Multi-parameter fusion dynamic optimization method for SDC industrial control parameters

Abstract

This application discloses a multi-parameter fusion-based dynamic optimization method for SDC industrial control parameters, belonging to the field of parameter optimization technology. The method includes: predicting the predicted coal seam dip angle sequence and the predicted coal mining speed sequence within a future preset time zone; predicting the roof pressure of the working face and obtaining the predicted roof pressure field sequence; analyzing and determining the suitable safe time zone and the suitable efficiency time zone; optimizing the hydraulic support parameters of the hydraulic support group, outputting a queue of adapted support parameter commands, and executing hydraulic support within the preset time zone. This solves the technical problems of existing hydraulic support parameters being unable to respond in real time to complex and changing underground environments, leading to insufficient roof pressure prediction accuracy, untimely support, insufficient initial support force, or excessive pressure relief.

Description

Technical Field

[0001] This application relates to the field of parameter optimization technology, specifically to a dynamic optimization method for SDC industrial control parameters based on multi-parameter fusion. Background Technology

[0002] In coal mining operations, hydraulic support groups are key equipment for ensuring the safety and efficient production of the working face. The support parameters affect the safety of roof management, coal mining efficiency, and the energy consumption and lifespan of the equipment.

[0003] In existing technologies, the setting of hydraulic support parameters largely relies on empirical values ​​or static models, making it difficult to respond in real time to complex and ever-changing downhole environments, resulting in insufficient accuracy in predicting roof pressure. Furthermore, during parameter optimization, the failure to dynamically define safety and efficiency time zones based on predicted future working conditions and to use these as rigid constraints for dynamic parameter optimization makes it difficult to achieve globally optimal production efficiency and energy consumption while ensuring safety. This leads to poor adaptability of hydraulic support parameters to actual working conditions, easily resulting in problems such as untimely support, insufficient initial support force, or excessive pressure relief, affecting both safety and efficiency. Summary of the Invention

[0004] This application provides a dynamic optimization method for SDC industrial control parameters by integrating multiple parameters, which solves the technical problems that existing hydraulic support parameters cannot respond in real time to complex and ever-changing downhole environments, resulting in insufficient accuracy in roof pressure prediction, untimely support, insufficient initial support force, or excessive pressure relief.

[0005] The technical solution to the above-mentioned technical problems in this application is as follows:

[0006] On the one hand, this application provides a method for dynamic optimization of SDC industrial control parameters through multi-parameter fusion, the method comprising:

[0007] Based on the coal mining site perception data obtained from multi-source sensor monitoring, the predicted coal seam dip angle sequence and predicted coal mining speed sequence are obtained within a future preset time zone.

[0008] Based on the predicted coal seam dip angle sequence and the predicted coal mining speed sequence, the roof pressure of the working face is predicted, and the predicted roof pressure field sequence is obtained.

[0009] Based on the analysis of the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence, the appropriate safe time zone and the appropriate efficiency time zone are determined.

[0010] Using the adapted safety time zone and the adapted efficiency time zone as rigid constraints, the hydraulic support parameters of the hydraulic support group are optimized based on the predicted coal seam dip angle sequence, the predicted coal mining speed sequence and the predicted roof pressure field sequence. The adapted support parameter instruction queue within the preset time zone is output, and the hydraulic support within the preset time zone is executed.

[0011] This application provides one or more technical solutions, which have at least the following technical effects or advantages:

[0012] This application provides a multi-parameter fusion-based dynamic optimization method for SDC industrial control parameters. First, based on coal mining site sensing data acquired from multi-source sensor monitoring, it captures real-time dynamic changes in the complex underground environment, providing raw data support for subsequent parameter prediction and optimization, thereby improving prediction accuracy. Second, using pre-constructed coal seam dip angle predictors and coal mining speed predictors, it predicts the trends of coal seam dip angle and coal mining speed changes within a preset time zone based on historical multi-source sensing data sequences. This better adapts to the complexity and variability of the underground environment, improving prediction accuracy. Third, it analyzes and determines suitable safety and efficiency time zones, effectively avoiding safety accidents caused by untimely support. The efficiency time zone can be dynamically optimized based on the stability and pressure distribution uniformity of actual working conditions, ensuring that production efficiency is maximized under safe conditions, avoiding inefficiency or resource waste that may result from static efficiency settings. Finally, the hydraulic support parameters of the hydraulic support group are optimized by using the adaptation of the safe time zone and the adaptation of the efficiency time zone as rigid constraints. Through iterative optimization, the support parameter instruction queue corresponding to the maximum support adaptation index is output, realizing the global optimal balance of support parameters between safety, production efficiency and energy consumption.

[0013] Through the above technical solution, this application realizes the real-time response of hydraulic support parameters to the complex and ever-changing underground environment based on multi-source sensing data fusion and dynamic parameter optimization. It improves the accuracy of roof pressure prediction, avoids problems such as untimely support, insufficient initial support force or excessive pressure relief, ensures that coal mining operations achieve the global optimal production efficiency and energy consumption under the premise of safety, and enhances the adaptability of hydraulic support parameters to actual working conditions. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart illustrating the multi-parameter fusion dynamic optimization method for SDC industrial control parameters provided in this application embodiment. Detailed Implementation

[0016] This application provides a multi-parameter fusion SDC industrial control parameter dynamic optimization method to address the technical problems that existing hydraulic support parameters cannot respond in real time to complex and changing downhole environments, resulting in insufficient accuracy of roof pressure prediction, untimely support, insufficient initial support force, or excessive pressure relief.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0019] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid unnecessarily obscuring the description of this application. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0020] Examples, such as Figure 1 As shown in the embodiments of this application, a method for dynamic optimization of SDC industrial control parameters by multi-parameter fusion is provided, including:

[0021] S10: Based on the coal mining site perception data obtained from multi-source sensor monitoring, predict the predicted coal seam dip angle sequence and predicted coal mining speed sequence within the future preset time zone;

[0022] In this embodiment of the application, firstly, the coal mining site sensing data includes real-time dip angle data of the coal seam collected by an inclination sensor, real-time operating speed data of the coal mining machine collected by a speed sensor, real-time support pressure data of the hydraulic support collected by a pressure sensor, support displacement data collected by a displacement sensor, and working face image data collected by a high-definition camera.

[0023] Then, the collected multi-source data is aggregated to the edge computing node via an industrial bus or wireless transmission module. After outlier removal, data smoothing filtering, and spatiotemporal registration, a historical multi-source sensing data sequence set is constructed. This sequence set is then input into a pre-built coal seam dip angle predictor and a coal mining speed predictor, outputting a predicted coal seam dip angle sequence.

[0024] Furthermore, the coal mining speed predictor combines the coal mining machine's operating commands, the current cutting load, and historical speed change trends to output a predicted coal mining speed sequence with the same time granularity. During the prediction process, online parameter fine-tuning is performed using newly acquired data in real time to ensure the dynamic adaptability of the predicted sequence.

[0025] Specifically, step S10 in the method includes:

[0026] During the coal mining process, multi-source sensors are used to monitor the coal mining site in real time and obtain a set of multi-source sensing data sequences within a historical time zone. The set of multi-source sensing data sequences includes roof environmental parameter sequences, equipment operating condition parameter sequences, and global production parameter sequences.

[0027] Using a pre-built coal seam dip angle predictor and a coal mining speed predictor, the predicted coal seam dip angle sequence and the predicted coal mining speed sequence within a preset time zone are obtained based on the source sensing data sequence set.

