Mine fan ventilation control method and system based on digital twinning

Abstract

The present application relates to the technical field of mine ventilation control, and particularly relates to a mine fan ventilation control method and system based on digital twinning. The method comprises data preprocessing on acquired mine monitoring data; feature extraction on the preprocessed mine monitoring data; prediction of the data feature and fan air volume relationship by using a random forest algorithm; calculation of the data feature and fan air volume relationship by using a digital twinning model to obtain a fan ventilation control strategy; and generation of a PLC control instruction based on the fan ventilation control strategy. Through multi-source data fusion and deep analysis, the real-time ventilation demand of the mine can be accurately grasped. Based on the random forest algorithm prediction and the digital twinning model calculation, dynamic and accurate adjustment of the fan air volume is realized, the harmful gas concentration in each region of the mine is maintained within a safety threshold, the oxygen content is sufficient, an ideal environment is created for personnel operation and equipment operation, and the safety and reliability of the ventilation system are significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mine ventilation control, in particular to a mine fan ventilation control method and system based on digital twinning. BACKGROUND

[0002] In today's mining industry, with the continuous deepening and expansion of mining activities, the mine ventilation system is facing many challenges that need to be solved, and the demand for advanced ventilation control technology is increasingly urgent.

[0003] From the perspective of environmental complexity, the depth and scope of mining continue to expand, and the underground environment becomes extremely complex. On the one hand, the risk of harmful gas production and accumulation increases significantly, such as carbon monoxide, nitrogen dioxide and other toxic and harmful gases, as well as exhaust gas from diesel equipment operation, which not only seriously threatens the health of miners, but also poses a major hidden danger to safety production. On the other hand, multi-section mining makes the gas distribution and ventilation demand of each section differ greatly, and the traditional ventilation mode is difficult to meet this complex and differentiated demand.

[0004] In terms of ventilation mode, the traditional mine ventilation system mostly adopts a fixed ventilation mode, lacking the ability to adjust in real time according to actual working conditions. This mode cannot adapt to the dynamic changes in the number of personnel entering the mine with the production progress, as well as the special requirements of different operation scenarios and personnel density on air quality and oxygen content, resulting in not only a large waste of energy, but also difficulty in ensuring good ventilation effect, and being unable to effectively cope with the complex and variable environment underground.

[0005] From the perspective of safety and energy saving, the safety, reliability and energy saving of the mine ventilation system are the key to ensuring production. Ensuring the safety and comfort of the underground working environment, avoiding safety accidents caused by poor ventilation, and reducing energy consumption to reduce operating costs are important goals for the industry. However, existing ventilation technology has obvious shortcomings in these areas and is difficult to meet the needs of actual production.

[0006] Therefore, in order to solve the above problems and improve the overall performance of the mine ventilation system, an innovative technical solution is needed that can consider multiple factors such as the number of personnel underground, equipment operating conditions, and gas concentration in each section, to achieve real-time dynamic adjustment of ventilation volume, and use advanced control methods and monitoring technology to achieve precise ventilation, providing strong support for mine safety production and sustainable development. SUMMARY

[0007] In order to solve the above-mentioned problems, the present application provides a mine fan ventilation control method and system based on digital twinning.

[0008] In a first aspect, the present application provides a mine fan air volume dynamic control method based on monitoring data, which adopts the following technical solution:

[0009] A mine fan ventilation control method based on digital twinning, comprising:

[0010] Obtaining mine monitoring data;

[0011] Data preprocessing is performed on the obtained mine monitoring data;

[0012] Feature extraction is performed on the preprocessed mine monitoring data;

[0013] The relationship between data features and fan air volume is predicted using a random forest algorithm;

[0014] The relationship between data features and fan air volume is calculated using a digital twinning model to obtain a fan ventilation control strategy;

[0015] PLC control instructions are generated based on the fan ventilation control strategy.

[0016] Further, the obtained mine monitoring data includes static monitoring data, dynamic monitoring data, and equipment monitoring data, wherein the static monitoring data includes personnel data entering the mine, simultaneous diesel engine power data, and air supply volume data; the dynamic data includes mine harmful gas concentration, oxygen concentration, and carbon dioxide concentration; and the equipment monitoring data includes motor bearing and motor temperature data, vibration data, and electrical parameter data.

