Mining wireless gateway adaptive network adjusting method and system based on deep learning

Abstract

The invention provides a mining wireless gateway adaptive network adjustment method and system based on deep learning, and relates to the technical field of industrial gateways, and the method comprises the steps: carrying out the feature analysis through collecting the network state parameters of a mining wireless gateway, and constructing a standardized network data set; constructing a hybrid deep learning branch, training by using the standardized network data set, and generating a network parameter prediction channel; according to the method, network state parameters are analyzed, circuit control is carried out in combination with an MOS tube switch array, gain amplification is carried out on the mining wireless gateway, and self-adaptive network adjustment is carried out on the mining wireless gateway. The technical problem that the network performance is further affected due to adjustment lag and passive response caused by the fact that the wireless gateway is easily interfered in a dynamic extreme environment in the prior art is solved, intelligent optimization is carried out on the network performance of the mining wireless gateway based on deep learning and network state feature analysis, and the network performance of the mining wireless gateway is improved. And the stability and the efficiency of wireless communication of the industrial gateway are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial gateway, in particular to a mine wireless gateway adaptive network regulation method and system based on deep learning. BACKGROUND

[0002] In the actual deployment of the mine wireless gateway, the tunnel environment where it is located is complex and changeable, and it faces complex signal interference and environmental changes, which often leads to unstable performance of wireless communication. Due to the particularity of the mine environment, such as electromagnetic interference in the deep underground, the influence of metal equipment, and the change of terrain, the existing network regulation method often relies on traditional static or simple dynamic adjustment strategy, and makes lagging adjustment through preset threshold or classic control algorithm, which is difficult to cope with the rapid mutation of channel state, often leading to that the adjustment instruction lags behind the environmental change, causing communication link oscillation or even interruption, which seriously restricts the stability and real-time performance of mine wireless communication.

[0003] In summary, the existing technology has the technical problem that the wireless gateway is easily disturbed in the dynamic extreme environment, leading to lagging adjustment and passive response, which further affects the network performance of the mine wireless gateway. SUMMARY

[0004] The purpose of the present application is to provide a mine wireless gateway adaptive network regulation method and system based on deep learning, to solve the technical problem in the prior art that the wireless gateway is easily disturbed in the dynamic extreme environment, leading to lagging adjustment and passive response, which further affects the network performance of the mine wireless gateway.

[0005] In order to achieve the above purpose, the present application provides a mine wireless gateway adaptive network regulation method and system based on deep learning.

[0006] In a first aspect, the present application provides a mine wireless gateway adaptive network regulation method based on deep learning, which is realized by a mine wireless gateway adaptive network regulation system based on deep learning, wherein the mine wireless gateway adaptive network regulation method based on deep learning comprises: performing feature analysis by collecting network state parameters of the mine wireless gateway, obtaining a multi-dimensional network feature data set for preprocessing, and constructing a standardized network data set; constructing a hybrid deep learning branch, training the hybrid deep learning branch using the standardized network data set, and generating a network parameter prediction channel; analyzing the network state parameters through the network parameter prediction channel, outputting network regulation parameters, combining a MOS tube switch array for circuit control, obtaining a network control strategy for gain amplification of the mine wireless gateway, and generating a gain signal for adaptive network regulation of the mine wireless gateway.

[0007] Optionally, the double-frequency sensing nodes are alternately arranged on the roof and sidewall of the mine roadway to construct a three-dimensional monitoring network; a data acquisition frequency is set to activate the three-dimensional monitoring network, and a plurality of sensing nodes are determined to perform sensing collision analysis; when there is data sensing collision among the plurality of sensing nodes, the plurality of sensing nodes are accessed to the mine wireless gateway according to a data transmission cycle in a time division multiple access method to obtain the network state parameters of the plurality of sensing nodes.

[0008] Optionally, the acquisition time data of the three-dimensional monitoring network is divided into equal-length time slots according to the data acquisition frequency; the equal-length time slots are matched with the plurality of sensing nodes to construct a time slot allocation table; the plurality of sensing nodes are allocated based on the time slot allocation table to obtain a plurality of transmission time slots; adjacent time slot overlap analysis is performed on the plurality of transmission time slots to obtain data transmission overlap parameters; data transmission conflicts are monitored according to the data transmission overlap parameters to obtain data collision results of the plurality of sensing nodes.

[0009] Optionally, multi-scale decomposition is performed based on the network state parameters to obtain multi-scale decomposition coefficients; noise recognition is performed based on the multi-scale decomposition coefficients to obtain noise coefficient components; the network state parameters are filtered according to the noise coefficient components to obtain effective signal features; signal propagation ray tracing is performed on the mine wireless gateway according to the effective signal features to obtain the multi-dimensional network feature data set.

[0010] Optionally, numerical analysis is performed based on the multi-dimensional network feature data set to obtain numerical distribution characteristic parameters; the numerical distribution characteristic parameters are mapped using a min-max standardization method to determine a target mapping interval; Z-score standardization calculation is performed based on the target mapping interval to obtain a standardized grid initial data set; the standardization effect of the standardized grid initial data set is verified, and the standardized grid initial data set is updated in reverse to obtain the standardized network data set.

[0011] Optionally, a plurality of convolutional layers and a plurality of pooling layers are alternately arranged to construct a convolutional neural network branch; a flattening layer is connected to the end of the convolutional neural network branch to generate a one-dimensional feature vector; a plurality of layers of gated recurrent unit networks are set, each layer of the plurality of layers of gated recurrent unit networks including a plurality of gated recurrent units, to construct a recurrent neural network branch; a fully connected layer is connected to the end of the recurrent neural network branch to extract time sequence features; and the one-dimensional feature vector and the time sequence features are spliced to construct the hybrid deep learning branch.

[0012] Optionally, the standardized network data set is divided into a training set, a validation set, and a test set; the training set, the validation set, and the test set are respectively used for segmented training of the convolutional neural network branch and the recurrent neural network branch to obtain a plurality of segmented training parameters; the convolutional neural network branch and the recurrent neural network branch are combined with the plurality of segmented training parameters to perform end-to-end joint training, and a joint training result is generated; the standardized network data set is optimized and monitored based on the joint training result to construct the network parameter prediction channel.

[0013] Optionally, the network state parameter is input into the network parameter prediction channel: S1: spatial sequence features of the network state parameter are extracted by the convolutional neural network branch; S2: time sequence features of the network state parameter are captured by the recurrent neural network branch; S3: the spatial sequence features and the time sequence features are fused to generate a network state comprehensive evaluation result; multi-dimensional network analysis is performed based on the network state comprehensive evaluation result to generate a network adjustment parameter; the MOS tube switch array is combined with the network adjustment parameter to analyze and generate a circuit control signal; the circuit control signal is decomposed into a plurality of adjustment steps, and the circuit control signal is gain-controlled according to the plurality of adjustment steps to generate a target gain value; the circuit control signal is gain-gradually adjusted based on the target gain value to generate the gain signal.

[0014] Optionally, the network state comprehensive evaluation result is analyzed to obtain a network power evaluation result and a channel quality evaluation result; power analysis is performed based on the network power evaluation result to determine a power to be adjusted level; channel analysis is performed based on the channel quality evaluation result to determine a channel to be switched parameter; the power to be adjusted level and the channel to be switched parameter are combined and adjusted based on the signal reception strength to generate the network adjustment parameter.

