Distributed multi-agent collaborative coal conveying control system and control method for long-distance scraper conveyor

Abstract

The invention belongs to the technical field of coal transportation of fully mechanized coal mining faces, and particularly relates to a multi-agent distributed collaborative coal conveying control system and method for a long-distance scraper conveyor. Comprising a machine head intelligent body, and the machine head intelligent body comprises a first permanent magnet motor and a machine head assembly driven by the first permanent magnet motor; each middle intelligent agent comprises a second permanent magnet motor and a middle chain wheel assembly driven by the second permanent magnet motor; the tail intelligent body comprises a third permanent magnet motor and a tail assembly driven by the third permanent magnet motor; the centralized control system comprises an industrial personal computer, a current sensor and a tension sensor; when tension control or motor driving of a single intelligent agent breaks down, coal can still be transported normally, and the mining efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of coal transportation in fully mechanized mining working faces, and specifically relates to a multi-agent distributed collaborative coal transportation control system and a control method for a long-distance scraper conveyor. Background Art

[0002] With the development of intelligent and unmanned coal mining, the coal mining length of the fully-mechanized mining face has gradually increased. The scraper conveyor is the only coal transportation equipment in the fully-mechanized mining face. As the fully-mechanized mining face increases, new requirements are also put forward for its own length. The long-distance scraper conveyor is the current development trend.

[0003] However, the current maximum length of the scraper conveyor is 480m. If the coal conveying length is simply increased, it will inevitably lead to chain stacking, chain breaking, insufficient motor power, and excessive motor size. It is mainly reflected in the following aspects. The current scraper conveyor only relies on the hydraulic cylinder at the tail of the machine to push the chain. With the development of long-distance scraper conveyors, the total length of chain tensioning will continue to increase, resulting in the burden on the hydraulic system at the tail of the machine increasing. The hydraulic cylinder extends too long and becomes stuck and cannot be retracted. The cross-lap telescopic structure of the scraper conveyor tail is worn and curled, making it impossible to work smoothly. Therefore, only single-end tensioning force control cannot meet the development trend. In addition, the current scraper conveyor adopts the method of driving the motor at both ends of the head and the tail. As the coal conveying distance of the scraper conveyor increases, it will lead to insufficient motor power or excessive motor size.

[0004] Therefore, the present invention adopts a strategy of coordinated intelligent coal transportation using distributed motors and tensioning systems to meet the problems existing in long-distance scraper conveyors, improve transportation efficiency and safety factors, and improve the level of intelligent mining in coal mines. Summary of the invention

[0005] In order to solve the problems existing in the motor drive and chain tail tension force control systems of long-distance scraper conveyors and improve the coal mining efficiency and stability of scraper conveyors, the present invention provides a distributed multi-agent collaborative coal transportation control system and control method for long-distance scraper conveyors.

[0006] The present invention adopts the following technical scheme: a distributed multi-agent collaborative coal transportation control system for a long-distance scraper conveyor, comprising: A machine head intelligent body, the machine head intelligent body comprising a first permanent magnet motor and a machine head assembly driven by the first permanent magnet motor, the machine head assembly being adjusted and tensioned by a machine head tensioning hydraulic cylinder; a plurality of intermediate agents, the intermediate agents comprising a second permanent magnet motor and a middle sprocket assembly driven by the second permanent magnet motor; A tail intelligent body, the tail intelligent body comprising a third permanent magnet motor and a tail assembly driven by the third permanent magnet motor, the tail assembly being adjusted and tensioned by a tail tensioning hydraulic cylinder; The head intelligent body is arranged at the head and connected to the cross side unloading part, the tail intelligent body is arranged at the tail, and a plurality of the intermediate intelligent bodies are arranged at intervals between the cross side unloading part and the tail intelligent body, two adjacent intermediate intelligent bodies are connected through the coal conveying intermediate part, the intermediate intelligent body is connected to the cross side unloading part through the coal conveying intermediate part, and the intermediate intelligent body is connected to the tail intelligent body through the coal conveying intermediate part; A centralized control system, the centralized control system comprising an industrial computer, a current sensor and a tension sensor; the current sensor is used to detect the current value of the first permanent magnet motor, the second permanent magnet motor and the third permanent magnet motor, and is used to determine the load of transporting coal and the output power of the motor; the tension sensor is used to detect the tension of the chain; the industrial computer receives information from the current sensor and the tension sensor and controls the first permanent magnet motor, the second permanent magnet motor and the third permanent magnet motor; A scraper conveyor chain system is arranged on the head intelligent body, the cross side unloading part, the middle intelligent body, the coal conveying middle part and the tail intelligent body.

[0007] In some embodiments, the nose assembly includes: Head bottom plate; A head frame, wherein the head frame is mounted on the head bottom plate, the front end of the head frame is hingedly connected to the end of the cross side unloading portion through a head rotating shaft, a sprocket I is arranged in the head frame, the sprocket I is mounted on a sprocket shaft I, and the sprocket shaft I is driven by a first permanent magnet motor; The machine head tensioning hydraulic cylinder is connected between the machine head bottom plate and the rear part of the machine head frame.

