Intelligent coal mining scheduling method

Abstract

The invention relates to the technical field of coal mining, in particular to an intelligent coal mining scheduling method. The intelligent coal mining scheduling method comprises the following steps: constructing a behavior prediction model, and training the behavior prediction model to preset precision by using a historical data layer; obtaining a current actual equipment state vector and a control instruction vector; inputting the control instruction vector and the current actual equipment state vector into a behavior prediction model to obtain a predicted equipment state vector; after the control instruction vector is sent to the main execution device, an actual device state vector is obtained; performing time sequence alignment on the actual equipment state vector and the predicted equipment state vector, and calculating to obtain an instant confidence coefficient; and adjusting a scheduling strategy according to the instant confidence coefficient. According to the intelligent coal mining scheduling method provided by the invention, the problem of a control efficiency verification blind area is fundamentally solved, whether the control instruction is accurately executed or not can be directly and instantly sensed, and the comprehensive state change caused after the instruction is executed is effectively evaluated.

Description

Technical Field

[0001] This invention relates to the field of coal mining technology, and more specifically, to a method for intelligent coal mining scheduling. Background Technology

[0002] Intelligent coal mining is a core direction for ensuring energy security and improving mine production efficiency. In modern fully mechanized mining faces, mining equipment with coal mining machines as the core is usually controlled by a centralized scheduling architecture.

[0003] However, the traditional centralized dispatching model suffers from a blind spot in verifying the control effectiveness of coal mining machines. Specifically, after the dispatching system issues control commands to the coal mining machine, the verification of the command execution effect must go through a lengthy loop of sensor acquisition → data transmission → central processing → state assessment before feedback can reach the decision-making end. This indirect verification method, relying on a long feedback loop, leads to severe control loop delays and state perception lags. The dispatching system always makes decisions based on a past, non-real-time system state image, making it difficult to cope with the complex dynamic environment underground. Essentially, there is a significant information gap between command issuance and effect verification, which puts the control process of the coal mining machine in a semi-open-loop uncertain state. This not only restricts further improvements in mining efficiency but also poses a potential threat to safe production at the working face. Summary of the Invention

[0004] The purpose of this invention is to provide a smart coal mining scheduling method to improve the technical problem of blind spots in the control effectiveness verification of the centralized scheduling mode in the prior art.

[0005] The intelligent coal mining scheduling method provided by this invention includes: Construct a behavior prediction model and train the behavior prediction model to a preset accuracy using a historical data layer; Obtain the current actual device state vector and control command vector; The control command vector and the current actual device state vector are input into the behavior prediction model to obtain the predicted device state vector; After sending the control command vector to the main execution device, the actual device state vector is obtained; The actual device state vector and the predicted device state vector are aligned over time, and the instantaneous confidence level is calculated. Adjust the scheduling strategy based on the instantaneous confidence level.

[0006] As one possible implementation, the method further includes: While sending the control command vector to the main execution device, the predicted device state vector is input into the digital twin of the coal mining system to obtain the predicted operating state. After obtaining the actual equipment state vector, the actual equipment state vector is input into the digital twin of the coal mining system to obtain the actual operating state.

[0007] As one possible implementation, the method further includes: If the instantaneous confidence level is above the high threshold, then the confidence level is determined to be high confidence. If the instantaneous confidence level is between the low threshold and the high threshold, then the confidence level is determined to be medium confidence. If the instantaneous confidence level is below a low threshold, then the confidence level is determined to be low confidence.

[0008] As one possible implementation, the method further includes: After determining the confidence level, the status verification report is sent to the scheduling system; If no confirmation signal is received from the scheduling system within the third preset time period, the status verification report is resent to the scheduling system. The status verification report includes the actual device status vector, the instantaneous confidence level, the confidence level, the control command vector, and the command execution timestamp.

[0009] As one possible implementation, adjusting the scheduling strategy based on the instantaneous confidence level includes: If the confidence level is high, the current scheduling strategy is maintained, and the effectiveness evaluation result of this control is recorded. If the confidence level is medium confidence, the sensor sampling frequency and the reporting frequency of the status verification report will be increased, and an early warning will be issued. If the confidence level is low, a step-down shutdown command is issued to the main execution device to start the backup sensor for data cross-validation, record abnormal events and generate a diagnostic report; the abnormal events include: the duration of the instantaneous confidence level below the low threshold exceeds a first preset duration, or the normalization error of the sensing data of the backup sensor exceeds a preset safety threshold and the duration exceeds a second preset duration.

