A Smart Anti-interference Method Based on Multi-Physics Coupling and Polarization Diversity

Abstract

This invention relates to the field of wireless communication technology, and in particular to an intelligent anti-interference method based on multi-physics coupling and polarization diversity. It integrates material shielding effects, dust scattering characteristics, and dynamic polarization optimization mechanisms to improve the reliability and stability of wireless communication in deep well environments. First, a modified path loss model is established based on the eddy current shielding effect of metallic support materials and Mie scattering theory of dust concentration. Second, an intelligent anti-interference layered framework is proposed, combining direct sequence spread spectrum (DSSS) technology with wavelet adaptive filtering, and incorporating game theory to optimize power allocation among multi-source nodes. Furthermore, a polarization-delay joint compensation strategy is designed, employing circularly polarized antennas and a channel equalizer based on convolutional neural networks and long short-term memory networks to suppress multipath fading and inter-symbol interference in non-line-of-sight scenarios. This invention can reduce the bit error rate by 42% compared to traditional methods in high dust concentration (200 g / m³) environments and improve the signal-to-interference-plus-noise ratio by 15 dB in areas with strong electromagnetic interference.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, specifically a smart anti-interference method based on multi-physics coupling and polarization diversity. Background Technology

[0002] With the surge in global demand for mineral resources and the advancement of intelligent mine construction, wireless communication technology in deep-well environments has become a core support for ensuring safe production and improving mining efficiency. Deep-well spaces are characterized by narrow tunnels, complex support structures, high dust concentrations, and dense electromagnetic interference sources (such as high-voltage equipment, frequency converters, and electromechanical switches), leading to problems such as multipath fading, increased path loss, and significant delay spread for wireless signals. For example, in emergencies such as coal mine collapses, traditional wired communication systems are prone to failure due to cable damage, while existing wireless technologies (such as WiFi and ZigBee) have significant shortcomings in penetration, anti-interference, and dynamic adaptability, making it difficult to achieve reliable two-way communication and real-time rescue command. Statistics show that the average mining depth of large and medium-sized coal mines in my country exceeds 500 meters, with some mines exceeding 1,000 meters in depth. The complex electromagnetic environment further exacerbates the challenges to signal transmission stability.

[0003] Existing wireless communication technologies in mines mainly include short-range wireless communication, acoustic wave transmission, electromagnetic wave transmission, and 5G technology. Among these, ZigBee is used in downhole sensor networks due to its low power consumption and network capacity advantages, but its anti-interference capability is limited. Acoustic wave transmission propagates through oil pipe strings and uses single-carrier frequency domain equalization (SC-FDE) technology to suppress multipath interference, but its hardware complexity is high and it is significantly affected by dust concentration. Electromagnetic wave transmission (such as MWD systems) has strong penetration, but it suffers from high bit error rates due to the shielding effect of metal supports and high-frequency noise interference. 5G technology reduces roadway attenuation by optimizing the 700-900MHz frequency band and introduces reconfigurable smart surfaces (RIS) to enhance coverage, but it still faces signal quality fluctuations caused by dynamic electromagnetic interference (such as transient noise from electromechanical equipment). Furthermore, existing anti-interference technologies such as adaptive filtering and blind source separation (BSS) suffer from high computational complexity or residual interference problems, while deep learning-based signal detection methods (such as dual-path network DPN) can reduce bit error rates, but their ability to jointly optimize complex multi-source interference is insufficient.

[0004] In general, existing mine wireless communication technologies suffer from the following core problems: First, insufficient modeling of environmental coupling effects. Existing research often uses simplified path loss models, failing to systematically integrate the electromagnetic shielding effect of support materials with the scattering and absorption of dust particles, resulting in limited channel prediction accuracy. Second, weak dynamic interference suppression capabilities. Traditional anti-interference technologies are difficult to adapt to the mixed characteristics of impulse noise and Gaussian noise in deep well environments, and lack real-time perception and compensation mechanisms for time-varying channel states. Third, a lack of coordinated optimization of polarization and multipath effects. Existing research often focuses on the independent performance analysis of a single polarization mode, neglecting the multipath suppression potential of circularly polarized antennas in non-line-of-sight (NLOS) dense reflection scenarios. At the same time, there is a lack of joint optimization strategies for dynamic equalization algorithms and polarization diversity techniques, making it difficult to achieve coordinated suppression of complex electromagnetic interference and multipath fading. Summary of the Invention