[0008] In some embodiments, the middle sprocket assembly includes: A supporting bottom plate, with supporting side plates arranged on both sides of the supporting bottom plate; Lifting hydraulic cylinders, wherein two groups of lifting hydraulic cylinders are provided and symmetrically arranged on both sides of the supporting bottom plate; An intermediate sprocket shaft, on which a sprocket is arranged, the intermediate sprocket shaft is installed between two lifting hydraulic cylinders, coal retaining plates are arranged on both sides of the intermediate sprocket shaft, and the intermediate sprocket shaft is driven by a second permanent magnet motor; The intermediate transition frame is fixedly connected to the front and rear ends of the supporting bottom plate and is respectively connected to the middle parts of the coal conveying at both ends.

[0009] In some embodiments, a first-level movable middle plate is symmetrically installed on both sides of the middle sprocket shaft, one end of the first-level movable middle plate is hingedly connected to the middle transition frame, the other end of the first-level movable middle plate is rotatably connected to the second-level movable middle plate through the middle secondary rotating shaft, the outer surface of the second-level movable middle plate is in tangential contact with the sprocket shaft, and a coal-blocking hydraulic cylinder is connected between the second-level movable middle plate and the middle transition frame. The clamping force provided by the coal-blocking hydraulic cylinder keeps the second-level movable middle plate and the middle sprocket shaft in a state of tangential contact at all times, thereby preventing the transported coal from falling to the lower side of the scraper conveyor.

[0010] In some embodiments, a plurality of U-shaped grooves are provided in the middle of the primary movable middle plate, the U-shaped grooves are used to allow the protruding portion of the middle sprocket shaft to pass through, and a derailleur is provided at the U-shaped grooves.

[0011] In some embodiments, the tail assembly includes: tail bottom plate; A tail frame, wherein the tail frame is mounted on the tail bottom plate, a sprocket III is arranged in the tail frame, the sprocket III is mounted on a sprocket shaft III, and the sprocket shaft III is driven by a third permanent magnet motor; A tail transition frame, one end of which is connected to the coal conveying middle part, and the other end is hingedly connected to the tail frame through a tail rotating shaft; The tail tensioning hydraulic cylinder is connected between the tail frame and the tail bottom plate.

[0012] A control method for a distributed multi-agent collaborative coal handling control system for a long-distance scraper conveyor, comprising: S1: The industrial computer receives the current values ​​In of the first permanent magnet motor, the second permanent magnet motor and the third permanent magnet motor and the signal Fn of the tension sensor in the scraper; S2: Based on the collected current signals, the industrial computer controls the power of multiple motors to balance the power output among the head intelligent body, the middle intelligent body and the tail intelligent body, that is, the current values ​​of the first permanent magnet motor, the second permanent magnet motor and the third permanent magnet motor are the same, and the drive of the scraper conveyor is globally controlled to ensure the smooth operation of the chain; S3: Based on the collected tension signal Fn, the industrial computer adjusts the local tension of the scraper conveyor in sections to ensure that all chain tension values ​​are always greater than 0; S4: Taking the maximum value of the function Y=0.3 Fn+0.4 In+0.3 M as the target, where M is the coal transport volume per hour, and the relationship between the motor power and the tensioning force is weighed, the driving power and tensioning force of the first permanent magnet motor, the second permanent magnet motor and the third permanent magnet motor are coordinated and controlled to ensure that the coal transport work reaches the best state; S5: When the first permanent magnet motor, the second permanent magnet motor, the third permanent magnet motor, the head tensioning hydraulic cylinder, or the tail tensioning hydraulic cylinder fails, the control quantity is transferred to the adjacent intelligent body to ensure that the motor power and the tensioning force meet the control requirements.

[0013] In some embodiments, in step S2, the process of controlling the power of multiple motors includes: S21: Build strategy network and value network, The input of the strategy network is the current value and the speed of the first permanent magnet motor, the second permanent magnet motor and the third permanent magnet motor, and the output is the probability distribution function of the speed; The input of the value network is the current value and speed of the first permanent magnet motor, the second permanent magnet motor and the third permanent magnet motor, and the output is the value of the current state, which is used to evaluate the quality of the current state; Both the policy network and the value network use feedforward neural networks; S22: Establish a three-dimensional model consistent with the real-world device in the software, add input interface and output interface to the three-dimensional model, the input interface is used to input the initial speed, and the output interface is used to output the "current value and speed of the first permanent magnet motor, the second permanent magnet motor and the third permanent magnet motor" simulated in real time by the three-dimensional model and the reward value given to the agent after the current agent takes action. According to the formula Calculate reward value; S23: input an initial value in the input interface, the three-dimensional model starts running, and a round of training data is generated in the three-dimensional model; S24: input the training data into the strategy network and the value network for training; S25: Repeat steps S23-S24 continuously to calculate the average reward change in the last six training rounds. If the change is less than the set threshold When R<0.01, the strategy is considered to have converged, and then the training is stopped. The current values ​​and speeds of the first permanent magnet motor, the second permanent magnet motor, and the third permanent magnet motor collected in real time are input through the trained strategy network, and the speed is output to operate each motor.

