Mine transportation scheduling method and system based on hybrid genetic algorithm

Abstract

The invention discloses a mine transportation scheduling method and system based on a hybrid genetic algorithm, and belongs to the technical field of intelligent optimization scheduling. The method comprises the following steps: S100, inputting mine tunnel network, facility and equipment parameters; s200, a Dijkstra algorithm and breadth-first search are combined, and an optimal path set containing a shortest path and a secondary short path is generated for each pair of ore unloading fields and mining sites so as to disperse traffic flows; s300, establishing a multi-objective optimization model for minimizing waste time and fuel consumption and maximizing transportation volume; s400, a non-dominated sorting genetic algorithm is adopted for solving, the core of the non-dominated sorting genetic algorithm is that a special triple coding mode is adopted, an execution vehicle is allocated for each transportation task, a going route is selected for each transportation task, a return route is selected for each transportation task, and complex constraints are fused into a chromosome structure; s500, performing dynamic simulation on each scheduling scheme, updating the vehicle state and speed based on the real-time road congestion degree, and accurately evaluating the performance of the scheme; s600, selecting a final scheme from the Pareto optimal solution set by using a multi-attribute decision-making method; and S700, when a fault occurs, automatically re-evaluating and allocating tasks. According to the method, the problems of single path, inaccurate model and uncomfortable coding of a traditional method are effectively solved, and high efficiency, energy conservation and high robustness of vehicle scheduling in a complex mine tunnel network are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent optimization scheduling and path planning, in particular to a mine transportation scheduling method and system based on improved hybrid genetic algorithm for multi-objective optimization in complex network environment. The method is suitable for, but not limited to, vehicle scheduling and path planning in closed or semi-closed scenes such as mines, ports and large logistics parks. BACKGROUND

[0002] Mine transportation scheduling is the core link of mine production and operation, and its efficiency directly affects the production cost and economic benefit. This problem is essentially a complex, multi-constrained and multi-objective vehicle path planning problem. The traditional scheduling method and the early intelligent method have the following inherent defects: Single and rigid path planning: Most existing systems default to using a unique shortest path between fixed origin and destination points for vehicles. This can easily cause path congestion when multiple vehicles are scheduled simultaneously, forming a system bottleneck and reducing overall efficiency.

[0003] The model is too simplified and deviates from reality: Many models consider vehicle speed as constant, ignoring the significant impact of road traffic dynamics on travel time. At the same time, insufficient consideration of key time consumption factors such as shovel loading and vehicle queuing leads to poor results in practical applications of theoretical optimization.

[0004] Algorithm encoding method does not match problem structure: Traditional genetic algorithms often use order-based encoding (e.g. "Vehicle 1-Task A-Vehicle 2-Task B") when solving such problems. This encoding method is difficult to effectively express and handle complex constraints such as "shovel exclusivity" (a shovel can only serve one truck at a time) and "independent selection of vehicle return paths", resulting in low search efficiency and the generation of invalid solutions.

[0005] Weak multi-objective processing capability: Mine scheduling naturally needs to balance "efficiency" (transportation volume), "economy" (fuel cost) and "time" (waiting and congestion) and other multiple objectives. Traditional methods are difficult to balance these conflicting objectives, lack the ability to effectively obtain and evaluate a set of balanced optimal solutions (Pareto optimal solution set), and the decision-making process for selecting the final solution from the solution set often relies on experience and lacks scientificity.

[0006] Therefore, there is an urgent need in the art for an intelligent scheduling method that can systematically solve the above problems and has high adaptability, high accuracy and strong robustness. SUMMARY

[0007] In view of the deficiencies of the prior art, the purpose of the present application is to provide a mine transportation scheduling method and system based on a hybrid genetic algorithm. The method aims to effectively solve the problems of single path, inaccurate model, unsuitable coding, and difficult multi-objective trade-off through innovative path planning, accurate dynamic simulation, genetic coding highly consistent with the problem structure, and scientific decision-making mechanism, thereby realizing efficient, energy-saving, and stable operation of mine transportation scheduling.