[0010] As one possible implementation, the step of calculating the instantaneous confidence level based on the actual device state vector and the predicted device state vector includes: Calculate the normalization error of each parameter in the predicted device state vector; Based on the normalized error and safety level of each parameter, the weighted comprehensive error is calculated; The instantaneous confidence level is calculated based on the weighted synthesis error.

[0011] As one possible implementation, the normalization error is calculated as follows: NEi=|Ai- Yi| / Ri, where NEi is the normalization error of the i-th parameter, Ai is the actual value of the i-th parameter in the actual device state vector, Yi is the predicted value of the i-th parameter in the predicted device state vector, and Ri is the reference value of the i-th parameter. The formula for calculating the weighted comprehensive error is: TE=∑(ωi×NEi), where ωi is the weight assigned to the security level of the i-th parameter; The formula for calculating the instantaneous confidence level is: IC = e^(-λ×TE).

[0012] As one possible implementation, the behavior prediction model employs a fully connected neural network containing one hidden layer; The behavior prediction model calculates the hidden layer features using the formula H=σ(W×[U,X]+b), and then generates the predicted device state vector Y using the formula Y=(Wout×H)+bout; where U is the control command vector and X is the current actual device state vector.

[0013] As one possible implementation, the control command vector includes traction speed command, drum rotation speed command, and / or cutting height command.

[0014] As one possible implementation, the values ​​of each parameter in the current actual equipment state vector and the values ​​of each parameter in the actual equipment state vector are both obtained by sensors in the coal mining system, and the parameters in the predicted equipment state vector correspond one-to-one with the parameters in the actual equipment state vector.

[0015] The sensors in the coal mining system include equipment status sensors and environmental parameter sensors. The equipment status sensors include several sensors, which are used to monitor the current of the cutting motor, the vibration amplitude of the rocker arm, and the temperature of the gearbox. The environmental parameter sensors include several sensors, which are used to monitor the gas concentration at the working face, the dust concentration in the roadway, and the resistance of the hydraulic support.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: The behavioral prediction model can perform localized "sandbox simulation" after inputting control command vectors and current actual equipment state vectors. It can generate predicted equipment state vectors before command execution and obtain actual equipment state vectors after command execution. It can compare the actual equipment state vectors with the predicted equipment state vectors in real time and calculate the instantaneous confidence level. This allows the scheduling system to perceive the command execution effect in near real time and realize instantaneous verification of the execution effect. On this basis, differentiated scheduling strategies can be triggered according to the instantaneous confidence level, realizing a paradigm shift from "lagging control" to "proactive and precise control". It transforms the traditional "semi-open-loop" uncertain control into a complete, fast, and precise "closed-loop" intelligent control from command prediction, instantaneous effect verification to dynamic strategy adjustment. It fundamentally overcomes the problem of blind spots in control effectiveness verification and can directly and instantly perceive whether control commands are executed accurately. It can effectively evaluate the comprehensive state changes caused by command execution, such as changes in coal seam thickness and geological structural disturbances. At the same time, it can also improve mining efficiency and safety, promptly detect mechanical failures caused by abnormal equipment operation, and provide early warning of safety accident risks. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 This is a first schematic flowchart of the intelligent coal mining scheduling method provided in an embodiment of the present invention; Figure 2 This is a second schematic flowchart of the intelligent coal mining scheduling method provided in an embodiment of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0021] Figure 1 This is a schematic flowchart of a coal intelligent mining scheduling method provided in an embodiment of the present invention. The method includes: S102, Construct a behavior prediction model and train the behavior prediction model to a preset accuracy using historical data layer; S104, obtain the current actual device state vector and control command vector; S106, Input the control command vector and the current actual device state vector into the behavior prediction model to obtain the predicted device state vector; S108, after sending the control command vector to the main execution device, obtain the actual device status vector; S110, perform time series alignment between the actual device state vector and the predicted device state vector, and calculate the instantaneous confidence level; S112, adjust the scheduling strategy based on the real-time confidence level.

