GA-SVR-based intelligent optimization control method and system for mine ventilator

Abstract

The invention relates to the technical field of ventilation regulation and control, and provides a GA-SVR-based mine ventilator intelligent optimization control method and system, and the method comprises the steps: generating an initial population, and coding a combination of a blade angle and a motor frequency for each individual in the population; based on the required ventilation quantity and wind pressure, calculating a second fitness value of each individual after predicting the total pressure efficiency and wind pressure of each individual by adopting the trained support vector regression model, and executing adaptive genetic manipulation and an elitist retention strategy until convergence to obtain the optimal total pressure efficiency and wind pressure so as to control the mine ventilator; wherein the support vector regression model performs hyper-parameter optimization by using a genetic algorithm in a training process. An intelligent regulation and control mechanism with on-demand air supply as a core is established, and the limitation of small sample data is effectively overcome.

Description

Technical Field

[0001] This invention belongs to the field of ventilation control technology, and particularly relates to an intelligent optimization control method and system for mine ventilation fans based on GA-SVR. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] As the core barrier to maintaining safety in underground operations, the main ventilation fan in a mine ventilation system needs to operate continuously for long periods, resulting in extremely high energy consumption. However, limited by traditional manual experience-based judgment or rigid scheduling modes, existing control methods are unable to accurately follow the complex and ever-changing ventilation load demands underground, causing the mine ventilation system to operate inefficiently on a regular basis, resulting in significant energy losses.

[0004] With the iteration of the Industrial Internet of Things and intelligent algorithms, data-driven wind turbine optimization has become a hot topic. Existing mainstream methods such as Computational Fluid Dynamics (CFD) simulation, machine learning, and metaheuristic optimization still face many constraints in practical deployment: the high computational cost of CFD hinders real-time applications, machine learning models suffer from drastic prediction fluctuations due to reliance on manual parameter tuning, and optimization algorithms are hampered by safety constraints, making it difficult to balance energy efficiency and safety goals. Summary of the Invention

[0005] To address the technical problems mentioned above, this invention provides an intelligent optimization control method and system for mine ventilation fans based on GA-SVR. It constructs a parameter inverse inversion framework, which can infer the optimal operating conditions based on the real-time changing ventilation load demand. It establishes an intelligent control mechanism with "on-demand air supply" as the core, and uses a genetic algorithm (GA) to realize the automated optimization of hyperparameters of the support vector regression (SVR) model, effectively overcoming the limitations of small sample data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of the present invention provides an intelligent optimization control method for mine ventilation fans based on GA-SVR, comprising: Obtain the required ventilation volume, air pressure, and selected mode; An initial population is generated, and each individual in the population encodes a combination of blade angle and motor frequency; Based on the required ventilation volume, air pressure and mode, a trained support vector regression model is used to predict the total pressure efficiency and air pressure of each individual. Then, the reverse optimization solution is started, the second fitness value of each individual is calculated, and adaptive genetic operation and elite retention strategy are executed until convergence, so as to obtain the optimal total pressure efficiency and air pressure to control the mine ventilation fan. Among them, the support vector regression model uses a genetic algorithm for hyperparameter optimization during the training process.

[0007] Furthermore, the calculation of the second fitness value includes the product of the wind pressure error and the penalty coefficient, as well as the total pressure efficiency of the wind turbine.

[0008] Furthermore, the penalty coefficient is set according to the selected optimization mode. The penalty coefficient in the air volume priority mode is less than the penalty coefficient in the balanced mode, and the penalty coefficient in the balanced mode is less than the penalty coefficient in the air pressure priority mode.

[0009] Furthermore, in adaptive genetic operations, when the number of generations in which the population stagnates exceeds a set value, the mutation probability is increased and the crossover probability is decreased.

[0010] Furthermore, in the adaptive genetic operation, a linear adjustment strategy is used to adjust the crossover probability during the normal convergence phase. and mutation probability : ; ; in, The initial mutation probability, The initial crossover probability, is the preset total number of generations, and g is the current iteration number.

