Two-stage optimization scheduling method for strip mine transport vehicles

Abstract

The invention belongs to the technical field of mine dispatching, and particularly relates to a two-stage optimization dispatching method for strip mine transport vehicles. The method comprises the following steps: firstly, grouping excavators in a mining area, taking areas in different ranges from an excavator group in the driving-in direction of a transport vehicle as a first key area and a second key area, and determining the first key area and the second key area of each excavator group; then, when the transport vehicle arrives at the first key area, allocating an excavator to the transport vehicle by using the scheduling model; when the second key area is reached, the scheduling model is utilized again to distribute the excavator for the second key area; when the transport vehicle arrives at the excavators, the distance between the transport vehicle and each excavator in the group is calculated according to the GPS data, the excavator corresponding to the minimum distance is the actually arrived excavator of the transport vehicle, and whether the actually arrived excavator is consistent with the target excavator allocated in the second stage or not is judged; if yes, the transport vehicles obey distribution; and if not, indicating that the transport vehicles do not obey distribution, dynamically adjusting a vehicle distribution sequence of each excavator in the group, and preparing for excavator distribution of subsequent transport vehicles. According to the method, the condition that the driver does not completely comply with the scheduling instruction is considered, the transport vehicles are redistributed in real time, and the scheduling capability is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent mine scheduling, and specifically is a two-stage optimization scheduling method for open-pit mine transport vehicles. Background Art

[0002] At present, the dispatch of transport vehicles in mines usually requires manual management of on-site vehicles. When unreasonable scheduling is found, the vehicles are directed to change the transportation route. Manual management is not only inefficient, but also related to the work level and experience of managers. It is difficult to make the optimal dispatch, which is not conducive to the development of mine production towards digitalization and intelligence.

[0003] With the development of scheduling technology, the transportation efficiency of mining vehicles has been improved. However, due to the complex and changeable actual environment of mines, it is still very difficult to achieve efficient vehicle scheduling. The current scheduling algorithm mainly considers the optimization of the vehicle transportation process. Although it can improve transportation efficiency, when the vehicle does not match the production capacity of the assigned loading equipment, it will cause the vehicle to queue for loading and unloading, which will extend the waiting time of the vehicle and affect the overall efficiency. In addition, there may be special circumstances where the vehicle does not obey the allocation, which will affect the scheduling of subsequent vehicles. Therefore, this application proposes a two-stage scheduling optimization method for open-pit mine transportation vehicles. Summary of the invention

[0004] In view of the shortcomings of the prior art, the technical problem to be solved by the present invention is to propose a two-stage optimization scheduling method for open-pit mine transport vehicles.

[0005] The present invention solves the technical problem by adopting the following technical solution:

[0006] A two-stage optimization scheduling method for open-pit mine transport vehicles, characterized in that the method comprises the following steps:

[0007] Step 1, grouping the excavators in the mining area, grouping the excavators with similar distances into one group, and obtaining multiple excavator groups; taking the area within the range of d1 and d2 (d1>d2) from the excavator group in the direction of entry of the transport vehicle as the first key area and the second key area, and determining the first key area and the second key area of ​​each excavator group;

[0008] Step 2: When the transport vehicle arrives at the first key area, the scheduling model is used to assign an excavator to it; when it arrives at the second key area, the scheduling model is used again to assign an excavator to it;

[0009] 2.1) Assign transport vehicles to each excavator group. When the transport vehicle arrives at the first key area of ​​the corresponding excavator group, use the scheduling model to assign an excavator to the transport vehicle to complete the first stage of excavator assignment. The transport vehicle goes to the target excavator assigned in the first stage according to the assignment result.