[0028] In this embodiment of the application, firstly, during the coal mining process, the coal mining site is monitored in real time by multi-source sensors to obtain a set of multi-source sensing data sequences within the historical time zone. The set of multi-source sensing data sequences includes roof environmental parameter sequences, equipment operating condition parameter sequences, and global production parameter sequences.

[0029] The roof environment parameter sequence includes continuous data on the change of coal seam dip angle over time, collected by multiple dip angle sensors arranged at different locations on the working face. The sensors acquire data at a preset sampling frequency, such as once per second, forming a sequence corresponding to timestamps and dip angle values. It also includes real-time roof pressure data collected by a stress sensor array and roof temperature distribution sequence obtained by an infrared thermal imager, which together constitute a multi-dimensional perception of the roof environment.

[0030] The equipment operating parameter sequence includes data on the changes in initial support force and working resistance over time recorded by the pressure sensors of each support in the hydraulic support group; records of column raising speed, column lowering speed and column moving time during the moving process; operating parameters such as cutting motor current, traction motor current, drum speed and rocker arm height of the coal mining machine; and data such as chain speed and motor power of the scraper conveyor. All of the above parameters are integrated into a continuous sequence according to a unified time base.

[0031] The global production parameter sequence is obtained through weighing sensors on the scraper conveyor or coal flow image analysis, including time-series data reflecting the overall production status such as gas safety monitoring of the entire working face, actual coal mining volume, production process, and cycle operation time.

[0032] Secondly, using pre-built coal seam dip angle predictors and coal mining speed predictors, predicted coal seam dip angle sequences and predicted coal mining speed sequences within a preset time zone are obtained based on the multi-source sensing data sequence set. The coal seam dip angle predictor employs a deep learning model based on a Long Short-Term Memory (LSTM) network. The input layer of this model receives multi-dimensional features from the historical multi-source sensing data sequence set, including historical coal seam dip angle sequences, roof pressure sequences, coal mining machine position sequences, and operating parameter sequences. Through the hidden layer, spatiotemporal correlation modeling is performed to capture the long-term dependencies and nonlinear characteristics of coal seam dip angle changes.

[0033] During the model training phase, supervised learning is performed using historical dip angle data with time labels and corresponding influencing factor data. The network weights are optimized through the backpropagation algorithm, enabling the model to accurately learn the evolution law of coal seam dip angle under different geological conditions and mining states.

[0034] In the prediction phase, preprocessed historical multi-source sensing data sequences are input into a trained LSTM model. The model outputs a sequence of predicted coal seam dip angles for a preset time period, such as one hour in the future, with a time granularity of 5 minutes. Each sequence element corresponds to a predicted dip angle value at a specific time point, along with a corresponding prediction confidence level. The coal mining speed predictor combines a physical rule model with a data-driven model.

[0035] Furthermore, the physical rule model, based on the design parameters of the coal mining machine, such as rated power, maximum traction force, coal seam hardness, and current cutting depth, calculates the theoretically maximum possible range of coal mining speeds. The data-driven model employs a gradient boosting decision tree, using historical coal mining speed sequences, coal mining machine operation commands, cutting loads, and predicted coal seam dip angles as input features to learn the adjustment patterns and influencing factor weights of coal mining speeds in actual production. The constraints of the physical rule model are used as the output boundary of the data-driven model. By weighted fusion of the results from both models, the final predicted coal mining speed sequence is generated, ensuring that the prediction results conform to both the physical characteristics of the equipment and the impact of actual operating habits and complex working conditions. During the prediction process, newly acquired real-time sensing data is continuously input into the predictor. The prediction model is dynamically adjusted through sliding window updates of model parameters or incremental training to adapt to the time-varying characteristics of the underground environment, improving the accuracy and timeliness of the predicted sequence.

[0036] S20: Based on the predicted coal seam dip angle sequence and the predicted coal mining speed sequence, predict the roof pressure of the working face and obtain the predicted roof pressure field sequence;

[0037] In this embodiment, an input feature vector for the roof pressure field prediction model is constructed based on the predicted coal seam dip angle sequence and the predicted coal mining speed sequence. Specifically, the dip angle value at each time point in the predicted coal seam dip angle sequence is used as a parameter reflecting changes in roof occurrence conditions, the changes of which affect the stress distribution state of the roof rock mass. The predicted coal mining speed sequence reflects the dynamic rhythm of the working face advancement; the speed changes the roof exposure time and the goaf filling progress, thus affecting the evolution of roof pressure.

[0038] Secondly, the roof pressure at the working face is predicted, and a predicted roof pressure field sequence is obtained. The roof pressure field prediction model adopts a hybrid modeling method combining three-dimensional elastic finite element method and data-driven approach, treating the coal seam as an anisotropic elastic body. The dip angle parameters at different time points in the predicted coal seam dip angle sequence are input to determine the initial stress field distribution of the roof rock mass. Simultaneously, the impact of the roof collapse process in the goaf on pressure transmission is simulated to obtain the theoretical distribution range of the roof pressure field, and finally, the predicted roof pressure field sequence is output.

[0039] The construction methods for the coal seam dip angle predictor and the coal mining speed predictor include:

[0040] Several sample multi-source sensing data sequence sets were collected, and historical coal seam dip angle sequences and historical coal mining speed sequences of different sample multi-source sensing data sequence sets within historical time zones were collected to obtain several sample coal seam dip angle sequences and several sample coal mining speed sequences.

[0041] Using the aforementioned multi-source sensing data sequence set as input, and the aforementioned coal seam dip angle sequence as supervised training of the long short-term memory network until convergence, a coal seam dip angle predictor is generated.

[0042] Using the aforementioned multi-source sensing data sequence set as input, and the aforementioned coal mining speed sequence as supervised training, a long short-term memory network is trained until convergence, thereby generating a coal mining speed predictor.

[0043] In this embodiment, firstly, several sample multi-source sensing data sequence sets are collected. The sample data originates from coal mining faces with different geological conditions, mining processes, and equipment model combinations. Each sample multi-source sensing data sequence set contains a complete sequence of roof environmental parameters, equipment operating parameters, and global production parameters within the corresponding historical time zone, with a time span of no less than 3 months to cover seasonal geological changes and production cycle fluctuations. Simultaneously, the historical coal seam dip angle sequence and historical coal mining speed sequence within the corresponding historical time zone of each sample multi-source sensing data sequence set are collected as tag data. The historical coal seam dip angle sequence is obtained periodically by high-precision inclinometers at fixed underground stations, with at least one full working face scan per week to form a baseline sequence; the historical coal mining speed sequence is extracted from the operation log of the coal mining machine control system, containing actual traction speed data accurate to the second.

[0044] Secondly, using the aforementioned multi-source sensing data sequence set as input and the aforementioned coal seam dip angle sequence as supervised training, a Long Short-Term Memory (LSTM) network is constructed to predict the coal seam dip angle. During model construction, the input layer performs feature fusion on the multi-source sensing data sequence, standardizing parameters of different dimensions and units, such as dip angle, pressure, and current, and arranging them into a three-dimensional tensor according to the time step, which is the number of samples × the number of time steps × the feature dimension. The LSTM network contains three hidden layers, with 128, 64, and 32 neurons per layer, respectively. Dropout is used to prevent overfitting, with a dropout rate of 0.2. The output layer is a fully connected layer, using a linear activation function to output the predicted dip angle value for a future preset time zone. During training, the root mean square error is used as the loss function, and the Adam optimizer is used for parameter updates. The initial learning rate is set to 0.001, and dynamically adjusted through a learning rate decay strategy, for example, decaying to 0.5 every 50 epochs. The model training iterations stop when the validation set loss no longer decreases for 10 consecutive epochs, and the network parameters at this point are saved as the final coal seam dip angle predictor.