[0014] In some embodiments, in step S3, the process of adjusting the local tension of the scraper conveyor in sections includes: S31: Build strategy network and value network, The input of the strategy network is the chain tension and the extension and contraction of the hydraulic cylinder, and the output is the correction value of the extension and contraction of the hydraulic cylinder; The input of the value network is the chain tension and the extension of the hydraulic cylinder, and the output is the value of the current state, which is used to evaluate the quality of the current state; Both the policy network and the value network use feedforward neural networks; S32: Establish a three-dimensional model consistent with the real-world device in the software, add input interface and output interface to the three-dimensional model, the input interface is used to input the chain tension, and the output interface is used to output the "chain tension and hydraulic cylinder extension and contraction amount" simulated in real time by the three-dimensional model and the reward value given after the current hydraulic cylinder takes action. If Fn>0, the reward is 0; if Fn≤0, the reward is −c; S33: input an initial value in the input interface, the three-dimensional model starts running, and a round of training data is generated in the three-dimensional model; S34: input the training data into the strategy network and the value network for training; S35: Repeat steps S33-S34 continuously. When the system reward value remains at a stable positive value after multiple trainings and the reward value fluctuation range is less than <10%, the system is considered to have reached convergence. Then stop training, input the chain tension and hydraulic cylinder extension amount collected in real time through the trained strategy network, and output the correction value of the hydraulic cylinder extension amount.

[0015] In some embodiments, in step S4, the process of coordinated control of driving power and tensioning force adjustment includes: S41: Build strategy network and value network, The input of the strategy network is the motor power balance state, chain tension state, coal delivery volume M per hour, and motor speed, and the output is motor speed and adjustment tension; The input of the value network is the motor power balance state, chain tension state, coal delivery volume M per hour, and motor speed. The output is the value of the current state, which is used to evaluate the quality of the current state. Both the policy network and the value network use feedforward neural networks; S42: Establish a 3D model consistent with the real-world device in the software, add input interface and output interface to the 3D model, the input interface is used to input the chain tension Fn, the motor current value In and the coal transport volume M, and the output interface is used to output the "motor speed" simulated in real time by the 3D model. xn, hydraulic cylinder extension ωn" and the reward value given after the current hydraulic cylinder takes action, ; S43: input an initial value in the input interface, the three-dimensional model starts running, and the three-dimensional model outputs a round of training data; S44: input the training data into the strategy network and the value network for training; S45: Repeat steps S43-S44 continuously. When the trade-off between chain tension, motor current, and coal transport capacity reaches a stable state, that is, the reward value growth in the iteration is <10% and the reward value fluctuation is small, stop training, and use the trained strategy network to input the real-time collected chain tension, hydraulic cylinder extension and motor current values, and output the correction value of the hydraulic cylinder extension and motor speed.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The invention integrates the chain tension control and chain motion drive into an intelligent unit, and adopts a distributed approach to divide it into a head intelligent unit, a middle intelligent unit, and a tail intelligent unit. When the tension control or motor drive of a single intelligent unit fails, coal can still be transported normally, thereby improving the efficiency of mining.

[0017] The head intelligent body and the tail intelligent body of the present invention design a tensioning device based on the theory of different chord lengths at a fixed point on a circle, which can effectively reduce the displacement of the head and tail when the chain is tensioned and reduce structural deformation.

[0018] The present invention designs an intermediate intelligent body, and adopts a hydraulic cylinder to lift and lower a sprocket to realize the control of the tensioning force.

[0019] The present invention adopts a distributed tensioning force control method, which can effectively reduce the burden of the tail tensioning force adjustment system and extend the service life of the tensioning device compared to the traditional single tail tensioning force control.

[0020] By adopting the method of multi-point drive and multi-point tension coordinated control and weighing the pros and cons between the two, the transportation process of the chain drive system is made more stable and the control scheme is more reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic diagram of the structure of the scraper conveyor of the present invention; Figure 2 A schematic diagram of the structure of the head intelligent body of the present invention; Figure 3 A schematic diagram of the structure of the tail intelligent body of the present invention; Figure 4 A schematic diagram of the structure of the coal conveying middle part of the present invention; Figure 5 The cross side unloading part of the present invention; Figure 6 Appearance of the intermediate agent of the present invention Figure 1 ; Figure 7 Appearance of the intermediate agent of the present invention Figure 2 ; Figure 8 The internal schematic diagram of the intermediate intelligent body of the present invention; Fig. 9 A partial schematic diagram of the intermediate intelligent agent of the present invention; Fig.10 The tensioning principle of the nose and tail intelligent bodies of the present invention; Fig.11 The intermediate intelligent body tensioning principle of the present invention; In the figure, 1-machine head intelligent body, 101-machine head rotating shaft, 102-machine head frame, 103-machine head bottom plate, 104-machine head tensioning hydraulic cylinder, 2-cross side unloading part, 201-transfer machine middle groove, 202-transfer machine chain, 203-transfer machine scraper, 204-transfer machine guide plate, 205-transfer machine transition frame, 3-coal conveying middle part, 4-middle intelligent body, 401-protection cover, 402-middle transition frame, 403-middle sprocket shaft, 404-lifting hydraulic cylinder, 405-support bottom plate, 406 -Support side plate, 407-coal blocking plate, 408-first-level movable middle plate, 409-second-level movable middle plate, 410-coal blocking hydraulic cylinder, 411-middle first-level rotating shaft, 412-middle second-level rotating shaft, 413-lifting slot, 414-chain derailleur, 5-tail intelligent body, 501-tail rotating shaft, 502-tail frame, 503-tail bottom plate, 504-tail tensioning hydraulic cylinder, 505-tail transition frame, 6-scraper, 7-chain, 8-sprocket shaft, 9-sprocket, 10-permanent magnet motor, 11-middle slot. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, a distributed multi-agent collaborative coal transportation control system for a long-distance scraper conveyor includes: A machine head intelligent body 1, wherein the machine head intelligent body 1 comprises a first permanent magnet motor 10.1 and a machine head assembly driven by the first permanent magnet motor 10.1, and the machine head assembly is adjusted and tensioned by a machine head tensioning hydraulic cylinder 104; A plurality of intermediate intelligent bodies 4, wherein the intermediate intelligent bodies 4 include a second permanent magnet motor 10.2 and a middle sprocket assembly driven by the second permanent magnet motor 10.2; The tail intelligent body 5 comprises a third permanent magnet motor 10.3 and a tail assembly driven by the third permanent magnet motor 10.3, and the tail assembly is adjusted and tensioned by a tail tensioning hydraulic cylinder 504; The head intelligent body 1 is arranged at the head and connected to the cross side unloading part 2, the tail intelligent body 5 is arranged at the tail, and the plurality of intermediate intelligent bodies 4 are arranged at intervals between the cross side unloading part 2 and the tail intelligent body 5, two adjacent intermediate intelligent bodies 4 are connected through the coal conveying intermediate part 3, the intermediate intelligent body 4 is connected to the cross side unloading part 2 through the coal conveying intermediate part 3, and the intermediate intelligent body 4 is connected to the tail intelligent body 5 through the coal conveying intermediate part 3; The centralized control system includes an industrial computer, a current sensor and a tension sensor; the current sensor is used to detect the current value of the first permanent magnet motor 10.1, the second permanent magnet motor 10.2 and the third permanent magnet motor 10.3, and is used to determine the load of transporting coal and the output power of the motor; the tension sensor is used to detect the tension of the chain; the industrial computer receives information from the current sensor and the tension sensor and controls the first permanent magnet motor 10.1, the second permanent magnet motor 10.2 and the third permanent magnet motor 10.3; A scraper conveyor chain system is arranged on the head intelligent body 1 , the cross side unloading part 2 , the middle intelligent body 4 , the coal conveying middle part 3 and the tail intelligent body 5 .