[0022] The behavioral prediction model can perform localized "sandbox simulation" after inputting control command vectors and current actual equipment state vectors. It can generate predicted equipment state vectors before command execution and obtain actual equipment state vectors after command execution. It can compare the actual equipment state vectors with the predicted equipment state vectors in real time and calculate the instantaneous confidence level. This allows the scheduling system to perceive the command execution effect in near real time and realize instantaneous verification of the execution effect. On this basis, differentiated scheduling strategies can be triggered according to the instantaneous confidence level, realizing a paradigm shift from "lagging control" to "proactive and precise control". It transforms the traditional "semi-open-loop" uncertain control into a complete, fast, and precise "closed-loop" intelligent control from command prediction, instantaneous effect verification to dynamic strategy adjustment. It fundamentally overcomes the problem of blind spots in control effectiveness verification and can directly and instantly perceive whether control commands are executed accurately. It can effectively evaluate the comprehensive state changes caused by command execution, such as changes in coal seam thickness and geological structural disturbances. At the same time, it can also improve mining efficiency and safety, promptly detect mechanical failures caused by abnormal equipment operation, and provide early warning of safety accident risks.

[0023] During the training phase of the behavior prediction model, 100,000 sets of historical operating data from the device over the past 6 months can be selected as samples. Each set of samples contains control commands and the corresponding actual device status. The predicted output is calculated through forward propagation, and then the network weight matrix and bias vector are iteratively adjusted using the backpropagation algorithm. The preset accuracy is set to the average absolute error between the predicted value and the actual value ≤ 5%. After 200 iterations, the model meets the standard, and then the behavior prediction model is deployed locally to the edge control cluster, which consists of the main execution device and sensors.

[0024] Once the model is deployed, the scheduling system sends the aforementioned control command vectors to the edge control clusters through the industrial ring network.

[0025] The method provided in this embodiment may further include the following steps: while sending the control command vector to the main execution device, the predicted device state vector is input into the digital twin of the coal mining system to update the state of the corresponding virtual device in the digital twin and obtain the predicted operating state. By synchronizing the predicted device state vector to the digital twin in advance, a pre-simulation image of the future system state can be obtained. This makes the system decision no longer solely dependent on lagging actual data, but combines predictions of future development based on the current state. Thus, when facing dynamic working conditions such as changes in coal seam thickness and geological structural disturbances, more forward-looking and adaptive scheduling decisions can be made, ensuring the smooth and efficient operation of the mining process.

[0026] The method provided in this embodiment may further include the following steps: after obtaining the actual equipment state vector, the actual equipment state vector is input into the digital twin of the coal mining system to correct the digital twin and obtain the actual operating state. In this way, the digital twin can be synchronized with the real state of the physical world in a theoretical sense. This correction action provides real and accurate data feedback for iterative optimization of the behavior prediction model based on historical actual data.

[0027] Preferably, the confidence level can be determined based on the specific value of the instantaneous confidence level, as follows: A. If the instantaneous confidence level is above the high threshold, then the confidence level is determined to be high confidence. B. If the instantaneous confidence level is between the low and high thresholds, then the confidence level is determined to be medium confidence. C. If the instantaneous confidence level is below the low threshold, then the confidence level is determined to be low confidence.

[0028] The high threshold and low threshold mentioned above are both preset values. Specifically, the high threshold can be 0.8 and the low threshold can be 0.5.

[0029] After determining the confidence level, the status verification report is sent to the scheduling system. If no acceptance confirmation signal is received from the scheduling system within the third preset time period, the status verification report is resent to the scheduling system for subsequent retrieval. The status verification report includes the actual device status vector, real-time confidence level, confidence level, control command vector, and command execution timestamp (i.e., the start time of the main execution device executing the control command, which is recorded when the main execution device executes the control command).

[0030] Different scheduling strategies can also be triggered based on different confidence levels, as follows: A. If the confidence level is high, it indicates that the current equipment is operating stably and the command execution is accurate. In this case, the current scheduling strategy can be maintained, and the effectiveness evaluation result of this control should be recorded (marked as "control command execution is accurate, equipment and environment status is stable, confidence level a, meets expectations"), to accumulate data for subsequent scheduling optimization. Here, 'a' is the specific value of the instantaneous confidence level calculated in step S110.

[0031] B. If the confidence level is medium, the scheduling system will immediately activate the monitoring mechanism: increase the sensor sampling frequency and reporting frequency, and issue an early warning. Specifically, the sensor sampling frequency will be increased to 20Hz (from 10Hz), the status verification report reporting frequency will be increased to once every 2 seconds (from once every 5 seconds), and an early warning message "There is a slight deviation in the equipment's operating status; please monitor" will be displayed on the operation interface to remind maintenance personnel to pay attention to the equipment's dynamics.