[0045] Finally, using the aforementioned multi-source sensing data sequence sets as input and the aforementioned coal mining speed sequences as supervised training, a separate LSTM network is generated to produce a coal mining speed predictor. The network structure of this predictor is similar to that of the coal seam dip angle predictor, but it focuses more on parameters directly related to the operation of the coal mining machine in its input feature selection, such as the cutting motor current sequence, the rocker arm height sequence, and the operation command encoding sequence. The output layer outputs the corresponding speed prediction value according to the time granularity of a preset time zone, such as 5 minutes / step. During training, considering that coal mining speed is greatly affected by human intervention, dynamic weights are introduced into the loss function, assigning higher penalty coefficients to speed adjustment periods that frequently occur in actual production, such as when encountering faults, to improve the model's predictive ability for special working conditions. Similarly, the optimal training rounds are determined using an early stopping method to ensure the model's generalization performance under complex working conditions. After training, both predictors were tested offline and debugged online. New data collected from the actual underground were used to calibrate the models to ensure that the prediction error was controlled within the allowable range of engineering, such as coal seam dip angle prediction error ≤ ±0.5° and coal mining speed prediction error ≤ ±0.2m / min.

[0046] Specifically, step S20 in the method includes:

[0047] The arithmetic mean and fluctuation analysis were performed on the predicted coal seam dip angle sequence and the predicted coal mining speed sequence, respectively, to calculate the mean value of the predicted coal seam dip angle, the fluctuation coefficient of the predicted coal seam dip angle, the mean value of the predicted coal mining speed, and the fluctuation coefficient of the predicted coal mining speed.

[0048] The roof pressure prediction complexity index is obtained by evaluating the mean predicted coal seam dip angle, the predicted coal seam dip angle fluctuation coefficient, the mean predicted coal mining speed, and the predicted coal mining speed fluctuation coefficient. The roof pressure prediction complexity index is positively correlated with the mean predicted coal seam dip angle, the predicted coal seam dip angle fluctuation coefficient, the predicted coal mining speed, and the predicted coal mining speed fluctuation coefficient.

[0049] Based on the complex index for roof pressure prediction, the roof pressure field prediction channel is invoked. The roof pressure of the working face is predicted according to the predicted coal seam dip angle sequence and the predicted coal mining speed sequence, and the predicted roof pressure field sequence is obtained.

[0050] In this embodiment, firstly, feature extraction and quantitative analysis are performed on the predicted coal seam dip angle sequence and the predicted coal mining speed sequence. For the predicted coal seam dip angle sequence, its arithmetic mean is calculated to reflect the average inclination of the coal seam within a future preset time zone. The larger the mean, the steeper the overall coal seam dip angle, and the greater the potential impact on roof stability. Simultaneously, the fluctuation coefficient is calculated to measure the drastic degree of dip angle change. The fluctuation coefficient is defined as the ratio of the sequence standard deviation to the mean. The higher the fluctuation coefficient, the more frequent and larger the changes in the coal seam dip angle within the predicted period, and the more likely dynamic fluctuations will occur in the roof pressure distribution. For the predicted coal mining speed sequence, its arithmetic mean is also calculated, representing the average advance speed of the coal mining machine within the future preset time zone. This mean is directly related to the advance rhythm of the working face and the length of roof exposure time. The coal mining speed fluctuation coefficient is obtained by the ratio of the standard deviation to the mean of the speed sequence, used to characterize the frequency of coal mining machine speed adjustments. A high fluctuation coefficient indicates that the coal mining machine may frequently change speed due to changes in geological conditions or operational adjustments, resulting in uneven roof load loading rates.

[0051] For example, if the mean of the predicted coal seam dip angle sequence is 8° and the standard deviation is 1.2° in the next hour, then its fluctuation coefficient is 1.2 / 8=0.15; if the mean of the predicted coal mining speed sequence is 5m / min and the standard deviation is 0.8m / min, then its fluctuation coefficient is 0.8 / 5=0.16.

[0052] Secondly, based on the above calculations, the predicted mean coal seam dip angle, predicted coal seam dip angle fluctuation coefficient, predicted mean coal mining speed, and predicted coal mining speed fluctuation coefficient are obtained, and a roof pressure prediction complexity index is constructed. This index is calculated using a weighted summation method, with the weights of each parameter determined through the analytic hierarchy process (AHP) combined with practical engineering experience to reflect the influence of different parameters on the complexity of roof pressure. For example, the mean coal seam dip angle and fluctuation coefficient may be assigned higher weights because the coal seam occurrence state is the fundamental factor determining roof pressure. The roof pressure prediction complexity index is positively correlated with all four parameters; that is, an increase in any parameter value leads to an increase in the complexity index. The higher the index value, the more complex the combination of factors affecting roof pressure within the future preset time zone, the more significant the dynamic changes in the pressure field, and the higher the requirements for the adaptability of the prediction model.

[0053] For example, if the weight of the average predicted coal seam dip angle is 0.35, the weight of the fluctuation coefficient is 0.3, the weight of the average coal mining speed is 0.2, and the weight of the fluctuation coefficient is 0.15, the following calculation can be made using the above example data: roof pressure prediction complexity index = 8 × 0.35 + 0.15 × 0.3 + 5 × 0.2 + 0.16 × 0.15 = 3.869.

[0054] Finally, different roof pressure field prediction channels are invoked based on the numerical range of the roof pressure prediction complexity index. For example, a complexity index threshold range is set: when the index ≤ 2.5, the simplified physical model channel is invoked, and the pressure field distribution is quickly calculated using the two-dimensional elasticity analytical method; when 2.5 < index ≤ 5, the standard hybrid modeling channel is invoked, and the three-dimensional finite element and data-driven basic fusion mode is enabled; when the index > 5, the enhanced hybrid modeling channel is invoked, and a geological anomaly identification module and a dynamic boundary condition adjustment mechanism are superimposed on the standard mode.

[0055] For example, the complexity index of 3.869 falls within the range of 2.5 < index ≤ 5, therefore the standard hybrid modeling channel is invoked. The predicted coal seam dip sequence and the predicted mining speed sequence are used as inputs. A geomechanical model including the coal seam, roof, and floor is constructed using the channel's built-in 3D elasticity finite element module. The model mesh is dynamically adjusted according to the complexity index; the higher the index, the denser the mesh. For example, a 5m × 5m × 3m element size is used. During finite element calculations, the dip angle value at each time point in the predicted coal seam dip sequence is used as a dynamic input parameter to update the initial stress field of the rock mass in real time. The predicted mining speed sequence is used to control the advance speed of the goaf boundary, simulating the change in the exposed roof area under different advance rhythms. Meanwhile, the data-driven module corrects the finite element calculation results through an error compensation network trained on historical pressure monitoring data. For example, when the predicted coal seam dip angle fluctuation coefficient is high, the correction weight for the pressure concentration area is automatically increased, and finally the predicted roof pressure field sequence with a spatiotemporal resolution of 5 minutes × 1m × 1m is output. Each sequence element contains the predicted pressure value of each monitoring point in three-dimensional space and the corresponding confidence interval.

[0056] Furthermore, based on the roof pressure prediction complexity index, the roof pressure field prediction channel is invoked, and the roof pressure of the working face is predicted according to the predicted coal seam dip angle sequence and the predicted coal mining speed sequence to obtain the predicted roof pressure field sequence, including:

[0057] The sample coal seam dip angle sequence set, sample coal mining velocity sequence set, and sample roof pressure field sequence set are collected as training data and P-fold cross-partitioned to obtain P sample training sets, where P is an integer greater than or equal to 10;

[0058] Using the sample coal seam dip angle sequence and sample coal mining speed sequence as input, and the sample roof pressure field sequence as supervision, the long short-term memory network is trained to convergence using the P sample training sets to generate P roof pressure field predictors, and the roof pressure field prediction channel is integrated and constructed according to the mean fusion strategy.