[0005] This invention aims to address the technical problems in wireless communication in complex electromagnetic environments of deep wells, such as insufficient path loss prediction accuracy, limited dynamic interference suppression capability, and lack of synergistic optimization of multipath fading and polarization effects. It proposes an intelligent anti-interference method based on multi-physics coupling and polarization diversity, which combines multi-physics mechanism modeling with intelligent anti-interference strategies to achieve highly reliable wireless communication with low bit error rate.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent anti-interference method based on multi-physics coupling and polarization diversity, comprising the following steps: constructing a multi-physics coupling path loss model based on the eddy current shielding effect of metal support materials and the Mie scattering theory of dust concentration; designing an intelligent anti-interference hierarchical architecture, including spread spectrum anti-narrowband interference, wavelet adaptive filtering to suppress impulse noise, and game theory-based distributed power control; utilizing polarization diversity and deep learning channel equalization strategies, suppressing multipath effects through circularly polarized antennas, and combining CNN-LSTM networks for dynamic channel equalization.

[0007] Furthermore, the construction of the multiphysics coupled path loss model includes: establishing a metal support shielding effect model, calculating the shielding attenuation factor based on eddy current loss and reflection loss; establishing a dust scattering effect model, calculating the extinction efficiency factor and total attenuation factor based on Mie scattering theory; and integrating the shielding attenuation factor and total attenuation factor into the path loss formula to form a modified path loss model.

[0008] Furthermore, the intelligent anti-interference layered architecture includes: the first layer adopts direct sequence spread spectrum (DSSS) technology, using Gold code sequences for spread spectrum modulation to suppress narrowband interference; the second layer adopts an adaptive threshold denoising method based on discrete wavelet transform (DWT) to filter impulse noise; and the third layer constructs a distributed power control model based on non-cooperative game theory, iteratively optimizing the transmit power of each node to maximize the system utility function.

[0009] Furthermore, the processing gain of the DSSS technology is no less than 20 dB, the code length of the Gold code is 63, and the maximum cross-correlation value is no greater than 7.

[0010] Furthermore, the adaptive threshold denoising method employs a hybrid soft and hard thresholding strategy, with the threshold dynamically determined based on the median absolute deviation (MAD).

[0011] Furthermore, the utility function of the distributed power control model considers channel gain, noise power, interference power, and energy consumption weight, and iteratively solves for the Nash equilibrium point using a gradient projection algorithm.

[0012] Furthermore, the polarization diversity and deep learning channel equalization strategy includes: using a circularly polarized antenna to receive the signal to reduce the correlation between multipath components and achieve polarization diversity; constructing a CNN-LSTM dynamic equalizer, including a one-dimensional convolutional layer, a max pooling layer, an LSTM layer and a fully connected layer, for performing time-series modeling and symbol recovery of the received signal.

[0013] Furthermore, the input to the CNN-LSTM equalizer is a time-domain signal sequence of length 256, and the output is the equalized symbol estimate.

[0014] Furthermore, a wireless communication system includes: a transmitting module for generating and transmitting a DSSS-modulated signal; a receiving module including a circularly polarized antenna and a CNN-LSTM equalizer; and a processing module for performing wavelet denoising, power control, and channel equalization operations.

[0015] Furthermore, the system is applicable to deep well safety monitoring, emergency communication, or intelligent mine wireless communication scenarios.

[0016] The beneficial effects of the present invention are: (1) systematically integrating the metal shielding effect, dust scattering characteristics and geometric structure influence, significantly improving the channel prediction accuracy; (2) achieving coordinated optimization of polarization diversity and dynamic equilibrium, effectively reducing the bit error rate in NLOS scenarios; (3) constructing a low-computational-overhead hierarchical anti-interference system, significantly improving SINR in mixed noise environments. Attached Figure Description

[0017] Figure 1This is a diagram of the three-level hierarchical anti-interference architecture of the present invention.