[0032] C. If the confidence level is low, a gradual downgrade shutdown command can be issued to the main execution device to activate the backup sensor for data cross-validation, record abnormal events, and generate a diagnostic report. Abnormal events include: the duration of the instantaneous confidence level below the low threshold exceeding a first preset duration; or, the normalization error of the backup sensor's sensing data exceeding a preset safety threshold (which can be 0.1) and lasting for more than a second preset duration (which can be 10 seconds). After receiving the gradual downgrade shutdown command, the main execution device can gradually reduce the traction speed to 0, and the drum speed will decrease synchronously to ensure a smooth shutdown within 30 seconds. The diagnostic report can be pushed to the operation interface to prompt maintenance personnel for timely repairs.

[0033] This embodiment, through edge-side prediction and verification, combined with central-side twin synchronization and confidence-driven decision-making, can provide early warning of potential faults and safety accidents caused by equipment anomalies (such as vibration and temperature rise) or environmental deterioration (such as excessive gas and dust concentrations). It also enables proactive safety intervention for equipment such as coal mining machines, enhancing the safety resilience of the coal mining process.

[0034] The above step S110 may specifically include the following steps: calculating the normalized error of each parameter in the predicted device state vector; calculating the weighted comprehensive error based on the normalized error of each parameter and the safety level, and realizing weighted fusion; and calculating the instantaneous confidence level based on the weighted comprehensive error.

[0035] The formula for calculating the normalization error is: NEi=|Ai- Yi| / Ri, where NEi is the normalization error of the i-th parameter, Ai is the actual value of the i-th parameter in the actual device state vector, Yi is the predicted value of the i-th parameter in the predicted device state vector, and Ri is the reference value of the i-th parameter.

[0036] The formula for calculating the weighted comprehensive error is: TE=∑(ωi×NEi), where ωi is the weight assigned to the safety level of the i-th parameter; The formula for calculating instant confidence level is: IC = e^(-λ×TE).

[0037] In this way, an instantaneous confidence level between 0 and 1 can be generated through nonlinear mapping, providing a unified and quantitative standard for evaluating the control performance of the system.

[0038] To ensure that the data is compatible with the model input format, the edge control cluster first preprocesses U and X: parameters such as traction speed and drum speed are uniformly converted into standardized numerical formats, and then all parameters are mapped to the [0,1] interval through min-max normalization.

[0039] Specifically, the behavior prediction model employs a fully connected neural network with one hidden layer. The model calculates the hidden layer features using the formula H = σ(W × [U, X] + b), and then generates the predicted device state vector Y using the formula Y = (Wout × H) + bout. Here, U is the control command vector, and X is the current actual device state vector. Finally, the edge control cluster encapsulates U and Y into a command-prediction data packet and reports it to the scheduling system.

[0040] The number of input layer nodes in the fully connected neural network can be set to 9 (3 control parameters + 6 state parameters), the number of hidden layer nodes is 12, and the number of output layer nodes is 6.

[0041] Specifically, the aforementioned control command vector includes at least one of the following: traction speed command, drum speed command, and cutting height command.

[0042] Specifically, the values ​​of each parameter in the current actual equipment state vector and the values ​​of each parameter in the actual equipment state vector are both obtained by sensors in the coal mining system, and the parameters in the predicted equipment state vector correspond one-to-one with the parameters in the actual equipment state vector.

[0043] The sensors in the coal mining system include equipment status sensors and environmental parameter sensors. Several equipment status sensors monitor parameters such as the cutting motor current, rocker arm vibration amplitude, and gearbox temperature, enabling data acquisition of equipment operating status. Several environmental parameter sensors monitor parameters such as the working face gas concentration, roadway dust concentration, and hydraulic support resistance, enabling data acquisition of the working environment. Based on this, the current actual equipment status vector may include parameters such as the cutting motor current, rocker arm vibration amplitude, gearbox temperature, working face gas concentration, roadway dust concentration, and hydraulic support resistance. The predicted equipment status vector contains predicted parameters corresponding to the current actual equipment status vector.

[0044] This embodiment integrates multi-sensor perception, behavior prediction models, and digital twin technology to construct a fully closed-loop intelligent scheduling process from instruction issuance to strategy adjustment. The main execution device and its associated sensors can be physically and logically integrated into an autonomous unit.