[0059] The ratio of the roof pressure prediction complexity index to the historical maximum roof pressure prediction complexity index recorded within the historical time range is multiplied by P and rounded to obtain Q. Then, Q roof pressure field predictors are randomly selected from the P roof pressure field predictors in the roof pressure field prediction channel. Based on the predicted coal seam dip angle sequence and the predicted coal mining speed sequence, the roof pressure of the working face is predicted, and the predicted roof pressure field sequence is output. Here, Q is greater than or equal to 1 and less than or equal to P.

[0060] In this embodiment, firstly, sample coal seam dip angle sequences, sample mining speed sequences, and corresponding sample roof pressure field sequences are collected as training data. The sample data includes actual monitoring data under different geological structures, coal seam thicknesses, roof lithology, and mining depths. Each sample data set contains time-synchronized coal seam dip angle sequences, mining speed sequences, and roof pressure field sequences. The roof pressure field sequence is collected using an array of pressure sensors arranged on the support columns of the working face, with a sampling interval of 30 seconds and a spatial resolution of one monitoring point for every five supports, forming a pressure distribution sequence along the length of the working face.

[0061] Subsequently, the processed dataset is subjected to P-fold cross-partitioning, where P is set to 10, that is, the dataset is randomly divided into 10 mutually exclusive sample training sets. Each sample training set contains samples of different geological conditions and mining conditions in a corresponding proportion, so as to ensure the consistency of the distribution of each subset and avoid overfitting and data bias in the model training process.

[0062] Secondly, an independent roof pressure field predictor is constructed for each sample training set. The sample coal seam dip angle sequence and sample mining speed sequence are used as the joint features of the input layer. The coal seam dip angle sequence is converted into a feature vector reflecting the dip angle change trend after standardization. The mining speed sequence is used to extract dynamic features through sliding window statistics, such as the maximum speed, minimum speed and average acceleration within a 5-minute window. The two are concatenated to form an input matrix with the dimension of "time steps × (dip angle feature dimension + speed feature dimension)".

[0063] Specifically, the predictor employs an improved LSTM network, introducing an attention mechanism on top of the traditional LSTM unit. This mechanism calculates the contribution weights of input features at different time steps to roof pressure prediction, enhancing the model's focus on key influencing periods, such as abrupt changes in coal seam dip angle and periods of drastic adjustments in mining speed. The network structure contains four hidden layers with 256, 128, 64, and 32 neurons respectively. The Dropout activation rate is set to 0.3, and a batch normalization layer is added before the output layer to stabilize the training process. Using the sample roof pressure field sequence as the supervision signal, the root mean square error is used as the loss function, and the AdamW optimizer is used for parameter updates. The initial learning rate is set to 0.0005, and the weight decay coefficient is 1e-5. Training iterations stop when the validation set loss shows no improvement for 20 consecutive epochs, and the current network parameters are saved as the roof pressure field predictor for that sample training set. After all 10 training sets have been trained, 10 independent roof pressure field predictors are obtained. A mean fusion strategy is used to construct a roof pressure field prediction channel, that is, the output results of the 10 predictors are arithmetically averaged over time steps to generate the final integrated prediction result, so as to reduce the prediction variance of a single model and improve the overall prediction stability.

[0064] Finally, the number of models participating in the prediction is dynamically selected based on the roof pressure prediction complexity index. Specifically, the ratio of the current prediction complexity index to the maximum roof pressure prediction complexity index recorded in the historical time range is calculated. For example, the ratio of the maximum roof pressure prediction complexity index in the past 3 months is denoted as α, and the value of α ranges from [0,1]. α is multiplied by 10 and rounded to obtain Q. The value of Q is the number of roof pressure field predictors called in this prediction, and Q must satisfy the constraint 1≤Q≤10.

[0065] For example, if the historical maximum complexity index is 6.2 and the current predicted complexity index is 3.869, then α = 3.869 / 6.2 ≈ 0.624, and Q = 0.624 × 10 ≈ 6. Subsequently, 6 roof pressure field predictors are randomly selected from 10, and the predicted coal seam dip angle sequence and the predicted coal mining speed sequence are input into the selected 6 predictors. Each predictor outputs its own predicted roof pressure field sequence. Then, the pressure values ​​of the corresponding time steps and spatial monitoring points of the 6 sequences are averaged using a mean fusion strategy to generate the final predicted roof pressure field sequence.

[0066] Furthermore, when the complexity index of roof pressure prediction is high, i.e., close to the historical maximum, more predictors are used in the fusion process to enhance adaptability to complex working conditions by leveraging the complementarity of multiple models. When the complexity index is low, fewer predictors are needed to meet the prediction accuracy requirements, reducing computational resource consumption and achieving a dynamic balance between prediction efficiency and accuracy. For example, when Q=1, the output of a single predictor is directly used; when Q=10, the outputs of all predictors are fused, which is suitable for extremely complex roof pressure prediction scenarios.

[0067] S30: Based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence, determine the appropriate safe time zone and the appropriate efficiency time zone;

[0068] In this embodiment, firstly, the suitable safety time zone and the suitable efficiency time zone are determined based on the analysis of the predicted coal seam dip angle sequence, the predicted coal mining speed sequence, and the predicted roof pressure field sequence. The suitable safety time zone refers to a continuous period within a future preset time zone where the roof pressure field distribution of the working face meets the preset safety threshold requirements, and the coal mining machine, operating at its current or adjusted speed, will not trigger safety risks such as roof support system overload, spalling, or roof collapse. The suitable efficiency time zone refers to a continuous period where, under the premise of ensuring safety, the coal mining machine can operate at a higher advance speed, enabling the working face production efficiency to reach a preset target value, such as a preset target value ≥ 90% of the designed capacity. Both need to be comprehensively determined based on the multi-dimensional characteristics of the predicted coal seam dip angle sequence, the predicted coal mining speed sequence, and the predicted roof pressure field sequence.

[0069] The determination of the appropriate safe time zone based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence includes:

[0070] Simulate the future trajectory of the coal mining machine based on the predicted coal mining speed sequence, and record the initial exposure time of the roof of each hydraulic support as the roof exposure time;

[0071] Based on the predicted coal seam dip angle sequence and the predicted roof pressure field sequence, the maximum allowable unsupported time of the roof of each support is calculated, and the end time of the adapted safe time zone is calculated in combination with the roof exposure time.

[0072] The start time of the adapted safe time zone is obtained by adding the minimum process delay to the time when the top plate is exposed, and the adapted safe time zone is determined by combining the end time of the adapted safe time zone.

[0073] In this embodiment, firstly, the future trajectory of the coal mining machine is simulated based on the predicted coal mining speed sequence. This trajectory starts from the current position of the coal mining machine and is calculated by integrating the speed values ​​at each time point in the predicted coal mining speed sequence to obtain the position coordinates of the coal mining machine corresponding to each time step within a preset future time zone. Based on these position coordinates, combined with the arrangement parameters of the hydraulic supports at the working face, such as the support spacing, quantity, and numbering order, the initial exposure time of the roof of each hydraulic support is recorded as the roof exposure time.

[0074] For example, if the coal mining machine is currently located at support number 5, and the predicted coal mining speed sequence for the next 10 minutes is 4 m / min, 5 m / min, and 4.5 m / min respectively, with a support spacing of 1.5 m, then the exposure time of the roof of support number 6 can be calculated as (1.5 m / 4 m / min) = 0.375 minutes later, or 22.5 seconds. For support number 7, the exposure time is calculated by adding (1.5 m / 5 m / min) = 0.3 minutes to the exposure time of support number 6, resulting in 18 seconds. This process continues until all supports are traversed, forming a complete sequence of roof exposure times for each support.

