node j at the kth layer; G represents the upper limit of the accumulated mileage of public transportation vehicles; N represents the upper limit of the fleet size; represents the upper limit of parking lot k capacity; Indicates charging plan T j The upper capacity limit of the charging station; L represents the peak load, which is the maximum number of vehicles that can be charged at each charging station in the same period of time;

[0043] In objective function (2), the optimization objective is to minimize the total cost; in objective function (3), the objective is to minimize the load peak; constraint (4) is the trip coverage constraint, which means that each trip is carried out by one and only one bus; constraint (5) is the flow conservation constraint, which means that all nodes in the same layer except the starting point and the end point must comply with the flow conservation constraint; constraint (6) is the cumulative mileage calculation formula, which means that the cumulative mileage of node j is equal to the cumulative mileage of the previous node i plus the distance between node i and node j; constraint (7) means that at the starting point or the charging station, the cumulative mileage is cleared. The constraint (8) indicates that the cumulative mileage cannot exceed the maximum mileage; constraint (9) is the fleet size limit, indicating that the number of vehicles cannot exceed the maximum fleet size; constraint (10) is the parking lot capacity limit, indicating that the number of vehicles sent from the parking lot cannot exceed the maximum parking lot capacity; constraint (11) is the charging station capacity limit, indicating that the number of vehicles in the same charging plan cannot exceed the maximum capacity of the charging station; constraint (12) is the load constraint, indicating that the load of each charging station does not exceed the load peak; constraint (13) indicates that the decision variable is a 0 or 1 variable. Constraint (14) indicates that the cumulative mileage of the vehicle is greater than or equal to 0;

[0044] For constraint (6) and constraint (8), since there are nonlinear terms, they are linearized. Here, the big M method is used. For constraint (6), the following linear formula is used for replacement:

[0045]

[0046]

[0047] Among them, M is a large positive constant; when hour, It's obvious Established; when hour, Constraints (15) and (16) are relaxed;

[0048] Similarly, for constraint (8), the following linear formula is used for substitution:

[0049]

[0050] Since then, a bi-objective integer programming model has been constructed.

[0051] Furthermore, the dual objectives of the dual objective integer programming model are processed by the lexicographic optimization method, and the dual objective problem is converted into the first objective and the second objective of the single objective model for solving, specifically: the first objective is to minimize the operating cost objective, and the second objective is to minimize the load peak objective; the dual objectives of the dual objective integer programming model are converted into the minimization of the operating cost objective of the minimization cost problem model and the minimization of the load peak objective of the minimization load peak problem model;

[0052] Among them, the cost minimization problem model is as follows:

[0053]

[0054] st

[0055]

[0056] Eqs:

[0057]

[0058]

[0059]

[0060]

[0061]

[0062]

[0063]

[0064]

[0065] When solving the first objective, the optimal solution is obtained After that, constrain Add to the constraints and solve the problem of minimizing the load peak with the objective function as the goal; the model is as follows:

[0066] MinL (28)

[0067] st

[0068]

[0069]

[0070]

[0071] Eqs:

[0072]

[0073]

[0074]

[0075]

[0076]

[0077]

[0078]

[0079]

[0080]

[0081] in, is a Boolean variable, indicating whether the arc from node i to node j in the kth layer in the time extension network model is selected; is a continuous variable, which represents the cumulative mileage from the k-th layer vehicle to the node i in the time extension network model. This variable is used to track whether the cumulative mileage of the vehicle exceeds the upper limit of the cruising range; represents the cost from node i to node j in the kth layer; represents the distance from node i to node j at the kth layer; G represents the upper limit of the accumulated mileage of public transportation vehicles; N represents the upper limit of the fleet size; represents the upper limit of parking lot k capacity; Indicates charging plan T j The upper capacity limit of the charging station; L represents the peak load, which is the maximum number of vehicles that can be charged at each charging station in the same period of time.

[0082] Furthermore, the branch pricing algorithm is designed to solve the first objective of the transformed single-objective model through the branch pricing algorithm; specifically, an efficient branch pricing algorithm is designed, a special heuristic rule and a train chain pool acceleration strategy are proposed, and the above algorithm is used to solve: the first objective is to minimize the operating cost objective;

[0083] The problem of minimizing operating costs is solved using a column generation algorithm, which transforms the original problem into a column generation main problem and sub-problems. The main problem is the set partitioning problem of the train chain, and the sub-problem is the shortest path problem with resource constraints:

[0084] For the main question:

[0085] Restricting the main problem, that is, solving the set partitioning problem of the train chain, the solution model RMP_COST is as follows:

[0086] (RMP_COST)

[0087] min∑ r∈R c r z r (41)

[0088] st

[0089]

[0090]

[0091] ∑ r∈R z r ≤N (45)

[0092]

[0093] In the RMP_COST model, the decision variable is z r , indicating whether the train chain r is selected; R is the set of train chains that restrict the main problem, c r is the cost of the train chain r∈R, It is a Boolean parameter, indicating whether the train chain starts from the k yard; is a Boolean parameter, indicating whether the train chain r covers the node s∈S, where S is the train set; is a Boolean parameter, indicating whether the train chain r covers the node t∈T, where T is the charging plan set; N represents the upper limit of the fleet size, represents the upper limit of parking lot k capacity, Indicates charging plan T t The capacity limit of the charging station you are at;

[0094] In order to generate the constraint conditions of the main problem, the objective function (41) represents the minimization of the operating cost, whose value is equal to the sum of the costs of the selected train chain; constraint (42) represents the parking lot capacity constraint; constraint (43) represents the train unique coverage constraint; constraint (44) represents the charging station capacity constraint; constraint (45) represents the maximum vehicle size limit; constraint (46) represents the value range of the decision variable, z r is a Boolean variable;

[0095] The dual variable is usually restricted to a linear relaxation of the master problem, that is, z r Change from a Boolean variable to a continuous variable with a value range of [0,1];

[0096] For the sub-problems:

[0097] For the minimization problem, when the test number of all feasible columns is greater than or equal to 0, the introduction of the column does not improve the problem, and the minimization problem obtains the optimal value; for the main problem of train chain set partitioning, let the dual variables of constraints (42)-(45) be: α k , β s , γ t ,π(k∈K,s∈S,t∈T);

[0098] The subproblem is to find the train chain with the smallest number of tests, that is, to find the shortest path in the network. Since the train chain mileage constraint needs to be considered, it is a shortest path problem with resource constraints. The subproblem model is as follows:

[0099] (SUBP_COST)

[0100]

[0101] st

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108] in for:

[0109]

[0110] In the SUB_COST model, the decision variables are and is a Boolean variable, representing whether the arc (i, j) at the kth layer is selected; is the cumulative mileage of node i at the kth layer; G is the maximum mileage of the bus;

[0111] Objective function (47) represents minimizing the number of train chain inspections, and its value is equal to the sum of the paths of all selected edges; constraint (48) represents that the train chain starts from the starting point and has only one path; constraint (49) is a flow conservation constraint, which means that all nodes in the same layer except the starting point and the end point must comply with the flow conservation constraint; constraint (50) is a cumulative mileage calculation formula, which represents the cumulative mileage of node j, which is equal to the cumulative mileage of the preceding node i plus the distance between node i and node j; constraint (51) represents that at the starting point or at the charging station, the cumulative mileage is cleared to 0, indicating that the vehicle starts from the station or is fully charged at the charging station; constraint (52) represents that the cumulative mileage cannot exceed the maximum mileage; constraint (53) represents that the decision variable is a 0 or 1 variable; constraint (54) represents that the vehicle's cumulative mileage is greater than or equal to 0;

[0112] The time expansion network has multiple layers. For each layer k∈K, the time expansion network can be independently calculated from the starting point θ k To the end point θ k ′, this process adopts parallel computing method; when the shortest path problem of each layer is solved, the path with the smallest cost in all layers is taken as the optimal solution, that is, the optimal solution of the sub-problem; the shortest path problem of each layer is the shortest path problem with resource constraints, and the label correction method is used to solve the shortest path problem with resource constraints.