[0017] Further, the data preprocessing of the obtained mine monitoring data includes missing data processing and abnormal data processing, wherein the K- nearest neighbor algorithm is used to fill in the missing data, and the Z- score method is used to identify abnormal data.

[0018] Further, the feature extraction of the preprocessed mine monitoring data includes first dimensionality reduction feature extraction through PCA principal component analysis method, and based on wavelet transform to extract time series features, and further extracting features using a multi-layer autoencoder, and finally obtaining the last data features by feature fusion and selection of the features extracted by PCA principal component analysis, wavelet transform, and autoencoder respectively.

[0019] Further, the relationship between data features and fan air volume is predicted using a random forest algorithm, including using the CART algorithm to construct a decision tree, using mean square error as a node splitting measure, forming a random forest by constructing N decision trees, and using out-of-bag data to evaluate each decision tree, using out-of-bag error as the performance estimate of the random forest model on unseen data, and finally optimizing the random forest parameters through random search to obtain the relationship between data features and fan air volume.

[0020] Further, the wind fan ventilation control strategy is obtained by using the digital twin model to solve the data characteristics and the fan air volume relationship, including respectively constructing a physical ventilation model and a data-driven model based on the digital twin, and fusing the physical ventilation model and the data-driven model by using a weighted fusion method, setting a target of mine air quality and ventilation energy consumption, defining an optimization objective function and setting a constraint condition to solve the data characteristics and the fan air volume relationship, so as to simulate the running state of the ventilation system.

[0021] Further, the PLC control instruction is generated based on the wind fan ventilation control strategy, including converting the air volume control requirement in the wind fan ventilation control strategy into a motor speed control requirement of the fan, establishing a relationship between the air volume and the motor speed according to the performance curve of the fan, converting the motor speed control requirement into a PLC recognizable control instruction, and finally sending the generated PLC control instruction to the PLC controller of the mine ventilation system through the industrial Ethernet to realize real-time control of the fan.

[0022] The second aspect is a mine fan ventilation control system based on digital twin, including:

[0023] The data acquisition module is configured to acquire mine monitoring data.

[0024] The preprocessing module is configured to perform data preprocessing on the acquired mine monitoring data.

[0025] The feature extraction module is configured to perform feature extraction on the preprocessed mine monitoring data.

[0026] The prediction module is configured to predict the data characteristics and the fan air volume relationship by using a random forest algorithm.

[0027] The solving module is configured to solve the data characteristics and the fan air volume relationship by using a digital twin model to obtain a wind fan ventilation control strategy.

[0028] The control module is configured to generate a PLC control instruction based on the wind fan ventilation control strategy.

[0029] The third aspect provides a computer readable storage medium, wherein a plurality of instructions are stored, the instructions are suitable for being loaded and executed by a processor of a terminal device.

[0030] The fourth aspect provides a terminal device, including a processor and a computer readable storage medium, the processor is used to implement each instruction; the computer readable storage medium is used to store a plurality of instructions, the instructions are suitable for being loaded and executed by the processor to execute the mine fan ventilation control method based on the digital twin.

[0031] In summary, the present application has the following beneficial technical effects:

[0032] 1. Through multi-source data fusion and deep analysis, the real-time ventilation demand of the mine can be accurately grasped. Based on the random forest algorithm prediction and the digital twin model solution, dynamic and accurate adjustment of the fan air volume is realized, the harmful gas concentration in each area of the underground mine is maintained within the safety threshold, the oxygen content is sufficient, an ideal environment is created for personnel operation and equipment operation, and the safety and reliability of the ventilation system are significantly improved.

[0033] 2. Goodbye to the energy waste of traditional fixed ventilation mode, intelligently control the fan speed according to the actual working condition. While ensuring the ventilation effect, the fan energy consumption is greatly reduced. Taking the centrifugal fan as an example, the speed is accurately adjusted according to the linear relationship between air volume and speed. Compared with the traditional ventilation mode, the mine operation cost is effectively reduced, and green mining is promoted.

[0034] 3. Real-time analysis of equipment monitoring data realizes comprehensive perception of the running state of key equipment such as fan. By monitoring the motor bearing temperature, vibration data and electrical parameters, potential equipment fault hidden dangers are found in time, early warning and maintenance are arranged, and production interruption caused by equipment sudden failure is avoided. According to statistics, after the application of the technology, the continuity and stability of mine production are effectively guaranteed. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 is a digital twin-based mine fan ventilation control method schematic diagram of embodiment 1 of the present application;

[0036] Figure 2 is a digital twin-based mine fan ventilation control system schematic diagram of embodiment 1 of the present application. DETAILED DESCRIPTION

[0037] The present application will be further described in detail below in combination with the drawings.