[0075] Secondly, based on the predicted coal seam dip angle sequence and the predicted roof pressure field sequence, the maximum allowable unsupported time for each support is calculated. For the predicted coal seam dip angle sequence, the dip angle value corresponding to each support is extracted from the sequence. Combined with the roof lithology parameters at the support location, such as uniaxial compressive strength, elastic modulus, and internal friction angle, the natural caving step of the roof strata under that dip angle condition is calculated using the limit equilibrium theory. Then, combined with the predicted pressure value at the support location in the predicted roof pressure field sequence, if the predicted pressure value is less than the ultimate bearing capacity of the roof strata, the maximum allowable unsupported time is mainly determined by the natural caving step and the mining speed; if the predicted pressure value is close to or exceeds the ultimate bearing capacity, a safety factor, usually 1.2-1.5, needs to be introduced to correct the natural caving step, thereby shortening the maximum allowable unsupported time.

[0076] Finally, a minimum process delay is added to the exposure time of the top plate of each support to obtain the start time of the appropriate safety time zone. The minimum process delay refers to the shortest time required from the exposure of the top plate to the time when the support equipment can complete the support action, including the entire process time of lowering the support column, moving the support, raising the support column and reaching the initial support force. This value is determined according to the model of the support used and the operating specifications, and is generally between 30 seconds and 2 minutes.

[0077] For example, if the roof exposure time of a certain support is at the 2nd minute, and the minimum process delay is set to 1 minute, then the start time of the adapted safety time zone for that support is the 2nd minute + 1 minute = 3rd minute. The start and end times of the adapted safety time zones for all supports are summarized, and the maximum value of the start times of all supports is taken as the unified start time of the adapted safety time zone for the entire working face, ensuring that all supports have met the process delay requirements. The minimum value of the end times of all supports is taken as the unified end time of the adapted safety time zone for the entire working face, ensuring that all supports have not exceeded the maximum allowable unsupported time. The continuous time period between the start and end times is the adapted safety time zone. If the coal mining machine speed needs to be adjusted within the preset time zone, the exposure time of each support under the adjusted speed sequence needs to be recalculated, and the start and end times of the adapted safety time zone updated accordingly.

[0078] Specifically, the time zone for determining the adaptation efficiency is determined based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence, including:

[0079] Perform pressure distribution uniformity analysis on several of the predicted top plate pressure field sequences, and output a pressure uniformity index sequence;

[0080] The mean value of the pressure uniformity index and the fluctuation coefficient of the pressure uniformity index are calculated based on the pressure uniformity index sequence.

[0081] An efficiency compensation coefficient is calculated based on the predicted coal seam dip angle fluctuation coefficient, the predicted coal mining speed fluctuation coefficient, the average pressure uniformity index, and the pressure uniformity index fluctuation coefficient. The efficiency compensation coefficient is negatively correlated with the predicted coal seam dip angle fluctuation coefficient, the predicted coal mining speed fluctuation coefficient, and the pressure uniformity index fluctuation coefficient, and positively correlated with the average pressure uniformity index.

[0082] The product of the efficiency compensation coefficient and the initial efficiency time zone is used as the adapted efficiency time zone.

[0083] In this embodiment, firstly, a pressure distribution uniformity analysis is performed on the pressure field data corresponding to each time step in the predicted roof pressure field sequence. Specifically, for the three-dimensional pressure field at each time point, the pressure values ​​of each monitoring point are extracted along the length of the working face to construct a pressure distribution vector. The distribution uniformity of this vector is calculated using the information entropy algorithm, with the formula H = -Σ(pi × ln pi), where pi is the proportion of the pressure value of the i-th monitoring point to the total pressure value. A larger H value indicates a more uniform pressure distribution, while a smaller H value indicates a significant pressure concentration. The H values ​​of each time step are arranged in chronological order to form a pressure uniformity index sequence.

[0084] Subsequently, the mean pressure uniformity index μH and the pressure uniformity index fluctuation coefficient σH are calculated based on the sequence. μH is the arithmetic mean of all elements in the sequence, reflecting the uniformity level of the overall pressure distribution. σH is the ratio of the standard deviation of the sequence to μH, which characterizes the stability of pressure uniformity over time. The smaller σH is, the more stable the pressure distribution is.

[0085] Furthermore, the efficiency compensation coefficient K is calculated using the formula K = k0 × (μH / (1+α+β+σH)), where k0 is the baseline compensation coefficient, calibrated based on the working face's designed capacity and efficiency parameters under historical optimal conditions, with a value ranging from 0.8 to 1.2. In the formula, μH is positively correlated with K, meaning the more uniform the pressure distribution, the higher the efficiency compensation coefficient. α, β, and σH are all negatively correlated with K, indicating that when the coal seam dip angle fluctuates greatly, the mining speed is unstable, or the pressure uniformity fluctuates drastically, the efficiency compensation coefficient decreases to reserve more safety margin.

[0086] Finally, the efficiency compensation coefficient K is multiplied by the initial efficiency time zone T0 to obtain the adaptive efficiency time zone, i.e., adaptive efficiency time zone = K × T0. The initial efficiency time zone T0 refers to the continuous high-efficiency operating period when the coal mining machine advances at its maximum design speed under ideal working conditions, i.e., when the coal seam dip angle is stable, the coal mining speed is constant, and the pressure distribution is uniform. It is calculated through the theoretical production capacity model.

[0087] S40: Using the adapted safety time zone and the adapted efficiency time zone as rigid constraints, optimize the hydraulic support parameters of the hydraulic support group based on the predicted coal seam dip angle sequence, the predicted coal mining speed sequence and the predicted roof pressure field sequence, output the adapted support parameter instruction queue in the preset time zone, and execute the hydraulic support in the preset time zone.

[0088] In this embodiment, the hydraulic support parameters of the hydraulic support group are optimized based on the predicted coal seam dip angle sequence, the predicted coal mining speed sequence, and the predicted roof pressure field sequence, with the adaptation of the safe time zone and the adaptation of the efficiency time zone as rigid constraints. Specifically, this includes the construction of the objective function, the setting of the constraint conditions, and the solution of the multi-objective optimization algorithm.

[0089] Specifically, using the adapted safety time zone and adapted efficiency time zone as rigid constraints, the hydraulic support parameters of the hydraulic support group are optimized based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence, and an adapted support parameter instruction queue within a preset time zone is output, including:

[0090] The hydraulic support parameter adjustment space of the hydraulic support group is obtained, wherein the hydraulic support group includes K hydraulic supports, and the hydraulic support parameters include the moving time point, the target initial support force, the lifting speed and the lowering speed.

[0091] Using the adaptive safety time zone and the adaptive efficiency time zone as rigid constraints, the hydraulic support parameters are randomly selected within the hydraulic support parameter adjustment space to generate several support parameter instruction queues, wherein each support parameter instruction queue includes K support parameter instruction sequences for K hydraulic supports.

[0092] Based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence, the hydraulic support parameters of the hydraulic support group are optimized by combining the several support parameter instruction queues, and the appropriate support parameter instruction queue within the preset time zone is output.

[0093] In this embodiment, firstly, the adjustment space of the hydraulic support parameters of the hydraulic support group is obtained. For a support group consisting of K hydraulic supports, the support parameters of each support must be within a specific range to ensure safety and efficiency. The time point for moving the support must be limited to the interval [start time, end time] of the appropriate safe time zone, and the time interval between moving any two adjacent supports must not be less than the minimum moving interval time. The minimum moving interval time is set according to the support model and the linkage requirements of the working face equipment, and is usually 15-30 seconds.