[0113] Furthermore, the label correction method is as follows:

[0114] 1) Tag definition

[0115] make Indicates that the mth item starts from the parking lot θ k The path to node i; for the path Label With path Associated, with two resources in Indicates the starting point θ from the parking lot k The sum of the costs of the mth path to node i, Indicates the starting point θ from the parking lot k The cumulative mileage of the mth path to node i;

[0116] 2) Tag extension

[0117] Label The maintenance of adopts forward dynamic programming; in the label correction method, for the edge (i, j) and the path Extend the inode label The label of the jth node The label expansion calculation rules are as follows:

[0118]

[0119]

[0120] The label extension conditions are as follows:

[0121]

[0122] When the cruising range condition is met, the label is extended, and the cumulative cruising range also needs to distinguish whether the j node is a charging station. If so, the cumulative cruising range needs to be cleared;

[0123] 3) Label dominance

[0124] The computational speed of the labeling algorithm depends on the number of labels. When a label is not part of the optimal solution, the label will be removed. The label dominance rule will be used to remove labels that are not the optimal solution to speed up the computation. Domination Tags If and only if the following conditions are met at the same time:

[0125]

[0126]

[0127] If the tag Domination Tags It represents the label Related Paths is not part of the optimal solution, because any A feasible extension also applies to and The cost is lower; therefore, and will be removed;

[0128] First, the label correction method needs to be initialized. k The label is set to (0,0), indicating that the cost and cumulative mileage at the starting point are both 0; the label correction method needs to record the node number of the label update and check all arcs emitted from a node in the next iteration. The FIFO node processing rule is adopted here, so the nodes are stored in a queue; when processing a node, all labels of the node are first compared pairwise, and the label dominance rule is used to delete redundant and inefficient labels; if the label extension condition (43) is met, the label of this node is extended; when the queue is empty, the algorithm backtracks to find the optimal path and the optimal solution.

[0129] Furthermore, the branch pricing algorithm adopts a branch strategy, specifically:

[0130] Branch the connection relationship between nodes:

[0131] When the solution of the linear relaxation restricted main problem is not an integer, the solution of the set partitioning model is transformed into the solution of the time extension network model, and the unbranched arcs (i, j) are branched; two sub-problems are obtained, each containing branch information x ij =0 and x ij =1; for branch information x ij = 0, delete arc (i, j) on the time extension network; and for branch information x ij =1, the arcs on the time extension network can be deleted according to the unique constraint of train number;

[0132] The branch-and-bound tree has a global upper bound to record the value of the best integer solution found; it uses a depth-first search rule and always prioritizes solving branch x. ij =1;

[0133] It is necessary to use a heuristic algorithm to construct an initial solution and convert the decimal solution into an integer solution;

[0134] The heuristic algorithm continuously sets the cost of arcs on the time expansion network and solves the shortest path problem on the time expansion network to obtain a train chain. The solution method is the same as the sub-problem. After the train chain is obtained, it is added to the train chain set, and the covered train nodes, parking lot nodes and charging nodes on the time expansion network are deleted. Repeat the above process until all trains are covered.

[0135] Specifically:

[0136] Heuristically construct an initial solution:

[0137] The purpose of heuristically constructing the initial solution is to make each train chain cover more trains while minimizing the cost as much as possible, while ensuring that the solution meets the capacity constraint and the vehicle quantity constraint. The steps of heuristically constructing the initial solution are as follows:

[0138] Step 1. Time-expanded network construction: Set the cost of arc ij at layer k to v is a constant, which is a very small number. is the cost of each arc ij at the kth layer; if node j is a train node, then μ = -1, otherwise μ = 0. By adding this item, the vehicle can cover more trains;

[0139] Step 2. Solve the shortest path problem on the time-expanded network, obtain a train chain, and delete the covered train nodes, parking lot nodes with capacity reaching the upper limit, and charging nodes; the obtained train chain meets the range constraint and will not conflict with the covered trains, parking lots with full capacity, and charging stations;

[0140] Step 3. Repeat Step 2 until all train nodes are covered;

[0141] The initial solution also adds an artificial train chain that meets all the constraints of the main problem but is very expensive to ensure the feasibility of the problem;

[0142] Heuristic fix for decimal solutions:

[0143] When solving the subproblems of each branch and bound tree, the solution obtained is often not an integer solution. In this case, a heuristic method is used to repair the decimal solution. The steps are as follows:

[0144] Step 1. Select the train chain with the largest decimal solution that has not been considered;

[0145] Step 2. Time-expanded network construction: The cost of arc ij is set to If node j is a node covered by the train chain, then λ=-10, otherwise λ=0;

[0146] Step 3. Solve the shortest path problem on the time-expanded network, obtain a train chain, and delete the covered train nodes, parking lot nodes with capacity limits, and charging nodes; the obtained train chain meets the range constraint and will not conflict with the covered trains, parking lots with full capacity, and charging stations;

[0147] Step 4. Repeat Step 1, Step 2, and Step 3 until all train nodes are covered.

[0148] Furthermore, the train chain pool acceleration strategy is specifically as follows:

[0149] The train number chain of the child node of the branch and bound tree uses the existing information of the parent node to speed up the solution of the child node; the train number chain pool R is designed pool , used to collect all the train chains generated by each node of branch and bound; let the branch and bound tree node currently being solved be p, then the initial train chain set R of the main problem of node p p Available from the train chain pool R pool However, the train number chains that do not meet the branch constraints need to be deleted:

[0150] R p =R pool -r (61)

[0151] Where r∈R pool , and r does not satisfy the branch constraint of p;

[0152] In particular, for the root node of the branch and bound tree, R rootIt consists of two parts: the train chain generated by the heuristic construction of the initial solution and the expensive manual train chain;

[0153] When the node p is solved, the train chain generated by the column generation process is merged with the train chain pool to obtain a new train chain pool:

[0154] R pool =R pool ∪R p (62).

[0155] The second object of the present invention is achieved by the following technical solutions:

[0156] An electric bus regional dispatching system, comprising:

[0157] The time extension network module is used to establish a time extension network model, and explain the multi-depot scheduling problem of pure electric buses through the time extension network model; and set the nodes and arcs of the time extension network model, as well as the cost of the arcs;

[0158] The dual-objective integer programming module is used to expand the nodes and arcs of the network model according to time and establish a dual-objective integer programming model that considers time-of-use electricity prices and energy consumption control;

[0159] A conversion module, used for processing the dual objectives of the dual objective integer programming model by using a lexicographic optimization method, and converting the dual objective problem into the first objective and the second objective of the single objective model for solving;

[0160] A first objective solving module, used for designing a branch-and-price algorithm, and solving the first objective of the transformed single objective model through the branch-and-price algorithm;

[0161] A second objective solving module, used to use the optimal solution obtained from the first objective as the initial solution of the second objective, and solve the second objective of the converted single objective model through a commercial solver;

[0162] The result output module is used to output the scheduling plan.

[0163] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0164] The present invention explains the multi-depot scheduling problem of pure electric buses by establishing a time expansion network model; setting the cost of arcs in the time expansion network; establishing a dual-objective integer programming model considering time-of-use electricity prices and energy consumption control; using a lexicographic optimization method to process the dual objectives, and converting the dual-objective problem into a single objective for solving; designing an efficient solution algorithm to solve the first objective of the converted single objective model, and using the optimal solution obtained by the first objective as the initial solution of the second objective, in order to find a solution with lower load peak and lower operating cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0165] Figure 1 It is a method flow chart of an electric bus regional dispatching method according to the present invention;

[0166] Figure 2 is a flowchart of the scheduling method in Embodiment 1 of the present invention;

[0167] Figure 3 is a schematic diagram of the time expansion network structure in Embodiment 1 of the present invention;

[0168] Figure 4 is a flow chart of the branch pricing algorithm in Embodiment 1 of the present invention;

[0169] Figure 5 This is a schematic diagram of the distribution of parking lots and charging stations in Example 1 of the present invention;

[0170] Figure 6 It is the peak and valley time-of-use electricity price diagram of Guangzhou in Embodiment 1 of the present invention;

[0171] Figure 7a The second goal in Embodiment 1 of the present invention is to reduce the load peak of the charging station in each time period during the load peak from 12:00 to 14:00;

[0172] Figure 7b is the load peak value of the charging station in each time period when the charging vehicles are completely removed in Embodiment 1 of the present invention;

[0173] Figure 8 The line conditions and bus line distribution map in Embodiment 1 of the present invention;

[0174] Fig. 9 This is a comparison chart of the results calculated by the extended FIFO rule and the branch pricing algorithm in Embodiment 1 of the present invention;

[0175] Fig.10a It is a charging period and electricity price distribution diagram obtained by the expanded FIFO rule in Embodiment 1 of the present invention;

[0176] Fig.10b The charging period and electricity price distribution diagram calculated by the branch pricing algorithm in Embodiment 1 of the present invention;

[0177] Fig.11 It is a structural diagram of an electric bus regional dispatching system in Example 2 of the present invention. DETAILED DESCRIPTION

[0178] The present invention is further described in detail below in conjunction with embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0179] Embodiment 1:

[0180] A method for regional dispatching of electric buses, such as Figure 1 As shown, the following steps are included:

[0181] Establish a time-extended network model and explain the multi-depot scheduling problem of pure electric buses through the time-extended network model;

[0182] Set the time to expand the nodes and arcs of the network model, and set the cost of the arcs;

[0183] According to the nodes and arcs of the time-expanded network model and the arc costs, a dual-objective integer programming model considering time-of-use electricity prices and energy consumption control is established;

[0184] A lexicographic optimization method is used to process the dual objectives of a dual objective integer programming model, wherein the dual objectives include a first objective problem and a second objective problem, and the dual objective problem is converted into the first objective and the second objective of a single objective model for solving;

[0185] Design a branch-and-price algorithm to solve the first objective of the transformed single-objective model;

[0186] The optimal solution obtained for the first objective is used as the initial solution for the second objective, and the second objective of the transformed single-objective model is solved by a commercial solver to complete the regional scheduling of electric buses.