[0008] To achieve the above purpose, the present application adopts the following technical solutions: In a first aspect, the present application provides a mine transportation scheduling method based on a hybrid genetic algorithm, executed by a computing device, comprising the following steps: S100: Input and initialization. Input the topological structure of the mine tunnel network (including all nodes and road segments), the locations of all unloading mines and mining points, the number of electric shovels at each mining point, the total number N of truck fleets, the performance parameters of each truck (such as empty / full load speed, empty / full load fuel consumption, idle fuel consumption, etc.), and the predetermined total working time T. At the same time, initialize the parameters of the genetic algorithm, such as population size, crossover probability, mutation probability, etc.

[0009] S200: Global optimal path set generation. For each origin-destination pair consisting of an unloading mine q and a mining point p, perform the following operations to generate diversified path options: S201: Calculate the shortest path between the origin-destination pair using Dijkstra's algorithm.

[0010] S202: Use the breadth-first search algorithm to traverse the mine tunnel network graph with a preset path quantity limit `limit` as a constraint to calculate a batch of suboptimal paths between the origin-destination pair. This algorithm systematically explores multiple feasible paths from the starting point to the ending point by maintaining a queue, ensuring the diversity of the paths.

[0011] S203: Merge the shortest path and the suboptimal paths to form the optimal path set for the origin-destination pair, and delete the duplicate paths. This path set provides a flexible and congestion-avoiding path selection basis for subsequent scheduling.

[0012] S300: Multi-objective optimization model construction. Establish a mathematical model containing the following three objective functions to comprehensively evaluate the pros and cons of the scheduling scheme: S301: Objective function f1: Minimize total wasted time. This time includes the additional travel time due to reduced speed caused by road congestion, the time spent waiting in line at the entrance of a road segment due to traffic saturation, and the time spent waiting in line at a mining point due to electric shovel occupation.

[0013] S302: Objective function f2: minimize total fuel consumption cost. The cost includes empty running fuel consumption, full load running fuel consumption, idling fuel consumption when truck is waiting at the spot, and extra fuel consumption due to frequent start-stop.

[0014] S303: Objective function f3: minimize production difference. That is, minimize the difference between theoretical maximum production F and actual transport amount, which is equivalent to maximizing actual transport amount. Wherein, theoretical maximum production F = T × Σ (number of electric shovels at each mining point) × (truck rated load u / time required for electric shovel to fill a truck t_m).

[0015] S400: Solution based on special triple encoding genetic algorithm. A non-dominated sorting genetic algorithm is used to solve the multi-objective optimization model. One of the core innovations of the present application is its unique triple real number encoding method for chromosomes: S401: First part of genes (task allocation): This part is composed of multiple gene points, each of which corresponds to a theoretical transport task. The total number of tasks is calculated based on the number of electric shovels, loading speed, and total working time T. The value of each gene point is decoded as the truck number that executes the task.

[0016] S402: Second part of genes (outbound route selection): This part of gene points is one-to-one corresponding to the first part. The value of each point is decoded as the truck that executes the task, and the selected outbound route number from the current unloading site to the target mining point. This number points to a specific path in the set of optimal paths corresponding to the origin-destination pair generated in step S2.

[0017] S403: Third part of genes (return route selection): This part of gene points is also one-to-one corresponding to the first part. The value of each point is decoded as the truck that executes the task, and the selected return route number from the mining point to any unloading site. This number is also selected from the corresponding set of optimal paths. This encoding method naturally embeds complex constraints such as vehicle task sequence, electric shovel exclusivity, and independent selection of round-trip paths, greatly improving the feasibility and search efficiency of the algorithm.

[0018] S500: Dynamic simulation and fitness evaluation. For each chromosome (i.e., a complete scheduling scheme) in the population of the genetic algorithm in S4, a dynamic simulation based on time stepping is performed to accurately calculate its objective function value (fitness): S501: Initialization: Set the simulation clock t = 0, and initialize all truck states (position in respective unloading site, state as "waiting to enter the line", load empty, etc.).