[0094] The target initial support force is determined based on the predicted pressure value at the corresponding support position in the predicted roof pressure field sequence. It is set to 1.1-1.3 times the predicted pressure value, with a lower limit of not less than 80% of the rated initial support force and an upper limit of not more than 120% of the rated initial support force, in order to avoid insufficient initial support force leading to roof subsidence or excessive height causing hydraulic system overload.

[0095] The lifting and lowering speeds must meet the mechanical performance limitations of the support operation. The lifting speed range is set to 0.05-0.2 m / s, and the lowering speed range is set to 0.08-0.25 m / s. The lowering speed should be slightly greater than the lifting speed to shorten the non-support time.

[0096] Secondly, using the adapted safety time zone and adapted efficiency time zone as rigid constraints, random sampling of support parameters is performed within the hydraulic support parameter adjustment space to generate a support parameter instruction queue set containing N candidate schemes. The relocation time of all supports falls within the time window of the adapted safety time zone, and the relocation time difference between any two adjacent supports is greater than or equal to the minimum relocation interval; the target initial support force, lifting speed, and column lowering speed of each support are strictly within the upper and lower limits of the corresponding parameters.

[0097] For example, if the working face contains 100 hydraulic supports, and the support parameters of each support include 4 dimensions, namely the time point of the support shift, the initial support force of the target, the lifting speed, and the lowering speed, then the dimension of a single instruction queue is 100×4, and the value of N is usually set to 50-200 according to the convergence requirements of the optimization algorithm.

[0098] Next, the hydraulic support parameters of the hydraulic support group are optimized, and a queue of adaptive support parameter instructions within a preset time zone is output, including the construction of a multi-objective optimization function. The construction of the multi-objective optimization function needs to comprehensively consider support safety, operation efficiency, and energy consumption costs.

[0099] Furthermore, based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence, and combined with the several support parameter instruction queues, the hydraulic support parameters of the hydraulic support group are optimized, and an adapted support parameter instruction queue within a preset time zone is output, including:

[0100] Constructing a simulation operating space for hydraulic support based on digital twins;

[0101] The predicted coal seam dip angle sequence, predicted coal mining speed sequence, predicted roof pressure field sequence, and several support parameter instruction queues are randomly combined to generate several hydraulic support schemes.

[0102] Within the hydraulic support simulation operating space, support simulations are performed according to the several hydraulic support schemes, outputting several total simulated frame movement times, several simulated roof support pressure uniformity coefficients, and several simulated support energy consumptions, and evaluating and determining several support adaptation indices. The support adaptation index is positively correlated with the simulated roof support pressure uniformity coefficient, and negatively correlated with the total simulated frame movement time and simulated support energy consumption.

[0103] Using the hydraulic support parameter adjustment space, the adaptation safety time zone, and the adaptation efficiency time zone as optimization constraints, the hydraulic support parameters are optimized according to the several support parameter instruction queues and several support adaptation indices, and the adaptation support parameter instruction queue within the preset time zone is output.

[0104] In this embodiment, firstly, a hydraulic support simulation operating space consistent with the actual physical environment of the working face is constructed based on digital twin technology. This space is based on a three-dimensional geological model of the working face and integrates a multibody dynamics model of the hydraulic support, a coupled mechanical model of the roof-support-surrounding rock, and a dynamic response model of the hydraulic system.

[0105] Secondly, the predicted coal seam dip angle sequence, predicted coal mining speed sequence, predicted roof pressure field sequence, and several generated support parameter command queues are randomly combined to form M independent hydraulic support schemes. Each scheme includes a parameter command queue and a corresponding predicted environmental sequence, such as the time-varying data of dip angle, speed, and pressure field. To avoid combination explosions, the support parameter command queues are first divided into three categories based on their clustering results: safety-oriented, efficiency-oriented, and balanced. Then, a certain proportion of queues are randomly selected from each category and combined with the predicted sequences to ensure the diversity of schemes covers different working conditions.

[0106] Next, parallel simulations of each scheme were performed within the hydraulic support simulation operating space, outputting the total simulated frame movement time, the simulated roof support pressure uniformity coefficient, and the simulated support energy consumption for each scheme. The simulated roof support pressure uniformity coefficient is the ratio of the standard deviation to the mean of the support pressure for each support; a smaller ratio indicates better uniformity. The simulated support energy consumption is the total power consumption of the hydraulic systems of all supports.

[0107] Then, the support adaptability index S is calculated using the formula S=ω1×(1-σp / μp)-ω2×(Ttotal / T0total)-ω3×(E / E0), where ω1, ω2, and ω3 are weighting coefficients, which are adjusted according to the working face management objectives, typically ω1>ω2≥ω3. σp / μp is the pressure uniformity coefficient, Ttotal / T0total is the relative frame relocation time (compared to the theoretical minimum total frame relocation time T0total under ideal working conditions), and E / E0 is the relative energy consumption (compared to the baseline energy consumption E0). This index is positively correlated with the pressure uniformity coefficient; the better the uniformity, the higher S. It is negatively correlated with the total frame relocation time and support energy consumption; shorter time and lower energy consumption result in a higher S, comprehensively reflecting the adaptability of the support scheme.

[0108] Finally, the hydraulic support parameter adjustment space, the adaptation to the safe time zone, and the adaptation to the efficiency time zone are used as optimization constraints. For example, the total frame relocation time should not exceed 90% of the adaptation to the efficiency time zone, and a buffer time is reserved. The support parameter command queue is then optimized.

[0109] For example, the algorithm sets the population size to 100, the crossover probability to 0.8, the mutation probability to 0.1, and the number of iterations to 50 generations. Each individual in each generation corresponds to a support parameter instruction queue, and its support fit index S is calculated through the simulation process described above. Based on the S value, selection, crossover, and mutation operations are performed to gradually screen out the solution that achieves Pareto optimality in terms of safety, efficiency, and energy consumption.

[0110] Finally, the solution with the largest support adaptation index S is selected from the optimal solution set as the output, and its corresponding support parameter instruction queue is the adaptation support parameter instruction queue in the preset time zone, which contains the specific action parameters of each stent at each time step.

[0111] Furthermore, using the hydraulic support parameter adjustment space, the adaptation safety time zone, and the adaptation efficiency time zone as optimization constraints, the hydraulic support parameters are optimized based on the plurality of support parameter instruction queues and the plurality of support adaptation indices, and the adaptation support parameter instruction queue within the preset time zone is output, including:

[0112] The comprehensive support fluctuation coefficient is calculated based on the predicted coal seam dip angle fluctuation coefficient, the predicted coal mining speed fluctuation coefficient, and the pressure uniformity index fluctuation coefficient.

[0113] The ratio of the comprehensive support fluctuation coefficient to the preset support standard fluctuation coefficient is set as the optimization complexity, and the product of the optimization complexity and the initial optimization convergence number is used as the adaptive optimization convergence number.

[0114] Using the particle swarm optimization algorithm, with the hydraulic support parameter adjustment space, the adaptation safety time zone, and the adaptation efficiency time zone as optimization constraints, the hydraulic support parameters are iteratively optimized according to the several support parameter instruction queues and several support adaptation indices until the adaptation optimization convergence number is reached. The support parameter instruction queue corresponding to the maximum support adaptation index in the optimization process is output as the adaptation support parameter instruction queue.