[0187] Specific as Figure 2 As shown, the content is as follows:

[0188] 1. Model establishment

[0189] The main idea of ​​the electric bus regional scheduling method considering time-of-use electricity prices and energy consumption control disclosed in the present invention is: establishing a time-expanded network model to explain the pure electric bus multi-depot scheduling problem; setting the cost of the arc in the time-expanded network; establishing a dual-objective integer programming model considering time-of-use electricity prices and energy consumption control; using a lexicographic optimization method to process the dual objectives, and converting the dual-objective problem into a single objective for solving; designing an efficient solution algorithm to solve the first objective of the converted single objective model; and calling a commercial solver to solve the second objective of the converted single objective model.

[0190] 1. Problem Description

[0191] A bus company uses a large number of electric buses to provide public transportation services. In daily operations, the vehicles need to be charged. How to reasonably arrange charging is an important part of the electric bus regional dispatch plan. In order to reduce the cost of the bus company and reduce the peak load of the power grid to improve the safety of the power grid, it is necessary to formulate a scientific and reasonable electric bus regional dispatch plan.

[0192] The present invention provides a solution for regional dispatching of electric buses by considering electricity prices at different time periods and peak load of the power grid during charging.

[0193] 2. Basic Assumptions

[0194] (1) Travel time assumption: It is assumed that the travel time of each train is a constant, while the travel time of trains in different time periods may be different.

[0195] (2) Departure status assumption: It is assumed that all electric buses can be fully charged at night and the cumulative mileage is 0 when they depart from the depot.

[0196] (3) Daytime charging assumption: Electric buses are allowed to use their free time to go to charging stations for charging during the operating hours. Here, it is assumed that electric buses use fast charging mode. Charging stations can be set up at bus stations, parking lots, or at a separate location. It is assumed that the charging time of all electric buses is equal to the full charging time of the electric buses, and the electric buses are charged at maximum power, so that each vehicle can be fully charged after charging.

[0197] 3. Establish a time-expanding network model

[0198] In the multi-depot dispatching problem of pure electric buses, in order to facilitate tracking of the depot to which a vehicle belongs, the network is layered according to the depot to establish a time-expanded network model G = (V k ,A k ), where V is the set of nodes at the kth layer, and A is the set of arcs at the kth layer; the time-expanded network is as follows Figure 3 shown.

[0199] In the time extension network, different layers represent different depots, and a trip can be specified to be completed by a certain depot.

[0200] Nodes represent the execution of a plan and are divided into three types: the starting point or end point of the parking lot, the train node, and the charging plan node. The train node and the charging plan node are the abstractions of the train and the charging plan on the time extension network. They both have the attributes of start time, start station, end time, and end station. Their meanings are the same as the train and the charging plan.

[0201] The arc represents a feasible connection between two plans, including the departure arc, the return arc, the train connection arc, and the charging arc. The upper capacity limit of these arcs is 1, ensuring that each train can only be executed by one vehicle. The departure arc is the arc connecting the starting point of the depot with each train node, indicating the process of the vehicle departure and execution of the train. The return arc is the arc connecting each train node with the end point of the depot, indicating the process of the vehicle returning. The train connection arc is the arc connecting the train node or the charging plan node with the train node, which corresponds to two situations in actual situations: 1) Waiting: The vehicle waits at the original station and executes the next train; 2) Empty driving: The vehicle drives empty to other stations, waits and executes the train. The charging arc is the arc connecting the train node with the charging plan node, indicating that the vehicle drives empty from the station where the current train is located to the charging station and completes charging.

[0202] In the time extension network, let S represent the set of train nodes, T represent the set of charging nodes, Θ k represents the starting point of the parking lot on the kth floor, Θ′ k Indicates the end point of the k-th floor. For a train connection arc or charging arc (i, j) ∈ A k |i∈S,j∈S∪T, the connection condition needs to be met: st j -et i -DH ij ≥0, where st j represents the starting time of node j, et i Indicates the end time of node i, DH ij Indicates the end position el from node i i To the starting position sl of node j j The feasible train chain starts from the starting point of the yard Θ k Departure, passing a series of train nodes and charging plan nodes and meeting the mileage limit, and finally returning to the parking lot end Θ′ k The multi-depot scheduling problem is to find several feasible train chains that can completely cover the train set S and meet the capacity constraints of the depot and charging station.

[0203] 4. Set the cost of arcs in the time-extended network

[0204] In the multi-depot scheduling problem, the cost items involve not only vehicle fixed costs, driving distance costs, waiting time costs, but also charging costs. The charging cost includes electricity price costs and charging fixed costs, and the electricity price costs are related to charging time, charging power, and charging costs. Due to the time-of-use electricity price mechanism, charging during peak hours will pay more electricity bills than charging during off-peak hours. Taking the peak and valley time-of-use electricity prices for commercial electricity in Guangzhou as an example, the electricity price is divided into three stages: peak, flat, and off-peak. The peak electricity price is 0.97 yuan / kWh, and the off-peak electricity price is 0.3 yuan / kWh. That is to say, charging during off-peak hours costs only one-third of that during peak hours. The fluctuation of electricity prices affects the charging cost of buses, so the factor of time-of-use electricity prices needs to be considered in the model. In addition, in charging activities, the maintenance of charging sites, personnel costs, and battery depreciation costs are also factors that need to be considered. These are fixed charging costs. For the charging plan t∈T, the charging cost The calculation formula is:

[0205]

[0206] Where P is the charging power of the electric bus, W(t) is the electricity price at time t, and st t is the starting time of the charging plan t, et t is the end time of charging plan t, c f Fixed cost for charging.

[0207] Therefore, the cost of an arc can be calculated as follows:

[0208]

[0209] in, For the starting arc, For train connection arcs, For the charging arc, For the starting arc, Indicates the end position el from node i i To the starting position sl of node j j The empty driving distance, c d represents the unit distance cost, c w Indicates the unit waiting time cost, where the waiting time ID ij =st j -et i -DH ij , st j represents the starting time of node j, et i Indicates the end time of node i, DH ij Indicates the end position el from node i i To the starting position sl of node j jThe idle time, c v represents the fixed cost per vehicle, Indicates the charging cost, which varies with the charging period.

[0210] S3. Establish a dual-objective integer programming model considering time-of-use electricity prices and energy consumption control.

[0211] In the regional dispatch model of pure electric buses, it is necessary not only to consider the operating costs of the bus companies, but also to take into account the minimum load peak of the power grid.

[0212]

[0213] Min L(4) st

[0214]

[0215]

[0216]

[0217]

[0218]

[0219]

[0220]

[0221]

[0222]

[0223]

[0224]

[0225] in, is a Boolean variable, indicating whether the arc from node i to node j in the kth layer in the time extension network model is selected; is a continuous variable, which represents the cumulative mileage from the k-th layer vehicle to the node i in the time extension network model. This variable is used to track whether the cumulative mileage of the vehicle exceeds the upper limit of the cruising range; represents the cost from node i to node j in the kth layer; represents the distance from node i to node j at the kth layer; G represents the upper limit of the accumulated mileage of public transportation vehicles; N represents the upper limit of the fleet size; represents the upper limit of parking lot k capacity; Indicates charging plan T jThe upper capacity limit of the charging station; L represents the peak load, which is the maximum number of vehicles that can be charged at each charging station in the same period of time.