[0024] like Figure 2 As shown, the head assembly includes: Head bottom plate 103; A machine head frame 102, wherein the machine head frame 102 is mounted on a machine head bottom plate 103, and a front end of the machine head frame 102 is hingedly connected to an end of the cross side unloading portion 2 through a machine head rotating shaft 101, and a sprocket wheel 19.1 is arranged in the machine head frame 102, and the sprocket wheel 19.1 is mounted on a sprocket wheel shaft 18.1, and the sprocket wheel shaft 18.1 is driven by a first permanent magnet motor 10.1; The machine head tensioning hydraulic cylinder 104 is connected between the machine head bottom plate 103 and the rear part of the machine head frame 102 .

[0025] like Figure 6-9 As shown, the middle sprocket assembly includes: A supporting bottom plate 405, with supporting side plates 406 disposed on both sides of the supporting bottom plate 405; Lifting hydraulic cylinders 404, wherein two groups of lifting hydraulic cylinders 404 are provided and symmetrically arranged on both sides of the supporting bottom plate 405; An intermediate sprocket shaft 403, on which a sprocket 9.2 is arranged, the intermediate sprocket shaft 403 is installed between two lifting hydraulic cylinders 404, coal retaining plates 407 are arranged on both sides of the intermediate sprocket shaft 403, and the intermediate sprocket shaft 403 is driven by a second permanent magnet motor 10.2; The intermediate transition frame 402 is fixedly connected to the front and rear ends of the support bottom plate 405 and is respectively connected to the coal conveying middle parts 3 at both ends.

[0026] A first-level movable middle plate 408 is symmetrically installed on both sides of the middle sprocket shaft 403. One end of the first-level movable middle plate 408 is hingedly connected to the middle transition frame 402, and the other end of the first-level movable middle plate 408 is rotatably connected to the second-level movable middle plate 409 through the middle-level secondary rotating shaft 412. The outer surface of the second-level movable middle plate 409 is in tangential contact with the sprocket shaft, and a coal-blocking hydraulic cylinder 410 is connected between the second-level movable middle plate 409 and the middle transition frame 402. The clamping force provided by the coal-blocking hydraulic cylinder 410 keeps the second-level movable middle plate 408 and the middle sprocket shaft in a tangential contact state at all times to prevent the transported coal from falling to the lower side of the scraper conveyor.

[0027] A plurality of U-shaped grooves are arranged in the middle of the primary movable middle plate 408 , and the U-shaped grooves are used for allowing the protruding portion of the middle sprocket shaft 403 to pass through, and a derailleur 414 is arranged at the U-shaped grooves.

[0028] like Figure 3 As shown, the tail assembly includes: Tail bottom plate 503; A tail frame 502, wherein the tail frame 502 is mounted on a tail bottom plate 503, a sprocket III9.3 is arranged in the tail frame 502, the sprocket III9.3 is mounted on a sprocket shaft III8.3, and the sprocket shaft III8.3 is driven by a third permanent magnet motor 10.3; A tail transition frame 505, one end of which is connected to the coal conveying middle part 3, and the other end is hingedly connected to the tail frame 502 through the tail rotating shaft 501; The tail tensioning hydraulic cylinder 504 is connected between the tail frame 502 and the tail bottom plate 503 .