[0038] Embodiment 1

[0039] Referring to Figure 1 , a digital twin-based mine fan ventilation control method of the present embodiment comprises:

[0040] obtaining mine monitoring data;

[0041] data preprocessing is performed on the obtained mine monitoring data;

[0042] feature extraction is performed on the preprocessed mine monitoring data;

[0043] the relationship between data features and fan air volume is predicted by using a random forest algorithm;

[0044] The data characteristics and fan air volume relationship are calculated by using a digital twin model to obtain a fan ventilation control strategy.

[0045] The PLC control instructions are generated based on the fan ventilation control strategy.

[0046] Specifically:

[0047] S1. Obtain mine monitoring data, including static monitoring data, dynamic monitoring data, and equipment monitoring data:

[0048] Static monitoring data:

[0049] Personnel data: records the number of personnel entering the mine, distribution area, and other information. It can be obtained through the attendance system or personnel positioning system at the mine entrance.

[0050] Simultaneous diesel engine power data: statistics the total power of diesel engines running simultaneously in the mine. Diesel engines consume oxygen and produce harmful gases when working in the mine, and their power is related to gas production. It can be obtained through the power monitoring device of the diesel engine.

[0051] Air supply volume data: measures the air supply volume of each area in the mine, reflecting the current working state of the ventilation system. It can be calculated using wind speed sensors and ventilation duct cross-sectional area.

[0052] Dynamic monitoring data:

[0053] Mine harmful gas concentration: such as carbon monoxide (CO), hydrogen sulfide (H2S), etc. These gases are extremely harmful to human health and need to be monitored in real time. Gas sensors can be used for detection.

[0054] Oxygen concentration: oxygen is essential for personnel to breathe, and low oxygen concentration in the mine can cause personnel to suffocate due to lack of oxygen. Data is obtained through oxygen sensors.

[0055] Carbon dioxide concentration: high carbon dioxide concentration can affect the respiratory function of personnel, and gas sensors are also used for monitoring.

[0056] Equipment monitoring data:

[0057] Motor bearing and motor temperature data: motors generate heat during operation, and high temperatures can cause motor damage. Temperature sensors are used to monitor the temperature of motor bearings and the motor body.

[0058] Vibration data: abnormal motor vibration may indicate that the equipment has a fault or is not installed stably. Vibration sensors are used to collect motor vibration data.

[0059] Electrical parameter data: This includes parameters such as motor current, voltage, and power. These parameters reflect the motor's operating status and energy consumption, and can be measured using electrical instruments.

[0060] S2. Data Cleaning

[0061] Missing data handling: The K-nearest neighbor algorithm is used for data imputation. For a sample with missing values... Calculate it with all other samples Euclidean distance between (in (Number of features). Select the nearest one. One sample, based on this The average of the feature values ​​corresponding to each sample To fill in missing values.

[0062] Outlier Handling: Outlier data is identified using the Z-score method. For each data point in the dataset... Calculate its Z-score value (in The mean of the dataset. (where the standard deviation of the dataset is denoted as ). If a data point is found to be outlier, it can be deleted or corrected by averaging adjacent normal data.

[0063] Data normalization: The min-max normalization method is used to map the data to the interval [0, 1]. For the original data... The result after normalization ,in and These are the minimum and maximum values ​​of the feature data, respectively.

[0064] S3. Feature Extraction:

[0065] Step 1: Before feature extraction, the preprocessed mine monitoring data needs to be divided into training and testing sets according to a certain ratio, typically 80:20 or 70:30. Let the preprocessed data matrix be... Its dimensions are ,in It is the sample size. It is the number of features. Divided into training set and test set At the same time, the corresponding labels (if there is a supervised learning scenario) are also divided accordingly.

[0066] Step 2: Preliminary dimensionality reduction based on principal component analysis (PCA)

[0067] Data centering: for training set , compute the mean vector of each column where is the number of training set samples, is the th sample vector. Subtract the mean vector from each row of to get the centered data matrix .

[0068] Covariance matrix computation: compute the covariance matrix of

[0069] .