[0018] Figure 2 The time delay spectrum comparison and cumulative distribution curves of the two polarization modes of the present invention are shown.

[0019] Figure 3 This is a diagram of the CNN-LSTM dynamic equalizer architecture.

[0020] Figure 4 It is a three-dimensional model of a mine tunnel.

[0021] Figure 5 Path loss results under high-concentration dust and non-conductive support.

[0022] Figure 6 Path loss results of high-concentration dust and highly conductive metal supports.

[0023] Figure 7 This relates to convergence and power balance in game theory power control.

[0024] Figure 8 This section compares the convergence performance of equalizers in static channels.

[0025] Figure 9 This section compares the bit error rate performance of equalizers in dynamic channels. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative effort are all within the scope of protection of the present invention. Improvements and modifications based on the present invention are all within the scope of protection of the present invention.

[0027] This invention proposes an intelligent anti-interference method based on multi-physics coupling and polarization diversity. This method systematically integrates material shielding effects, dust scattering characteristics, and dynamic anti-interference mechanisms to achieve reliability and stability in complex deep well environments. Specifically, it quantifies the eddy current shielding effect of metallic support materials using finite element method (FEM) simulation, establishes a nonlinear relationship between dust concentration and signal attenuation using Mie scattering theory, and proposes a modified path loss model. This paper breaks through the simplistic assumptions of traditional models regarding the coupling effects of multiple factors. Based on this, an intelligent anti-interference hierarchical architecture is designed. Direct sequence spread spectrum (DSSS) technology is used to expand the signal bandwidth to suppress narrowband interference. Wavelet adaptive threshold denoising algorithm is used to eliminate impulse noise, and game theory is introduced to optimize the power allocation of multi-source nodes, achieving a 15dB improvement in signal-to-interference-plus-noise ratio (SINR) in mixed noise environments. Simultaneously, combining polarization diversity and deep learning techniques, a collaborative architecture of circularly polarized antenna and CNN-LSTM dynamic equalizer is proposed. Circularly polarized waves are used to suppress multipath reflection phase reversal, and time-varying response is predicted based on historical channel state information (CSI), reducing the bit error rate (BER) by 42% in non-line-of-sight (NLOS) scenarios. This model is verified through simulation and field tests in coal mines, demonstrating significant robustness in high dust (200g / m³) and strong interference (85dBμV / m) environments, providing a theoretical and technological breakthrough for the design of deep-well communication systems.

[0028] The specific steps of this invention are as follows: 1) Constructing a multi-physics coupling path loss model; 2) Designing an intelligent anti-interference hierarchical architecture; 3) Utilizing polarization diversity and deep learning channel equalization strategies.

[0029] In step 1), wireless signal propagation in a deep well environment is subject to multiple physical field couplings. The deep well communication environment not only exhibits NLOS multipath propagation but also displays strong nonlinear coupling characteristics due to the support eddy current effect and the time-varying dust concentration. The densely distributed metal support structures (such as steel arches and metal mesh) within the tunnel form an electromagnetic shielding layer through the eddy current effect, significantly altering the propagation path and energy distribution of electromagnetic waves. Simultaneously, high-concentration dust (mainly composed of coal dust and rock fragments) produces Mie scattering and absorption effects on electromagnetic waves, further exacerbating signal attenuation. Furthermore, the multiple reflections and diffractions caused by the narrow tunnel geometry lead to severe multipath fading and delay spread, especially in the non-line-of-sight (NLOS) region, where the signal energy distribution exhibits high non-uniformity. However, existing channel models (such as COST231 and logarithmic distance models) are mostly based on simplification assumptions, treating the above factors as independent variables and neglecting the nonlinear superposition effect of multiphysical field coupling on signal propagation. Therefore, this invention proposes a multiphysical field coupling path loss model to optimize the above problems. To quantify the aforementioned multiphysics coupling effect, this application establishes a metal support shielding effect model and a dust scattering effect model, and finally integrates a multiphysics coupling path loss model.