[0015] In a second aspect, the application further provides a deep learning-based adaptive network adjustment system for mine wireless gateway, which is used to execute the deep learning-based adaptive network adjustment method for mine wireless gateway as described in the first aspect. The deep learning-based adaptive network adjustment system for mine wireless gateway comprises: a parameter analysis module, which is used to perform feature analysis by collecting network state parameters of the mine wireless gateway, obtain a multi-dimensional network feature dataset for preprocessing, and construct a standardized network dataset; a channel construction module, which is used to construct a hybrid deep learning branch, train the hybrid deep learning branch using the standardized network dataset, and generate a network parameter prediction channel; and a network adjustment module, which is used to analyze the network state parameters through the network parameter prediction channel, output network adjustment parameters, combine a MOS tube switch array for circuit control, obtain a network control strategy for gain amplification of the mine wireless gateway, and generate a gain signal for adaptive network adjustment of the mine wireless gateway.

[0016] The one or more technical solutions provided in the application have at least the following technical effects or advantages: The network state parameters of the mine wireless gateway are collected for feature analysis, a multi-dimensional network feature dataset is obtained for preprocessing, a standardized network dataset is constructed, a hybrid deep learning branch is constructed, the hybrid deep learning branch is trained using the standardized network dataset, a network parameter prediction channel is generated, the network state parameters are analyzed through the network parameter prediction channel, network adjustment parameters are output, a MOS tube switch array is combined for circuit control, a network control strategy is obtained for gain amplification of the mine wireless gateway, and a gain signal is generated for adaptive network adjustment of the mine wireless gateway. That is, by collecting the network state parameters of the mine wireless gateway and performing multi-dimensional feature analysis, various interference factors in a complex environment are identified based on a deep learning model, circuit control is performed in combination with a MOS tube switch array, intelligent gain amplification is performed, changes in the network state are predicted in advance and active adjustment is performed, thereby improving the stability and efficiency of wireless communication of the mine industrial gateway.

[0017] The above description is only a summary of the technical solutions of the application. In order to enable the technical means of the application to be more clearly understood, the application can be implemented in accordance with the contents of the specification, and in order to enable the above and other purposes, features and advantages of the application to be more apparent and easy to understand, the following specific embodiments of the application are described. It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the application, nor are they intended to limit the scope of the application. Other features of the application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the accompanying drawings in the following description are only exemplary and, for those skilled in the art, other drawings can be obtained without creative effort on the basis of the provided drawings.

[0019] Figure 1 The flowchart of the adaptive network regulation method of the mine wireless gateway based on deep learning of the application.

[0020] Figure 2 The structure diagram of the adaptive network regulation system of the mine wireless gateway based on deep learning of the application.

[0021] Legend: parameter analysis module 11, channel construction module 12, network regulation module 13. DETAILED DESCRIPTION

[0022] The application provides an adaptive network regulation method and system of a mine wireless gateway based on deep learning, which solves the technical problem in the prior art that the wireless gateway is easily disturbed in a dynamic extreme environment, resulting in lagging regulation and passive response, which further affects the network performance of the mine wireless gateway. By collecting network state parameters of the mine wireless gateway and performing multi-dimensional feature analysis, various interference factors in a complex environment are identified based on a deep learning model, circuit control is performed in combination with a MOS tube switch array, intelligent gain amplification is performed, changes in the network state are predicted in advance and active regulation is performed, thereby improving the stability and efficiency of wireless communication of the mine industrial gateway.

[0023] The technical solutions in the application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. It should be understood that the application is not limited by the exemplary embodiments described herein. Based on the embodiments of the application, all other embodiments obtained by those skilled in the art without creative effort fall within the scope of protection of the application. In addition, it should be noted that, for convenience of description, only parts related to the application are shown in the drawings, not all.

[0024] Embodiment one, please refer to the accompanying Figure 1 The application provides an adaptive network regulation method of a mine wireless gateway based on deep learning, which is applied to an adaptive network regulation system of a mine wireless gateway based on deep learning. The adaptive network regulation method of the mine wireless gateway based on deep learning specifically comprises the following steps: A multi-dimensional network feature dataset is obtained by collecting network state parameters of the mine wireless gateway, and preprocessed to construct a standardized network dataset.

[0025] Further, the application further includes the following steps: deploying double-frequency sensing nodes alternately on the roof and sidewall of the mine roadway to construct a three-dimensional monitoring network; setting a data acquisition frequency to activate the three-dimensional monitoring network and determine a plurality of sensing nodes for sensing collision analysis; when there is data sensing collision among the plurality of sensing nodes, the plurality of sensing nodes are accessed to the mine wireless gateway for sensing according to a time division multiple access method according to a data transmission cycle, and the network state parameters of the plurality of sensing nodes are obtained.

[0026] Further, the application further includes the following steps: dividing the collection time data of the three-dimensional monitoring network into equal-length time slots according to the data acquisition frequency; matching the equal-length time slots with the plurality of sensing nodes to construct a time slot allocation table; allocating the plurality of sensing nodes based on the time slot allocation table to obtain a plurality of transmission time slots; performing adjacent time slot overlap analysis on the plurality of transmission time slots to obtain data transmission overlap parameters; monitoring data transmission conflicts according to the data transmission overlap parameters to obtain data collision results of the plurality of sensing nodes.

[0027] Specifically, in the mine roadway, signal transmission is affected by factors such as terrain, obstacles and electromagnetic interference, so it is necessary to ensure the accuracy and stability of the monitoring data through scientific sensing node deployment and data acquisition strategy. In the roof and sidewall position of the mine, double-frequency sensing nodes are alternately deployed to form a three-dimensional monitoring network covering the entire mine roadway. The mine roadway is an underground passage for passing and transporting ore in the mining process of the mine, the roof of the roadway refers to the rock layer above the roadway, and the sidewall is the rock wall on both sides of the roadway. In the mine, the roof and sidewall of the roadway are the key areas for signal propagation and sensing device deployment. The roof node is mainly responsible for wide-area coverage, while the sidewall node can effectively detect the signal blind area close to the ground, which maximizes the reduction of monitoring dead angles. Double-frequency sensing nodes refer to sensors that use two frequency bands simultaneously, commonly 2.4GHz and 5GHz frequency bands. Double-frequency sensing nodes provide frequency diversity, automatically switching to another frequency when one frequency is interfered, enhancing the robustness of communication.

[0028] In the mine roadway, a plurality of sensing nodes need to synchronously collect environmental data and transmit to the wireless gateway. Due to the limited bandwidth of wireless communication, if a plurality of sensing nodes simultaneously transmit data at the same time, data collision may occur, affecting the integrity and transmission efficiency of the data. In order to avoid this problem, time slot allocation is needed to ensure that each sensing node transmits data at different time periods, thereby avoiding data collision.