[0070] Eigenvalue decomposition: perform eigenvalue decomposition on the covariance matrix , i.e. where is the orthogonal matrix composed of eigenvectors, is the diagonal matrix composed of eigenvalues.

[0071] Principal component selection: sort the eigenvalues in descending order and compute the cumulative variance contribution rate . When reaches 0.9, determine that the eigenvectors corresponding to the first largest eigenvalues form the projection matrix .

[0072] Data projection: multiply the centered training set by the projection matrix to get the reduced dimension training set . For the test set , also perform centering (using the mean vector computed using the training set) and then multiply by the projection matrix to get the reduced dimension test set .

[0073] Step 3: Wavelet transform to extract time series features (for dynamic monitoring data)

[0074] Data screening: screen out dynamic monitoring data (such as time series data of changes of mine harmful gas concentration over time, changes of oxygen concentration over time, etc.) from the reduced dimension data and , and denote them as and , respectively.

[0075] Wavelet basis selection: Daubechies wavelet is chosen as the wavelet basis because of its good compact support and vanishing moment properties, which can effectively capture the local features of the signal.

[0076] Wavelet decomposition: 3-layer wavelet decomposition is performed on each time series data in , resulting in approximation coefficients and detail coefficients , , . The wavelet decomposition formula is , where is the scaling function, is the wavelet function.

[0077] Feature extraction: features are extracted from the approximation coefficients and detail coefficients, including mean, variance, and energy. For example, the energy of the layer detail coefficient is . The extracted features are combined into a new feature vector, resulting in the wavelet-transformed training set features . The same wavelet decomposition and feature extraction operations are performed on , resulting in the wavelet-transformed test set features .

[0078] Step 4: Autoencoder further extracts features

[0079] Model construction: a multi-layer autoencoder is constructed, with the input layer dimension being the number of features (after PCA dimension reduction), and the hidden layer dimension set to half of the input layer dimension. The output of the encoder is , where is the weight matrix of the encoder, is the bias vector, with ReLU activation function. The output of the decoder is , where is the weight matrix of the decoder, is the bias vector, with linear activation function.

[0080] Model training: the mean square error loss function

[0081] is used to train the autoencoder, with the Adam optimization algorithm updating the parameters of the model, and the training number of rounds set to 100 rounds.

[0082] Feature extraction: after training, input into the encoder to get the output of the hidden layer, which is the training set features extracted by the autoencoder . Input Input into the encoder, get the test set features extracted from the autoencoder .

[0083] Step 5: Feature fusion

[0084] Fuse the PCA dimension-reduced features, wavelet transform extracted features and autoencoder extracted features. For the training set, concatenate , and by column into a new feature matrix . For the test set, concatenate , and by column into a new feature matrix .

[0085] Step 6: Feature selection

[0086] Use chi-square test to select features from the fused feature matrix (hypothesis is a supervised learning scenario, with corresponding class labels ).

[0087] Calculate chi-square value: for each feature in , calculate the chi-square value between it and the class label , where is the actual observation value, is the expected observation value.

[0088] Feature screening: select the top 80% of features with the highest chi-square values as the final training set features . According to the feature indices selected from the training set, select the corresponding features from to get the final test set features .

[0089] S4. Predict the relationship between data features and fan flow rate using random forest algorithm

[0090] 1. Before using the random forest algorithm, the data after feature extraction needs to be prepared. Assume that the data matrix obtained after feature extraction is , with dimensions , where is the number of samples, is the number of features; the corresponding fan flow rate label vector is , with dimensions .

[0091] Divide the data set into training set and ​Test set. To ensure the model's generalization ability, stratified sampling can be used, especially when the data has class imbalance, to ensure that the proportion of each class in the training and test sets is similar to that in the original dataset.

[0092] 2. Random Forest Model Construction and Training

[0093] 2.1 Basic Principles of Random Forest

[0094] Random forests are ensemble learning models composed of multiple decision trees. Each decision tree employs bootstrap sampling and random feature selection during its construction to increase the model's diversity and generalization ability.

[0095] 2.2 Self-service sampling

[0096] From the training set Random selection with replacement These samples form a new training subset. The corresponding tag subset is Because sampling with replacement was used, approximately [number missing] The samples will appear in In this context, the samples that are not selected constitute the out-of-bag (OOB) data.