[0115] In this embodiment, the comprehensive support fluctuation coefficient is first calculated. The comprehensive support fluctuation coefficient is a comprehensive quantitative representation of coal seam conditions, the mining process, and roof pressure stability. Its calculation formula is set as γ=λ1×α+λ2×β+λ3×σH, where λ1, λ2, and λ3 are the weights of the predicted coal seam dip angle fluctuation coefficient α, the predicted mining speed fluctuation coefficient β, and the pressure uniformity index fluctuation coefficient σH, respectively, and λ1+λ2+λ3=1. The weight values ​​are determined based on the degree of influence of each fluctuation factor on the stability of the support system. For example, when the roof pressure fluctuation has the greatest impact on support safety, λ3>λ1≥λ2 can be set. Specific values ​​can be calibrated through the analytic hierarchy process or regression analysis based on historical accident statistics.

[0116] Subsequently, the ratio of the overall support fluctuation coefficient γ to the preset standard support fluctuation coefficient γ0 is defined as the optimization complexity C, i.e., C = γ / γ0. The preset standard support fluctuation coefficient γ0 is the benchmark value of the overall fluctuation coefficient of the working face under typical stable working conditions, determined by field measurement data or theoretical calculation. The optimization complexity C reflects the complexity of the current predicted working condition relative to the standard working condition. The larger the C value, the more severe the working condition fluctuation, and the more thorough the optimization process needs to avoid getting trapped in local optima.

[0117] Furthermore, the number of iterations N for adaptive optimization is calculated based on the optimization complexity C, using the formula N = N0 × C, where N0 is the initial number of iterations required for convergence, i.e., the number of iterations required for the algorithm to converge under standard conditions, typically set to 30-100 iterations based on experience. The number of iterations N for adaptive optimization increases linearly with the optimization complexity C, ensuring that the quality of the solution is improved by increasing the number of iterations under complex conditions. For example, if C = 1.5 and N0 = 50, then N = 75, meaning the algorithm needs to iterate 75 times before stopping the optimization.

[0118] Finally, a particle swarm optimization algorithm is used for parameter optimization. Each support parameter instruction queue is encoded as a particle in the particle swarm, and the particle's position vector corresponds to a multi-dimensional combination of support parameters. During the algorithm initialization phase, N particles and N candidate support parameter instruction queues are randomly generated, and the initial velocity of each particle is set according to the scale of the parameter adjustment space. In each iteration, the particle updates its velocity and position based on its individual historical best position and the group's historical best position, continuously searching the parameter space. Each generation of particles needs to calculate its support fitness index S through the hydraulic support simulation runtime space, which serves as the particle's fitness value. The iteration process continues until the fitness optimization convergence number N is reached. At this point, the support parameter instruction queue corresponding to the maximum support fitness index S obtained in the entire optimization process is output, which is the final adapted support parameter instruction queue. This queue enables dynamic optimization control of the hydraulic support group under the current predicted working conditions.

[0119] In summary, this application achieves dynamic optimization of hydraulic support group support parameters by constructing a rigid constraint framework that adapts to both safe and efficient time zones, combined with digital twin simulation and multi-objective optimization algorithms. Based on the dynamic changes in coal seam dip angle, mining speed, and roof pressure, a more adaptable support parameter command queue is output in real time, improving the support safety and operational efficiency of the hydraulic support group under unsteady conditions.

[0120] In summary, the embodiments of this application have at least the following technical effects:

[0121] This application provides a multi-parameter fusion-based dynamic optimization method for SDC industrial control parameters. First, based on coal mining site sensing data acquired from multi-source sensor monitoring, it captures real-time dynamic changes in the complex underground environment, providing raw data support for subsequent parameter prediction and optimization, thereby improving prediction accuracy. Second, using pre-constructed coal seam dip angle predictors and coal mining speed predictors, it predicts the trends of coal seam dip angle and coal mining speed changes within a preset time zone based on historical multi-source sensing data sequences. This better adapts to the complexity and variability of the underground environment, improving prediction accuracy. Third, it analyzes and determines suitable safety and efficiency time zones, effectively avoiding safety accidents caused by untimely support. The efficiency time zone can be dynamically optimized based on the stability and pressure distribution uniformity of actual working conditions, ensuring that production efficiency is maximized under safe conditions, avoiding inefficiency or resource waste that may result from static efficiency settings. Finally, the hydraulic support parameters of the hydraulic support group are optimized using the adaptation to safe and efficient time zones as rigid constraints. Through iterative optimization, the support parameter instruction queue corresponding to the maximum support adaptation index is output, achieving a globally optimal balance between safety, production efficiency, and energy consumption in the support parameters. Through the above technical solution, this application, based on multi-source sensing data fusion and dynamic parameter optimization, achieves real-time response of hydraulic support parameters to the complex and variable underground environment, improves the accuracy of roof pressure prediction, avoids problems such as untimely support, insufficient initial support force, or excessive pressure relief, ensures that coal mining operations achieve globally optimal production efficiency and energy consumption under the premise of safety, and enhances the adaptability of hydraulic support parameters to actual working conditions.

[0122] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0123] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0124] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A multi-parameter fusion-based dynamic optimization method for SDC industrial control parameters, characterized in that the method... include: Based on the coal mining site perception data obtained from multi-source sensor monitoring, the predicted coal seam dip angle sequence and predicted coal mining speed sequence are obtained within a future preset time zone. Based on the predicted coal seam dip angle sequence and the predicted coal mining speed sequence, the roof pressure of the working face is predicted, and the predicted roof pressure field sequence is obtained. Based on the analysis of the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence, the appropriate safe time zone and the appropriate efficiency time zone are determined. Using the adaptive safety time zone and the adaptive efficiency time zone as rigid constraints, the hydraulic support parameters of the hydraulic support group are optimized based on the predicted coal seam dip angle sequence, the predicted coal mining speed sequence and the predicted roof pressure field sequence. The adaptive support parameter instruction queue within the preset time zone is output, and the hydraulic support within the preset time zone is executed. The determination of the appropriate safe time zone based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence includes: Simulate the future trajectory of the coal mining machine based on the predicted coal mining speed sequence, and record the initial exposure time of the roof of each hydraulic support as the roof exposure time; Based on the predicted coal seam dip angle sequence and the predicted roof pressure field sequence, the maximum allowable unsupported time of the roof of each support is calculated, and the end time of the adapted safe time zone is calculated in combination with the roof exposure time. The start time of the adapted safe time zone is obtained by adding the minimum process delay to the time when the top plate is exposed, and the adapted safe time zone is determined by combining the end time of the adapted safe time zone. The determination of the adaptation efficiency time zone based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence includes: Perform pressure distribution uniformity analysis on several of the predicted top plate pressure field sequences, and output a pressure uniformity index sequence; The mean value of the pressure uniformity index and the fluctuation coefficient of the pressure uniformity index are calculated based on the pressure uniformity index sequence. The efficiency compensation coefficient is calculated based on the predicted coal seam dip angle fluctuation coefficient, the predicted coal mining speed fluctuation coefficient, the average pressure uniformity index, and the pressure uniformity index fluctuation coefficient. The efficiency compensation coefficient is negatively correlated with the predicted coal seam dip angle fluctuation coefficient, the predicted coal mining speed fluctuation coefficient, and the pressure uniformity index fluctuation coefficient, and positively correlated with the average pressure uniformity index. The product of the efficiency compensation coefficient and the initial efficiency time zone is used as the adapted efficiency time zone.