[0011] Furthermore, the hyperparameters include regularization parameters, kernel parameters, and insensitive loss parameters.

[0012] Furthermore, the training process of the support vector regression model includes: Generate an initial population, where each individual encodes a combination of hyperparameters; For each individual, three support vector regression models are configured using hyperparameters to predict total pressure efficiency, wind pressure, and input power, respectively. The support vector regression models are trained on the training set, and the first fitness value of each individual is calculated on the test set. Genetic operations and an elite retention strategy are performed until convergence to obtain the optimal combination of hyperparameters. Using all the training data, retrain the support vector regression model with the optimal combination of hyperparameters and evaluate its performance. If the performance meets the requirements, the support vector regression model training is complete.

[0013] A second aspect of the present invention provides an intelligent optimization control system for mine ventilation fans based on GA-SVR, comprising: The model training module is configured to use a genetic algorithm to optimize the hyperparameters of the support vector regression model during training. The ventilation demand acquisition module is configured to acquire the required ventilation volume, air pressure, and selected mode. The initialization module is configured to generate an initial population, where each individual in the population encodes a combination of blade angle and motor frequency. The inverse optimization control module is configured to: based on the required ventilation volume, air pressure and mode, use a trained support vector regression model to predict the total pressure efficiency and air pressure of each individual, then start the inverse optimization solution, calculate the second fitness value of each individual, and execute adaptive genetic operations and elite retention strategies until convergence, so as to obtain the optimal total pressure efficiency and air pressure to control the mine ventilation fan.

[0014] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the intelligent optimization control method for mine ventilation fans based on GA-SVR as described above.

[0015] A fourth aspect of the present invention provides a computer device including a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein the processor executes the program to implement the steps of the intelligent optimization control method for mine ventilation fans based on GA-SVR as described above.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention constructs a parameter inverse inversion framework, which can deduce the optimal operating conditions based on the real-time changing ventilation load demand, and establishes an intelligent control mechanism with "on-demand air supply" as the core.

[0017] The genetic algorithm (GA) was used to automatically optimize the hyperparameters of the support vector regression (SVR) model, which effectively overcame the limitations of small sample data and ensured excellent prediction accuracy while significantly improving the generalization ability of the SVR model.

[0018] This invention develops multiple optimization modes covering airflow guidance, air pressure guidance, and balanced control, and introduces an adaptive penalty function mechanism to dynamically handle safety boundaries, which greatly enhances the adaptability to operating conditions. Attached Figure Description

[0019] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0020] Figure 1 This is the GA of Embodiment 1 of the present invention. SVR agent model training flowchart; Figure 2 This is a framework diagram of the multi-mode reverse optimization algorithm of Embodiment 1 of the present invention; Figure 3 This is a schematic diagram comparing the reverse optimization process of different modes under normal ventilation requirements in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram comparing the reverse optimization process of different modes under extreme ventilation requirements in Embodiment 1 of the present invention; Figure 5 This is a block diagram of the control system according to Embodiment 2 of the present invention; Figure 6 This is a schematic diagram of the core function interface of the host computer software in Embodiment 2 of the present invention; Figure 7 This is a schematic diagram of the structure of a computer device according to Embodiment 4 of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0022] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0023] Example 1 This embodiment provides an intelligent optimization control method for mine ventilation fans based on GA-SVR.

[0024] Current ventilation control methods mainly remain at the stage of forward mapping from operating parameters to performance results, and do not have the "reverse solution" function of locking the optimal control parameters based on the given air volume and air pressure targets. Therefore, on the basis of strictly adhering to the bottom line of underground safety, establishing a control mechanism that can achieve dynamic optimization of fan energy efficiency has become the core technical bottleneck restricting the intelligent upgrading of mines.