[0029] like Figure 5 As shown, the cross side unloading part 2 adopts the existing structure. The cross side unloading part 2 is composed of a transfer machine guide plate 204, a transfer machine transition frame 205, a transfer machine scraper 203, a transfer machine chain 202, and a transfer machine middle groove 201. The coal transported by the scraper conveyor is transferred to the transfer machine at the cross side unloading part, and then the coal is transported out.

[0030] A control method for a distributed multi-agent collaborative coal handling control system for a long-distance scraper conveyor, comprising: S1: The industrial computer receives the current values ​​In of the first permanent magnet motor 10.1, the second permanent magnet motor 10.2 and the third permanent magnet motor 10.3 and the tension sensor Fn signal in the scraper.

[0031] S2: The industrial computer controls the power of multiple motors based on the collected current signals, so that the power output between the head intelligent body 1, the middle intelligent body 4 and the tail intelligent body 5 is balanced, that is, the current values ​​of the first permanent magnet motor 10.1, the second permanent magnet motor 10.2 and the third permanent magnet motor 10.3 are the same, and the drive of the scraper conveyor is globally controlled to ensure the smooth operation of the chain; The process of controlling the power of multiple motors includes: S21: Build strategy network and value network, The input of the strategy network is the current value and the speed of the first permanent magnet motor 10.1, the second permanent magnet motor 10.2 and the third permanent magnet motor 10.3, and the output is the probability distribution function of the speed; The input of the value network is the current value and speed of the first permanent magnet motor 10.1, the second permanent magnet motor 10.2 and the third permanent magnet motor 10.3, and the output is the value of the current state, which is used to evaluate the quality of the current state; Both the policy network and the value network use feedforward neural networks; S22: Establish a three-dimensional model consistent with the real-world device in the software, add input interface and output interface to the three-dimensional model, the input interface is used to input the initial speed, and the output interface is used to output the "current value and speed of the first permanent magnet motor 10.1, the second permanent magnet motor 10.2 and the third permanent magnet motor 10.3" simulated in real time by the three-dimensional model and the reward value given to the agent after the current agent takes action, according to the formula Calculate reward value; S23: input an initial value in the input interface, the three-dimensional model starts running, and a round of training data is generated in the three-dimensional model; S24: input the training data into the strategy network and the value network for training; The advantage function Â(t) is calculated at each time step using the generalized advantage estimation (GAE).

[0032] Â(t) = δ t +(γ×λ)×δ t+1 +(γ×λ) 2 ×δ t+2 + …, where δ t =R t +γ×V(S t+1 )−V(S t ).

[0033] Calculate the cumulative discount reward G t : G t =R t +γ×R t+1 +γ 2 ×R t+2+…, in, .

[0034] By optimizing the strategy, it updates its policy network and value network, enabling it to obtain higher rewards in future interactions.

[0035] The PPO (Proximal Policy Optimization) algorithm is used to optimize the strategy. The objective function of PPO is designed as a clipping objective function, which prevents drastic changes in the strategy by limiting the amplitude of the strategy update: in, , which represents the ratio between the current strategy and the old strategy.

[0036] Update the value network parameters using mean squared error (MSE) Use an optimizer (Adam) to update the policy network and value network parameters simultaneously.

[0037] S25: Repeat steps S23-S24 continuously to calculate the average reward change in the last six training rounds. If the change is less than the set threshold When R<0.01, the strategy is considered to have converged, and then the training is stopped. The current value and speed of the first permanent magnet motor (10.1), the second permanent magnet motor (10.2), and the third permanent magnet motor (10.3) collected in real time are input through the trained strategy network, and the speed is output to operate each motor.

[0038] S3: Industrial computer based on the collected tension signal F n, adjust the local tension of the scraper conveyor in sections to ensure that all chain tension values ​​are always greater than 0.

[0039] The process of segmented local tension adjustment of the scraper conveyor includes: S31: Build strategy network and value network, The input of the strategy network is the chain tension and the extension and contraction of the hydraulic cylinder, and the output is the correction value of the extension and contraction of the hydraulic cylinder; The input of the value network is the chain tension and the extension of the hydraulic cylinder, and the output is the value of the current state, which is used to evaluate the quality of the current state; Both the policy network and the value network use feedforward neural networks; S32: Establish a three-dimensional model consistent with the real-world device in the software, add input interface and output interface to the three-dimensional model, the input interface is used to input the chain tension, and the output interface is used to output the "chain tension and hydraulic cylinder extension and contraction amount" simulated in real time by the three-dimensional model and the reward value given after the current hydraulic cylinder takes action. If Fn>0, the reward is 0; if Fn≤0, the reward is −c; S33: input an initial value in the input interface, the three-dimensional model starts running, and a round of training data is generated in the three-dimensional model; S34: input the training data into the strategy network and the value network for training; The policy network and value network are updated by the MAPPO algorithm. In order to reduce variance and improve learning stability, the generalized advantage estimation (GAE) is used: in, is the discount factor, V(S t ) is the value estimate of the state.

[0040] Use the PPO algorithm for optimization and update the strategy by cutting the objective function: , in, , which represents the ratio between the current strategy and the old strategy. is the advantage function.

[0041] S5: Use mean squared error to optimize the value network, minimizing the following loss function: Among them, G t is the cumulative return calculated based on the discount factor.