[0030] Metal Support Shielding Effect Model: The attenuation of electromagnetic waves by a metal support structure stems from the combined effects of eddy current loss and reflection loss. When an electromagnetic wave penetrates the metal mesh, an alternating magnetic field induces eddy currents on the conductor surface, and its power loss density can be expressed as: Where σ is the electrical conductivity of the metal, and ω is the angular frequency of the electromagnetic wave. ρ is the effective value of magnetic induction intensity, d is the thickness of the support material, and ρ is the resistivity.

[0031] Furthermore, the present invention defines the shielding attenuation factor as: ,in, To improve shielding efficiency.

[0032] Dust scattering effect model: The attenuation of electromagnetic waves by deep well dust particles (particle size r = 1-100μm) is due to scattering ( ) and absorption ( (Joint contributions). Based on Mie scattering theory, extinction efficiency factor It can be represented as: ,in, These are dimensional parameters; , Here, represents the Michaelis coefficients, and n is the index of the order of the series expansion. This indicates taking the real part of a complex number.

[0033] Assuming the dust particles are uniformly distributed, the particle number density per unit volume ,in Mass concentration (g / m³) =2.2 g / cm 3 r is the density of SiC eff =10μm is the equivalent particle size.

[0034] Total attenuation factor It can be represented as: Where N is the particle number density per unit volume. The extinction cross section of a single particle. Extinction efficiency factor For mass concentration, r is the density of SiC eff The equivalent particle size.

[0035] Multiphysics Coupled Path Loss Model: Based on the single-factor model described above, a multiphysics coupled path loss formula is proposed. Specifically, considering support shielding, dust scattering, and tunnel geometric effects, the path loss model is modified as follows: ,in, Let represent the free-space reference loss (d0=1 m, λ=0.125 m@2.4GHz), and n represent the path loss exponent. Indicates the shadow fading component. To shield the attenuation factor, This is the total attenuation factor.

[0036] In step 2), electromagnetic interference in the deep well environment is characterized by a mixture of impulse noise (transient interference from electromechanical switches) and Gaussian noise, and time-varying multipath fading. For example... Figure 1 As shown, this application proposes a three-level hierarchical anti-interference architecture of "spread spectrum-filtering-game optimization", which achieves coordinated optimization of interference suppression and signal recovery through staged processing.

[0037] Spread Spectrum for Narrowband Interference Suppression: In the first layer, to suppress narrowband interference in the deep well electromagnetic environment, Direct Sequence Spread Spectrum (DSSS) technology is used to extend the bandwidth of the original signal to 20MHz, thereby improving anti-interference capability. This technology utilizes the low cross-correlation characteristics of pseudo-random codes (PN codes) to achieve despreading suppression of interference signals at the receiver.

[0038] This system constructs a Gold code sequence based on the IEEE 802.15.4 standard, with a code length of 63, a bit rate of 10 Mcps, and a processing gain of: The code has an autocorrelation peak value N of 63 and a maximum cross-correlation value ≤9, which can effectively suppress co-frequency interference signals and achieve a balance between spread spectrum modulation and signal fidelity.

[0039] Wavelet Adaptive Threshold Filtering: For second-level transient impulse interference, an adaptive denoising method based on Discrete Wavelet Transform (DWT) is proposed. This method performs multi-scale decomposition of the signal in the wavelet domain and applies dynamic thresholds to different frequency bands to enhance the suppression of sudden non-stationary noise.

[0040] First, the original signal is decomposed into three levels using wavelet basis functions to extract detail coefficients. Then, a hybrid soft-hard thresholding strategy is used for the high-frequency coefficients, that is, a hard threshold is used in the high-intensity pulse segment to suppress spikes, and a soft threshold is used in the medium and low intensity segment to preserve signal details.

[0041] Finally, the threshold is determined using an adaptive formula based on the median absolute deviation (MAD): , For adaptive threshold, is the median absolute deviation of the high-frequency wavelet coefficients, used for robust estimation of noise amplitude; the constant 0.6745 is a correction factor under the standard normal distribution, used to adjust... The signal is converted to a noise standard deviation estimate; N is the number of high-frequency coefficients after wavelet decomposition, and the 2logN term comes from the maximum likelihood threshold theory to ensure the statistical optimality of the threshold. Finally, the signal is reconstructed through inverse wavelet transform to achieve effective filtering of impulse interference.