[0029] Set the data collection frequency, that is, the number of times the sensor collects data per second. During the monitoring process, reasonable setting of the collection frequency can ensure that sufficient data is obtained while avoiding excessive frequency that brings too much useless information. At the same time, activate all nodes of the stereoscopic monitoring network. According to the data sampling frequency, divide the collection time data of the stereoscopic monitoring network into equal-length time slots, that is, divide the time into small segments of equal length. Match the equal-length time slots with multiple sensor nodes to construct a time slot allocation table, which clearly records the correspondence between each sensor node and its exclusive transmission time slot. For example, sensor node 1 transmits data in time slot 1, sensor node 2 transmits data in time slot 2, and so on. In this way, different sensor nodes can transmit data at different time periods to avoid signal conflicts.

[0030] Based on the time slot allocation table, allocate multiple sensor nodes, each of which is allocated to a transmission time slot of its own, to obtain multiple transmission time slots. When allocating time slots, it is necessary to ensure that the time slots of adjacent nodes do not overlap. Traverse the multiple transmission time slots to analyze the overlap of adjacent time slots and check whether there is an overlap. By calculating the node clock drift, signal propagation delay, and transceiver switching time, the probability and degree of signal overlap at the boundary of adjacent time slots are simulated and evaluated to obtain data transmission overlap parameters. The data transmission overlap parameter is an index for quantifying the risk of inter-slot interference, which can be understood as the length of time that the two data packets before and after may overlap at the time slot boundary, usually in units of microseconds. For example, in a simulated tunnel with a cross-section of 20 double-frequency sensor nodes are alternately deployed, the data collection frequency is 10 Hz, the collection time data of the stereoscopic monitoring network is divided into 20 equal-length time slots, and each time slot is 5 ms long. After the time slot allocation table is established, to verify its reliability, an overlap analysis is performed. Assuming that the maximum drift of the node crystal oscillator is ±50ppm, the maximum signal propagation delay is 1 μs, and the transceiver switching time is 10 μs. Calculate the time accumulation error in the worst case, traverse all adjacent time slots (such as the end of time slot 1 and the beginning of time slot 2) for analysis, and find that due to clock drift, after continuous transmission for 1 minute, the time slot boundary of the first node and the last node produces a deviation of up to 28 μs. Combined with other factors, the data transmission overlap parameter in the worst case can reach 28 μs. According to this parameter, it is monitored that there is a high risk of data transmission conflict, and the data collision result that the current allocation scheme will collide under long-time operation is obtained.

[0031] According to the data transmission overlap parameter, potential data transmission conflict risks are monitored, and data collision results of all sensor nodes are obtained. If the overlap parameter exceeds the safety threshold, it means that the current allocation scheme has potential risks, and the time slot allocation needs to be adjusted or a protection interval needs to be introduced. When the data collision result shows that multiple sensor nodes have data sensing collisions, a scheduling mechanism is started, and a time division multiple access method is used to allocate a dedicated and fixed transmission time slot for each node according to the data transmission period. That is, the time slots of the sensor nodes are re-allocated according to the set data transmission period. For example, node 1 can be allocated to the first 0.5s of period 1, and node 2 can be allocated to the last 0.5s of period 1. In this way, the transmission time slots of all nodes will be optimized, avoiding time overlap and reducing the possibility of data collision. Time division multiple access is a communication technology that divides time into multiple time slots and allocates these time slots to different devices for communication, i.e., the time axis is divided into periodic frames, each frame is further divided into several non-overlapping time slots, and each node can only transmit data in the specific time slot allocated to it, thereby fundamentally avoiding data collision.

[0032] Each node orderly transmits data to the mine wireless gateway as if it has a fixed speaking time. According to the new time slot allocation, each sensor node transmits data according to the allocated time slot. The conflict during data transmission is monitored, and the network state parameters of each sensor node are obtained, including signal strength, transmission delay, bandwidth, packet loss rate, and other indicators reflecting network quality and stability. By alternately deploying dual-frequency sensor nodes and using time division multiple access for data transmission, data transmission conflicts can be effectively avoided, the access efficiency of the mine wireless gateway can be improved, and the accurate acquisition of network state parameters is ensured, real-time monitoring of wireless communication quality in the mine roadway is ensured.

[0033] Further, the application further includes the following steps: performing multi-scale decomposition based on the network state parameters to obtain multi-scale decomposition coefficients; performing noise identification based on the multi-scale decomposition coefficients to obtain noise coefficient components; filtering the network state parameters according to the noise coefficient components to obtain effective signal features; and performing signal propagation ray tracing on the mine wireless gateway according to the effective signal features to obtain the multi-dimensional network feature data set.

[0034] Specifically, the network status parameters are subjected to multi-scale decomposition, which decomposes the original data into components of multiple scales. Multi-scale decomposition is a signal processing technique that decomposes an original signal into a set of components of different frequency scales. Wavelet transform is a typical method that can analyze a signal in both time and frequency domains, thereby revealing short-term fluctuations (high-frequency details) and long-term trends (low-frequency approximations) in the signal. For example, the original signal can contain some high-frequency noise or low-frequency stable signals, and through multi-scale decomposition, these signals can be separated to obtain multi-scale decomposition coefficients, which reflect the characteristics of the signal at different scales.

[0035] Noise identification is performed according to the multi-scale decomposition coefficients. Noise is usually manifested as high-frequency or irregular fluctuations unrelated to the signal. The goal of noise identification is to separate these interference components from the original signal. The noise identification algorithm identifies the noise components in the signal according to the decomposed coefficients and calculates the noise coefficient components, which reflect the influence of noise on the signal. Noise usually comes from environmental electromagnetic interference or sensor errors, and through noise identification, the useful signal can be separated from the invalid noise in the original data. The noise coefficient components obtained through noise identification reflect the size and influence of noise in the signal.

[0036] The network status parameters are filtered according to the noise coefficient components, i.e., the noise coefficient components are discarded or reduced when reconstructing the signal, thereby obtaining the effective signal features that reflect the real channel changes after denoising. The effective signal features can include signal strength, delay, bandwidth, and other parameters, which accurately reflect the performance of the wireless network. In the filtering process, the noise coefficient components are subtracted, thereby improving the quality of the signal.

[0037] According to the effective signal characteristics, the signal propagation ray tracing of the mine wireless gateway is performed, and the propagation path of the wireless signal from the sensor to the mine wireless gateway is simulated. The ray tracing simulation not only considers the straight path of the signal, but also considers the obstacles, reflections, refractions and other factors that may exist in the environment. Through ray tracing, the propagation of the signal in the actual environment can be obtained, so as to evaluate the coverage range and transmission quality of the wireless network. In wireless communication, signal propagation ray tracing is a simulation technique used to calculate the propagation path and process of wireless signals from the transmitting end to the receiving end. For example, assume that 50 sensor nodes are deployed in a mine, the network acquisition frequency is 20 times per second, and the signal strength data of the sensor nodes contains noise within a period. The network state parameters are decomposed into multiple scales, and the decomposition scale is set to 5 frequency bandwidth segments. After decomposition, the decomposition coefficients are as follows: the average value of the signal strength of scale 1 (low frequency) is -60 dBm, and the delay is 50 ms; the average value of the signal strength of scale 2 (medium frequency) is -62 dBm, and the delay is 55 ms; the average value of the signal strength of scale 3 (high frequency) is -65 dBm, and the delay is 60 ms. Through noise identification, it is identified that the high frequency part contains about 5% noise interference, and the noise coefficient is 0.2. Then, by filtering the signal, the high frequency noise component is removed, and the effective signal characteristics are obtained, the signal strength is -62 dBm, and the delay is 52 ms. By using signal propagation ray tracing for simulation, there are multiple obstacles in the mine, and the ray tracing result shows that the signal strength of the signal reaching the mine wireless gateway after reflection and refraction is -58 dBm, and the delay is 54 ms.