[0097] 2.3 Random Feature Selection

[0098] When constructing each node of each decision tree, all [the nodes] are no longer considered. Instead of individual features, they are randomly selected. Features ,generally From this Select the optimal feature from the available features to split the node.

[0099] 2.4 Decision Tree Construction

[0100] The decision tree is constructed using the CART (Classification and Regression Trees) algorithm. For regression problems, the mean squared error (MSE) is commonly used as a measure of node splitting.

[0101] Let the sample set of the current node be... For a certain feature and the dividing point, Divided into and Two parts. The mean square error after partitioning is:

[0102] where and are and the number of samples in , and are and the mean squared error of sample labels in

[0103]

[0104]

[0105] where and are and the mean of sample labels in

[0106] The feature and split point that minimizes is chosen for node splitting, and the process is repeated until a stopping condition is met (e.g., the number of samples in a node is less than a certain threshold, the depth of the tree reaches a maximum, etc.).

[0107] 2.5 Random Forest Training

[0108] The process of self-sampling, random feature selection, and decision tree construction is repeated to build trees, forming a random forest .

[0109] 3. Model Evaluation and Optimization

[0110] 3.1 Out-of-Bag (OOB) Error Evaluation

[0111] During training, each decision tree is evaluated using out-of-bag data. For each sample , find those decision trees that did not use this sample for training, and use these trees to make predictions for , obtaining predicted values . The out-of-bag error is then:

[0112]

[0113] The out-of-bag error can be used as an estimate of the model's performance on unseen data, serving as a preliminary assessment of the model's generalization ability.

[0114] 3.2 Test Set Evaluation

[0115] The test set The trained random forest model is evaluated using common metrics including mean squared error (MSE), mean absolute error (MAE), and coefficient of determination (R²). 2 ).

[0116]

[0117]

[0118]

[0119] in It is the number of test set samples. The model is for the first The predicted value for each sample, It is the mean of the labels of the test set samples.

[0120] 3.3 Model Optimization

[0121] Parameter tuning: Parameters of a random forest, such as the number of decision trees, can be tuned using methods such as grid search or random search. The number of features randomly selected for each node The maximum depth of the tree, etc.

[0122] Feature importance analysis: Random forests can calculate the importance of each feature, rank them by importance, remove unimportant features, reduce model complexity, and improve the model's generalization ability. Importance It can be calculated in the following ways:

[0123] For each decision tree Calculate the prediction error of out-of-bag data under the original features. Then, the features in the out-of-bag data are randomly shuffled. The value is used to recalculate the prediction error. Then the characteristics In decision tree The importance of is

[0124] . All features in the decision tree The importance is averaged to obtain the features. Ultimate importance:

[0125] .

[0126] 4. Combining Bayesian optimization and feature interaction analysis

[0127] 4.1 Bayesian Optimization of Random Forest Parameters

[0128] Traditional grid search and random search are inefficient for parameter tuning, especially in high-dimensional parameter spaces. Bayesian optimization, a global optimization method based on a probabilistic model, can adaptively select the next set of parameters to be evaluated based on historical parameter evaluation results, thus finding the optimal parameter combination more quickly.

[0129] Bayesian optimization approximates the objective function (such as out-of-bag error) by constructing a surrogate model (such as a Gaussian process) and selects the next set of parameters based on the posterior distribution of the surrogate model. The specific steps are as follows:

[0130] Initialize a set of parameters, train and evaluate the random forest model, and obtain the objective function value.

[0131] Based on this historical data, a Gaussian process model is constructed to predict the mean and variance of the objective function under different parameter combinations.

[0132] Use the acquisition function (such as the expected improvement function) to select the next most promising combination of parameters.

[0133] The Gaussian process model is updated by training and evaluating new parameter combinations.

[0134] Repeat the above steps until the stopping condition is met (such as reaching the maximum number of iterations).

[0135] 4.2 Feature Interaction Analysis

[0136] In addition to considering the importance of individual features, we can also analyze the impact of interactions between features on wind turbine airflow prediction. Feature interaction detection methods based on random forests can be used, such as discovering interactions between features by calculating the importance of conditional features.

[0137] Let features and characteristics Calculation on fixed features Under the value of , feature Importance .if With features General importance If there is a significant difference, it indicates a characteristic. and characteristics There are interactions between them. In this way, more information hidden in the data can be uncovered, further improving the predictive performance of the model.