[0025] The intelligent optimization control method for mine ventilation fans based on GA-SVR provided in this embodiment uses a data-driven surrogate model to replace traditional simulation, realizing millisecond-level prediction of fan performance. It also coordinates with a multi-mode genetic algorithm to perform inverse optimization of operating parameters, and can dynamically lock the optimal energy-efficiency operating condition while strictly adhering to underground safety constraints, thereby significantly reducing the operating energy consumption of the mine ventilation system.

[0026] The intelligent optimization control method for mine ventilation fans based on GA-SVR provided in this embodiment mainly consists of two parts: the first part is to use a genetic algorithm to optimize the support vector regression (GA-SVR) algorithm and build a fast predictive proxy model for the performance of the ventilation fan; the second part is to apply a multi-mode adaptive genetic algorithm to perform inverse inversion and optimization control of the operating parameters.

[0027] The intelligent optimization control method for mine ventilation fans based on GA-SVR provided in this embodiment, such as Figure 1 As shown, the specific steps include: Step 1: Data Acquisition and Preprocessing.

[0028] Step 101: On the ventilation fan test platform, establish a three-factor, multi-level orthogonal experimental scheme, and adjust the following three key variables of the mine ventilation system: Blade installation angle As a core geometric adjustment parameter, it is precisely controlled by a servo adjustment mechanism, and the measurement angle range needs to cover the commonly used working range; motor operating frequency The operating frequency of the motor determines the speed of the blades. The speed of the fan blades is controlled by adjusting the motor frequency through a frequency converter. Pipeline resistance characteristics: By adjusting the opening of electric dampers, the ventilation resistance changes in different underground roadways are simulated, and the air volume under different operating conditions is recorded. Wind pressure .

[0029] In addition to collecting data on the above-mentioned ventilators at different blade angles Motor frequency Air volume corresponding to different operating conditions Wind pressure The operating data also needs to record the input power. and the calculated total pressure efficiency The dataset D = {( )}.

[0030] Step 102: Clean, interpolate, and normalize the data to eliminate the influence of dimensions; divide the data into training and test sets according to the blade angle, with 80% of the data used as the training set and 20% of the data used as the test set.

[0031] Step 2, GA SVR performance prediction model construction.

[0032] Step 201: Build a Support Vector Machine (SVR) model, with the blade angle as the input. Motor frequency Air volume ,Right now The output is the total pressure efficiency. Wind pressure and input power ,Right now The RBF radial basis kernel function is used.

[0033] This embodiment uses a MultiOutputRegressor wrapper model to simultaneously predict the total pressure efficiency of the wind turbine. Wind pressure and input power This method trains three independent SVR regression models, but uses the same hyperparameter configuration.

[0034] The mathematical expression for each independent SVR training model is as follows: ; in, It is the first Output The corresponding high-dimensional feature space weight vector; It is the feature map corresponding to the RBF kernel; It is the first The bias term corresponding to each output.

[0035] Specifically, the kernel function corresponding to the RBF kernel satisfies:

[0036] In the formula, For hyperparameters, This represents the currently input sample data. This represents another set of input sample data. Represents the first in the training set One sample data, Represents the first in the training set Sample data.

[0037] Step 202: Set the genetic algorithm (GA) parameters, including population size, maximum number of generations, selection operator, crossover operator, and mutation operator.

[0038] To optimize SVR hyperparameters using real-number encoding, the three-dimensional vector of each individual is [ The parameter optimization range is set, and in this embodiment, the optimization range of the SVR hyperparameter is [missing information]. , , .

[0039] Step 203: Define the first fitness function as the negative R² value of the prediction results on the test set: ; In the formula: The coefficient of determination is represented by , where n is the total number of samples in the test set. The sample index is used to encode individuals for the genetic algorithm; The output variable index has values ​​of 1, 2, and 3, which correspond to total pressure efficiency, fan pressure, and input power, respectively. The true value of the sample; For using parameters ( The SVR model prediction value; Let be the sample mean of the j-th output variable; R² is the coefficient of determination, a larger value indicates a more accurate prediction, so a negative value is taken as the minimization objective.