[0038] Through the multi-scale decomposition technology, different levels of signal characteristics are extracted from complex network state parameters, and then noise in the signal is identified and removed. The noise identification and filtering steps effectively improve the quality of the signal, remove interference factors, and the signal propagation ray tracing technology can simulate the signal propagation path in the actual environment, helping to evaluate the coverage range and performance of the wireless network.

[0039] Further, the application further includes the following steps: performing numerical analysis based on the multi-dimensional network feature data set to obtain numerical distribution characteristic parameters; performing data mapping on the numerical distribution characteristic parameters using a min-max standardization method to determine a target mapping interval; performing Z-score standardization calculation based on the target mapping interval to obtain a standardized grid initial data set; verifying the standardization effect of the standardized grid initial data set, updating the standardized grid initial data set in reverse, and obtaining the standardized network data set.

[0040] Specifically, numerical analysis is performed on the multi-dimensional network feature dataset to obtain numerical distribution characteristic parameters of each feature dimension, including global minimum value, maximum value, mean value, and standard deviation. The numerical distribution characteristic parameters are parameters describing the data distribution characteristics, such as the minimum value, maximum value, mean value, standard deviation, and the like of the data. The min-max normalization method is used to map the numerical distribution characteristic parameters, determine the target mapping interval, set the normalized numerical range, usually [0, 1], by calculating the minimum value and maximum value of each feature parameter, and convert the original data to the target mapping interval through a linear transformation formula, handle the data boundary condition, and prevent the standardized data from overflowing. For example, assume that the multi-dimensional network feature dataset contains three features, signal strength (-100 dBm to -50 dBm), multipath time delay spread (10 ns to 500 ns), and channel load rate (0% to 100%). Numerical analysis is performed on the multi-dimensional network feature dataset, and the minimum value of the signal strength is -98, the maximum value is -52, the mean value is -75, and the standard deviation is 10; the minimum value of the time delay spread is 15, the maximum value is 480, the mean value is 150, and the standard deviation is 100; the minimum value of the load rate is 0, the maximum value is 95, the mean value is 30, and the standard deviation is 20. All of the three features are linearly mapped to the [0, 1] interval. A value -80 dBm in the signal strength value is converted to (-80-(-98)) / (-52-(-98))=18 / 46=0.39.

[0041] According to the Z-score normalization calculation according to the target mapping interval, the distribution mean value of each feature is 0 and the standard deviation is 1, so as to obtain a standardized grid initial dataset that is dimensionless and conforms to the standard normal distribution. The Z-score normalization is a data standardization method, which makes the mean value of the processed data 0 and the standard deviation 1. For example, the new mean value and standard deviation of the data in the [0, 1] interval are calculated, and assume that the mean value of the signal strength is 0.5 and the standard deviation is 0.2, then 0.39 is further converted to (0.39-0.5) / 0.2=-0.55.

[0042] The standardized grid initial data set is verified to check whether the data meets the expected standards, i.e., whether the data has been sufficiently standardized and not distorted due to calculation errors in the standardization process. For example, if the numerical range of certain features is too narrow during the standardization process, resulting in poor standardization of the data, further adjustment or remapping of the data can be performed. After verification, the standardized grid initial data set is updated as necessary, including adjusting those abnormal data points or adjusting the data mapping according to new calculation methods to ensure that the data accurately reflects the characteristics of the original data. For example, check whether the standard deviation of a feature is close to 0 (meaning that the feature is invalid), or whether there is a distribution anomaly caused by extreme values; according to the verification result, the standardized grid initial data set is updated in reverse, the feature is first logarithmically transformed before the min-max step, and then the two-level standardization process is re-executed, finally all features in the standardized network data set are presented as a good distribution with a mean of 0 and a standard deviation of 1. Through the min-max standardization and Z-score standardization methods, the original network state feature data is converted into a unified standard format, eliminating the scale differences between features.

[0043] A hybrid deep learning branch is constructed, and the standardized network data set is used to train the hybrid deep learning branch to generate a network parameter prediction channel.

[0044] Further, the present application further comprises the following steps: alternately arranging a plurality of convolutional layers and a plurality of pooling layers to construct a convolutional neural network branch; connecting a flattening layer to the end of the convolutional neural network branch to generate a one-dimensional feature vector; setting a multi-layer gated recurrent unit network, each layer of the multi-layer gated recurrent unit network comprising a plurality of gated recurrent units, to construct a recurrent neural network branch; connecting a fully connected layer to the end of the recurrent neural network branch to extract time series features; and splicing the one-dimensional feature vector and the time series features to construct the hybrid deep learning branch.

[0045] Specifically, a plurality of convolutional layers and pooling layers are alternately arranged to construct a convolutional neural network branch, such as Conv1D→BN→ReLU→MaxPooling→Conv1D→BN→ReLU→MaxPooling. The convolutional neural network branch includes a plurality of convolutional layers and pooling layers, which are alternately arranged. Each convolutional layer uses a different size of convolution kernel to extract spatial features of different scales. The larger the convolution kernel, the more extensive the features that can be extracted, which is suitable for capturing spatial information in a larger area. Through multiple layers of convolution, multi-scale features from local to global can be captured.

[0046] Different sizes of convolutional layers are responsible for extracting multi-scale spatial features, batch normalization layers are placed after convolution and before activation functions to speed up the model training process and improve stability, rectified linear units are introduced as activation functions to introduce nonlinearity, and the subsequent max pooling retains significant features and reduces dimensionality, thereby reducing computational complexity and overfitting risk. The end of the convolutional neural network branch is connected to the flattening layer to convert the last feature map into a one-dimensional feature vector. The flattening layer is to flatten the multi-dimensional input data into one-dimensional data in order to input into the fully connected layer. In the convolutional neural network, the flattening layer is usually located between the convolutional layer and the fully connected layer.

[0047] A multi-layer gated recurrent unit network is set up to build a recurrent neural network branch, where the gated recurrent unit finely controls the information flow through its reset gate and update gate calculation mechanism, a Dropout layer is added to prevent overfitting, and a tanh activation function is used to process the candidate state of the gated recurrent unit to generate new candidate memories. The gated recurrent unit is a variant of recurrent neural network, which can effectively capture long-term dependencies in time series by introducing the mechanism of reset gate and update gate, and at the same time alleviate the gradient vanishing / explosion problem of traditional recurrent neural network. The reset gate determines how to combine new input with previous memory, and the update gate determines how much past memory to retain. The multi-layer gated recurrent unit network controls the reset gate and update gate, so that the model can remember useful information and forget irrelevant information.

[0048] A fully connected layer is connected at the end of the recurrent neural network branch to map the time series data to the target dimension and extract the time series features in the network. The one-dimensional feature vector output by the convolutional neural network branch is spliced with the time series features output by the recurrent neural network branch to form a comprehensive feature vector that integrates spatial and temporal information, thereby constructing a hybrid deep learning branch based on the attention mechanism of convolutional neural network-long short-term memory network.