[0042] The Adam optimizer is used to update the parameters of the policy network and the value network to reduce the loss function and improve the stability of the chain pulling force.

[0043] S35: Repeat steps S33-S34 continuously. When the system reward value remains at a stable positive value after multiple trainings and the reward value fluctuation range is less than <10%, the system is considered to have reached convergence. Then stop training, input the chain tension and hydraulic cylinder extension amount collected in real time through the trained strategy network, and output the correction value of the hydraulic cylinder extension amount.

[0044] S4: Taking the maximum value of the function Y=0.3 Fn+0.4 In+0.3 M as the target, where M is the coal transport volume per hour, and the relationship between the motor power and the tensioning force is weighed, the driving power and tensioning force of the first permanent magnet motor 10.1, the second permanent magnet motor 10.2 and the third permanent magnet motor 10.3 are coordinated and controlled to ensure that the coal transport work reaches the best state; The process of coordinated control of drive power and tension force adjustment includes: S41: Build strategy network and value network, The input of the strategy network is the motor power balance state, chain tension state, coal delivery volume M per hour, and motor speed, and the output is motor speed and adjustment tension; The input of the value network is the motor power balance state, chain tension state, coal delivery volume M per hour, and motor speed. The output is the value of the current state, which is used to evaluate the quality of the current state. Both the policy network and the value network use feedforward neural networks; S42: Establish a 3D model consistent with the real-world device in the software, add input interface and output interface to the 3D model, the input interface is used to input the chain tension Fn, the motor current value In and the coal transport volume M, and the output interface is used to output the "motor speed" simulated in real time by the 3D model. xn, hydraulic cylinder extension ωn" and the reward value given after the current hydraulic cylinder takes action, ; S43: input an initial value in the input interface, the three-dimensional model starts running, and the three-dimensional model outputs a round of training data; S44: input the training data into the strategy network and the value network for training; Calculate the advantage function for each agent , the generalized advantage estimation (GAE) can be used to reduce the variance in is the discount factor, is a parameter that balances bias and variance, and V(St) is an estimate of the value network.

[0045] Use PPO's clipping objective function for policy optimization: in is the ratio of the current strategy to the old strategy, is a hyperparameter that limits the magnitude of policy updates.

[0046] Use the mean squared error loss function to optimize the value network: Among them G t The Adam optimization algorithm is used to update the parameters of the policy network and the value network, minimize the loss function, and gradually improve the accuracy of system control.

[0047] S45: When the trade-off between the three (chain tension, motor current, and coal delivery) reaches a stable state, that is, the reward value growth in iteration is <10% and the reward value fluctuation is small, stop training.

[0048] S45: Repeat steps S43-S44 continuously. When the trade-off between chain tension, motor current, and coal transport capacity reaches a stable state, that is, the reward value growth in the iteration is <10% and the reward value fluctuation is small, stop training, and use the trained strategy network to input the real-time collected chain tension, hydraulic cylinder extension and motor current values, and output the correction value of the hydraulic cylinder extension and motor speed.

[0049] S5: When the first permanent magnet motor 10.1 or the second permanent magnet motor 10.2 or the third permanent magnet motor 10.3 or the head tensioning hydraulic cylinder 104 or the tail tensioning hydraulic cylinder 504 fails, the control quantity is transferred to the adjacent intelligent body to ensure that the motor power and the tensioning force meet the control requirements.

[0050] In the present invention, steps S2 and S3 are responsible for their own training, and step S4 takes care of the overall situation. It can be considered that the training is carried out in stages and layers to avoid one-time training without convergence.

[0051] Steps S2 and S3 are to respectively manage themselves and adjust their own current value or chain tension value to the best.

[0052] S4 is adjusted again. S4 takes into account the overall impact and weighs the pros and cons according to the weight. (For example, S2 adjusts the current to the best, but in s4, in order to take into account the overall situation, after another adjustment, the current value may not be the best state in S2.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distributed multi-agent collaborative coal handling control system for a long-distance scraper conveyor, characterized in that: include: A machine head intelligent body (1), the machine head intelligent body (1) comprising a first permanent magnet motor (10.1) and a machine head assembly driven by the first permanent magnet motor (10.1), the machine head assembly being adjusted and tensioned by a machine head tensioning hydraulic cylinder (104); A plurality of intermediate intelligent bodies (4), each of which comprises a second permanent magnet motor (10.2) and a middle sprocket assembly driven by the second permanent magnet motor (10.2); A tail intelligent body (5), the tail intelligent body (5) comprising a third permanent magnet motor (10.3) and a tail assembly driven by the third permanent magnet motor (10.3), the tail assembly being adjusted and tensioned by a tail tensioning hydraulic cylinder (504); The head intelligent body (1) is arranged at the head and connected to the cross side unloading part (2), the tail intelligent body (5) is arranged at the tail, a plurality of the intermediate intelligent bodies (4) are arranged at intervals between the cross side unloading part (2) and the tail intelligent body (5), two adjacent intermediate intelligent bodies (4) are connected via a coal conveying intermediate part (3), the intermediate intelligent body (4) and the cross side unloading part (2) are connected via the coal conveying intermediate part (3), and the intermediate intelligent body (4) and the tail intelligent body (5) are connected via the coal conveying intermediate part (3); A centralized control system, the centralized control system comprising an industrial computer, a current sensor and a tension sensor; the current sensor is used to detect the current value of the first permanent magnet motor (10.1), the second permanent magnet motor (10.2) and the third permanent magnet motor (10.3), and is used to determine the load of transported coal and the output power of the motor; the tension sensor is used to detect the tension of the chain; the industrial computer receives information from the current sensor and the tension sensor and controls the first permanent magnet motor (10.1), the second permanent magnet motor (10.2) and the third permanent magnet motor (10.3); A scraper conveyor chain system is provided on the head intelligent body (1), the cross-side unloading part (2), the middle intelligent body (4), the coal conveying middle part (3) and the tail intelligent body (5).