[0040] Step 204: Optimize SVR hyperparameters (including regularization parameter C, kernel parameter γ, and insensitive loss parameter) using a genetic algorithm. The specific process is as follows: (1) Initialize the GA population and randomly generate 50 groups. combination; (2) For each individual in each generation, configure 3 SVR models using the hyperparameters of this group, and train the SVR models on the training set (using 3-fold cross-validation to improve training speed). Then, calculate the output variables on the test set. The value is used to obtain the individual's first fitness. (3) Perform the bidding selection operation, simulate binary crossover, and polynomial mutation to generate the next generation population; (4) Repeat process (2)-(3) until the maximum number of generations or fitness convergence is reached; (5) Selecting the optimal individual parameters ; (6) Use all training data to retrain the final SVR model with optimal hyperparameters.

[0041] It should be noted that in the data preprocessing stage, normalization eliminates the differences in the units of different variables. After the original data and the SVR model are trained, the output prediction results are normalized values. It is necessary to use denormalization to restore the normalized values ​​to units that can be directly used in actual engineering (for example, restoring 0.5 to 500Pa wind pressure) through the inverse operation of normalization (i.e., denormalization).

[0042] Step 205: SVR model evaluation.

[0043] The performance of the SVR model was evaluated on an independent test set, with key metrics including the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE).

[0044] In this embodiment, the final SVR model achieved a comprehensive R² of 0.992, an RMSE of 35.9042, and a MAE of 8.5916 on the test set, meeting the engineering accuracy requirements; the optimal hyperparameters found were... , , .

[0045] Step 3: Multi-mode reverse optimization modeling.

[0046] Step 301: Optimize problem modeling.

[0047] To address the performance control requirements of mine ventilation systems, this study conducts inverse optimization research on ventilation operating parameters based on the established forward performance prediction model (GA-SVR). The specific objective is to optimize the target ventilation volume and target fan pressure requirements under different ventilation scenarios, using the total pressure efficiency of the fan as the metric. To maximize the optimization objective, we solve for the optimal wind turbine operating parameters that satisfy this requirement: motor current and frequency. With blade angle The optimization algorithm selected is the genetic algorithm (GA), which simulates the iterative optimization mechanism of natural selection and genetic mutation to find the global optimal solution to this nonlinear optimization problem.

[0048] For a given target air volume and target wind pressure The following optimization problem is constructed: ; in, This represents the minimum angle value used in the experimental data for training the SVR model. This represents the maximum angle value from the experimental data used to train the SVR model. This represents the minimum frequency value of the experimental data used to train the SVR model. This represents the maximum frequency value of the experimental data used to train the SVR model. This indicates the total pressure efficiency of the fan (which is determined by the blade angle). Motor frequency Target air volume The function is also the target to be maximized in this optimization.

[0049] Step 302: Adaptive genetic algorithm design.

[0050] To improve the optimization efficiency and robustness of the standard genetic algorithm (GA), this embodiment introduces an adaptive mechanism to improve its core operators and constraint handling strategies. The specific design is as follows: The crossover probability in generation g is dynamically adjusted based on the population characteristics at different evolutionary stages. and mutation probability , For the current generation: When the population stagnates for more than 5 generations, increase the mutation probability to 0.2 and appropriately decrease the crossover probability to 0.7. During the normal convergence phase, a linear adjustment strategy is adopted: ; ; in, The initial mutation probability, The initial crossover probability, This is the preset total number of generations.

[0051] At a fixed target air volume Based on this, the constrained optimization problem is transformed into an unconstrained optimization problem using a penalty term function. The second fitness function of the genetic algorithm is: ; in, The original objective function value, The actual wind pressure value predicted for the decision variable. To constrain the relaxation factor (to avoid excessive punishment for small deviations). This is the penalty coefficient.

[0052] Step 303: Implementation of three optimization modes.