[0049] Exemplarily, to predict the signal strength in the next 1 second, the convolutional neural network-long short-term memory network input is the normalized network features in the past 3 seconds (30 time points), such as signal strength, signal-to-noise ratio, and bit error rate. The input data shape is (30, 3); first, it passes through a Conv1D layer containing 16 convolution kernels with a size of 3 to extract local correlation features, followed by BN and ReLU, and then down-samples through MaxPooling (pool_size=2); then it is another Conv1D layer containing 32 convolution kernels with a size of 5, followed by BN, ReLU, and MaxPooling. Finally, a one-dimensional feature vector containing about 100 elements is obtained through the flattening layer. The same input sequence is sent to a two-layer multi-layer gated recurrent unit network, each layer has 50 units, and the Dropout=0.2 is set between layers. The GRU learns the time sequence dependence of the sequence through its gating mechanism. The output of the last GRU layer is connected to a fully connected layer to extract a 50-dimensional time sequence feature. The 100-dimensional vector of the convolutional neural network branch and the 50-dimensional vector of the recurrent neural network branch are spliced into a 150-dimensional comprehensive feature vector. An attention mechanism is introduced on the fusion feature, and appropriate weights are assigned to the outputs of the recurrent neural network branch at different time steps, so that the convolutional neural network-long short-term memory network pays more attention to the most critical historical time for prediction.

[0050] By combining the convolutional neural network and the recurrent neural network, the hybrid deep learning model can extract spatial features and time sequence features at the same time. The convolutional neural network effectively captures the spatial structure of the input data, and the recurrent neural network processes the time dependence in the data. Through the attention mechanism, the model can focus on key information, further improving the extraction ability of spatial and temporal features, and thus improving the performance of the network on complex spatial and temporal data.

[0051] Further, the application also includes the following steps: dividing the standardized network data set into a training set, a validation set, and a test set; using the training set, the validation set, and the test set to perform segmented training on the convolutional neural network branch and the recurrent neural network branch respectively, to obtain a plurality of segmented training parameters; unfreezing the convolutional neural network branch and the recurrent neural network branch to perform end-to-end joint training combined with the plurality of segmented training parameters, to generate a joint training result; based on the joint training result, performing optimization monitoring training on the standardized network data set, to construct the network parameter prediction channel.

[0052] Specifically, the standardized network dataset is divided into a training set, a validation set, and a test set in a certain proportion (such as 70: 15: 15). The training set is used to directly train the model parameters; the validation set is used to monitor the model performance during the training process, adjust the hyperparameters, and select the best model to prevent overfitting; the test set is used to finally evaluate the generalization performance of the model, reflecting its real performance on unknown data.

[0053] The training set, the validation set, and the test set are used to train the convolutional neural network branch and the recurrent neural network branch respectively. The segmented training refers to dividing the entire training process into multiple stages, training different parts separately in each stage, and obtaining multiple segmented training parameters respectively. When training the convolutional neural network branch, the training process mainly focuses on the extraction of spatial features. The data in the training set will pass through the convolutional layer and the pooling layer, and be processed by batch normalization and activation function, etc., to obtain the representation of spatial features. When training the recurrent neural network branch, the training process focuses on the extraction of time series features. The recurrent neural network branch captures the time series patterns in the data through the gating mechanism and learns the long and short term dependencies. That is, the training set is used to train the convolutional neural network branch and the recurrent neural network branch independently or semi-independently, such as fixing the recurrent neural network branch first and training the convolutional neural network branch only with the training set; then fixing the convolutional neural network branch and training the recurrent neural network branch only with the training set, so that the two branches can initially master the ability of spatial and temporal feature extraction, and obtain their respective segmented training parameters.

[0054] The two branches are unfrozen, and the previously frozen network layers will participate in the training again, so that they can adjust the parameters in the joint training process and improve the performance of the model. The convolutional neural network branch and the recurrent neural network branch are combined with multiple segmented training parameters for end-to-end joint training, and the convolutional neural network branch and the recurrent neural network branch are trained together to optimize their parameters. Through end-to-end training, the two networks can influence each other and learn together, thereby further improving the comprehensive ability of the model. During the joint training process, the spatial features extracted by the convolutional neural network and the time series features captured by the recurrent neural network will be fused in the model to form a more powerful joint network. During the training process, the model will gradually adjust the parameters according to the feedback of the training set and the validation set, and finally obtain the optimized joint training result.

[0055] Based on the joint training result, the standardized network dataset is optimized and monitored in the subsequent training period to further optimize the performance of the network in the dynamic environment, improve its real-time prediction and adjustment ability, and construct a network parameter prediction channel to receive the input network state parameters and generate prediction values about the network status.

[0056] The training strategy of phased and post-association solves the problems of complex mixed model training difficulty, easy overfitting and unstable convergence. The segmented training ensures that the two functionally different sub-networks of the convolutional neural network branch and the recurrent neural network branch can lay a good foundation for feature extraction, avoiding the unstable gradient flow or the lazy situation of one branch at the initial stage of end-to-end training. Starting from this point, joint training is carried out, so that the network parameter prediction channel can fine-tune at a very high starting point, accelerate the convergence speed, and more easily find a global optimal solution or a strong local optimal solution with better performance.

[0057] The network state parameters are analyzed by the network parameter prediction channel, and network adjustment parameters are output. Circuit control is performed in combination with the MOS tube switch array to obtain a network control strategy for gain amplification of the mine wireless gateway, and a gain signal is generated for adaptive network adjustment of the mine wireless gateway.

[0058] Further, the application further includes the following steps: inputting the network state parameters into the network parameter prediction channel: S1: extracting the spatial sequence features of the network state parameters by the convolutional neural network branch; S2: capturing the time sequence features of the network state parameters by the recurrent neural network branch; S3: fusing the spatial sequence features and the time sequence features to generate a network state comprehensive evaluation result; performing multi-dimensional network analysis based on the network state comprehensive evaluation result to generate network adjustment parameters; analyzing the network adjustment parameters by the MOS tube switch array to generate a circuit control signal; decomposing the circuit control signal into multiple adjustment steps, and performing gain control on the circuit control signal according to the multiple adjustment steps to generate a target gain value; performing gain progressive adjustment on the circuit control signal based on the target gain value to generate the gain signal.

[0059] Further, the application further includes the following steps: analyzing the network state comprehensive evaluation result to obtain a network power evaluation result and a channel quality evaluation result; performing power analysis based on the network power evaluation result to determine a power to be adjusted level; performing channel analysis based on the channel quality evaluation result to determine a channel to be switched parameter; combining the power to be adjusted level and the channel to be switched parameter to generate the network adjustment parameters.

[0060] Specifically, the collected network state parameters of the mine wireless gateway are input into the network parameter prediction channel for parallel processing. The spatial sequence features of the network state parameters are extracted by the convolutional neural network branch. The convolution kernel of the convolutional neural network mainly slides in the time step dimension, focusing on analyzing the mutual relationship and local combination mode among the signal strength, signal-to-noise ratio and bit error rate at the same time point, such as identifying the typical instantaneous interference features of the signal-to-noise ratio being acceptable but the bit error rate suddenly rising. The time sequence features of the network state parameters are captured by the recurrent neural network branch. The multi-layer gated recurrent unit network processes the input of each time step in turn, and remembers and forgets information through the internal gating mechanism, thereby capturing the evolution law in the time dimension, such as the signal strength continuously and slowly declining in the past 5 time steps or the signal quality presenting regular fluctuations with a period of 2 seconds.