[0019] S502: Step advance: In each simulation step, calculate the expected time of the next state change for all trucks, and take the minimum value as the step size Δt.

[0020] S503: Dynamic update: Real-time calculation of traffic volume of each section, and calculation of maximum traffic capacity C of the section according to the length of the section and the safety distance of the vehicle, and then obtain the real-time congestion degree (crowd = current traffic volume / C).

[0021] S504: According to the congestion degree, the actual driving time of the vehicle on the section is dynamically calculated by using the BPR function `t_r = t_f * [1 + α * (crowd)^β]`, where t_f is the free flow time, and α and β are calibration parameters.

[0022] S505: Track and update the state attributes (position, load, current task progress, state, etc.) of each truck. The state includes: waiting to enter the line, driving, waiting for a shovel, loading, unloading, and completed.

[0023] S506: Target accumulation: During the simulation process, the values of three target functions f1, f2, and f3 are accumulated in real time according to the state of the truck.

[0024] S507: Termination and output: When the simulation clock t ≥ T, terminate the simulation, and output the accumulated target function value as the fitness of the chromosome (scheduling scheme).

[0025] S600: Optimal scheduling scheme decision. After the genetic algorithm is run, a Pareto optimal solution set is output. A multi-attribute decision-making method is used to comprehensively evaluate the solution set to select the final implementation scheme. The method can be: S601: Objective weighting method: such as entropy weight method, automatically calculate the weight according to the dispersion degree of each target function value in the solution set, and then perform weighted scoring.

[0026] S602: Combination of subjective and objective methods: first get the objective weight by entropy weight method, then combine the subjective weight (such as more emphasis on transportation volume) set by the decision maker in advance, use game theory model to get the comprehensive weight, finally use TOPSIS method to calculate the closeness of each solution to the ideal solution, and sort to select the best solution.

[0027] S700: Real-time fault response steps: S701: In the simulation or actual operation, if a truck is detected to have a fault, remove all uncompleted tasks from its task list.

[0028] S702: Determine whether the task load of the remaining trucks is saturated. If not, dynamically reassign the removed tasks to the idle truck with the least current task load in chronological order.

[0029] S703: Based on the new task allocation, quickly re-execute steps S4 to S6 to generate a new, feasible real-time scheduling scheme in a very short time, thereby ensuring the robustness of the system.

[0030] In a second aspect, the present application provides a mine transportation scheduling system, comprising at least one processor and a memory, the memory storing computer program instructions, when the program instructions are executed by the processor, the system is configured to execute the method of any one of the first aspect.

[0031] In a third aspect, the present application provides a computer readable storage medium, which stores a computer program, when the computer program is executed by a processor, the method of any one of the first aspect is realized.

[0032] Advantages Compared with the prior art, the present application has the following obvious advantages: 1. Avoid congestion from the source: By generating a "set of optimal paths" instead of a single shortest path, the system provides a rich choice of paths for vehicle scheduling, which can intelligently disperse traffic and effectively prevent systemic congestion.

[0033] 2. High model accuracy, close to reality: The introduction of a dynamic congestion model based on the BPR function and a refined time stepping simulation fully considers complex factors such as traffic interaction and queuing, making the optimization results have very high practical value.

[0034] 3. Algorithm perfectly fits the problem: The special triple coding design is a major innovation of the present application, which cleverly integrates complex physical and operational constraints into the chromosome structure, enabling the genetic algorithm to efficiently search the feasible region and significantly improving the solution speed and quality.

[0035] 4. Scientific global optimization and decision-making: The NSGA-III algorithm is used to process high-dimensional target space, which can obtain a well-distributed Pareto optimal solution set. Combined with advanced multi-attribute decision-making methods, the most suitable solution for current production needs can be selected from multiple equilibrium solutions, achieving truly intelligent decision-making.

[0036] 5. Excellent system robustness: The built-in fault response mechanism can quickly respond to unexpected situations (such as vehicle failure) and re-plan, ensuring the continuous and stable operation of the scheduling system and reducing the risk of production interruption. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 is a schematic diagram of the special triple chromosome coding structure in the present application.