[0053] For the above constraints, different priorities will lead to different optimal solutions. Now, an algorithm framework based on Python is designed, including three modes: air volume priority mode, air pressure priority mode, and balanced mode. The penalty coefficient of the fitness function corresponding to these three modes is different, depending on the weight setting of fan air volume and fan air pressure. If air volume is prioritized, the fitness function will penalize the air pressure error less; if air pressure is prioritized, the fitness function will penalize the air pressure error more. Dynamic optimization is achieved by adjusting the strictness of the air pressure constraint. Finally, the optimal solution is selected according to the actual working conditions.

[0054] The specific design of the three modes is as follows: (1) Airflow priority mode: This mode is suitable for periods with high ventilation demand and high safety margin, and sets a penalty coefficient. This mode is used to weaken wind pressure constraints and focus on improving efficiency. It is suitable for non-production periods or periods with low methane concentrations. (2) Wind pressure priority mode: This mode sets a penalty coefficient. Used to enhance wind pressure constraint, suitable for dangerous periods such as high gas and high temperature, prioritizing ventilation safety, with efficiency as a secondary consideration; (3) Balanced mode: This mode sets a penalty coefficient. It balances efficiency and safety.

[0055] Step 4: Optimize the execution process online, such as... Figure 2 As shown, the specific process is as follows: Step 401: Specify the current required ventilation volume Wind pressure Select the optimization mode (air volume priority mode / air pressure priority mode / balance mode) and set the algorithm parameters (e.g., population size 20, maximum number of generations 100).

[0056] Step 402: Generate the initial population, with each individual encoded as follows: The angle and frequency parameters are normalized to the range of [0,1] for processing.

[0057] Step 403: Use the SVR model trained in Step 2 to predict each individual. , and .

[0058] Step 404: Calculate the second fitness value of each individual using formula (7) and perform adaptive genetic operations, including selection (roulette wheel selection), crossover (simulated binary crossover), mutation (polynomial mutation), and adopt an elite retention strategy.

[0059] Step 405: Iterate through steps 403 and 404 until the maximum algebra is reached to obtain the final convergent value. and .

[0060] Test condition 1: Normal ventilation requirements.

[0061] Set target parameters 2500 m³ / min 1500Pa, this setting meets the main ventilation or auxiliary fan requirements of most medium and large-sized mines. Based on the method of this embodiment, the iterative process of the three modes with respect to blade angle, motor frequency, total pressure efficiency, fan pressure, input power, and fitness function value is as follows: Figure 3As shown in the figure. The maximum efficiency of the three modes was 79.63 (airflow priority mode), 79.62 (air pressure priority mode), and 79.62 (balanced mode). The comparison of the final results of the three optimization modes under this test condition is shown in Table 1.

[0062] Table 1 2500 m³ / min Comparison of optimization mode results at 1500Pa

[0063] Test Condition 2: Extreme ventilation requirements.

[0064] Set target parameters 2500 m³ / min 2400Pa is a setting commonly used in applications such as high-resistance ventilation networks, long-distance tunneling, or deep mining. Based on the method of this embodiment, the optimization and iteration process of each indicator under different modes, such as... Figure 4 As shown in the figure. The maximum efficiency of the three modes was 77.04 (airflow priority mode), 75.21 (air pressure priority mode), and 75.21 (balanced mode). The comparison of the final results of the three optimization modes under this test condition is shown in Table 2.

[0065] Table 2 2500 m³ / min Comparison of optimization mode results at 2400Pa

[0066] This test mainly demonstrates the difference between the air volume priority mode and the air pressure priority mode. When there is an extreme ventilation demand, the air volume priority mode will aim to maximize efficiency and thus weaken the constraint of air pressure.

[0067] The intelligent optimization control method for mine ventilation fans based on GA-SVR provided in this embodiment uses a genetic algorithm (GA) to realize the automatic optimization of hyperparameters of the support vector regression (SVR) model, effectively overcoming the limitations of small sample data. It can significantly improve the generalization ability of the SVR model while ensuring excellent prediction accuracy (the surrogate model can achieve an R²>0.97 prediction accuracy).