[0042] Game theory-driven distributed power allocation: In deep-well communication systems, when multiple source nodes (such as sensors and device controllers) autonomously adjust their transmission power, co-channel interference and energy waste can easily occur. Traditional centralized power control methods rely on a central controller, which poses a single point of failure risk and has poor adaptability in dynamic environments.

[0043] To address this, this application models the multi-node power allocation problem based on non-cooperative game theory, endowing each node with autonomous decision-making capabilities. In this model, each communication node in the network is considered a game participant, whose goal is to maximize its own utility function while considering channel quality and energy consumption. Through distributed optimization, the nodes eventually converge to a Nash equilibrium, thereby achieving global performance optimization without centralized coordination.

[0044] The utility function for the i-th node is defined as follows: ,in, Let be the transmit power of node i. Its corresponding channel gain, For noise power, This indicates interference from other nodes. This is the energy consumption weighting coefficient.

[0045] To find the Nash equilibrium point, this application employs the gradient projection algorithm to iteratively update the node power. The main steps are as follows: Iterative update The formula is as follows: Where α is the step size coefficient. This indicates a non-negative projection operation.

[0046] Through multiple iterations, all nodes can converge to a stable power configuration without communication negotiation, ensuring communication quality while reducing interference and energy consumption.

[0047] In step 3), in the complex deep-well communication environment, signal propagation is severely affected by multipath effects and time-varying interference. To improve the channel quality and demodulation reliability at the receiver, this invention designs a joint strategy of "physical layer anti-interference + deep learning equalizer". First, a circularly polarized (CP) antenna is used as the front-end receiving device. By utilizing its polarization orthogonality for reflected and scattered waves, the correlation between multipath components is effectively reduced, thereby compressing the RMS delay spread and mitigating inter-symbol interference (ISI).

[0048] Building upon this foundation, to further compensate for residual ISI effects, this application constructs a dynamic channel equalizer based on a CNN-LSTM architecture. This network integrates the local perturbation awareness capability of CNNs with the temporal modeling advantages of LSTMs, adaptively adjusting the equalization coefficients under varying dust concentrations and material reflection conditions to achieve modeling and symbol recovery of dynamic channel state information (CSI). Ultimately, this joint strategy significantly improves convergence speed and equalization accuracy while maintaining high receiver reliability.

[0049] Multipath suppression mechanism of circularly polarized antennas: In NLOS scenarios of mobile communication, signals reach the receiver through multiple paths, increasing delay spread and leading to severe inter-symbol interference (ISI). Circularly polarized (CP) antennas, due to their equivalent radiation of left- and right-hand circular polarization components, generate orthogonal polarization components from reflected and scattered waves from different paths, thus significantly reducing the correlation between different path components and achieving polarization diversity.

[0050] To quantitatively evaluate the performance of the CP antenna in suppressing multipath delay spread and improving diversity gain, this application uses the WirelessInSite ray tracing simulation tool to compare the channel delay characteristics of the CP and traditional linear polarization (LP) antennas in the same NLOS urban street environment. Simulation results show that the RMS delay spread of the CP antenna is on average 47% of that of the LP antenna, a reduction of approximately 53%; under the same link conditions, the CP antenna can achieve a diversity gain of approximately 8.3 dB.

[0051] Figure 2 The comparison of the time delay spectrum and cumulative distribution curves of the two polarization methods in the simulation scenario is shown. It can be seen that the circularly polarized antenna significantly compresses the multipath time delay distribution by polarization orthogonalizing the multipath components, effectively reducing ISI, and providing a more "flat" input channel response with less time delay spread for the subsequent deep learning channel equalization module.

[0052] like Figure 3 As shown, the CNN-LSTM dynamic equalizer design takes the input time-domain received signal as a sequence: Where T=256 represents the input sampling length. The sequence is first processed by a one-dimensional convolutional layer, where the convolutional kernel scans the signal sequence using a sliding window to effectively extract local perturbation features between adjacent symbols.

[0053] The convolution process can be formally represented as: ,in, This represents the output of the c-th channel at time step t. For the parameters of the c-th convolutional kernel, For bias terms, The ReLU activation function is used, where k represents the kernel index, and the input sequence is at time step [number missing]. The value of .