[0038] Figure 2 is a schematic diagram of the mine route network distribution used in Example 1 of the present application.

[0039] Figure 3 Figure 1 is a Pareto optimal solution set distribution diagram obtained by a genetic algorithm in Embodiment 1 of the present application. DETAILED DESCRIPTION

[0040] The present application will be described in detail below with reference to the drawings and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present application, and detailed implementation modes and specific operation processes are given, but the protection scope of the present application is not limited to the following embodiments.

[0041] Embodiment 1: In this embodiment, a mine transportation scenario is taken as the background, and the present application is implemented by using software programming on a computer equipped with an Intel i7 processor and 16 GB of memory.

[0042] S100: Input and initialization: mine network: as shown in Figure 1, the network contains a plurality of nodes and road segments. Among them, there are 3 unloading sites (nodes 1, 2, and 3), and 5 mining points (nodes 31, 34, 36, 38, and 41), and the number of electric shovels B_p of each mining point is [2, 3, 1, 4, 3] respectively. Figure 2

[0043] Vehicle fleet parameters: total number of trucks N = 25, all of which are 20-ton mine trucks. Performance parameters are set as follows: empty / full load average speed xv = 30 km / h, empty load fuel consumption x_l = 10 L / h, full load fuel consumption x_l' = 15 L / h, idle fuel consumption x_lt = 1 L / h, and restart fuel consumption x_ld = 0.05 L / occasion. Rated load u = 20 tons.

[0044] Time parameters: total working time T = 1 hour (one shift). Electric shovel loading time t_m = 0.0833 hours, and unloading time t_out = 0.02 hours.

[0045] Algorithm parameters: NSGA-III algorithm population size is set to 200, iteration number is 500, crossover probability is 0.4, and mutation probability is 0.1. Path number limit `limit = 5`. BPR function parameters take the classic values α = 0.15, β = 4.

[0046] S200: Global optimal path set generation: for each unloading site-mining point pair (such as unloading site 1 -> mining point 31), run algorithm 1 (breadth-first search + Dijkstra). Taking unloading site 1 -> mining point 31 as an example, Dijkstra algorithm obtains a shortest path (such as 1-4-10-15-31), and BFS algorithm finds 4 sub-shortest paths (such as 1-5-10-15-31, 1-4-10-14-30-31, etc.). The 5 paths are combined to form the optimal path set of the OD pair. Repeat this operation for all OD pairs.​

[0047] S300: Establish three objective functions: f1 = Σ(All trucks' deceleration waste time + queuing waiting time) f2 = Σ(All trucks' driving fuel consumption + idling fuel consumption + restarting fuel consumption) f3 = F - Σ(All trucks' total tonnage carried), where theoretical maximum production F = 1 × (2+3+1+4+3) × (20 / 0.0833) ≈ 3120 tons.

[0048] S400 and S500: Genetic algorithm solution and dynamic calculation: Encoding: Theoretical task number = (2+3+1+4+3) × (1 / 0.0833) ≈ 156. Therefore, the chromosome length is 156 (task allocation) + 156 (outbound route) + 156 (return route) = 468. Each gene point is a random decimal between (0,1).

[0049] Decoding and simulation: For each individual in the population, first decode the task list of each truck and the outbound and return routes for each task. Then start the dynamic simulation. The simulation starts at t=0 and progresses step by step. For example, a truck starts from unloading site 1 and chooses route 1-4-10-15-31 to go to mining point 31. The simulation system will calculate the traffic and congestion on route 4-10 in real time and dynamically adjust its travel time. After arriving at node 31, if the shovel is idle, it enters the loading state, which takes time t_m; if it is busy, it waits in line. After loading is completed, it returns to the unloading site according to its return gene. The entire process continues to accumulate f1, f2, and f3 until t=T.