[0061] The spatial sequence features representing the instantaneous state output by the convolutional neural network branch and the time sequence features representing the historical law output by the recurrent neural network branch are fused to generate a network state comprehensive evaluation result through splicing operation. For example, the spatial features obtained by the convolutional neural network and the time sequence features obtained by the recurrent neural network are spliced into a new feature vector, and then further processed through a fully connected layer to finally output the comprehensive evaluation result of the network state. The network state comprehensive evaluation result contains various aspects of network performance, such as signal strength, delay, bandwidth utilization, etc.

[0062] The network state comprehensive evaluation result is analyzed to obtain a network power evaluation result and a channel quality evaluation result. The network power evaluation result is an evaluation of the signal transmission power level required to maintain or achieve the target communication quality in the current and future period of time, which is usually related to the signal strength, transmission distance and communication quality of the network, and reflects the power demand of the network equipment at a certain moment; the channel quality evaluation result is a comprehensive judgment of the quality of the wireless channel used at present and in the future, not only considering the current state, but also predicting its future availability, including the prediction of interference level, multipath effect, bandwidth occupancy, etc., which helps to determine whether the channel is suitable for current data transmission, and is usually quantified by signal-to-noise ratio or bit error rate, etc.

[0063] Based on the network power evaluation result, power analysis is performed, and according to the severity of the power gap, a specific power adjustment level is determined. For example, a required 5dB power boost is mapped to a level 3 power boost instruction. The power adjustment level is divided into several discrete levels according to the power evaluation result, such as fine tuning, medium rise and strong rise, which helps to simplify the control logic and improve stability.

[0064] Based on the channel quality evaluation results, channel analysis is performed, and from the pre-set clean channel list, an optimal channel to switch parameter is determined. When it is determined that channel switching is needed, the channel to switch parameter indicates the specific identity of the target channel, i.e., whether channel switching is needed and to which channel. The triggering conditions for channel switching usually include that the signal-to-noise ratio is lower than a certain threshold, the bit error rate is too high, etc. If the quality of the current channel is lower than the preset standard, the model will automatically recommend switching to a channel with higher quality.

[0065] The signal reception strength is retrieved in real time, and the power to be adjusted level and the channel to switch parameter are combined for comprehensive adjustment, i.e., the signal strength at the receiving end is taken as input to help judge the current advantages and disadvantages of the network state. Based on the comprehensive analysis results of power adjustment and channel switching, corresponding network adjustment parameters are generated, including transmission power adjustment value, channel switching instruction and PGA gain control value. The transmission power adjustment value represents the network transmission power that needs to be adjusted, and according to the network evaluation results, it is determined whether the power output needs to be increased or decreased; the channel switching instruction is used to instruct the network to switch to another channel, which is usually triggered when the current channel quality is poor; and the PGA gain control value is the value for controlling the signal gain, which is usually used to adjust the gain of the signal amplifier to optimize the signal transmission quality.

[0066] The received network adjustment parameters are sent to the control logic unit of the MOS transistor switch array for analysis, and corresponding gate control voltages are generated through the MOS transistor driving circuit to accurately control the conduction state of each switch transistor in the MOS transistor switch array, thereby changing the configuration of the radio frequency front-end matching network, attenuation network or amplification circuit. That is, the MOS transistor switch array is used to control power adjustment and gain adjustment. According to the adjustment parameters, the gate voltage of the MOS transistor will be adjusted to control the conduction and non-conduction of the switch. Through the control of the gate voltage, the switch state of the MOS transistor can be adjusted in real time. Different switch states correspond to different power outputs and signal gain levels. In order to ensure the adjustment accuracy, the output power value is monitored in real time, and the monitoring result is fed back to the control logic to form a closed-loop power control loop, in other words, dynamic adjustment is performed according to the difference between the actual power value and the target power value to form a closed-loop power control loop. The circuit control signal is a signal generated based on the network adjustment parameters, which is used to control the working state of the circuit and directly affects the power output, signal gain, channel selection, etc.

[0067] The circuit control signal is decomposed into multiple adjustment steps, and the circuit control signal is gain-controlled according to the multiple adjustment steps to generate a target gain value. The multiple adjustment steps are the steps of gain or power adjustment, which are usually adjusted step by step by a set increment. The smaller the step, the higher the accuracy of the adjustment. The target gain value determines the final amplification multiple of the signal. After the target gain value is generated, gradual gain adjustment is performed step by step, so that the gain value gradually transitions to the target gain value, avoiding excessive adjustment and ensuring smooth improvement of signal transmission quality. Through gradual gain adjustment, the signal quality is stably optimized in real-time network environment.

[0068] According to the target gain value, the circuit control signal is gradually adjusted in gain to generate a gain signal, which reflects the final adjustment result of power and gain in the network and is transmitted to the network device as a feedback signal to optimize its signal transmission.

[0069] Exemplarily, the network adjustment parameters include power adjustment +7dB, channel switching: channel 36, PGA gain +3dB. The analysis process of the MOS transistor switch array is as follows: for power adjustment, the driving circuit generates a specific gate voltage combination to control the different conduction states of the four MOS transistor switches, switches the operating bias point of the transmitter power amplifier from class A to class AB, and realizes +5dB adjustment; at the same time, the other two MOS transistors switch the π-type attenuation network on the radio frequency path to provide an additional +2dB adjustment. Real-time monitoring of the power meter reading forms a closed-loop control to ensure that the synthesized output power is accurately +7dB. For +3dB adjustment of the PGA gain, to avoid instantaneous mutation, it is decomposed into 6 adjustment steps of 0.5dB, and one step is executed every 10ms. According to this step sequence, gain control is performed to generate a gradual change curve from the current value to the target gain value. Based on this target value, gradual gain adjustment is performed to smoothly generate the required gain signal within 60ms, and the received signal strength is smoothly transitioned without any glitches or oscillations.

[0070] According to the gain signal, the mine wireless gateway is adaptively adjusted to ensure that the communication link maintains the optimal state in the complex mine environment, thereby maintaining stable and efficient communication quality. The transmission power, channel selection and gain control of the network are automatically adjusted according to the real-time network state. With the precise control of the MOS transistor switch array, the network signal strength and quality can be effectively adjusted to ensure stable network performance in a dynamically changing complex environment. The adaptive adjustment of the gain signal enables the network to be optimized in real time according to the changes in the environment, improving the coverage, stability and overall communication quality of the network.

[0071] In summary, the adaptive network adjustment method for the mine-used wireless gateway based on deep learning provided in the application has the following technical effects: the network state parameters of the mine-used wireless gateway are collected for feature analysis, a multi-dimensional network feature dataset is obtained for preprocessing to construct a standardized network dataset; a hybrid deep learning branch is constructed, the standardized network dataset is used to train the hybrid deep learning branch to generate a network parameter prediction channel; the network state parameters are analyzed through the network parameter prediction channel to output network adjustment parameters, circuit control is performed in combination with a MOS tube switch array to obtain a network control strategy to perform gain amplification on the mine-used wireless gateway to generate a gain signal to perform adaptive network adjustment on the mine-used wireless gateway. That is, by collecting the network state parameters of the mine-used wireless gateway and performing multi-dimensional feature analysis, various interference factors in a complex environment are identified based on a deep learning model, circuit control is performed in combination with a MOS tube switch array to perform intelligent gain amplification, the change of the network state is predicted in advance and active adjustment is performed, thereby improving the stability and efficiency of the wireless communication of the mine-used industrial gateway.