[0045] Specifically, the behavior prediction model can be deployed to the corresponding edge control cluster, which consists of the main execution device (such as the coal mining machine body), equipment status sensors and environmental parameter sensors, and can synchronously collect equipment operating status and working environment data.

[0046] In this architecture, various sensors distributed across the coal mining machine and the working face continuously collect data on the equipment's operating status and environmental conditions, transmitting this data to the ground-based dispatching system via an industrial ring network or wireless network. The dispatching system then fuses and analyzes the multi-source data to generate control commands for the coal mining machine, which are then issued to the machine for execution.

[0047] This embodiment realizes real-time perception of command execution effect, quantitative assessment of equipment and environmental risks, and intelligent adaptation of scheduling strategies, effectively solving the delay problem of traditional centralized scheduling and providing technical support for the safety and efficiency of coal mining.

[0048] The technical solution of the present invention will be described in detail below through specific embodiments.

[0049] The current actual equipment state vector X is [350A (cutting motor current), 0.4mm (rocker arm vibration amplitude), 75℃ (gearbox temperature), 0.6% (working face gas concentration), 8mg / m³ (roadway dust concentration), 28MPa (hydraulic support resistance)]. The control command vector U issued by the dispatching system is [5m / min (traction speed), 30r / s (drum speed), 3.5m (cutting height)]. After preprocessing, U' is [0.5, 0.6, 0.7] and X' is [0.35, 0.4, 0.75, 0.6, 0.4, 0.56].

[0050] After preprocessing, the vector [U',X'] is input into the behavior prediction model of the edge control cluster, and the model is calculated through the internal fully connected neural network layer. Assume the preset input-to-hidden layer weight matrix W is a 9×12 matrix (example elements: W[0][0]=0.12, W[1][0]=0.08,...,W[8]

[11] =0.05), the bias vector b is a 12-dimensional vector (b=[0.03,0.05,...,0.02]), and the activation function is the sigmoid function. According to the formula H=ơ(W×[U,X]+b), the hidden layer feature vector H (12 dimensions) can be calculated, with example values ​​of [0.62,0.58,0.71,0.49,0.65,0.53,0.74,0.61,0.55,0.68,0.52,0.63]. Subsequently, the hidden layer feature vector H is transformed by the output layer to generate the predicted device state vector Y. If the output layer weight matrix Wout is a 12×6 matrix (example elements: Wout[0][0]=0.21, Wout[1][0]=0.18,...,Wout

[11] [5]=0.16), and the output layer bias vector bout is a 6-dimensional vector (bout=[0.04,0.06,0.03,0.05,0.02,0.07]), according to the formula Y=(Wout×H)+bout, the final value of Y is [348A,0.42mm,76℃,0.65%,8.5mg / m³,28.5MPa]. The edge control cluster associates U and Y and encapsulates them into an instruction-prediction data packet, which is then sent to the coal mining machine scheduling system via a wireless network.

[0051] According to the preset parameter mapping relationship, the scheduling system writes each parameter value in Y into the corresponding data field of the virtual equipment model one by one—for example, the predicted cutting motor current of 348A is written into the "Motor Operation Parameter - Current" field, and the predicted gas concentration of 0.65% is written into the "Environmental Parameter - Gas Concentration" field, ensuring that each parameter corresponds one-to-one with the field of the virtual model. At the same time, based on the update timestamp of Y "2024-05-2010:00:03", the scheduling system sets a valid state time stamp (1 minute, that is, the predicted state is valid until 10:01:03) in the virtual equipment model, and marks the current state of the virtual model as "predicted operating state".

[0052] Almost simultaneously with the update of the digital twin by the scheduling system, the edge control cluster, after sending the predicted equipment state vector Y, immediately converted the control command vector U into an executable electrical signal. This signal then drove the coal mining machine to perform the following operations via the equipment control interface: the traction system was adjusted to 5 m / min, the drum speed was stabilized at 30 r / s, and the cutting mechanism was positioned at a height of 3.5 m. During execution, the edge control cluster continuously monitored the coal mining machine's readiness status. When it detected that the rocker arm vibration amplitude was ≤0.5 mm and the gearbox temperature was ≤80℃ (meeting the equipment's safe operating threshold), it determined that the equipment was in an operable state and started the command execution timer at 10:00:05, recording the start time of command execution. Subsequently, it tracked the equipment's response status in real time. At 10:00:07, it detected that the coal mining machine's traction system had started and the drum had begun to rotate, and immediately updated the local equipment status record to "command execution in progress."