[0226] In objective function (3), the optimization objective is to minimize the total cost. In objective function (4), the objective is to minimize the load peak. Constraint (5) is the trip coverage constraint, which means that each trip is executed by only one bus. Constraint (6) is the flow conservation constraint, which means that all nodes in the same layer (except the starting point and the end point) must comply with the flow conservation constraint. Constraint (7) is the cumulative mileage calculation formula, which means that the cumulative mileage of node j is equal to the cumulative mileage of the previous node i plus the distance between node i and node j. Constraint (8) means that at the starting point or the charging station, the cumulative mileage is cleared to 0, indicating that the vehicle departs from the station or the vehicle is fully charged at the charging station. Constraint (9) means that the cumulative mileage cannot exceed the maximum mileage. Constraint (10) is the fleet size limit, which means that the number of vehicles cannot exceed the maximum fleet size. Constraint (11) is the parking lot capacity limit, which means that the number of vehicles sent from the parking lot cannot exceed the maximum parking lot capacity. Constraint (12) is the charging station capacity limit, which means that the number of vehicles in the same charging plan cannot exceed the maximum capacity of the charging station. Constraint (13) is a load constraint, which means that the load of each charging station does not exceed the peak load. Constraint (14) means that the decision variable is a 0 or 1 variable. Constraint (15) means that the vehicle's cumulative mileage is greater than or equal to 0.

[0227] For constraint (7) and constraint (9), since there are nonlinear terms, they will be linearized. Here, the big M method is used, and for constraint (7), the following linear formula is used for replacement:

[0228]

[0229]

[0230] Among them, M is a large normal quantity. hour, It's obvious Established; when hour, Constraints (16) and (17) are relaxed. Similarly, for constraint (9), the following linear formula is used for replacement:

[0231]

[0232] So far, a mixed integer linear programming bi-objective model has been constructed.

[0233] S4. Use the lexicographic optimization method to deal with dual objectives and transform the dual objective problem into a single objective for solution.

[0234] The mixed integer linear programming bi-objective model consists of two objectives: minimizing the operating cost and minimizing the load peak. High-level information is known before optimization, and reducing the operating cost of bus companies is usually more important than reducing the peak load of the power grid. Therefore, the lexicographic optimization method is used, that is, the bi-objective problem is converted into a single objective for solution given the target preference information.

[0235] First, we ignore the objective function (4) and the load constraint (13) and only solve the cost minimization problem. Obviously, this is a single-objective problem. Next, we further optimize the second objective when the first objective reaches the optimal value. That is, when the operating cost is not worse than the solution obtained in the first objective, we solve the problem with minimizing the load peak as the secondary objective. The cost minimization problem model is as follows:

[0236]

[0237] st

[0238]

[0239] Eqs:(5)-(12)

[0240] When solving the first objective, the optimal solution is obtained After that, constrain Add to the constraints and solve the problem of minimizing the load peak with the objective function as the goal. The model is as follows:

[0241] MinL (21)

[0242] st

[0243]

[0244]

[0245]

[0246] Eqs:(5)-(13)

[0247] in, is a Boolean variable, indicating whether the arc from node i to node j in the kth layer in the time extension network model is selected; is a continuous variable, which represents the cumulative mileage from the k-th layer vehicle to the node i in the time extension network model. This variable is used to track whether the cumulative mileage of the vehicle exceeds the upper limit of the cruising range; represents the cost from node i to node j in the kth layer; represents the distance from node i to node j at the kth layer; G represents the upper limit of the accumulated mileage of public transportation vehicles; N represents the upper limit of the fleet size; represents the upper limit of parking lot k capacity; Indicates charging plan T j The upper capacity limit of the charging station; L represents the peak load, which is the maximum number of vehicles that can be charged at each charging station in the same period of time;

[0248] S5. Design an efficient solution algorithm to solve the first objective of the transformed single-objective model.

[0249] For the first goal of minimizing operating costs, an efficient branch pricing algorithm is designed, a special heuristic rule and a train chain pool acceleration strategy are proposed, and the above algorithm is used to solve the problem. Figure 4 This is the flow chart of the branch pricing algorithm.

[0250] The problem of minimizing operating costs is expressed as a problem of selecting train chains. Each train chain satisfies the mileage constraint and the train connection constraint. All train chains satisfy the vehicle number constraint, the parking lot capacity constraint, the charging station capacity constraint, and the train unique constraint. However, the number of feasible train chains is huge. If all train chains are expressed, it will take a long time. Therefore, a column generation algorithm is used. This algorithm does not process all train chains at the same time, but only optimizes and solves the current train chain by restricting the main problem. The remaining train chains will only be selected into the train chain pool when the column generation subproblem is judged to improve the current optimal solution of the restricted main problem, otherwise, they will be delayed. Therefore, the column generation algorithm only considers the columns that can improve the optimal solution of the restricted main problem instead of all possible columns. The problem of minimizing operating costs is solved by the column generation algorithm, which can transform the original problem into the column generation main problem and subproblems, where the main problem is the set partition problem of the train chain and the subproblem is the shortest path problem with resource constraints.

[0251] ■Main Question

[0252] The restricted main problem is to solve the set partitioning problem of the train chain. The solution model RMP_COST is as follows:

[0253] (RMP_COST)

[0254] min∑ r∈R c r z r (25)

[0255] st

[0256]

[0257]

[0258] ∑ r∈R zr ≤N (29)

[0259]

[0260] In the RMP_COST model, the decision variable is z r , indicating whether the train chain r is selected; R is the set of train chains that restrict the main problem, c r is the cost of the train chain r∈R, It is a Boolean parameter, indicating whether the train chain starts from the k yard; is a Boolean parameter, indicating whether the train chain r covers the node s∈S, where S is the train set; is a Boolean parameter, indicating whether the train chain r covers the node t∈T, where T is the charging plan set; N represents the upper limit of the fleet size, represents the upper limit of parking lot k capacity, Indicates charging plan T t The maximum capacity of the charging station you are at.

[0261] Objective function (25) represents minimizing the operating cost, and its value is equal to the sum of the costs of the selected train chain. Constraint (26) represents the parking lot capacity constraint. Constraint (27) represents the train unique coverage constraint. Constraint (28) represents the charging station capacity constraint. Constraint (29) represents the maximum vehicle size limit. Constraint (30) represents the value range of the decision variable, z r A Boolean variable.

[0262] In order to generate the dual variables of the constraints of the master problem, the master problem is usually linearly relaxed, that is, z r Change from a Boolean variable to a continuous variable with a value range of [0,1].

[0263] Sub-questions

[0264] For the minimization problem, when the test number of all feasible columns is greater than or equal to 0, the introduction of the column does not improve the problem, and the minimization problem obtains the optimal value. For the main problem of train chain set partitioning, let the dual variables of constraints (26)-(29) be: α k , β s , γ t ,π(k∈K,s∈S,t∈T), then the test number of the train chain r∈R is:

[0265]

[0266] Train chain cost c r It can be expressed by extending the edges of the network in time:

[0267]

[0268] Boolean parameters It can also be expressed in a similar way:

[0269]

[0270]

[0271]

[0272] Substituting formula (32) to formula (35) into formula (31), we can obtain the time extension network expression of the train chain r∈R test number:

[0273]

[0274] Let the test number of each arc be

[0275]

[0276] In this way, subproblems can be constructed based on the number of checks on each edge. The subproblem is to find a train chain with the smallest number of checks, that is, to find the shortest path in the network. Since the mileage constraint of the train chain needs to be considered, it is a shortest path problem with resource constraints. The subproblem model is as follows:

[0277] (SUBP_COST)

[0278]

[0279] st

[0280]

[0281]

[0282]

[0283]

[0284]

[0285]

[0286] In the SUB_COST model, the decision variables are and is a Boolean variable, representing whether the arc (i, j) at the kth layer is selected; is the cumulative mileage of node i at the kth layer; G is the maximum mileage of the bus;.

[0287] Objective function (38) represents minimizing the number of train chain checks, and its value is equal to the sum of the paths of each selected edge. Constraint (39) indicates that the train chain starts from the starting point and there is only one of them. Constraint (40) is a flow conservation constraint, which means that all nodes in the same layer (except the starting point and the end point) must comply with the flow conservation constraint. Constraint (41) is a cumulative mileage calculation formula, which means that the cumulative mileage of node j is equal to the cumulative mileage of the previous node i plus the distance between node i and node j. Constraint (42) indicates that at the starting point or at the charging station, the cumulative mileage is cleared to 0, indicating that the vehicle starts from the station or the vehicle is fully charged at the charging station. Constraint (43) indicates that the cumulative mileage cannot exceed the maximum mileage. Constraint (44) indicates that the decision variable is a 0 or 1 variable. Constraint (45) indicates that the vehicle's cumulative mileage is greater than or equal to 0.

[0288] Note that the time extension network has multiple layers. For each layer k∈K, we can independently calculate k To the end point θ k ′, this process can be solved in parallel. After solving the shortest path problem of each layer, the path with the lowest cost in all layers is taken as the optimal solution, which is the optimal solution of the sub-problem. The shortest path problem of each layer is a shortest path problem with resource constraints. The label setting method or the label correction method can be used to solve the shortest path problem with resource constraints.

[0289] But due to It may be negative, so there may be arcs with negative weights in the network, which makes the label setting method infeasible. Therefore, the extended label correction method is used.