2. The multi-parameter fusion dynamic optimization method for SDC industrial control parameters according to claim 1, characterized in that, Based on coal mining site sensing data acquired through multi-source sensor monitoring, the predicted coal seam dip angle sequence and predicted coal mining speed sequence are obtained within a preset time zone, including: During the coal mining process, multi-source sensors are used to monitor the coal mining site in real time and obtain a set of multi-source sensing data sequences within a historical time zone. The set of multi-source sensing data sequences includes roof environmental parameter sequences, equipment operating condition parameter sequences, and global production parameter sequences. Using a pre-built coal seam dip angle predictor and a coal mining speed predictor, the predicted coal seam dip angle sequence and the predicted coal mining speed sequence within a preset time zone are obtained based on the source sensing data sequence set.

3. The multi-parameter fusion dynamic optimization method for SDC industrial control parameters according to claim 2, characterized in that, The method for constructing the coal seam dip angle predictor and the coal mining speed predictor includes: Several sample multi-source sensing data sequence sets were collected, and historical coal seam dip angle sequences and historical coal mining speed sequences of different sample multi-source sensing data sequence sets within historical time zones were collected to obtain several sample coal seam dip angle sequences and several sample coal mining speed sequences. Using the aforementioned multi-source sensing data sequence set as input, and the aforementioned coal seam dip angle sequence as supervised training of the long short-term memory network until convergence, a coal seam dip angle predictor is generated. Using the aforementioned multi-source sensing data sequence set as input, and the aforementioned coal mining speed sequence as supervised training, a long short-term memory network is trained until convergence, thereby generating a coal mining speed predictor.

4. The multi-parameter fusion dynamic optimization method for SDC industrial control parameters according to claim 1, characterized in that, Based on the predicted coal seam dip angle sequence and predicted coal mining speed sequence, the roof pressure of the working face is predicted, and the predicted roof pressure field sequence is obtained, including: The arithmetic mean and fluctuation analysis were performed on the predicted coal seam dip angle sequence and the predicted coal mining speed sequence, respectively, to calculate the mean value of the predicted coal seam dip angle, the fluctuation coefficient of the predicted coal seam dip angle, the mean value of the predicted coal mining speed, and the fluctuation coefficient of the predicted coal mining speed. The roof pressure prediction complexity index is obtained by evaluating the mean predicted coal seam dip angle, the predicted coal seam dip angle fluctuation coefficient, the mean predicted coal mining speed, and the predicted coal mining speed fluctuation coefficient. The roof pressure prediction complexity index is positively correlated with the mean predicted coal seam dip angle, the predicted coal seam dip angle fluctuation coefficient, the predicted coal mining speed, and the predicted coal mining speed fluctuation coefficient. Based on the complex index for roof pressure prediction, the roof pressure field prediction channel is invoked. The roof pressure of the working face is predicted according to the predicted coal seam dip angle sequence and the predicted coal mining speed sequence, and the predicted roof pressure field sequence is obtained.

5. The multi-parameter fusion dynamic optimization method for SDC industrial control parameters according to claim 4, characterized in that, Based on the aforementioned roof pressure prediction complexity index, the roof pressure field prediction channel is invoked. The roof pressure of the working face is predicted according to the predicted coal seam dip angle sequence and the predicted coal mining speed sequence, obtaining the predicted roof pressure field sequence, including: The sample coal seam dip angle sequence set, sample coal mining velocity sequence set, and sample roof pressure field sequence set are collected as training data and P-fold cross-partitioned to obtain P sample training sets, where P is an integer greater than or equal to 10; Using the sample coal seam dip angle sequence and sample coal mining speed sequence as input, and the sample roof pressure field sequence as supervision, the long short-term memory network is trained to convergence using the P sample training sets to generate P roof pressure field predictors, and the roof pressure field prediction channel is integrated and constructed according to the mean fusion strategy. The ratio of the roof pressure prediction complexity index to the historical maximum roof pressure prediction complexity index recorded within the historical time range is multiplied by P and rounded to obtain Q. Then, Q roof pressure field predictors are randomly selected from the P roof pressure field predictors in the roof pressure field prediction channel. Based on the predicted coal seam dip angle sequence and the predicted coal mining speed sequence, the roof pressure of the working face is predicted, and the predicted roof pressure field sequence is output. Here, Q is greater than or equal to 1 and less than or equal to P.

6. The multi-parameter fusion dynamic optimization method for SDC industrial control parameters according to claim 1, characterized in that, Using the adapted safety time zone and adapted efficiency time zone as rigid constraints, the hydraulic support parameters of the hydraulic support group are optimized based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence. The resulting sequence of adapted support parameter instructions within a preset time zone is output, including: The hydraulic support parameter adjustment space of the hydraulic support group is obtained, wherein the hydraulic support group includes K hydraulic supports, and the hydraulic support parameters include the moving time point, the target initial support force, the lifting speed and the lowering speed. Using the adaptive safety time zone and the adaptive efficiency time zone as rigid constraints, the hydraulic support parameters are randomly selected within the hydraulic support parameter adjustment space to generate several support parameter instruction queues. Each support parameter instruction queue includes K support parameter instruction sequences for K hydraulic supports. Based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence, the hydraulic support parameters of the hydraulic support group are optimized by combining the several support parameter instruction queues, and the appropriate support parameter instruction queue within the preset time zone is output.

7. The multi-parameter fusion dynamic optimization method for SDC industrial control parameters according to claim 6, characterized in that, Based on the predicted coal seam dip angle sequence, predicted coal mining speed sequence, and predicted roof pressure field sequence, and combined with the several support parameter instruction queues, the hydraulic support parameters of the hydraulic support group are optimized, and an adapted support parameter instruction queue within a preset time zone is output, including: Constructing a simulation operating space for hydraulic support based on digital twins; The predicted coal seam dip angle sequence, predicted coal mining speed sequence, predicted roof pressure field sequence, and several support parameter instruction queues are randomly combined to generate several hydraulic support schemes. Within the hydraulic support simulation operating space, support simulations are performed according to the several hydraulic support schemes, outputting several total simulated frame movement times, several simulated roof support pressure uniformity coefficients, and several simulated support energy consumptions, and evaluating and determining several support adaptation indices. The support adaptation index is positively correlated with the simulated roof support pressure uniformity coefficient, and negatively correlated with the total simulated frame movement time and simulated support energy consumption. Using the hydraulic support parameter adjustment space, the adaptation safety time zone, and the adaptation efficiency time zone as optimization constraints, the hydraulic support parameters are optimized according to the several support parameter instruction queues and several support adaptation indices, and the adaptation support parameter instruction queue within the preset time zone is output.

8. The multi-parameter fusion dynamic optimization method for SDC industrial control parameters according to claim 7, characterized in that, Using the hydraulic support parameter adjustment space, the adaptation safety time zone, and the adaptation efficiency time zone as optimization constraints, the hydraulic support parameters are optimized according to the plurality of support parameter command queues and the plurality of support adaptation indices, and the adaptation support parameter command queue within the preset time zone is output, including: The comprehensive support fluctuation coefficient is calculated based on the predicted coal seam dip angle fluctuation coefficient, the predicted coal mining speed fluctuation coefficient, and the pressure uniformity index fluctuation coefficient. The ratio of the comprehensive support fluctuation coefficient to the preset support standard fluctuation coefficient is set as the optimization complexity, and the product of the optimization complexity and the initial optimization convergence number is used as the adaptive optimization convergence number. Using the particle swarm optimization algorithm, with the hydraulic support parameter adjustment space, the adaptation safety time zone, and the adaptation efficiency time zone as optimization constraints, the hydraulic support parameters are iteratively optimized according to the several support parameter instruction queues and several support adaptation indices until the adaptation optimization convergence number is reached. The support parameter instruction queue corresponding to the maximum support adaptation index in the optimization process is output as the adaptation support parameter instruction queue.

Images