[0010] 2.2) When the transport vehicle arrives at the second key area of ​​the corresponding excavator group, the vehicle queuing status of each excavator in the group is considered, and the scheduling model is used again to allocate excavators to the transport vehicle to achieve the second stage of excavator allocation;

[0011] 2.3) When the transport vehicle arrives at the excavator, the vehicle status changes to queuing or loading; the distance between the transport vehicle and each excavator in the group is calculated based on the GPS data, and the excavator corresponding to the minimum distance is the excavator actually reached by the transport vehicle, and it is determined whether the excavator actually reached is consistent with the target excavator allocated in the second stage; if they are consistent, it indicates that the transport vehicle obeys the allocation and there is no need to adjust the number of allocated vehicles for the target excavator allocated in the second stage; if they are inconsistent, it indicates that the transport vehicle does not obey the allocation, then the number of allocated vehicles for the excavator actually reached and the target excavator allocated in the second stage is adjusted, and then the vehicle allocation sequence of each excavator in the group is dynamically adjusted to prepare for the subsequent excavator allocation of the transport vehicle.

[0012] Furthermore, the process of allocating excavators to transport vehicles using the scheduling model includes:

[0013] Construct two Q tables and initialize them according to the following formula;

[0014]

[0015] Where Q(i) is the Q value of excavator i, T assign (i) is the current number of assigned vehicles for excavator i, t load (i) is the loading time of excavator i, randn is the random disturbance factor;

[0016] Action selection: Generate a random number between 0 and 1. If the random number is less than the exploration rate, randomly select an excavator from the group as the optimal excavator for the current iteration; if the random number is greater than the exploration rate, select the excavator with the largest sum of Q values ​​in the two Q tables as the optimal excavator for the current iteration;

[0017] The reward of the optimal excavator of the current iteration is calculated according to the following formula:

[0018] R(s t ,a t )=T assign (f)·r assign +t load (f)·r load +t 1to2 (f)·r 1to2 +t 2toex (f)·r 2toex (3)

[0019] In the formula, R(s t ,a t ) is the optimal excavator f in the current iteration in the current state s t Next, perform action a t Rewards, T assign (f) is the current number of assigned vehicles of the optimal excavator f in the current iteration, r assign is the weight of the current number of assigned vehicles in the reward, t load (f) is the loading operation time of the optimal excavator f in the current iteration, r load is the weight of loading operation time in the reward, t 1to2 (f) is the time it takes for the transport vehicle to cross the first key area and the second key area to reach the optimal excavator f of the current iteration, r 1to2 is the weight of the time taken by the transport vehicle to cross the first key area and the second key area to reach the optimal excavator f of the current iteration in the reward, t 2toex (f) is the time it takes for the transport vehicle to reach the optimal excavator f of the current iteration from the second key area, r 2toex is the weight of the time it takes for the transport vehicle to reach the optimal excavator f of the current iteration from the second key area in the reward;

[0020] The Q table is updated according to the reward, that is, a random number between 0 and 1 is generated. If the random number is less than or equal to 0.5, the first Q table is updated; if the random number is greater than 0.5, the second Q table is updated, which effectively reduces the valuation deviation; the update formula of the two Q tables is:

[0021]

[0022]

[0023] In the formula, Q1(s t ,a t )、Q2(s t ,a t ) represent the excavator in the current state s in the first Q table and the second Q table respectively. t Next, perform action a t Q value, α is the learning rate, γ is the reward decay factor, Q1(s t+1 ,a t+1 )、Q2(s t+1 ,a t+1 ) represent the excavator in the next state s in the first Q table and the second Q table respectively. t+1 Next, perform action a t+1 Q value, A represents the action set;

[0024] Attenuate the exploration rate and complete the current iteration;

[0025] The above process is repeated for multiple iterations. After reaching the maximum number of iterations, the Q values ​​of the same excavators in the two Q tables are summed up, and the excavator with the largest sum of Q values ​​is taken as the target excavator to complete the allocation of the target excavator.