[0290] The linear relaxation of the restricted master problem is not necessarily integer feasible, and often requires a branching strategy. Directly r Branching is difficult because the decision variable z is fixed r It will destroy the structure of the pricing problem. r When it is fixed to 0, the subproblem may generate new columns similar to this column, making the branching strategy inefficient. Therefore, the connection relationship between nodes can be branched.

[0291] Specifically, when the solution to the linear relaxation restricted main problem is not an integer, the solution of the set partitioning model is transformed into the solution of the time extension network model, and the unbranched arcs (i, j) are branched; two sub-problems are obtained, each containing branch information x ij =0 and x ij =1; for branch information xij = 0, delete arc (i, j) on the time extension network; and for branch information x ij = 1, the arcs on the time-extended network can be deleted according to the unique constraint of the train number. This branching strategy can reflect the branch information on the time-extended network, which is convenient and feasible.

[0292] The branch and bound tree has a global upper bound to record the value of the best integer solution found. These integer solutions come from heuristic rules or RMP solutions are integer branch and bound nodes. In order to obtain integer solutions faster, the depth-first search rule is used, and the branch x is always solved first. ij =1.

[0293] Heuristic algorithms can usually obtain feasible solutions at a lower time cost. When the problem scale is large, the solution to the relaxed main problem is usually integer infeasible, which makes it difficult to obtain an integer solution. Only when an integer solution is obtained can the upper bound be updated. In addition, the root node of the branch and bound tree often needs to provide an initial solution, which can keep the main problem feasible and reduce the number of iterations of the column generation algorithm. Therefore, it is necessary to use a heuristic algorithm to construct an initial solution and convert the decimal solution into an integer solution.

[0294] Here, a new heuristic algorithm is proposed. This heuristic algorithm can be operated on the time-extended network so that the generated solution satisfies the branching conditions defined on the time-extended network. This method obtains a train chain by continuously setting the cost of the arcs on the time-extended network and solving the shortest path problem on the time-extended network. The solution method is the same as the sub-problem. After the train chain is obtained, it is added to the train chain set, and the train nodes that have been covered on the time-extended network, the parking lot nodes that have reached the upper limit of capacity, and the charging nodes are deleted. Repeat the above process until all trains have been covered.

[0295] Therefore, how to set a reasonable arc cost is very important. The basic idea of ​​parameter setting is: ① Try to be as close to the decimal solution as possible (decimal solution repair); ② Each train chain covers as many trains as possible; ③ The cost of the train chain is the least (the cost of empty trips is as little as possible).

[0296] ■ Heuristic construction of initial solution

[0297] At the root node of the branch and bound tree, an initial solution is often required as a start. A good initial solution can keep the main problem feasible, thereby reducing the number of iterations for column generation. Generally speaking, the fixed cost of vehicles accounts for a large proportion of the total cost. A good solution needs to reduce the number of vehicles as much as possible, which means that each vehicle needs to cover more trips. The purpose of heuristically constructing the initial solution is to make each train chain cover more trains while making the cost as small as possible, while ensuring that the solution meets the capacity constraint and the vehicle number constraint. The steps for heuristically constructing the initial solution are as follows:

[0298] Step 1. Time-expanded network construction: Set the cost of arc ij at layer k to v is a constant, which is a very small number. is the cost of each arc ij at the kth layer; if node j is a train node, then μ = -1, otherwise μ = 0. By adding this item, the vehicle can cover more trains;

[0299] Step 2. Solve the shortest path problem on the time-expanded network, obtain a train chain, and delete the covered train nodes, parking lot nodes with capacity limits, and charging nodes. The resulting train chain satisfies the range constraint and does not conflict with covered trains, parking lots with full capacity, and charging stations;

[0300] Step 3. Repeat the process in step 2 until all train nodes are covered.

[0301] The heuristic initial solution may fail because of the strict constraints on the number of vehicles or the parking lot capacity. Therefore, the initial solution also adds an artificial train chain that meets all the constraints of the main problem but is very expensive to ensure the feasibility of the problem.

[0302] ■Heuristic repair of decimal solutions

[0303] When solving the subproblems of each branch and bound tree, the solution obtained is often not an integer solution. At this time, a heuristic rule is needed to convert the decimal solution into an integer solution. The benefit of repairing is that the integer solution can be obtained faster, thereby providing an upper bound for the algorithm and reducing the size of the branch and bound tree. The constructed train chain needs to be more "like" the decimal solution. At the same time, similar to the heuristic construction of the initial solution, the train chain still needs to cover more trains while making the cost as small as possible. The steps for heuristically repairing the decimal solution are as follows:

[0304] Step 1. Select the train chain with the largest decimal solution that has not been considered;

[0305] Step 2. Time-expanded network construction: The cost of arc ij is set to If the j node is a node covered by the train chain, then λ=-10, otherwise λ=0.

[0306] Step 3. Solve the shortest path problem on the time-expanded network, obtain a train chain, and delete the covered train nodes, parking lot nodes with capacity limits, and charging nodes. The resulting train chain satisfies the range constraint and does not conflict with covered trains, parking lots with full capacity, and charging stations;

[0307] Step 4. Repeat 1, 2, and 3 until all train nodes are covered.

[0308] When solving with branch and bound, it is necessary to frequently call the column generation process at each node to generate the train chain, which consumes a lot of time. The train chain of the child node of the branch and bound tree uses the existing information of the parent node to speed up the solution of the child node; the train chain pool R is designed pool , used to collect all the train chains generated by each node of branch and bound; let the branch and bound tree node currently being solved be p, then the initial train chain set R of the main problem of node p p Available from the train chain pool R pool However, the train number chains that do not meet the branch constraints need to be deleted:

[0309] R p =R pool -r (43)

[0310] Where r∈R pool , and r does not satisfy the branch constraint of p;

[0311] In particular, for the root node of the branch and bound tree, R root It consists of two parts: the train chain generated by the heuristically constructed initial solution and the expensive manual train chain.

[0312] When the node p is solved, the train chain generated by the column generation process is merged with the train chain pool to obtain a new train chain pool:

[0313] R pool =R pool ∪R p (44)

[0314] S6. Call a commercial solver to solve the second objective of the transformed single-objective model.

[0315] For the second objective of minimizing the load peak, a commercial solver is called, and the optimal solution obtained for the first objective is used as the initial solution for the second objective, in order to find a solution with a lower load peak and lower operating cost.

[0316] 2. Example Analysis

[0317] ■Example 1

[0318] like Figure 5 There are 5 parking lots and 4 charging stations. The coordinates of the parking lots are: (26,23), (33,36), (43,29), (56,12), (58,41), and the coordinates of the charging stations are: (4,23), (12,2), (43,29), (57,39), among which the point to be selected (43,29) is both a parking lot and a charging station.

[0319] For electric buses, we refer to the data of BYD K7 buses and set the maximum range G to 250km and the charging time to 60min. For costs, we set the vehicle fixed cost G v is 700 yuan, and the distance cost is C d is 2.8 yuan, and the waiting cost is C w The fixed cost of charging is 0.7 yuan. f The electricity price is based on the peak and valley time-of-use electricity price in Guangzhou. Figure 6 It shows the fluctuation of electricity prices over time. It can be seen that the electricity prices are high and low in different periods, the low electricity price period is from 24:00 to 08:00, and the high electricity price period is from 14:00 to 17:00 and 19:00 to 22:00.

[0320] In order to test data of different scales, the number of trains is set to 50, 100, 200, 400, the number of parking lots is set to 2, 5, 8, and the number of charging stations is set to 2, 4, 6, 8, for a total of 4*3*4=48 data. For all data, the upper limit of the number of node accesses is set to 200. For the case where the number of trains is 50, 100, 200, the upper limit of the solution time is set to 12h. For the case where the number of trains is 400 and the number of parking lots is 2, the upper limit of the solution time is set to 24h. For the case where the number of trains is 400 and the number of parking lots is 5 and 8, the upper limit of the solution time is set to 48h.

[0321] For these example data, we use direct CPLEX and branch-and-price algorithm to calculate the first objective, and the results are shown in Table 1 below:

[0322] Table 1

[0323]

[0324] The first column |S| in the table represents the number of trips, the second column |K| in the table represents the number of parking lots, the third column l in the table represents the number of charging stations, the fourth to eighth columns in the table are the results of CPLEX calculations, the fourth column is the number of variables in the dual-objective integer programming model, the fifth column is the lower bound, the sixth column is the best integer solution found, and the seventh column is the calculation formula The eighth column, CPU, is the CPU time of the program calculation, in seconds. The ninth to thirteenth columns are the results of the branch pricing calculation, the ninth column is the number of node visits, and the tenth to thirteenth columns have the same meaning as the fifth to eighth columns. The fifth to seventh columns are '-', which means that the call to CPLEX failed to find a feasible solution. The tenth and twelfth columns are '-', which means that the main problem of the root node has not been solved, so the lower bound and gap value of the problem cannot be obtained. However, the branch pricing algorithm runs the heuristic rule at the root node, so a feasible solution can be obtained.