[0068] The intelligent optimization control method for mine ventilation fans based on GA-SVR provided in this embodiment constructs a parameter inverse inversion framework, which can deduce the optimal operating conditions based on the real-time changing ventilation load demand, and establishes an intelligent control mechanism with "on-demand air supply" as the core.

[0069] The intelligent optimization control method for mine ventilation fans based on GA-SVR provided in this embodiment has developed multiple optimization modes covering air volume guidance, air pressure guidance and balanced control, and introduced an adaptive penalty function mechanism to dynamically handle safety boundaries, which greatly enhances the adaptability of operating conditions.

[0070] The intelligent optimization control method for mine ventilation fans based on GA-SVR provided in this embodiment has extremely high computational efficiency. The computation time for a single optimization based on the trained prediction model can be up to 2 seconds. It can be seamlessly embedded into the existing ventilation monitoring system and achieves significant energy consumption reduction and operational efficiency improvement through online real-time optimization.

[0071] Example 2 The intelligent optimization control system for mine ventilation fans based on GA-SVR provided in this embodiment, such as Figure 5 As shown, a master-slave distributed control architecture is adopted, which includes three parts: a host computer control subsystem, a PLC control subsystem, and an actuator. It aims to achieve intelligent prediction and reverse control of the wind turbine's operating status through an offline trained GA-SVR (Genetic Optimization Algorithm-Support Vector Regression) model.

[0072] The host computer and the slave computer are physically connected via Ethernet and use the Modbus TCP industrial communication protocol for data exchange.

[0073] The host computer (PC) is equipped with the intelligent optimization control method for mine ventilation fans based on GA-SVR described in Example 1, and needs to have an Ethernet interface. Its core functional interface is as follows: Figure 6 As shown, the host computer software is mainly responsible for the calculation of complex algorithm models, human-computer interaction (HMI), target parameter setting, and control command generation.

[0074] The lower-level machine uses a programmable logic controller (PLC) as its core execution unit, responsible for receiving instructions from the upper-level machine and directly controlling the field devices. As a Modbus TCP server, it needs to have an Ethernet communication module and internally has pre-set holding registers for storing control parameters.

[0075] The actuators include a motor frequency converter and a blade angle adjustment mechanism. Both communicate with the PLC via the Modbus RTU protocol. The frequency converter receives frequency commands from the PLC to control and adjust the speed of the fan blades; the blade angle adjustment mechanism receives angle commands from the PLC to dynamically adjust the rotation angle of the fan blades.

[0076] The host computer is equipped with: The data analysis and preprocessing module is configured to: collect the operating data of the mine ventilation fan, perform preprocessing, and then divide the data into training and testing sets; The model training module is configured to: generate an initial population of hyperparameters, obtain the optimal SVR hyperparameters through a genetic algorithm, and train a GA-SVR positive prediction model. The ventilation demand acquisition module is configured to acquire the required ventilation volume and air pressure. The initialization module is configured to generate an initial population, where each individual in the population encodes a combination of blade angle and motor frequency. The control module is configured to: based on the required ventilation volume and air pressure, use a trained support vector regression model to predict the total pressure efficiency and air pressure of each individual, calculate the second fitness value of each individual, and execute adaptive genetic operations and elite retention strategies until convergence to obtain the optimal total pressure efficiency and air pressure to control the mine ventilation fan. Among them, the support vector regression model uses a genetic algorithm for hyperparameter optimization during the training process.

[0077] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0078] Example 3 This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the intelligent optimization control method for mine ventilation fans based on GA-SVR as described in Embodiment 1 above.

[0079] Example 4 This embodiment provides a computer device, such as... Figure 7 As shown, the system includes a computer-readable storage medium 1003, a processor 1001, a communication interface 1002, and a computer program stored on the computer-readable storage medium 1003 and executable on the processor 1001. The processor 1001, communication interface 1002, and computer-readable storage medium 1003 can be connected via a bus or other means. The communication interface 1002 is used to receive and send data. When the processor 1001 executes the program, it implements the steps in the intelligent optimization control method for mine ventilation fans based on GA-SVR as described in Embodiment 1 above.