[0138] 5. Predict fan air volume

[0139] The preprocessed and feature-extracted new data is input into the trained random forest model, and for each decision tree... , the predicted value The final prediction of the random forest is the average of all the decision tree predictions: .

[0140] Through the above steps, the relationship between data features and fan air volume can be accurately predicted using the random forest algorithm, and the performance and interpretability of the model can be further improved by expansion.

[0141] S5. Utilize the digital twin model to solve the relationship between data features and fan air volume, and obtain the fan ventilation control strategy, wherein

[0142] 1. Digital twin model construction

[0143] Physical model construction

[0144] Ventilation network topology modeling: The mine ventilation system is abstracted as a network topology structure composed of nodes and branches. Nodes represent the intersection of ventilation roadways, fan locations, chambers, etc., and branches represent ventilation roadways. Using graph theory, let be the ventilation network topology graph, where is the node set, is the branch set. For each branch , define its length , cross-sectional area , friction resistance coefficient , etc.

[0145] Fan characteristic modeling: The performance curve of the fan can usually be fitted with a polynomial function. Assume that the relationship between the fan pressure and the air volume can be expressed as

[0146] , where is a coefficient obtained through fan performance testing. At the same time, considering the relationship between the efficiency of the fan and the air volume , it can be fitted as

[0147] .

[0148] Airflow modeling: According to the principles of fluid mechanics, the airflow in the ventilation roadway follows the laws of energy conservation and mass conservation. For branch , its resistance can be calculated using the Darcy-Weisbach formula: , where is the resistance coefficient of the branch, is the air volume in the branch. The mass conservation law is expressed as , where and These are the sets of branches flowing into and out of a node.

[0149] Data-driven model building

[0150] Machine learning model selection: Long Short-Term Memory (LSTM) network is used to capture the dynamic characteristics of the mine ventilation system. The input gate of the LSTM unit... Forgotten Gate Output gate and cell state The calculation formula is as follows:

[0151]

[0152]

[0153]

[0154]

[0155]

[0156]

[0157] in, It is a moment The input vector, It is the hidden state from the previous moment. It is a weight matrix. It is a bias vector. It is the sigmoid function. It is the hyperbolic tangent function. It is element-wise multiplication.

[0158] Model Training: Historical monitoring data was used as the training set, with data features (such as harmful gas concentration and number of people) as input and fan airflow and other ventilation parameters as output to train the LSTM model. The mean squared error loss function was employed. ,in It is the sample size. It is the actual value. These are predicted values. The model parameters are updated using stochastic gradient descent (SGD) or improved algorithms such as the Adam algorithm.

[0159] Model fusion

[0160] A weighted fusion method is used to integrate the physical model and the data-driven model. Let the output of the physical model be... The output of the data-driven model is , the output after fusion wherein and are weight coefficients, and The weight coefficients can be optimized by a cross-validation method so that the error of the fusion model on the validation set is minimized.

[0161] 2. Solve the relationship between data characteristics and fan air volume

[0162] Optimization objective function definition

[0163] To ensure the air quality in the mine and reduce the energy consumption of ventilation, the optimization objective function is defined Let be the concentration of the th harmful gas, be the safety concentration threshold of the harmful gas; be the air volume of the th fan, be the power of the th fan, and be the penalty coefficient. The objective function can be expressed as:

[0164]

[0165] wherein is the number of types of harmful gases, is the number of fans.

[0166] Constraint condition setting

[0167] Air volume balance constraint: for each node in the ventilation network, the air volume flowing into the node is equal to the air volume flowing out of the node, i.e. .

[0168] Fan performance constraint: the air volume and air pressure of the fan need to meet its performance curve, i.e. , and wherein is the maximum air volume of the th fan.

[0169] Gas concentration constraint: the harmful gas concentration in each area of the mine cannot exceed the safety threshold, i.e.

[0170] Optimization algorithm solution

[0171] The particle swarm optimization algorithm (PSO) is used to solve the optimization problem. Let the position vector of the particle represent the air volume of each fan, and the velocity vector . The update formula of the particle is as follows: .

[0172]

[0173] where, is the inertia weight, and is the acceleration constant, and is a random number between 0 and 1, is the historical best position of the th particle, is the global best position, is the iteration number, denotes the dimension.