[0325] As can be seen from the table, CPLEX can quickly obtain the optimal solution in small-scale examples. However, when the number of variables exceeds 20,000, CPLEX finds it difficult to obtain the optimal solution, or even fails to find a feasible solution. In contrast, although the branch-and-price algorithm is slower than CPLEX in small-scale examples, it can still find feasible solutions in large-scale examples, and the average gap is also below 10%. However, in problems with more than 1 million variables, both CPLEX and the branch-and-price algorithm find it difficult to find a feasible solution. It is worth noting that the lower bound obtained by the branch-and-price algorithm is usually better than the lower bound obtained by CPLEX.

[0326] The calculation result of 400 trains is used as the initial solution for the second target calculation, and CPLEX is still called for calculation. In order to obtain a better solution, it is decided to set the frequency of RINS to 20. The actual CPLEX calculation time for all data is 48 hours. The results obtained after calculation are as follows Table 2:

[0327] Table 2 Dual-objective calculation results

[0328]

[0329] Figure 7 shows the peak loads of the charging station in different time periods. Figure 7(a) shows that the second goal reduces the peak load from 12:00 to 14:00. Figure 7(b) shows that the number of charging vehicles is completely reduced.

[0330] As can be seen from the table, the second goal optimizes the load peak and also slightly optimizes the cost. This shows that the calculation of the second goal is effective. As can be observed from the table, when the problem scale is relatively small, the second goal can achieve a good optimization effect.

[0331] ■Example 2

[0332] Guangzhou University City is located in Xiaoguwei Street, Panyu District, Guangzhou. The urban area is distributed on both sides of the Pearl River, with a total area of ​​34.4 square kilometers. It is the center of high-level talent training, scientific research and exchange in South China, a national first-class university park with integrated development of learning, research and production, and the "information port" and "intellectual center" of southern China. Guangzhou University City has more than 20 bus routes, and the test was carried out with the island routes Ring Line 1, Ring Line 2, Pan 201, and Pan 202 as the background. The route conditions and bus route distribution are as follows Figure 8 The bus route situation is shown in Table 3.

[0333] Table 3 Bus routes

[0334]

[0335] Among them, the line length is measured by Baidu Maps, and the one-way driving time is calculated based on the average bus operating speed of 20km / h. The first and last bus times and the departure intervals can be obtained by querying the bus network, with a total of 379 trips. The operating hours are 6:00-23:00, of which the peak hours are from 7:00 to 9:00 in the morning and from 16:00 to 18:00, and the low peak hours are after 21:00 in the evening. There are 5 stations in the four lines, namely: the South Terminal of the Municipal State Archives, the Terminal of the University Town Outer Ring West Road, the Terminal of the University Town Sports Center, the Terminal of the University Town Suishi Village, and the Terminal of the University Town Science Center. Vehicles depart and collect from these stations. In addition, a pure electric bus charging station is set up at the South Terminal of the Municipal State Archives for charging pure electric buses. Assume that the vehicle capacity of each terminal and charging station is 30 vehicles, and the maximum vehicle capacity is 100 vehicles. The driving schedule of the line is shown in Table 4 below:

[0336] Table 4 Timetable

[0337]

[0338] After searching Baidu Maps, the distance and time relationship between each site is shown in Table 5 and Table 6:

[0339] Table 5 Distance between stations

[0340]

[0341] Table 6 Time between stations

[0342]

[0343] In terms of charging, the BYD K6 pure electric bus is used as a prototype, and the vehicle has a range of 260 kilometers. In actual operation, in order to extend the battery life and avoid the vehicle's insufficient range, the range is usually artificially reduced to avoid deep discharge. Here, the vehicle's range is set to 130 kilometers. Taking BYD's fast-charging charging equipment EVA080KG as an example, the fast charging time of the vehicle is set to 30 minutes. In terms of cost, it is assumed that the fixed cost of the vehicle is 2,000 yuan / vehicle, the driving distance cost is 3 yuan / km, the waiting cost is 0.1 yuan / min, and the fixed cost of charging is 50 yuan / time. The computer configuration and parameter settings are the same as the basic data of the simulation example.

[0344] Based on the principle of simplicity and ease of implementation, bus companies mostly adopt a single-line dispatching mode. In this mode, the main resources of bus companies are organized as a single bus route. Vehicles are divided according to the route and fixedly configured. Charging stations are assigned according to the route or set up to take into account the convenience of single-line operation. For the single-line dispatching problem of electric buses, the extended FIFO rule can usually be used. The extended FIFO rule still uses the FIFO rule to connect the trains, but when connecting to the next train, it is necessary to determine whether the remaining range is safe. The safe range must allow the vehicle to return to the depot or charging station, that is, for The following equations are also true:

[0345]

[0346] g i +DH(iθ′)≤DMAX (46)

[0347] Formula (45) indicates that the vehicle's cruising range is sufficient to return to the nearest charging station, and formula (46) indicates that the vehicle's cruising range is sufficient to return to the parking lot. Therefore, the extended FIFO rule is used when connecting trains, as shown in Table 7 below:

[0348] Table 7 Extended FIFO rules and train connection rules

[0349]

[0350] For the real example problem, the extended FIFO rule and the branch pricing algorithm were used to solve it. Among them, the upper limit of the solution time of the branch pricing algorithm was set to 72 hours, and the computer configuration was the same as the simulation example. The solution obtained by the extended FIFO rule used a total of 39 vehicles, with a total cost of 97,188.50 yuan. The branch pricing algorithm explored a total of 88 nodes, with a gap value of 6.14%. The solution obtained used a total of 28 vehicles, with a total cost of 76,062.78 yuan. It can be seen that the results obtained by the branch pricing algorithm are better than those obtained by the FIFO rule.

[0351] Table 8 Vehicle dispatch plan

[0352]

[0353] The table lists the vehicle dispatch plan obtained using the branch pricing algorithm. For convenience, the South Terminal of the National Archives of the City, the West Outer Ring Road Terminal of the University Town, the Sports Center Terminal of the University Town, the Central Terminal of Suishi Village of the University Town, and the Central Terminal of Suishi Village of the University Town are simply recorded as A, B, C, D, and E. DH means empty driving, and Charge means charging at the charging station. Fig. 9 The comparison of the cost and number of charging vehicles calculated by the extended FIFO rule and the branch pricing algorithm is listed. Fig. 9 It can be seen that the number of vehicles in the results obtained by the branch pricing algorithm is much better than that of the extended FIFO rule, which is achieved by increasing the number of charging vehicles and appropriately increasing the number of idle vehicles. It is also observed that although the number of vehicles charged by the branch pricing algorithm is more than that of the extended FIFO rule, most of the vehicles are charged during the off-peak period of electricity prices, so the charging cost has only increased slightly.

[0354] Figure 10 shows the distribution of charging time periods and electricity prices. As can be seen from Figure 10(a), the result obtained by the extended FIFO rule shows that the charging time period selected by the vehicle is exactly during the peak electricity price period. Of course, there are differences in the charging time periods between different lines. Ring Line 1 is mainly charged between 13:00 and 14:00 and 20:00 and 21:00, Pan 202 and Ring Line 2 are mainly charged between 14:00 and 15:00, and Pan 201 is mainly charged around 16:00. Figure 10(b) is the result calculated by the branch pricing algorithm. Due to the regional scheduling mode, it is impossible to distinguish which line the charging vehicle belongs to. As can be seen from Figure 10(b), the vehicle is more inclined to choose the off-peak period for charging, and the number of vehicles charging at the same time is also lower than the result obtained by the extended FIFO rule.

[0355] The branch-and-price algorithm can obtain more cost-optimal results in actual examples, which shows that the algorithm has good characteristics and can be extended to practical applications.

[0356] Example 2

[0357] An electric bus regional dispatching system, such as Fig.11 As shown, including:

[0358] The time extension network module is used to establish a time extension network model, and explain the multi-depot scheduling problem of pure electric buses through the time extension network model; and set the nodes and arcs of the time extension network model, as well as the cost of the arcs;

[0359] The dual-objective integer programming module is used to expand the nodes and arcs of the network model according to time and establish a dual-objective integer programming model that considers time-of-use electricity prices and energy consumption control;

[0360] A conversion module, used for processing the dual objectives of the dual objective integer programming model by using a lexicographic optimization method, and converting the dual objective problem into the first objective and the second objective of the single objective model for solving;

[0361] A first objective solving module, used for designing a branch-and-price algorithm, and solving the first objective of the transformed single objective model through the branch-and-price algorithm;

[0362] A second objective solving module, used to use the optimal solution obtained from the first objective as the initial solution of the second objective, and solve the second objective of the converted single objective model through a commercial solver;

[0363] The result output module is used to output the scheduling plan.