[0050] S600: Schedule decision-making: After 500 generations of genetic algorithm, the Pareto frontier is obtained (e.g. Figure 3 ). Entropy weight method combined with TOPSIS is used for decision-making. After calculation, an optimal scheme is obtained, with target values: f1=0.854 hours, f2=146.6 liters, f3=1680 tons (i.e. actual transportation F-f3=1440 tons). The specific driving routes of all trucks under this scheme are output, for example, the route of truck 6 is: `1-5-10-9-26-

[38] -26-9-3-9-8-6-(2)-...`.

[0051] S700: Fault handling (only simulation test in this embodiment): To test the robustness of the system, assume that at t=0.5 hours, trucks 5, 14, and 19 fail simultaneously. The system immediately triggers S7 steps: Remove the three trucks unfinished task; detect the remaining truck task amount is not saturated, dynamically allocate these tasks to the current most idle No. 1 and No. 7 trucks; quickly re-optimize to generate a new scheme. The solution time is only 0.202 seconds. Although the total transportation volume of the new scheme decreases slightly, compared with not taking any measures, the yield loss is significantly recovered, and the fuel consumption and time cost are not significantly increased, which proves the good robustness of the present application.

[0052] Conclusion: This example fully verifies the effectiveness, accuracy and practicability of the method of the present application, and can provide a scientific and reliable optimal solution for vehicle scheduling in a complex mine environment.

[0053] The above describes only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or replacements within the technical range disclosed by the present application can be easily thought of by those skilled in the art, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A mine transportation scheduling method based on a hybrid genetic algorithm, characterized in that, The method is executed by a computing device and includes the following steps: S100: Input and initialization steps, input the topology of the mine network, the location information of all unloading yards and mining points, the number of electric shovels at each mining point, the total number of trucks in the truck fleet and the performance parameters of each truck, and the planned total working time T; S200: Global optimal path set generation step, for each origin-end point pair consisting of an unloading farm and a mining point, perform the following operations: Use Dijkstra's algorithm to calculate the shortest path between the origin and destination points; Using a breadth-first search algorithm, with a preset path count limit as a constraint, the topology of the mine network is traversed to calculate a batch of optimal paths between the origin and destination pairs. The shortest path is merged with the second-best path to form a set of better paths for the origin-end point pair, and duplicate paths are removed. S300: Steps for constructing a multi-objective optimization model, establishing a mathematical model containing the following three objective functions: The objective function f1 is used to minimize the total wasted time of the truck fleet during the total working time T, which includes the extra travel time due to road congestion, the time spent waiting in line to enter the road segment, and the time spent waiting for the electric shovel to be idle at the mining site. The objective function f2 is used to minimize the total fuel consumption cost of the truck fleet during the total working time T. The total fuel consumption cost includes fuel consumption when driving unloaded, fuel consumption when driving fully loaded, fuel consumption when the truck is idling, and additional fuel consumption caused by frequent start-stop of the truck. The objective function f3 is used to minimize the difference between the theoretical maximum output and the actual transport volume, where the theoretical maximum output is the total amount of ore that can be produced by all electric shovels working continuously for the total working time T. S400: The solution steps of the genetic algorithm based on special encoding are as follows: the multi-objective optimization model is solved using a non-dominated sorting genetic algorithm, wherein the chromosome encoding uses real number encoding, and each chromosome consists of three sequentially arranged genetic parts: The first part of the gene consists of multiple gene points, each of which corresponds to a theoretical transportation task. Its gene value is decoded into the truck number that performs the task. The second part of the gene consists of the same number of gene points as the first part of the gene. Each gene point corresponds to a task, and its gene value is decoded as the outbound route number selected by the truck performing the task from the unloading yard to the designated mining point. This number points to a path in the set of better paths of the corresponding origin-end point pair generated in step S200. The third part of the gene consists of the same number of gene points as the first part of the gene. Each gene point corresponds to a task, and its gene value is decoded as the return route number selected by the truck performing the task from the mining point back to the unloading site. This number points to a path in the set of better paths of the corresponding origin-end point pair generated in step S200. S500: Dynamic simulation and fitness evaluation step. For the scheduling scheme represented by each chromosome in the genetic algorithm population from step S400, dynamic simulation based on discrete event simulation or time-step simulation is performed. During the simulation: Real-time calculation of traffic flow on each road segment, and dynamic updates of congestion level on each road segment based on the maximum capacity of the road; Based on the congestion level, the truck's speed and travel time on the road segment are dynamically adjusted using the road traffic impedance function; Track and update the status attributes of each truck, including location, load, current task progress and status, which includes at least driving, waiting to enter the road section, waiting for electric shovel, loading, and unloading. When the simulation clock reaches the total working time T, the simulation is terminated, and the values ​​of the three objective functions f1, f2, and f3 corresponding to the scheduling scheme are accumulated and calculated as the fitness value of the chromosome. S600: Optimal scheduling scheme decision steps. After the genetic algorithm converges, the Pareto optimal solution set output by it is obtained. A multi-attribute decision method is used to comprehensively evaluate and rank each solution in the Pareto optimal solution set, and the solution with the highest comprehensive score is selected as the final mine transportation scheduling scheme output. S700: Fault response procedures. During simulation or actual scheduling, when a truck is detected to have a fault, the current task allocation scheme is dynamically adjusted.