[0053] Based on the recorded instruction execution start time (10:00:05) and the "instruction execution" status, the edge control cluster determined the data acquisition start time to be after the equipment started executing the instruction (10:00:07), set the sampling frequency to 10Hz and the acquisition period to 5 seconds, and collected a total of 50 sets of actual operating data. Through the equipment status sensor and environmental parameter sensor, the actual operating parameters of the coal mining machine were obtained—cutting motor current 352A, rocker arm vibration amplitude 0.45mm, gearbox temperature 77℃, and operating environment parameters—working face gas concentration 0.7%, roadway dust concentration 9mg / m³, and hydraulic support resistance 29MPa. Then, according to the same parameter dimension format as the predicted equipment state vector Y, these actual data were organized into the actual equipment state vector A=[352A,0.45mm,77℃,0.7%,9mg / m³,29MPa].

[0054] The edge control cluster retrieves the predicted device state vector Y corresponding to the current control command U from local storage. Based on the command execution start time of 10:00:05, it aligns A and Y in time series (ensuring comparison within the same time dimension), and then calculates the normalized error of each parameter. According to coal mine safety regulations and equipment technical specifications, the following reference values ​​Ri are preset for each parameter: cutting motor current Ri = 400A (equipment rated current), rocker arm vibration amplitude Ri = 1.0mm (maximum allowable vibration value), gearbox temperature Ri = 100℃ (equipment maximum withstand temperature), working face gas concentration Ri = 1.0% (safety concentration upper limit), roadway dust concentration Ri = 20mg / m³ (hygiene standard upper limit), and hydraulic support resistance Ri = 40MPa (support rated resistance).

[0055] Based on the normalized error formula NEi=|Ai-Yi| / Ri, the errors of each parameter are calculated as follows: cutting motor current NE1=|352-348| / 400=0.01, rocker arm vibration amplitude NE2=|0.45-0.42| / 1.0=0.03, gearbox temperature NE3=|77-76| / 100=0.01, working face gas concentration NE4=|0.7-0.65| / 1.0=0.05, roadway dust concentration NE5=|9-8.5| / 20=0.025, hydraulic support resistance NE6=|29-28.5| / 40=0.0125. These error values ​​are combined to form a normalized error vector NE=[0.01,0.03,0.01,0.05,0.025,0.0125]. Finally, the edge control cluster associates and stores the actual device state vector A, the predicted device state vector Y, the normalized error vector NE, and the instruction execution time information (10:00:05-10:00:12) in the local database to provide data support for subsequent confidence calculation.

[0056] In the confidence calculation stage, the edge control cluster first establishes a confidence calculation model, using the normalized error vector NE and the preset weight vector ω as inputs, and outputs the instantaneous confidence IC value in the 0-1 interval. The allocation of the weight vector ω follows the safety priority principle: safety-related parameters have higher weights than performance parameters, and environmental safety parameters have higher weights than equipment status parameters. Finally, ω is set as [0.15 (cutting motor current), 0.1 (rocker arm vibration amplitude), 0.1 (gearbox temperature), 0.3 (working face gas concentration), 0.2 (tunnel dust concentration), 0.15 (hydraulic support resistance)], and satisfies ∑ωi=1.

[0057] The confidence model adopts a two-level processing architecture: The first level performs parameter importance weighting, and calculates the weighted comprehensive error according to the formula TE=∑(ωi×NEi). Substituting the data, we get TE=0.15×0.01+0.1×0.03+0.1×0.01+0.3×0.05+0.2×0.025+0.15×0.0125=0.0015+0.003+0.001+0.015+0.005+0.001875=0.027375; The second level maps the weighted comprehensive error to confidence through the nonlinear transformation function IC=e(-λ×TE). With the sensitivity adjustment coefficient λ=2, we calculate IC=e(-2×0.027375)=e(-0.05475)≈0.947.