[0364] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for regional dispatching of electric buses, characterized in that: The following steps are involved: The time extension network model is established through the time extension network module, and the multi-depot scheduling problem of pure electric buses is explained through the time extension network model; Set the time to expand the nodes and arcs of the network model, and set the cost of the arcs; According to the nodes and arcs of the time-expanded network model and the arc costs, a dual-objective integer programming model considering time-of-use electricity prices and energy consumption control is established; A lexicographic optimization method is used to process the dual objectives of a dual objective integer programming model, wherein the dual objectives include a first objective problem and a second objective problem, and the dual objective problem is converted into the first objective and the second objective of a single objective model for solving; Design a branch-and-price algorithm to solve the first objective of the transformed single-objective model; The optimal solution of the first objective is used as the initial solution of the second objective, and the second objective of the transformed single-objective model is solved by a commercial solver to complete the regional dispatch of electric buses. The nodes and arcs of the time-expanded network model and the arc costs are used to establish a dual-objective integer programming model that considers time-of-use electricity prices and energy consumption control, specifically: The dual-objective integer programming model needs to consider the operating costs of the bus company and the minimum load peak of the power grid; then: Min L(2)st in, is a Boolean variable, indicating whether the arc from node i to node j in the kth layer in the time extension network model is selected; is a continuous variable, which represents the cumulative mileage from the k-th layer vehicle to the node i in the time extension network model. This variable is used to track whether the cumulative mileage of the vehicle exceeds the upper limit of the cruising range; represents the cost from node i to node j in the kth layer; represents the distance from node i to node j at the kth layer; G represents the upper limit of the accumulated mileage of public transportation vehicles; N represents the upper limit of the fleet size; represents the upper limit of parking lot k capacity; Indicates charging plan T j The upper capacity limit of the charging station; L represents the peak load, which is the maximum number of vehicles that can be charged at each charging station in the same period of time; In objective function (2), minimizing the total cost is taken as the optimization objective; in objective function (3), minimizing the load peak is taken as the objective; constraint (4) is the trip coverage constraint, which means that each trip is carried out by one and only one bus; constraint (5) is the flow conservation constraint, which means that all nodes in the same layer except the starting point and the end point must comply with the flow conservation constraint; constraint (6) is the cumulative mileage calculation formula, which means that the cumulative mileage of node j is equal to the cumulative mileage of the previous node i plus the distance between node i and node j; constraint (7) means that at the starting point or the charging station, the cumulative mileage is cleared to 0, which means that the vehicle leaves the station. departure or the vehicle is fully charged at the charging station; constraint (8) indicates that the cumulative mileage cannot exceed the maximum mileage; constraint (9) is the fleet size limit, indicating that the number of vehicles cannot exceed the maximum fleet size; constraint (10) is the parking lot capacity limit, indicating that the number of vehicles sent from the parking lot cannot exceed the maximum capacity of the parking lot; constraint (11) is the charging station capacity limit, indicating that the number of vehicles in the same charging plan cannot exceed the maximum capacity of the charging station; constraint (12) is the load constraint, indicating that the load of each charging station does not exceed the load peak; constraint (13) indicates that the decision variable is a 0 or 1 variable; constraint (14) indicates that the cumulative mileage of the vehicle is greater than or equal to 0; For constraint (6) and constraint (8), since there are nonlinear terms, they are linearized. Here, the big M method is used. For constraint (6), the following linear formula is used for replacement: Among them, M is a large positive constant; when hour, It's obvious Established; when hour, Constraints (15) and (16) are relaxed; Similarly, for constraint (8), the following linear formula is used for substitution: Since then, a bi-objective integer programming model has been constructed.

2. The electric bus regional dispatching method according to claim 1, characterized in that: The time-extended network model is established as follows: The network is layered according to the parking lot to establish a time-expanded network model G = (V k ,A k ), where V is the set of nodes at the kth level, and A is the set of arcs at the kth level; In the time extension network, different layers represent different parking lots, and a train is designated to be completed by a certain parking lot; a node represents the execution of a certain plan, and the nodes include: the starting or ending point of the parking lot, the train node, and the charging plan node; an arc represents a feasible connection between two plans, and the arc includes: a departure arc, a return arc, a train connection arc, and a charging arc.

3. The electric bus regional dispatching method according to claim 2, characterized in that: The cost of setting the arc is specifically: The cost in the multi-depot scheduling problem is reflected in the arc cost, and the arc cost is set as: in, For the starting arc, For train connection arcs, For the charging arc, To close the arc, Indicates the end position el from node i i To the starting position of node j The empty driving distance, c d represents the unit distance cost, c w Indicates the unit waiting time cost, where the waiting time ID ij =st j -et i -DH ij , st j represents the starting time of node j, et i Indicates the end time of node i, DH ij Indicates the end position el from node i i To the starting position sl of node j j The idle time, c v represents the fixed cost per vehicle, Indicates the charging cost; Charging costs The calculation formula is: Where P is the charging power of the electric bus, W(t) is the electricity price at time t, and st t is the starting time of the charging plan t, et t is the end time of charging plan t, c f Fixed cost for charging.

4. The electric bus regional dispatching method according to claim 1, characterized in that: The dual objectives of the dual objective integer programming model are processed by the lexicographic optimization method, and the dual objective problem is converted into the first objective and the second objective of the single objective model for solving, specifically: the first objective is to minimize the operating cost objective, and the second objective is to minimize the load peak objective; the dual objectives of the dual objective integer programming model are converted into the minimization of the operating cost objective of the minimization cost problem model and the minimization of the load peak objective of the minimization load peak problem model; Among them, the cost minimization problem model is as follows: st Eqs: When solving the first objective, the optimal solution is obtained After that, constrain Add to the constraints and solve the problem of minimizing the load peak with the objective function as the goal; the model is as follows: MinL(28)st Eqs: in, is a Boolean variable, indicating whether the arc from node i to node j in the kth layer in the time extension network model is selected; is a continuous variable, which represents the cumulative mileage from the k-th layer vehicle to the node i in the time extension network model. This variable is used to track whether the cumulative mileage of the vehicle exceeds the upper limit of the cruising range; represents the cost from node i to node j in the kth layer; represents the distance from node i to node j at the kth layer; G represents the upper limit of the accumulated mileage of public transportation vehicles; N represents the upper limit of the fleet size; represents the upper limit of parking lot k capacity; Indicates charging plan T j The upper capacity limit of the charging station; L represents the peak load, which is the maximum number of vehicles that can be charged at each charging station in the same period of time.