2. The method according to claim 1, characterized in that, In step S200, the specific execution process of the breadth-first search algorithm includes: Initialize a queue, and add the starting point and the list of empty paths to the queue; Loop through the queue and retrieve the nodes and their corresponding path lists; If the current node is the endpoint and the length of the path list meets the requirements, then save the current path to the result set; Iterate through all neighboring nodes of the current node. If a neighboring node has not been visited or meets a specific condition, update the queue. After the loop ends, the paths in the result set are merged with the shortest paths obtained by Dijkstra's algorithm to form the final set of better paths.

3. The method according to claim 1, characterized in that, In step S400, the total number of theoretical transportation tasks is determined by the following formula: Total number of tasks = Σ(number of electric shovels at each mining site) × (total working time T / standard time t_m required to fill a truck with an electric shovel).

4. The method according to claim 1, characterized in that, In step S500, the formula for calculating the road congestion is: crowd = current traffic flow on the road segment / maximum design traffic flow of the road segment; the road traffic impedance function adopts the BPR function form of the U.S. Highway Bureau, specifically: t_r = t_f * [1 + α * (crowd)^β], where t_r is the actual travel time, t_f is the travel time in free flow state, and α and β are model parameters calibrated according to the actual situation of the mine road.

5. The method according to claim 1, characterized in that, In step S600, the multi-attribute decision-making method is an objective weighting method based on information entropy, and its specific process includes: The decision matrix formed by the Pareto solution set is standardized. Calculate the information entropy of each objective function; Calculate the objective weights of each objective function based on information entropy; The standardized decision matrix is ​​weighted using the objective weights, and the comprehensive score for each solution is calculated.

6. The method according to claim 5, characterized in that, In step S600, the multi-attribute decision-making method is a comprehensive decision-making method that combines subjective weights and objective weights, and its specific process includes: Pre-set a subjective weight vector that reflects the decision-maker's preferences; The objective weight vector is calculated using the information entropy method; By using a game theory model or a weighted average method, the subjective weight vector and the objective weight vector are combined to obtain a comprehensive weight vector; The TOPSIS method is used to calculate the proximity of each Pareto solution to the ideal solution based on the comprehensive weight vector, and the solutions are sorted according to the proximity.

7. The method according to claim 1, characterized in that, The method also includes a fault handling step S700: during simulation operation or actual scheduling, when a truck is detected to have a fault, all unfinished tasks of the faulty truck are immediately removed from the current scheduling plan; Determine if the total task load of the remaining trucks is saturated; If the task is not saturated, the removed tasks will be dynamically reassigned to the truck with the shortest task queue or the one that is idle, according to the task's time sequence. Based on the updated task allocation, steps S4 to S6 are quickly re-executed to generate a new, real-time scheduling scheme to deal with failures.

8. A mine transportation scheduling system, characterized in that, The system includes at least one processor and a memory, the memory storing computer program instructions that, when executed by the processor, configure the system to perform the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images