[0026] Furthermore, in step 1, the starting position of each key area is calculated according to the following formula:

[0027]

[0028] Where lat1 and lon1 are the latitude and longitude of the excavator group, which are obtained by averaging the real-time GPS data of all excavators in the group; lat2 and lon2 are the latitude and longitude of the starting position of the key area, d is the distance between the starting position of the key area and the excavator group, and R is the radius of the earth.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] First, considering the situation that drivers do not fully comply with dispatch instructions in actual situations, the dispatch is divided into two stages. When the transport vehicle reaches the first key area, a preliminary excavator allocation is made to it. By real-time monitoring of the vehicle queue status or the actual arrival position of the loading vehicle, the vehicle allocation number of the excavator is dynamically updated, and secondary dispatch optimization is performed in the second key area. The transport vehicles that have not reached the destination are reallocated in real time. This not only effectively alleviates the system disturbance caused by human factors, but also significantly improves the dispatch system's adaptability to on-site emergencies, thereby ensuring the robustness and reliability of the entire dispatch system.

[0031] Second, in order to achieve the optimal allocation of excavators, the traditional Q-learning algorithm is improved. Through the Q-table initialization strategy based on the number of excavator allocation vehicles and loading time, excavators with lighter task loads are given priority, thereby improving the rationality of resource allocation; the dual Q-table update mechanism is adopted to reduce the dependence on a single Q-table, reduce the risk of overestimation of the Q value, and improve the accuracy of the algorithm in selecting the best excavator; by designing a gradually decaying exploration rate strategy, the algorithm is more inclined to utilize existing knowledge during the iteration process, which helps to improve the speed of training convergence, meet real-time requirements, and effectively alleviate the problem of long waiting time for excavators. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is the overall flow chart of the present invention;

[0033] Figure 2 A flow chart of allocating target excavators using the Q-learning algorithm of the present invention;

[0034] Figure 3 It is a scatter plot of the excavator waiting time of the present invention. DETAILED DESCRIPTION

[0035] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present application is not limited thereto.

[0036] The present invention provides a two-stage optimization scheduling method for open-pit mine transport vehicles (hereinafter referred to as the method, see Figures 1 to 3 ), including the following steps:

[0037] Step 1: Group the excavators in the mining area, group the excavators with similar distances into one group, and obtain multiple excavator groups; take the area within the range of d1 and d2 (d1>d2) from the excavator group in the direction of entry of the transport vehicle as the first key area and the second key area, and determine the first key area and the second key area of ​​each excavator group; calculate the starting position of each key area according to the following formula:

[0038]

[0039] Where lat1 and lon1 are the latitude and longitude of the excavator group, which are obtained by averaging the real-time GPS data of all excavators in the group; lat2 and lon2 are the latitude and longitude of the starting position of the key area, d is the distance between the starting position of the key area and the excavator group, and R is the radius of the earth.

[0040] Step 2: When the transport vehicle arrives at the first key area, the scheduling model is used to assign an excavator to it; when it arrives at the second key area, the scheduling model is used again to assign an excavator to it;

[0041] 2.1) Assign transport vehicles to each excavator group. For transport vehicle P, when it arrives at the first key area of ​​the corresponding excavator group (at this time, transport vehicle P just passes the starting position of the first key area), use the scheduling model to assign an excavator to transport vehicle P, completing the first stage of excavator assignment. At the same time, transport vehicle P goes to the target excavator assigned in the first stage according to the assignment result;

[0042] 2.2) In the actual operation environment, there is a system deviation caused by the fact that the transport vehicle drivers do not fully comply with the excavator allocation results of the first stage, which makes the actual number of allocated excavators inconsistent with the calculated number of allocated vehicles, thus affecting the excavator allocation of subsequent vehicles. Therefore, it is necessary to dynamically correct and optimize the allocation of excavators in the first stage;

[0043] When the transport vehicle P arrives at the second key area of ​​the corresponding excavator group, the actual queuing status of each excavator in the group is considered, and the scheduling model is used again to allocate excavators to the transport vehicle P to achieve the second stage of excavator allocation; since the transport vehicle driver may not fully comply with the excavator allocation result of the first stage, resulting in the actual queuing status of the excavator vehicle being inconsistent with the allocation result, the number of allocated excavators will be adjusted, so that the target excavators allocated by the transport vehicle P in the first stage and the second stage are inconsistent, but the actual number of allocated excavators is considered in the excavator allocation process of the second stage, and the number of allocated excavators in the second stage is adjusted, so the allocation result of the second stage is more reasonable than that of the first stage;