[0080] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent optimization control of mine ventilation fans based on GA-SVR, characterized in that, include: Obtain the required ventilation volume, air pressure, and selected mode; An initial population is generated, and each individual in the population encodes a combination of blade angle and motor frequency; Based on the required ventilation volume, air pressure and mode, a trained support vector regression model is used to predict the total pressure efficiency and air pressure of each individual. Then, the reverse optimization solution is started, the second fitness value of each individual is calculated, and adaptive genetic operation and elite retention strategy are executed until convergence, so as to obtain the optimal total pressure efficiency and air pressure to control the mine ventilation fan. Among them, the support vector regression model uses a genetic algorithm for hyperparameter optimization during the training process.

2. The intelligent optimization control method for mine ventilation fans based on GA-SVR as described in claim 1, characterized in that, The calculation of the second fitness value includes the product of wind pressure error and penalty coefficient, as well as the total pressure efficiency of the wind turbine.

3. The intelligent optimization control method for mine ventilation fans based on GA-SVR as described in claim 2, characterized in that, The penalty coefficient is set according to the selected optimization mode. The penalty coefficient in the air volume priority mode is less than that in the balanced mode, and the penalty coefficient in the balanced mode is less than that in the air pressure priority mode.

4. The intelligent optimization control method for mine ventilation fans based on GA-SVR as described in claim 1, characterized in that, In adaptive genetic operations, when the number of generations in which the population stagnates exceeds a set value, the mutation probability is increased and the crossover probability is decreased.

5. The intelligent optimization control method for mine ventilation fans based on GA-SVR as described in claim 1, characterized in that, In adaptive genetic operations, during the normal convergence phase, a linear adjustment strategy is used to adjust the crossover probability. and mutation probability : ； ； in, The initial mutation probability, The initial crossover probability, is the preset total number of generations, and g is the current iteration number.

6. The intelligent optimization control method for mine ventilation fans based on GA-SVR as described in claim 1, characterized in that, The hyperparameters include regularization parameters, kernel parameters, and insensitive loss parameters.

7. The intelligent optimization control method for mine ventilation fans based on GA-SVR as described in claim 1, characterized in that, The training process of the support vector regression model includes: Generate an initial population, where each individual encodes a combination of hyperparameters; For each individual, three support vector regression models are configured using hyperparameters to predict total pressure efficiency, wind pressure, and input power, respectively. The support vector regression models are trained on the training set, and the first fitness value of each individual is calculated on the test set. Genetic operations and an elite retention strategy are performed until convergence to obtain the optimal combination of hyperparameters. Using all the training data, retrain the support vector regression model with the optimal combination of hyperparameters and evaluate its performance. If the performance meets the requirements, the support vector regression model training is complete.

8. A mine ventilation fan intelligent optimization control system based on GA-SVR, characterized in that, include: The model training module is configured to use a genetic algorithm to optimize the hyperparameters of the support vector regression model during training. The ventilation demand acquisition module is configured to acquire the required ventilation volume, air pressure, and selected mode. The initialization module is configured to generate an initial population, where each individual in the population encodes a combination of blade angle and motor frequency. The inverse optimization control module is configured to: based on the required ventilation volume, air pressure and mode, use a trained support vector regression model to predict the total pressure efficiency and air pressure of each individual, then start the inverse optimization solution, calculate the second fitness value of each individual, and execute adaptive genetic operations and elite retention strategies until convergence, so as to obtain the optimal total pressure efficiency and air pressure to control the mine ventilation fan.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the intelligent optimization control method for mine ventilation fans based on GA-SVR as described in any one of claims 1-7.

10. A computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the intelligent optimization control method for mine ventilation fans based on GA-SVR as described in any one of claims 1-7.

Images