2. The distributed multi-agent collaborative coal handling control system for long-distance scraper conveyors according to claim 1 is characterized in that: The head assembly comprises: Head bottom plate (103); A machine head frame (102), the machine head frame (102) being mounted on a machine head bottom plate (103), the front end of the machine head frame (102) being hingedly connected to the end of the cross side unloading portion (2) via a machine head rotating shaft (101), a sprocket wheel I (9.1) being arranged in the machine head frame (102), the sprocket wheel I (9.1) being mounted on a sprocket wheel shaft I (8.1), and the sprocket wheel shaft I (8.1) being driven by a first permanent magnet motor (10.1); A machine head tensioning hydraulic cylinder (104), wherein the machine head tensioning hydraulic cylinder (104) is connected between the machine head bottom plate (103) and the rear part of the machine head frame (102).

3. The distributed multi-agent collaborative coal handling control system for long-distance scraper conveyors according to claim 1 is characterized in that: The middle sprocket assembly comprises: A supporting bottom plate (405), with supporting side plates (406) being arranged on both sides of the supporting bottom plate (405); Lifting hydraulic cylinders (404), wherein two groups of the lifting hydraulic cylinders (404) are arranged symmetrically on both sides of the supporting bottom plate (405); An intermediate sprocket shaft (403), a sprocket (9.2) being arranged on the intermediate sprocket shaft (403), the intermediate sprocket shaft (403) being installed between two lifting hydraulic cylinders (404), coal retaining plates (407) being arranged on both sides of the intermediate sprocket shaft (403), and the intermediate sprocket shaft (403) being driven by a second permanent magnet motor (10.2); An intermediate transition frame (402) is fixedly connected to the front and rear ends of the support bottom plate (405) and is respectively connected to the coal conveying middle parts (3) at the two ends.

4. The distributed multi-agent collaborative coal handling control system for long-distance scraper conveyors according to claim 3 is characterized in that: A first movable middle plate (408) is symmetrically installed on both sides of the intermediate sprocket shaft (403), one end of the first movable middle plate (408) is hingedly connected to the intermediate transition frame (402), and the other end of the first movable middle plate (408) is rotationally connected to the second movable middle plate (409) via the intermediate secondary rotating shaft (412), the outer surface of the second movable middle plate (409) is in tangential contact with the sprocket shaft, and a coal blocking hydraulic cylinder (410) is connected between the second movable middle plate (409) and the intermediate transition frame (402), and the clamping force provided by the coal blocking hydraulic cylinder (410) keeps the second movable middle plate (408) and the intermediate sprocket shaft in a tangential contact state, thereby preventing the transported coal from falling to the lower side of the scraper conveyor.

5. The distributed multi-agent collaborative coal handling control system for long-distance scraper conveyors according to claim 4 is characterized in that: A plurality of U-shaped grooves are arranged in the middle of the primary movable middle plate (408), the U-shaped grooves are used to allow the protruding portion of the middle sprocket shaft (403) to pass through, and a derailleur (414) is arranged at the U-shaped grooves.

6. The distributed multi-agent collaborative coal handling control system for long-distance scraper conveyors according to claim 1 is characterized in that: The tail assembly comprises: Tail bottom plate (503); A tail frame (502), the tail frame (502) being mounted on a tail bottom plate (503), a sprocket III (9.3) being arranged in the tail frame (502), the sprocket III (9.3) being mounted on a sprocket shaft III (8.3), and the sprocket shaft III (8.3) being driven by a third permanent magnet motor (10.3); A tail transition frame (505), one end of which is connected to the coal conveying middle part (3), and the other end of which is hingedly connected to the tail frame (502) via a tail rotating shaft (501); A tail tensioning hydraulic cylinder (504), wherein the tail tensioning hydraulic cylinder (504) is connected between the tail frame (502) and the tail bottom plate (503).

7. A control method for a distributed multi-agent collaborative coal handling control system for a long-distance scraper conveyor according to claim 1, characterized in that: include: S1: The industrial computer receives the current value In of the first permanent magnet motor (10.1), the second permanent magnet motor (10.2) and the third permanent magnet motor (10.3) and the tension sensor Fn signal in the scraper; S2: The industrial computer controls the power of multiple motors based on the collected current signals, so that the power output between the head intelligent body (1), the middle intelligent body (4) and the tail intelligent body (5) is balanced, that is, the current values ​​of the first permanent magnet motor (10.1), the second permanent magnet motor (10.2) and the third permanent magnet motor (10.3) are the same, and the drive of the scraper conveyor is globally controlled to ensure the smooth operation of the chain; S3: Industrial computer based on the collected tension signal F n, adjust the local tension of the scraper conveyor in sections to ensure that all chain tension values ​​are always greater than 0; S4: Taking the maximum value of the function Y=0.3 Fn+0.4 In+0.3 M as the target, where M is the coal transport volume per hour, and weighs the relationship between the motor power and the tensioning force, the driving power and tensioning force of the first permanent magnet motor (10.1), the second permanent magnet motor (10.2) and the third permanent magnet motor (10.3) are coordinated and controlled to ensure that the coal transport work reaches the optimal state; S5: When the first permanent magnet motor (10.1) or the second permanent magnet motor (10.2) or the third permanent magnet motor (10.3) or the head tensioning hydraulic cylinder (104) or the tail tensioning hydraulic cylinder (504) fails, the control quantity is transferred to the adjacent intelligent body to ensure that the motor power and the tensioning force meet the control requirements.