[0174] 3. Developing a fan ventilation control strategy

[0175] Dynamic zoning control: According to the regional functions and harmful gas distribution in the mine, the mine is divided into different ventilation zones. For each zone, an independent ventilation control strategy is developed based on the calculated optimal air volume. For example, personnel operation areas, coal storage areas, etc. are divided into different zones, and the personnel operation area is focused on oxygen supply and harmful gas dilution, and the coal storage area is focused on dust concentration control.

[0176] Real-time feedback regulation: Use the digital twin model to simulate the running state of the ventilation system in real time, and compare the simulation results with the actual monitoring data. If there is a deviation, adjust the air volume and running state of the fan in time. For example, when the actual monitoring of the harmful gas concentration in a certain area is higher than the simulation value, increase the air supply volume of the area and re-optimize the calculation.

[0177] Emergency response strategy: In view of the possible emergencies in the mine (such as fire, gas explosion, etc.), an emergency ventilation control strategy is developed. When an emergency occurs, according to the type and location of the event, quickly adjust the running mode of the fan to ensure that harmful gases can be discharged in time, and provide good ventilation conditions for personnel escape and rescue work. For example, when a fire occurs, use the reverse wind technology to introduce fresh air into the fire area, while discharging smoke and harmful gases out of the mine.

[0178] S6. Strategy conversion to motor parameters

[0179] The fan ventilation control strategy usually gives the target air volume, wind pressure and other ventilation parameters, while the PLC controls the running parameters of the fan motor, such as speed, voltage, current, etc. First, the mapping relationship between ventilation parameters and motor parameters needs to be established.

[0180] Air volume-speed relationship: For centrifugal fans, according to the similarity law, when the geometric shape of the fan and the air density are constant, the air volume is proportional to the speed , that is . Assume the current fan speed and air volume are known , and the target air volume required by the control strategy is known, then the target speed can be calculated.

[0181] Wind pressure-speed relationship: Wind pressure is proportional to the square of the speed, that is . According to this relationship, it can be further verified whether the target speed can meet the wind pressure requirement.

[0182] 2. Motor parameter conversion to PLC signal

[0183] PLC usually outputs voltage or current signals through analog output modules to control the speed regulating device of the motor (such as frequency converter). It is necessary to determine the signal range that the motor speed regulating device can accept, such as 0-10V voltage signal or 4-20mA current signal, and the corresponding motor speed range of the signal range.

[0184] Assuming that the motor speed regulating device accepts 0-10V voltage signal, and the corresponding motor speed range is 0-1500r / min. If the calculated target speed is , then the corresponding PLC output voltage .

[0185] 3. Instruction encoding and issuing

[0186] Encode the calculated PLC output signal value to convert it into a digital quantity that PLC can recognize and process. If the resolution of the analog output module of the PLC is 12 bits (i.e. 0-4095), then convert the above calculated voltage value to the corresponding digital quantity . Finally, through industrial communication network (such as Ethernet, Profibus, etc.), the encoded instruction is issued to the PLC controller.

[0187] Example 2

[0188] Figure 2 This example provides a mine fan ventilation control system based on digital twinning, as shown in Figure 2 , which is a mine ventilation control scheme, covering three parts of ventilation stage division and related parameter setting, ventilation equipment control architecture and each section gas monitoring.

[0189] In the ventilation stage division and related parameter setting, it is divided into four stages, each stage corresponds to different main fan air supply (from 40% to 100% increment), different interval of people into the well (from 0-39% to 80-100% increment), also gives the parameters of diesel equipment , and the corresponding PIC frequency (from 20HZ to 50HZ increment).

[0190] The ventilation equipment control architecture part, the main fan is connected with the core module through the PLC, the core module is connected with two local fans (local fan 1#, local fan 2#) respectively, each local fan also communicates with the core module through the PLC, which embodies the centralized control of the core module to the main fan and the local fan.

[0191] In terms of gas monitoring in each section, O2, CO, CO2 and NO2 are monitored in 1-4 sections, and these monitoring data can be used to assist the adjustment of the ventilation control strategy.

[0192] A computer readable storage medium, wherein a plurality of instructions are stored, the instructions are suitable for being loaded and executed by a processor of a terminal device, and the instructions implement a mine fan ventilation control method based on digital twinning.