[0044] 2.3) When the transport vehicle P arrives at the excavator, the vehicle status changes to queuing or loading, indicating that the transport vehicle P is already queuing or loading at one of the excavators; the distance between the transport vehicle P and each excavator in the group is calculated based on the GPS data, and the excavator corresponding to the minimum distance is the excavator actually arrived at by the transport vehicle P, and it is determined whether the excavator actually arrived is consistent with the target excavator allocated in the second stage; if they are consistent, it indicates that the transport vehicle obeys the allocation and there is no need to adjust the number of allocated vehicles of the target excavator allocated in the second stage, then the allocation sequence of the excavator actually arrived and the target excavator allocated in the second stage does not change; if they are inconsistent, it indicates that the transport vehicle does not obey the allocation, then the number of allocated vehicles of the excavator actually arrived and the target excavator allocated in the second stage are adjusted, that is, the number of allocated vehicles of the target excavator allocated in the second stage is reduced by 1, and the number of allocated vehicles of the excavator actually arrived is increased by 1, and then the vehicle allocation sequence of each excavator in the group is adjusted in real time and dynamically, so as to prepare for the subsequent excavator allocation of the transport vehicle and realize refined scheduling.

[0045] The process of assigning excavators to transport vehicles using the scheduling model includes:

[0046] Construct a double Q table, and initialize the Q table of the Q-learning algorithm according to formula (2) by introducing different random perturbation factors, so that the double Q table has obvious differences in the initial state;

[0047]

[0048] Where Q(i) is the Q value of excavator i; T assign (i) is the current number of assigned vehicles for excavator i. By adding 1, we avoid the problem of infinite denominator due to the number of assigned vehicles being 0; t load (i) is the loading time of excavator i, randn is the random disturbance factor;

[0049] In the iteration phase, the epsilon-greedy strategy is first used for action selection, that is, a random number between 0 and 1 is generated. If the random number is less than the exploration rate ε, an excavator is randomly selected from the group as the optimal excavator for the current iteration to achieve exploration of different excavators; if the random number is greater than the exploration rate ε, the excavator with the largest sum of Q values ​​in the two Q tables is selected as the optimal excavator for the current iteration;

[0050] The reward of the optimal excavator of the current iteration is calculated according to the following reward function:

[0051] R(s t ,a t )=T assign (f)·r assign +t load (f)·r load +t 1to2 (f)·r 1to2 +t 2toex (f)·r 2toex (3)

[0052] In the formula, R(s t ,a t ) is the optimal excavator f in the current iteration in the current state s t Next, perform action a t Rewards, T assign (f) is the current number of assigned vehicles of the optimal excavator f in the current iteration, r assign is the weight of the current number of assigned vehicles in the reward, t load (f) is the loading operation time of the optimal excavator f in the current iteration, r load is the weight of loading operation time in the reward, t 1to2 (f) is the time it takes for the transport vehicle to cross the first key area and the second key area to reach the optimal excavator f of the current iteration, r 1to2 is the weight of the time taken by the transport vehicle to cross the first key area and the second key area to reach the optimal excavator f of the current iteration in the reward, t 2toex (f) is the time it takes for the transport vehicle to reach the optimal excavator f of the current iteration from the second key area, r 2toex is the weight of the time it takes for the transport vehicle to reach the optimal excavator f of the current iteration from the second key area in the reward;

[0053] The Q table is updated according to the reward, that is, a random number between 0 and 1 is generated. If the random number is less than or equal to 0.5, the first Q table is updated; if the random number is greater than 0.5, the second Q table is updated, which effectively reduces the valuation deviation; the update formula of the two Q tables is:

[0054]

[0055]