8. The control method of the distributed multi-agent collaborative coal handling control system for a long-distance scraper conveyor according to claim 7 is characterized in that: In step S2, the process of controlling the power of multiple motors includes: S21: Build strategy network and value network, The input of the strategy network is the current value and the speed of the first permanent magnet motor (10.1), the second permanent magnet motor (10.2) and the third permanent magnet motor (10.3), and the output is the probability distribution function of the speed; The input of the value network is the current value and speed of the first permanent magnet motor (10.1), the second permanent magnet motor (10.2) and the third permanent magnet motor (10.3), and the output is the value of the current state, which is used to evaluate the quality of the current state; Both the policy network and the value network use feedforward neural networks; S22: Establish a three-dimensional model consistent with the real-world device in the software, add input interface and output interface to the three-dimensional model, the input interface is used to input the initial speed, and the output interface is used to output the "current value and speed of the first permanent magnet motor (10.1), the second permanent magnet motor (10.2) and the third permanent magnet motor (10.3)" simulated in real time by the three-dimensional model and the reward value given to the agent after the current agent takes action. According to the formula Calculate reward value; S23: input an initial value in the input interface, the three-dimensional model starts running, and a round of training data is generated in the three-dimensional model; S24: input the training data into the strategy network and the value network for training; S25: Repeat steps S23-S24 continuously, calculate the average reward change of the last 6 training rounds, if the change is less than the set threshold ΔR < 0.01, the strategy is considered to have converged, and then stop training. Through the trained strategy network, input the current value and speed of the first permanent magnet motor (10.1), the second permanent magnet motor (10.2) and the third permanent magnet motor (10.3) collected in real time, and output the speed for operating each motor.

9. The control method of the distributed multi-agent collaborative coal handling control system for a long-distance scraper conveyor according to claim 7 is characterized in that: In step S3, the process of adjusting the local tension of the scraper conveyor in sections includes: S31: Build strategy network and value network, The input of the strategy network is the chain tension and the extension and contraction of the hydraulic cylinder, and the output is the correction value of the extension and contraction of the hydraulic cylinder; The input of the value network is the chain tension and the extension of the hydraulic cylinder, and the output is the value of the current state, which is used to evaluate the quality of the current state; Both the policy network and the value network use feedforward neural networks; S32: Establish a three-dimensional model consistent with the real-world device in the software, and add input and output interfaces to the three-dimensional model. The input interface is used to input the chain tension, and the output interface is used to output the "chain tension and hydraulic cylinder extension and contraction amount" simulated in real time by the three-dimensional model and the reward value given after the current hydraulic cylinder takes action. If Fn>0, the reward is 0; if Fn≤0, the reward is −c; S33: input an initial value in the input interface, the three-dimensional model starts running, and a round of training data is generated in the three-dimensional model; S34: input the training data into the strategy network and the value network for training; S35: Repeat steps S33-S34 continuously. When the system reward value remains at a stable positive value after multiple trainings and the reward value fluctuation range is less than <10%, the system is considered to have reached convergence. Then stop training, input the chain tension and hydraulic cylinder extension amount collected in real time through the trained strategy network, and output the correction value of the hydraulic cylinder extension amount.

10. The control method of the distributed multi-agent collaborative coal handling control system for a long-distance scraper conveyor according to claim 7, characterized in that: In step S4, the process of coordinated control of driving power and tensioning force adjustment includes: S41: Build strategy network and value network, The input of the strategy network is the motor power balance state, chain tension state, coal delivery volume M per hour, and motor speed, and the output is motor speed and adjustment tension; The input of the value network is the motor power balance state, chain tension state, coal delivery volume M per hour, and motor speed. The output is the value of the current state, which is used to evaluate the quality of the current state. Both the policy network and the value network use feedforward neural networks; S42: Establish a three-dimensional model consistent with the real-world device in the software, add input interface and output interface to the three-dimensional model, the input interface is used to input the chain tension Fn, the motor current value In and the coal transport volume M, and the output interface is used to output the "motor speed" simulated in real time by the three-dimensional model. xn, hydraulic cylinder extension ωn" and the reward value given after the current hydraulic cylinder takes action, ; S43: input an initial value in the input interface, the three-dimensional model starts running, and the three-dimensional model outputs a round of training data; S44: input the training data into the strategy network and the value network for training; S45: Repeat steps S43-S44 continuously. When the trade-off between chain tension, motor current, and coal transport capacity reaches a stable state, that is, the reward value growth in the iteration is <10% and the reward value fluctuation is small, stop training, and use the trained strategy network to input the real-time collected chain tension, hydraulic cylinder extension and motor current values, and output the correction value of the hydraulic cylinder extension and motor speed.

Images