[0193] A terminal device, comprising a processor and a computer readable storage medium, the processor is used to implement instructions, and the computer readable storage medium is used to store a plurality of instructions, the instructions are suitable for being loaded and executed by the processor, and the instructions implement a mine fan ventilation control method based on digital twinning.

[0194] The above are preferred embodiments of the present application, not limited to the protection scope of the present application, therefore: any equivalent changes made according to the structure, shape and principle of the present application should be covered in the protection scope of the present application.

Claims

1. A mine ventilation control method based on digital twin, characterized in that, include: Acquire mine monitoring data; Data preprocessing is performed on the acquired mine monitoring data; Feature extraction is performed on the preprocessed mine monitoring data; Using the random forest algorithm to predict the relationship between data characteristics and wind turbine air volume; By using a digital twin model to solve the relationship between data characteristics and fan air volume, a fan ventilation control strategy is obtained; PLC control instructions are generated based on the fan ventilation control strategy; The acquisition of mine monitoring data includes static monitoring data, dynamic monitoring data, and equipment monitoring data. Static monitoring data includes data on personnel entering the mine, simultaneous operating power data of diesel engines, and air supply volume data. Dynamic monitoring data includes the concentration of harmful gases, oxygen, and carbon dioxide in the mine. Equipment monitoring data includes motor bearing and motor temperature data, vibration data, and electrical parameter data. The acquired mine monitoring data is preprocessed, including missing data processing and abnormal data processing. Specifically, the K-nearest neighbor algorithm is used to fill in missing data, and the Z-score method is used to identify abnormal data. The feature extraction of the preprocessed mine monitoring data includes first performing dimensionality reduction feature extraction using PCA principal component analysis, then extracting time series features based on wavelet transform, and further extracting features using a multilayer autoencoder. The final data features are obtained by fusing and selecting the features extracted by PCA principal component analysis, wavelet transform, and autoencoder respectively.

2. The mine ventilation control method based on digital twin according to claim 1, characterized in that, The method of using the random forest algorithm to predict the relationship between data features and wind turbine air volume includes constructing a decision tree using the CART algorithm, using the mean squared error as a measure of node splitting, forming a random forest by constructing N decision trees, evaluating each decision tree using out-of-bag data, using the out-of-bag error as a performance estimate of the random forest model on unseen data, and finally optimizing the parameters of the random forest through random search to obtain the relationship between data features and wind turbine air volume.

3. The mine ventilation control method based on digital twin according to claim 2, characterized in that, The method of using a digital twin model to solve the relationship between data characteristics and fan air volume to obtain a fan ventilation control strategy includes constructing a physical ventilation model and a data-driven model based on the digital twin, and using a weighted fusion method to merge the physical ventilation model and the data-driven model. The target is set with mine air quality and ventilation energy consumption. The relationship between data characteristics and fan air volume is solved by defining an optimization objective function and setting constraints, thereby simulating the operating state of the ventilation system.

4. The mine ventilation control method based on digital twin according to claim 3, characterized in that, The process of generating PLC control instructions based on the fan ventilation control strategy includes converting the air volume control requirements in the fan ventilation control strategy into fan motor speed control requirements, establishing the relationship between air volume and motor speed according to the fan performance curve, converting the motor speed control requirements into control instructions that can be recognized by the PLC, and finally sending the generated PLC control instructions to the PLC controller of the mine ventilation system via industrial Ethernet to achieve real-time control of the fan.

5. A mine ventilation control system based on digital twins, executing the mine ventilation control method based on digital twins as described in claim 1, characterized in that, include: The data acquisition module is configured to acquire mine monitoring data; The preprocessing module is configured to preprocess the acquired mine monitoring data; The feature extraction module is configured to extract features from the preprocessed mine monitoring data. The prediction module is configured to use the random forest algorithm to predict the relationship between data features and wind turbine air volume; The solution module is configured to use a digital twin model to solve the relationship between data features and fan air volume to obtain a fan ventilation control strategy; The control module is configured to generate PLC control instructions based on the fan ventilation control strategy.

6. A computer-readable storage medium storing a plurality of instructions, characterized in that, The instructions are adapted to be loaded by the processor of the terminal device and executed as described in claim 1.

7. A terminal device, comprising a processor and a computer-readable storage medium, wherein the processor implements instructions; and the computer-readable storage medium stores multiple instructions, characterized in that, The instructions are adapted to be loaded by a processor and executed as described in claim 1.

Images