[0056] In the formula, Q1(s t ,a t )、Q2(s t ,a t ) represent the excavator in the current state s in the first Q table and the second Q table respectively. t Next, perform action a t Q value, α is the learning rate, γ is the reward decay factor, Q1(s t+1 ,a t+1 )、Q2(s t+1 ,a t+1 ) represent the excavator in the next state s in the first Q table and the second Q table respectively. t+1 Next, perform action a t+1 Q value, A represents the action set;

[0057] The Q-table update adopts a dual-Q cross-learning mechanism, that is, when estimating the Q value of the next state in a Q-table, the Q value of another Q-table is used, which effectively alleviates the over-estimation problem in traditional Q-learning and improves the stability and convergence performance of the algorithm.

[0058] The exploration rate is dynamically optimized, that is, the exploration rate is decayed according to ε=max(0.1,0.995ε) in each iteration, so that the algorithm gradually shifts from exploration-oriented to experience-oriented during the iteration process. At the same time, a lower limit of 0.1 is set for the exploration rate to maintain continuous exploration capability and prevent falling into the local optimal solution.

[0059] The current iteration is now completed; repeat the above process to explore the optimal excavator through multiple iterations. After reaching the maximum number of iterations, sum the Q values ​​of the same excavators in the two Q tables, and take the excavator with the largest sum of Q values ​​as the target excavator to complete the allocation of the target excavator.

[0060] Figure 3In order to compare the impact of the method of the present invention and not using the scheduling algorithm on the waiting time of the excavator, the scheduling algorithm was not used before 150 seconds, and the method of the present invention was used for scheduling at 150 seconds. As can be seen from the figure, the use of the method of the present invention can effectively shorten the waiting time of the excavator and improve the utilization efficiency of the excavator and the transport vehicle. The method of the present invention dynamically updates the number of vehicles assigned to the excavator by real-time monitoring the queuing status of the vehicles or the actual arrival position of the loading vehicles, and performs secondary scheduling optimization in the second key area to reallocate the transport vehicles that have not reached the destination in real time, which helps to shorten the waiting time of the excavator. Through the Q table initialization strategy constructed based on the number of excavator assigned vehicles and the loading time, the excavator with a lighter task load is given priority, which improves the rationality of resource allocation; the double Q table update mechanism is used to effectively suppress the over-estimation problem of the Q value; the dynamically decayed exploration rate is designed to ensure the real-time response capability of the system while ensuring the convergence of the algorithm, so that the excavator is more inclined to use existing knowledge, accelerate the training convergence speed, and meet the real-time requirements; therefore, the scheduling effect of the method of the present invention is better than that of not using the algorithm.

[0061] Any matters not described in the present invention are applicable to the prior art.

Claims

1. A two-stage optimization scheduling method for open-pit mine transport vehicles, characterized in that: The method comprises the following steps: Step 1, grouping the excavators in the mining area, grouping the excavators with similar distances into one group, and obtaining multiple excavator groups; taking the area within the range of d1 and d2 (d1>d2) from the excavator group in the direction of entry of the transport vehicle as the first key area and the second key area, and determining the first key area and the second key area of ​​each excavator group; Step 2: When the transport vehicle arrives at the first key area, the scheduling model is used to assign an excavator to it; when it arrives at the second key area, the scheduling model is used again to assign an excavator to it; 2.1) Assign transport vehicles to each excavator group. When the transport vehicle arrives at the first key area of ​​the corresponding excavator group, use the scheduling model to assign an excavator to the transport vehicle to complete the first stage of excavator assignment. The transport vehicle goes to the target excavator assigned in the first stage according to the assignment result. 2.2) When the transport vehicle arrives at the second key area of ​​the corresponding excavator group, the vehicle queuing status of each excavator in the group is considered, and the scheduling model is used again to allocate excavators to the transport vehicle to achieve the second stage of excavator allocation; 2.3) When the transport vehicle arrives at the excavator, the vehicle status changes to queuing or loading; the distance between the transport vehicle and each excavator in the group is calculated based on the GPS data, and the excavator corresponding to the minimum distance is the excavator actually reached by the transport vehicle, and it is determined whether the excavator actually reached is consistent with the target excavator allocated in the second stage; if they are consistent, it indicates that the transport vehicle obeys the allocation and there is no need to adjust the number of allocated vehicles for the target excavator allocated in the second stage; if they are inconsistent, it indicates that the transport vehicle does not obey the allocation, then the number of allocated vehicles for the excavator actually reached and the target excavator allocated in the second stage is adjusted, and then the vehicle allocation sequence of each excavator in the group is dynamically adjusted to prepare for the subsequent excavator allocation of the transport vehicle.