[0054] After the convolution operation, the output features are downsampled using 2×1 max pooling to compress the feature dimension, and expressed as follows: ,in, These represent the outputs before and after pooling, respectively. Pooling reduces computational complexity while preserving key features, which helps improve the generalization performance of the network.

[0055] Next, the sequence features extracted by the CNN module are fed into the LSTM for temporal modeling. The internal update mechanism of the LSTM is as follows: ,in, For the current input, For output, This is the hidden state from the previous time step. For memory state, The memory state of the previous step, Forget gate vector, For the input gate vector, For the output gate vector, For the input weight matrix, The hidden layer weight matrix is... Bias vector, ⊙ represents the Sigmoid function, and ⊙ represents element-wise multiplication.

[0056] The aforementioned gating mechanism captures the evolution characteristics of Channel State Information (CSI) over time and outputs the hidden state hth_tht as the basis for equalization output.

[0057] Finally, the network feeds the output features of the LSTM into the fully connected layer to predict the equalization compensation coefficients: Where fc represents a fully connected layer, That is, the sign estimate after equilibrium. Here is the weight matrix of the fully connected (fc) layer. For the bias term of fc, This is the hidden state output vector of the LSTM at time step t.

[0058] Through three-stage optimization—multi-physics coupling modeling, intelligent anti-interference layered processing, and polarization diversity dynamic equalization—the method proposed in this application can achieve stable and efficient wireless communication in complex electromagnetic environments in deep wells. The transmission performance results obtained have important reference value in mine safety monitoring and emergency command.

[0059] In this application, in order to systematically evaluate the performance of the proposed "multi-physics coupled path loss model + CP + CNN-LSTM joint anti-interference strategy" in a complex deep well communication environment, the present invention builds the following three-layer simulation platform and executes it in sequence according to the process.

[0060] First, a 100m×4m×3m three-dimensional mine tunnel model was constructed in COMSOL Multiphysics. Figure 4 ), respectively embedded in steel (5mm thick, conductivity) ) and fiberglass (10mm thick, A support structure was constructed to simulate the shielding effect of different materials on 2.4 GHz electromagnetic waves. The outer boundary was set as a perfectly matched layer (PML) to absorb outgoing waves and avoid boundary reflections. A 20 dBm dipole antenna was installed at the transmitting end, and a probe was placed every 10 m along the centerline of the tunnel at the receiving end. Subsequently, through joint simulation of the "electromagnetic wave-frequency domain" module and the "particle tracking" module, multi-physics field coupling calculations of Mie scattering and absorption of 10 μm dust (concentration 50-300 g / m³) were realized, thereby obtaining the spatial field distribution and power delay spectrum.

[0061] Next, on the MATLAB / Simulink platform, a 1Msps, 256-symbol QPSK modulated baseband signal was generated. AWGN (SNR 5-15dB) and impulse noise with a 10% duty cycle and a peak power of -10dBm were injected into the channel to simulate complex mixed interference in a deep well environment. The obtained received waveforms and the corresponding COMSOL simulated impulse responses (100,000 sets in total) were used to form a training set. A CNN-LSTM dynamic equalizer was built in the PyTorch framework and trained end-to-end, with the optimization objective being a weighted sum of MSE and cross-entropy.

[0062] Finally, to verify the multipath suppression and diversity gain of the circularly polarized (CP) antenna, ray tracing simulations were performed in WirelessInSite using the same geometric and material parameters as in COMSOL. A wall roughness of 0.1 m was set to simulate irregular reflections, and transmission links for both linearly polarized (LP) and circularly polarized (CP) antennas were simulated. The RMS delay spread and diversity gain at each receiving point were then statistically analyzed.

[0063] To further highlight the modeling advantages of multi-physics coupled channel models in deep well environments, this invention systematically compares and analyzes path loss in three representative scenarios.

[0064] In scenario ① (open alleyway), due to the absence of dust and the extremely low conductivity of the support material, electromagnetic wave propagation mainly follows free space geometric attenuation. Both the traditional log-DistancePathLossModel and the coupling model of this application can reproduce the COMSOL multiphysics simulation results well in this scenario.