[0072] In the second embodiment, based on the same inventive concept as the adaptive network adjustment method for the mine-used wireless gateway based on deep learning in the foregoing first embodiment, the application further provides an adaptive network adjustment system for a mine-used wireless gateway based on deep learning. Please refer to the accompanying drawings Figure 2 The adaptive network adjustment system for the mine-used wireless gateway based on deep learning comprises: A parameter analysis module 11 is configured to collect the network state parameters of the mine-used wireless gateway for feature analysis, obtain a multi-dimensional network feature dataset for preprocessing to construct a standardized network dataset; a channel construction module 12 is configured to construct a hybrid deep learning branch, use the standardized network dataset to train the hybrid deep learning branch to generate a network parameter prediction channel; and a network adjustment module 13 is configured to analyze the network state parameters through the network parameter prediction channel to output network adjustment parameters, perform circuit control in combination with a MOS tube switch array to obtain a network control strategy to perform gain amplification on the mine-used wireless gateway to generate a gain signal to perform adaptive network adjustment on the mine-used wireless gateway.

[0073] Further, the parameter analysis module 11 in the adaptive network adjustment system for the mine-used wireless gateway based on deep learning is further configured to: alternately deploy double-frequency sensing nodes on the roof and sidewall of a mine roadway to construct a three-dimensional monitoring network; set a data collection frequency to activate the three-dimensional monitoring network to determine a plurality of sensing nodes for sensing collision analysis; when there is data sensing collision among the plurality of sensing nodes, access the mine-used wireless gateway for sensing according to a time division multiple access method according to a data transmission period to obtain the network state parameters of the plurality of sensing nodes.

[0074] Further, the parameter analysis module 11 in the deep learning-based adaptive network regulation system for mine wireless gateway is further configured to: divide the collection time data of the stereoscopic monitoring network into equal-length time slots according to the data collection frequency; match the equal-length time slots with the plurality of sensing nodes to construct a time slot allocation table; allocate the plurality of sensing nodes based on the time slot allocation table to obtain a plurality of transmission time slots; perform adjacent time slot overlap analysis on the plurality of transmission time slots to obtain data transmission overlap parameters; and monitor data transmission conflicts according to the data transmission overlap parameters to obtain data collision results of the plurality of sensing nodes.

[0075] Further, the parameter analysis module 11 in the deep learning-based adaptive network regulation system for mine wireless gateway is further configured to: perform multi-scale decomposition based on the network state parameters to obtain multi-scale decomposition coefficients; perform noise identification based on the multi-scale decomposition coefficients to obtain noise coefficient components; filter the network state parameters according to the noise coefficient components to obtain effective signal features; and perform signal propagation ray tracing on the mine wireless gateway according to the effective signal features to obtain the multi-dimensional network feature data set.

[0076] Further, the parameter analysis module 11 in the deep learning-based adaptive network regulation system for mine wireless gateway is further configured to: perform numerical analysis based on the multi-dimensional network feature data set to obtain numerical distribution characteristic parameters; perform data mapping on the numerical distribution characteristic parameters using a min-max normalization method to determine a target mapping interval; perform Z-score standardization calculation based on the target mapping interval to obtain a standardized grid initial data set; verify the standardization effect of the standardized grid initial data set, update the standardized grid initial data set in reverse, and obtain the standardized network data set.

[0077] Further, the channel construction module 12 in the deep learning-based adaptive network regulation system for mine wireless gateway is further configured to: alternately arrange a plurality of convolutional layers and a plurality of pooling layers to construct a convolutional neural network branch; connect a flattening layer to the end of the convolutional neural network branch to generate a one-dimensional feature vector; set a multi-layer gated recurrent unit network, each layer of which contains a plurality of gated recurrent units, to construct a recurrent neural network branch; connect a fully connected layer to the end of the recurrent neural network branch to extract time sequence features; and splice the one-dimensional feature vector and the time sequence features to construct the hybrid deep learning branch.

[0078] Further, the channel construction module 12 in the deep learning-based adaptive network regulation system for mine wireless gateway is further used for: dividing the standardized network data set into a training set, a validation set and a test set; performing segmented training on the convolutional neural network branch and the recurrent neural network branch respectively by using the training set, the validation set and the test set, to obtain a plurality of segmented training parameters; performing end-to-end joint training on the convolutional neural network branch and the recurrent neural network branch in combination with the plurality of segmented training parameters, to generate a joint training result; and performing optimization monitoring training on the standardized network data set based on the joint training result, to construct the network parameter prediction channel.

[0079] Further, the network regulation module 13 in the deep learning-based adaptive network regulation system for mine wireless gateway is further used for: inputting the network state parameters into the network parameter prediction channel; S1: extracting spatial sequence features of the network state parameters by the convolutional neural network branch; S2: capturing time sequence features of the network state parameters by the recurrent neural network branch; S3: fusing the spatial sequence features and the time sequence features to generate a network state comprehensive evaluation result; performing multi-dimensional network analysis based on the network state comprehensive evaluation result to generate a network regulation parameter; performing analysis on the network regulation parameter by the MOS tube switch array to generate a circuit control signal; decomposing the circuit control signal into a plurality of regulation steps, performing gain control on the circuit control signal according to the plurality of regulation steps to generate a target gain value; and performing gain progressive regulation on the circuit control signal based on the target gain value to generate the gain signal.

[0080] Further, the network regulation module 13 in the deep learning-based adaptive network regulation system for mine wireless gateway is further used for: performing analysis according to the network state comprehensive evaluation result to obtain a network power evaluation result and a channel quality evaluation result; performing power analysis based on the network power evaluation result to determine a power to be regulated level; performing channel analysis based on the channel quality evaluation result to determine a channel to be switched parameter; and combining and regulating the power to be regulated level and the channel to be switched parameter based on the signal receiving strength to generate the network regulation parameter.

[0081] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The foregoing Figure 1The deep learning based adaptive network adjustment method for mine wireless gateway in the first embodiment and the specific examples are also applicable to the deep learning based adaptive network adjustment system for mine wireless gateway in the present embodiment. Through the foregoing detailed description of the deep learning based adaptive network adjustment method for mine wireless gateway, those skilled in the art can clearly know the deep learning based adaptive network adjustment system for mine wireless gateway in the present embodiment. Therefore, for the sake of brevity of the description, the deep learning based adaptive network adjustment system for mine wireless gateway will not be described in detail herein.

[0082] The above description of disclosed embodiments enables one skilled in the art to make or use the application. Numerous modifications to these embodiments will be apparent to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0083] Obviously, various modifications and changes can be made to the present application by those skilled in the art without departing from the spirit and scope of the present application. Thus, it is intended that the present application cover the modifications and changes as long as they come within the scope of the application, and its equivalents.