2. The two-stage optimization scheduling method for open-pit mine transport vehicles according to claim 1 is characterized in that: The process of assigning excavators to transport vehicles using the scheduling model includes: Construct two Q tables and initialize them according to the following formula; Where Q(i) is the Q value of excavator i, T assign (i) is the current number of assigned vehicles for excavator i, t load (i) is the loading time of excavator i, randn is the random disturbance factor; Action selection: Generate a random number between 0 and 1. If the random number is less than the exploration rate, randomly select an excavator from the group as the optimal excavator for the current iteration; if the random number is greater than the exploration rate, select the excavator with the largest sum of Q values ​​in the two Q tables as the optimal excavator for the current iteration; The reward of the optimal excavator of the current iteration is calculated according to the following formula: R(s t ,a t )=T assign (f)·r assign +t load (f)·r load +t 1to2 (f)·r 1to2 +t 2toex (f)·r 2toex (3) In the formula, R(s t ,a t ) is the optimal excavator f in the current iteration in the current state s t Next, perform action a t Reward, T assign (f) is the current number of assigned vehicles of the optimal excavator f in the current iteration, r assign is the weight of the current number of assigned vehicles in the reward, t load (f) is the loading operation time of the optimal excavator f in the current iteration, r load is the weight of loading operation time in the reward, t 1to2 (f) is the time it takes for the transport vehicle to cross the first key area and the second key area to reach the optimal excavator f of the current iteration, r 1to2 is the weight of the time taken by the transport vehicle to cross the first key area and the second key area to reach the optimal excavator f of the current iteration in the reward, t 2toex (f) is the time it takes for the transport vehicle to reach the optimal excavator f of the current iteration from the second key area, r 2toex is the weight of the time it takes for the transport vehicle to reach the optimal excavator f of the current iteration from the second key area in the reward; The Q table is updated according to the reward, that is, a random number between 0 and 1 is generated. If the random number is less than or equal to 0.5, the first Q table is updated; if the random number is greater than 0.5, the second Q table is updated, which effectively reduces the valuation deviation; the update formula of the two Q tables is: In the formula, Q1(s t ,a t )、Q2(s t ,a t ) represent the excavator in the current state s in the first Q table and the second Q table respectively. t Next, perform action a t Q value, α is the learning rate, γ is the reward decay factor, Q1(s t+1 ,a t+1 )、Q2(s t+1 ,a t+1 ) represent the excavator in the next state s in the first Q table and the second Q table respectively. t+1 Next, perform action a t+1 Q value, A represents the action set; Decrease the exploration rate and complete the current iteration; The above process is repeated for multiple iterations. After reaching the maximum number of iterations, the Q values ​​of the same excavators in the two Q tables are summed up, and the excavator with the largest sum of Q values ​​is taken as the target excavator to complete the allocation of the target excavator.

3. The two-stage optimization scheduling method for open-pit mine transport vehicles according to claim 1 or 2, characterized in that: In step 1, the starting position of each key area is calculated according to the following formula: Where lat1 and lon1 are the latitude and longitude of the excavator group, which are obtained by averaging the real-time GPS data of all excavators in the group; lat2 and lon2 are the latitude and longitude of the starting position of the key area, d is the distance between the starting position of the key area and the excavator group, and R is the radius of the earth.

Images