[0058] Compare the instantaneous confidence level IC with a preset threshold (high confidence threshold θ). high =0.8, medium confidence threshold θ mid =0.5), since 0.947≥0.8, the current operating status is determined to be high confidence. The edge control cluster packages the actual equipment status vector A, the instantaneous confidence IC=0.947, the confidence level "high", the control command vector U, and the command execution timestamp "2024-05-2010:00:05" into a status verification report and reports it to the coal mining machine dispatch system. After receiving the report at 10:00:15, the dispatch system immediately returns a reception confirmation signal. After receiving the signal, the edge control cluster updates its local status record to "data reporting completed". If no confirmation signal is received within the preset time (10 seconds), the retransmission mechanism is activated to resend the report.

[0059] After receiving the status verification report, the coal mining machine scheduling system first extracts the actual equipment status vector A and the instantaneous confidence level IC=0.947. Then, based on the equipment identifier "coal mining machine-01", it locates the corresponding virtual equipment model in the global digital twin. According to the preset mapping relationship, the scheduling system replaces the predicted parameters in the virtual model one by one with the actual parameters in A—for example, replacing 348A (cutting motor current) with 352A and replacing 0.42mm (rocker arm vibration amplitude) with 0.45mm, thus completing the status correction of the global digital twin. At the same time, based on the update timestamp of A "2024-05-20 10:00:12", the system updates the status validity time stamp of the virtual equipment model (to 10:01:12) to ensure that the digital twin and the physical equipment status are synchronized in real time.

[0060] Subsequently, the scheduling system analyzes the instantaneous confidence level (IC) and its "high confidence" level, and matches it with the preset scheduling strategy rule base. Since IC ≥ θ high =0.8, the system determines that the current equipment is operating stably and the command is executed accurately, and decides to maintain the current scheduling strategy - the coal mining machine continues to operate at a traction speed of 5m / min, a drum speed of 30r / s, and a cutting height of 3.5m. At the same time, the effectiveness evaluation result of this control is recorded (marked "the control command is executed accurately, the equipment and environment are stable, the confidence level is 0.947, which meets the expectations"), to accumulate data for subsequent scheduling optimization.

[0061] If the IC calculated in this case is 0.65 (at θ) mid =0.5 and θ high If IC = 0.8 (below θ), the scheduling system will immediately activate the auxiliary monitoring mechanism: increasing the sensor sampling frequency to 20Hz (from 10Hz), increasing the status verification report reporting frequency to once every 2 seconds (from once every 5 seconds), and simultaneously displaying a warning message on the operation interface: "The equipment operating status has a slight deviation; please pay attention to monitoring," reminding maintenance personnel to monitor the equipment's dynamics; if IC = 0.4 (below θ), the ... mid If the value is 0.5 (which is considered low confidence), a safety intervention strategy is triggered: a gradual downgrade shutdown command is issued to the edge control cluster, the traction speed is gradually reduced from 5 m / min to 0, and the drum speed is reduced synchronously to ensure a smooth shutdown within 30 seconds; at the same time, the backup gas sensor and dust sensor are activated to cross-validate the data. If the gas concentration normalization error is detected to exceed 0.1 (preset safety threshold) for 10 seconds, the abnormal event is automatically recorded and a diagnostic report is generated and pushed to the operation interface to prompt maintenance personnel to carry out timely repairs.

[0062] In this scheduling process, the coal mining machine scheduling system feeds back the execution effect of "maintaining the current scheduling strategy" to the edge control cluster at 10:00:20. The edge control cluster updates its local record to "scheduling process closed loop completed", and the process ends.

[0063] Figure 2 A schematic flowchart of a coal intelligent mining scheduling method provided in an embodiment of the present invention, the method comprising: S201, Begin; S202, Build a behavior prediction model and deploy it to the edge control cluster; S203, issue control command vector; S204, Generate the predicted device state vector; S205, drives the physical device to execute instructions; S206, Collect the actual device status vector; S207, calculate the normalized error and synthesize the confidence level; whereby the confidence level can be regarded as the above-mentioned instantaneous confidence level.

[0064] S208, determine whether the confidence level meets the standard; S209, if so, then correct the digital twin and complete the scheduling; S210, if not, then trigger the scheduling control strategy; S211, End.

[0065] An embodiment of the present invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the above-described intelligent coal mining scheduling method embodiment and achieves the same technical effects. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0066] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by computer-controlled devices. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. The storage medium can be a memory, a disk, an optical disk, etc.