5. The electric bus regional dispatching method according to claim 1, characterized in that: The branch pricing algorithm is designed to solve the first objective of the transformed single-objective model through the branch pricing algorithm; specifically, an efficient branch pricing algorithm is designed, a special heuristic rule and a train chain pool acceleration strategy are proposed, and the above algorithm is used to solve: the first objective is to minimize the operating cost objective; The problem of minimizing operating costs is solved using a column generation algorithm, which transforms the original problem into a column generation main problem and sub-problems. The main problem is the set partitioning problem of the train chain, and the sub-problem is the shortest path problem with resource constraints: For the main question: Restricting the main problem, that is, solving the set partitioning problem of the train chain, the solution model RMP_COST is as follows: min∑ r∈R c r z r (41) ∑ r∈R With r ≤N (45) In the RMP_COST model, the decision variable is z r , indicating whether the train chain r is selected; R is the set of train chains that restrict the main problem, c r is the cost of the train chain r∈R, It is a Boolean parameter, indicating whether the train chain starts from the k yard; is a Boolean parameter, indicating whether the train chain r covers the node s∈S, where S is the train set; is a Boolean parameter, indicating whether the train chain r covers the node t∈T, where T is the charging plan set; N represents the upper limit of the fleet size, represents the upper limit of parking lot k capacity, Indicates charging plan T t The capacity limit of the charging station you are at; In order to generate the objective function (41) that constrains the various constraints of the main problem, it represents the minimization of the operating cost, whose value is equal to the sum of the costs of the selected train chains; Constraint (42) represents the parking lot capacity constraint; constraint (43) represents the unique coverage constraint; constraint (44) represents the charging station capacity constraint; constraint (45) represents the maximum size limit of vehicles; constraint (46) represents the value range of the decision variable, z r is a Boolean variable; The dual variable is usually restricted to a linear relaxation of the master problem, that is, z r Change from a Boolean variable to a continuous variable with a value range of [0,1]; For the sub-problems: For the minimization problem, when the test number of all feasible columns is greater than or equal to 0, the introduction of the column does not improve the problem, and the minimization problem obtains the optimal value; for the main problem of train chain set partitioning, let the dual variables of constraints (42)-(45) be: α k , β s , γ t ,π(k∈K,s∈S,t∈T); The subproblem is to find the train chain with the smallest number of tests, that is, to find the shortest path in the network. Since the train chain mileage constraint needs to be considered, it is a shortest path problem with resource constraints. The subproblem model SUB_COST is as follows: in for: In the SUB_COST model, the decision variables are and is a Boolean variable, representing whether the arc (i, j) at the kth layer is selected; is the cumulative mileage of node i at the kth layer; G is the maximum mileage of the bus; Objective function (47) represents minimizing the number of train chain checks, and its value is equal to the sum of the paths of all selected edges; constraint (48) indicates that the train chain starts from the starting point and has only one train chain; constraint (49) is a flow conservation constraint, which means that all nodes in the same layer except the starting point and the end point must comply with the flow conservation constraint; constraint (50) is a cumulative mileage calculation formula, which means that the cumulative mileage of node j is equal to the cumulative mileage of the preceding node i plus the distance between node i and node j; Constraint (51) indicates that at the starting point or when the vehicle leaves the charging station, the accumulated mileage is cleared to 0, which means that the vehicle departs from the station or is fully charged at the charging station; Constraint (52) indicates that the accumulated mileage cannot exceed the maximum mileage; Constraint (53) indicates that the decision variable is a 0 or 1 variable; constraint (54) indicates that the vehicle's cumulative mileage is greater than or equal to 0; The time expansion network has multiple layers. For each layer k∈K, the time expansion network can be independently calculated from the starting point θ k To the end point θ k ′’s shortest path problem, this process uses parallel computing; After solving the shortest path problem of each layer, the path with the lowest cost in all layers is taken as the optimal solution, that is, the optimal solution to the subproblem; the shortest path problem of each layer is the shortest path problem with resource constraints, and the label correction method is used to solve the shortest path problem with resource constraints.

6. The electric bus regional dispatching method according to claim 5, characterized in that: The label correction method is as follows: 1) Tag definition make Indicates that the mth item starts from the parking lot θ k The path to node i; for the path Label With path Associated, with two resources in Indicates the starting point θ from the parking lot k The sum of the costs of the mth path to node i, Indicates the starting point θ from the parking lot k The cumulative mileage of the mth path to node i; 2) Tag extension Label The maintenance of adopts forward dynamic programming; in the label correction method, for the edge (i, j) and the path Extend the inode label The label of the jth node The label extension calculation rules are as follows: The label extension conditions are as follows: When the cruising range condition is met, the label is extended, and the cumulative cruising range also needs to distinguish whether the j node is a charging station. If so, the cumulative cruising range needs to be cleared; 3) Label dominance The computation speed of the labeling algorithm depends on the number of labels. When a label is not part of the optimal solution, the label will be removed. The label dominance rule will be used to remove labels that are not the optimal solution to speed up the computation. Domination Tags If and only if the following conditions are met at the same time: If the tag Domination Tags It represents the label Related Paths is not part of the optimal solution, because any A feasible extension also applies to and The cost is lower; therefore, and will be removed; First, the label correction method needs to be initialized. k The label is set to (0,0), indicating that the cost and cumulative mileage are both 0 at the starting point. The label correction method needs to record the node number of the updated label and check all arcs emitted from a node in the next iteration. The FIFO node processing rule is adopted here, so the nodes are stored in a queue. When processing a node, all labels of the node are first compared pairwise, and the label dominance rule is used to delete redundant and inefficient labels. If the label extension condition (43) is met, the label of this node is extended. When the queue is empty, the algorithm backtracks to find the optimal path and the optimal solution.

7. The electric bus regional dispatching method according to claim 5, characterized in that: The branch pricing algorithm adopts a branch strategy, specifically: Branch the connection relationship between nodes: When the solution of the linear relaxation restricted main problem is not an integer, the solution of the set partitioning model is transformed into the solution of the time extension network model, and the unbranched arcs (i, j) are branched; two sub-problems are obtained, each containing branch information x ij =0 and x ij =1; for branch information x ij = 0, delete arc (i, j) on the time extension network; and for branch information x ij =1, the arcs on the time extension network can be deleted according to the unique constraint of train number; The branch-and-bound tree has a global upper bound to record the value of the best integer solution found; it uses a depth-first search rule and always prioritizes solving branch x. ij =1; It is necessary to use a heuristic algorithm to construct an initial solution and convert the decimal solution into an integer solution; The heuristic algorithm continuously sets the cost of arcs on the time expansion network and solves the shortest path problem on the time expansion network to obtain a train chain. The solution method is the same as the sub-problem. After the train chain is obtained, it is added to the train chain set, and the covered train nodes, parking lot nodes and charging nodes on the time expansion network are deleted. The above process is repeated until all trains are covered, specifically: Heuristically construct an initial solution: The purpose of heuristically constructing the initial solution is to make each train chain cover more trains while reducing costs, while ensuring that the solution meets the capacity constraints and vehicle quantity constraints; the steps of heuristically constructing the initial solution are as follows: Step 1. Time-expanded network construction: Set the cost of arc ij at layer k to v is a constant, which is a very small number. is the cost of each arc ij at the kth layer; if node j is a train node, then μ = -1, otherwise μ = 0. By adding this item, the vehicle can cover more trains; Step 2. Solve the shortest path problem on the time-expanded network, obtain a train chain, and delete the covered train nodes, parking lot nodes with capacity reaching the upper limit, and charging nodes; the obtained train chain meets the range constraint and will not conflict with the covered trains, parking lots with full capacity, and charging stations; Step 3. Repeat Step 2 until all train nodes are covered; The initial solution also adds an artificial train chain that meets all the constraints of the main problem but is very expensive to ensure the feasibility of the problem; Heuristic fix for decimal solutions: When solving the subproblems of each branch and bound tree, the solution obtained is often not an integer solution. In this case, a heuristic method is used to repair the decimal solution. The steps are as follows: Step 1. Select the train chain with the largest decimal solution that has not been considered; Step 2. Time-expanded network construction: The cost of arc ij is set to If node j is a node covered by the train chain, then λ=-10, otherwise λ=0; Step 3. Solve the shortest path problem on the time-expanded network, obtain a train chain, and delete the covered train nodes, parking lot nodes with capacity limits, and charging nodes; the obtained train chain meets the range constraint and will not conflict with the covered trains, parking lots with full capacity, and charging stations; Step 4. Repeat Step 1, Step 2, and Step 3 until all train nodes are covered.

8. The electric bus regional dispatching method according to claim 5, characterized in that: The train chain pool acceleration strategy is specifically as follows: The train number chain of the child node of the branch and bound tree uses the existing information of the parent node to speed up the solution of the child node; the train number chain pool R is designed pool , used to collect all the train chains generated by each node of branch and bound; let the branch and bound tree node currently being solved be p, then the initial train chain set R of the main problem of node p p Available from the train chain pool R pool However, the train number chains that do not meet the branch constraints need to be deleted: R p =R pool -r (61) Where r∈R pool , and r does not satisfy the branch constraint of p; In particular, for the root node of the branch and bound tree, R root It consists of two parts: the train chain generated by the heuristic construction of the initial solution and the expensive manual train chain; When the node p is solved, the train chain generated by the column generation process is merged with the train chain pool to obtain a new train chain pool: R pool =R pool ∪R p (62)。 9. An electric bus regional dispatching system, used to implement the electric bus regional dispatching method according to any one of claims 1 to 8, characterized in that: include: The time extension network module is used to establish a time extension network model and explain the multi-depot scheduling problem of pure electric buses through the time extension network model; And set the time to expand the nodes and arcs of the network model, and set the cost of the arc; The dual-objective integer programming module is used to expand the nodes and arcs of the network model according to time and establish a dual-objective integer programming model that considers time-of-use electricity prices and energy consumption control; A conversion module, used for processing the dual objectives of the dual objective integer programming model by using a lexicographic optimization method, and converting the dual objective problem into the first objective and the second objective of the single objective model for solving; A first objective solving module, used for designing a branch-and-price algorithm, and solving the first objective of the transformed single objective model through the branch-and-price algorithm; A second objective solving module, used to use the optimal solution obtained from the first objective as the initial solution of the second objective, and solve the second objective of the converted single objective model through a commercial solver; The result output module is used to output the scheduling plan.

Images