Claims

1. A deep learning-based adaptive network adjustment method for mining wireless gateways, characterized in that, include: By collecting network status parameters of mining wireless gateways for feature analysis, a multi-dimensional network feature dataset is obtained, which is then preprocessed to construct a standardized network dataset. Construct a hybrid deep learning branch, train the hybrid deep learning branch using the standardized network dataset, and generate network parameter prediction channels; The network state parameters are analyzed through the network parameter prediction channel, and network adjustment parameters are output. Combined with the MOS transistor switching array for circuit control, a network control strategy is obtained to amplify the gain of the mining wireless gateway and generate a gain signal for adaptive network adjustment of the mining wireless gateway.

2. The adaptive network adjustment method for mining wireless gateways based on deep learning as described in claim 1, characterized in that, The process and methods for collecting network status parameters of a mining wireless gateway include: Dual-frequency sensor nodes are alternately deployed on the roof and sidewalls of mine roadways to construct a three-dimensional monitoring network; Set the data acquisition frequency to activate the three-dimensional monitoring network, and identify multiple sensor nodes for sensor collision analysis; When multiple sensor nodes experience data sensing collisions, the multiple sensor nodes are connected to the mining wireless gateway using time division multiple access according to the data transmission cycle to obtain the network status parameters of the multiple sensor nodes.

3. The adaptive network adjustment method for mining wireless gateways based on deep learning as described in claim 2, characterized in that, The method includes setting a data acquisition frequency to activate the three-dimensional monitoring network and identifying multiple sensor nodes for sensor collision analysis. The data acquisition time data of the three-dimensional monitoring network is divided into equal-length time slots according to the data acquisition frequency. The equal-length time slots are matched with multiple sensing nodes to construct a time slot allocation table; Multiple sensor nodes are allocated based on the time slot allocation table to obtain multiple transmission time slots; By traversing the multiple transmission time slots, adjacent time slot overlap analysis is performed to obtain data transmission overlap parameters; Based on the data transmission overlap parameter, data transmission conflicts are monitored to obtain data collision results from multiple sensor nodes.

4. The adaptive network adjustment method for mining wireless gateways based on deep learning as described in claim 1, characterized in that, Feature analysis was performed by collecting network status parameters from mining wireless gateways to obtain a multi-dimensional network feature dataset. The methods included: Multi-scale decomposition is performed based on the network state parameters to obtain multi-scale decomposition coefficients. Noise is identified based on the multi-scale decomposition coefficients to obtain noise figure components; The network state parameters are filtered according to the noise figure components to obtain effective signal characteristics; Based on the effective signal characteristics, signal propagation ray tracing is performed on the mining wireless gateway to obtain the multi-dimensional network feature dataset.

5. The adaptive network adjustment method for mining wireless gateways based on deep learning as described in claim 1, characterized in that, Methods include: obtaining multi-dimensional network feature datasets, preprocessing them, and constructing standardized network datasets. Numerical analysis is performed on the multi-dimensional network feature dataset to obtain numerical distribution characteristic parameters. The min-max normalization method is used to map the numerical distribution characteristic parameters to determine the target mapping interval. Z-score standardization calculation is performed based on the target mapping interval to obtain the initial standardized grid dataset; Verify the standardization effect of the initial standardized grid dataset, and update the initial standardized grid dataset in reverse to obtain the standardized network dataset.

6. The adaptive network adjustment method for mining wireless gateways based on deep learning as described in claim 1, characterized in that, The process of constructing hybrid deep learning branches includes the following methods: Multiple convolutional layers and multiple pooling layers are arranged alternately to construct branches of a convolutional neural network; A flattening layer is connected to the end of the branch of the convolutional neural network to generate a one-dimensional feature vector; A multi-layer gated recurrent unit network is configured, wherein each layer of the multi-layer gated recurrent unit network contains multiple gated recurrent units, thereby constructing a recurrent neural network branch; A fully connected layer is connected to the end of the recurrent neural network branch to extract temporal features; The one-dimensional feature vector is concatenated with the temporal features to construct the hybrid deep learning branch.

7. The adaptive network adjustment method for mining wireless gateways based on deep learning as described in claim 6, characterized in that, The method for training the hybrid deep learning branch using the standardized network dataset to generate network parameter prediction channels includes: The standardized network dataset is divided into a training set, a validation set, and a test set; The training set, the validation set, and the test set are used to train the convolutional neural network branch and the recurrent neural network branch in segments to obtain multiple segmented training parameters; Unfreeze the convolutional neural network branch, combine the recurrent neural network branch with the multiple segmented training parameters, and perform end-to-end joint training to generate joint training results; Based on the joint training results, the standardized network dataset is optimized and monitored for training to construct the network parameter prediction channel.

8. The adaptive network adjustment method for mining wireless gateways based on deep learning as described in claim 6, characterized in that, The network state parameters are analyzed through the network parameter prediction channel to output network adjustment parameters. Combined with a MOS transistor switching array for circuit control, a network control strategy is obtained to amplify the gain of the mining wireless gateway and generate a gain signal. The method includes: The network state parameters are input into the network parameter prediction channel: S1: Extract spatial sequence features of network state parameters through the branches of the convolutional neural network; S2: Capture the time-series features of the network state parameters through the branches of the recurrent neural network; S3: The spatial sequence features and the time sequence features are fused to generate a comprehensive evaluation result of the network state; Based on the comprehensive evaluation results of the network state, multidimensional network analysis is performed to generate network adjustment parameters; The circuit control signal is generated by analyzing the MOS transistor switching array in conjunction with the network adjustment parameters. The circuit control signal is decomposed into multiple adjustment steps, and the gain of the circuit control signal is controlled according to the multiple adjustment steps to generate a target gain value. The gain signal is generated by progressively adjusting the gain of the circuit control signal based on the target gain value.

9. The adaptive network adjustment method for mining wireless gateways based on deep learning as described in claim 8, characterized in that, Based on the comprehensive network state evaluation results, multidimensional network analysis is performed to generate network adjustment parameters. The method includes: The network power assessment result and the channel quality assessment result are obtained by analyzing the comprehensive network status assessment result. Based on the network power assessment results, power analysis is performed to determine the power adjustment level. Based on the channel quality assessment results, channel analysis is performed to determine the parameters to be switched to the channel. The signal reception strength is retrieved, and the power level to be adjusted and the channel switching parameters to be combined and adjusted to generate the network adjustment parameters.

10. A deep learning-based adaptive network adjustment system for mining wireless gateways, characterized in that: The step of implementing the deep learning-based adaptive network adjustment method for mining wireless gateways according to any one of claims 1 to 9, wherein the deep learning-based adaptive network adjustment system for mining wireless gateways comprises: The parameter analysis module is used to perform feature analysis by collecting network status parameters of the mining wireless gateway, obtain a multi-dimensional network feature dataset for preprocessing, and construct a standardized network dataset. The channel construction module is used to construct a hybrid deep learning branch, train the hybrid deep learning branch using the standardized network dataset, and generate network parameter prediction channels. The network adjustment module is used to analyze the network status parameters through the network parameter prediction channel, output network adjustment parameters, combine with the MOS transistor switching array for circuit control, obtain a network control strategy to amplify the gain of the mining wireless gateway, and generate a gain signal to perform adaptive network adjustment of the mining wireless gateway.

Images