[0067] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

[0068] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0069] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for intelligent coal mining scheduling, characterized in that, include: Construct a behavior prediction model and train the behavior prediction model to a preset accuracy using a historical data layer; Obtain the current actual device state vector and control command vector; The control command vector and the current actual device state vector are input into the behavior prediction model to obtain the predicted device state vector; After sending the control command vector to the main execution device, the actual device state vector is obtained; The actual device state vector and the predicted device state vector are aligned over time, and the instantaneous confidence level is calculated. Adjust the scheduling strategy based on the instantaneous confidence level.

2. The intelligent coal mining scheduling method according to claim 1, characterized in that, The method further includes: While sending the control command vector to the main execution device, the predicted device state vector is input into the digital twin of the coal mining system to obtain the predicted operating state. After obtaining the actual equipment state vector, the actual equipment state vector is input into the digital twin of the coal mining system to obtain the actual operating state.

3. The intelligent coal mining scheduling method according to claim 1, characterized in that, The method further includes: If the instantaneous confidence level is above the high threshold, then the confidence level is determined to be high confidence. If the instantaneous confidence level is between the low threshold and the high threshold, then the confidence level is determined to be medium confidence. If the instantaneous confidence level is below a low threshold, then the confidence level is determined to be low confidence.

4. The intelligent coal mining scheduling method according to claim 3, characterized in that, The method further includes: After determining the confidence level, the status verification report is sent to the scheduling system; If no confirmation signal is received from the scheduling system within the third preset time period, the status verification report is resent to the scheduling system. The status verification report includes the actual device status vector, the instantaneous confidence level, the confidence level, the control command vector, and the command execution timestamp.

5. The intelligent coal mining scheduling method according to claim 4, characterized in that, The step of adjusting the scheduling strategy based on the instantaneous confidence level includes: If the confidence level is high, the current scheduling strategy is maintained, and the effectiveness evaluation result of this control is recorded. If the confidence level is medium confidence, the sensor sampling frequency and the reporting frequency of the status verification report will be increased, and an early warning will be issued. If the confidence level is low, a step-down shutdown command is issued to the main execution device to start the backup sensor for data cross-validation, record abnormal events and generate a diagnostic report; the abnormal events include: the duration of the instantaneous confidence level below the low threshold exceeds a first preset duration, or the normalization error of the sensing data of the backup sensor exceeds a preset safety threshold and the duration exceeds a second preset duration.

6. The intelligent coal mining scheduling method according to claim 1, characterized in that, The step of calculating the instantaneous confidence level based on the actual device state vector and the predicted device state vector includes: Calculate the normalization error of each parameter in the predicted device state vector; Based on the normalized error and safety level of each parameter, the weighted comprehensive error is calculated; The instantaneous confidence level is calculated based on the weighted synthesis error.

7. The intelligent coal mining scheduling method according to claim 6, characterized in that, The formula for calculating the normalization error is: NEi=|Ai- Yi| / Ri, where NEi is the normalization error of the i-th parameter, Ai is the actual value of the i-th parameter in the actual device state vector, Yi is the predicted value of the i-th parameter in the predicted device state vector, and Ri is the reference value of the i-th parameter. The formula for calculating the weighted comprehensive error is: TE=∑(ωi×NEi), where ωi is the weight assigned to the security level of the i-th parameter; The formula for calculating the instantaneous confidence level is: IC = e^(-λ×TE).

8. The intelligent coal mining scheduling method according to claim 1, characterized in that, The behavior prediction model employs a fully connected neural network with one hidden layer. The behavior prediction model calculates the hidden layer features using the formula H=σ(W×[U,X]+b), and then generates the predicted device state vector Y using the formula Y=(Wout×H)+bout; where U is the control command vector and X is the current actual device state vector.

9. The intelligent coal mining scheduling method according to claim 1, characterized in that, The control command vector includes traction speed command, drum rotation speed command, and / or cutting height command.

10. The intelligent coal mining scheduling method according to any one of claims 1-9, characterized in that, The values ​​of each parameter in the current actual equipment state vector and the values ​​of each parameter in the actual equipment state vector are both obtained by the sensors in the coal mining system. The parameters in the predicted equipment state vector correspond one-to-one with the parameters in the actual equipment state vector. The sensors in the coal mining system include equipment status sensors and environmental parameter sensors. The equipment status sensors include several sensors, which are used to monitor the current of the cutting motor, the vibration amplitude of the rocker arm, and the temperature of the gearbox. The environmental parameter sensors include several sensors, which are used to monitor the gas concentration at the working face, the dust concentration in the roadway, and the resistance of the hydraulic support.

Images