[0065] In scenario ② (high concentration of dust, non-conductive support), the Mie scattering and absorption effects of high-concentration particles begin to significantly enhance the nonlinear growth of path loss. Figure 5 This results in a mean error of approximately 4.2 dB for the traditional logarithmic distance model, which only considers geometric attenuation. In contrast, the dust attenuation term introduced in the coupled model accurately characterizes the contributions of scattering and absorption, keeping the mean error within 1.5 dB (a reduction of approximately 64%).

[0066] To more clearly present the comparison results of the most complex scenario ③ (high concentration of dust + highly conductive metal support), Figure 6 The path loss curve under this specific operating condition was selected. The three broken lines in the figure represent the COMSOL multiphysics simulation results, the predicted values ​​of the traditional logarithmic distance model, and the output of the multiphysics coupling model proposed in this application, respectively. This allows for a direct comparison: In the most complex environment, the traditional model severely underestimates the attenuation by neglecting the two major physical mechanisms of dust scattering and metal eddy current shielding; while the coupling model achieves high-precision fitting of the COMSOL simulation points by superimposing Mie scattering and eddy current loss terms into the logarithmic distance formula. Its curve almost coincides with the simulation results, with an average fitting error of only 2.1 dB.

[0067] As shown in Table 1, comparing the average error of the traditional model and the coupled model in three scenarios, the coupled model outperforms the traditional model in all operating conditions, especially under complex coupling conditions.

[0068] Table 1. Comparison of average path loss model errors in different scenarios

[0069]

[0070] In scenario ①, due to the simple environment and lack of additional physical coupling, the coupled model can naturally degenerate to the level of the traditional model.

[0071] In scenario ②, the dust scattering effect is prominent. The coupled model reduces the error from 4.2dB to 1.5dB by adding a dust attenuation term.

[0072] In scenario ③, the most complex physical mechanisms coexist, yet the coupled model still accurately reproduces the actual attenuation, significantly outperforming the traditional model. This result fully demonstrates that by incorporating the physical loss mechanisms of metal shielding and dust scattering into the path loss model, higher-precision channel modeling can be achieved in deep well environments with intertwined multi-physics fields, providing a reliable theoretical basis for subsequent channel simulation and equalization algorithm design.

[0073] To fully verify the effectiveness of the proposed three-level intelligent anti-interference architecture of "spread spectrum anti-interference - wavelet filtering - game theory power control", this application conducted experiments from three aspects: anti-interference performance, signal recovery quality and power consumption control, covering narrowband interference, impulse noise and multi-node co-channel interference scenarios commonly encountered in typical deep well communication environments.

[0074] (1) Comparative Analysis of Spread Spectrum Anti-Interference Performance: Under the test condition of SIR (Signal-to-Interference Ratio) of -5dB, the bit error rate performance of traditional baseband communication and the adopted DSSS (Direct Sequence Spread Spectrum) system under spectral interference conditions was compared. As shown in Table 2, the DSSS system, relying on Gold code spreading (processing gain up to 20dB), significantly reduced the bit error rate, and the BER remained at 10 under strong interference conditions. -3 The order of magnitude is far superior to that of the unspread system (10). -1 Level 1 performance.

[0075] Table 2. BER comparison under narrowband interference (SIR = -5dB)

[0076]

[0077] (2) Comparative Analysis of Wavelet Filtering Performance Against Impulse Interference: For typical non-Gaussian impulse interference in deep wells, a sudden impulse signal interference scenario with a 10% duty cycle was designed. Three wavelet denoising strategies were adopted: no filtering, fixed threshold denoising, and hybrid threshold (hard + soft) denoising. Their mean square error (MSE) performance under 1000 samples was compared. As shown in Table 3, the proposed hybrid threshold method effectively suppresses spike interference while maintaining low distortion, with an average MSE of 1.2 × 10⁻⁶. -3 It outperforms other methods, demonstrating stronger robustness and adaptability.

[0078] Table 3. Comparison of mean square error of different wavelet denoising methods under impulse noise (10% duty cycle)

[0079]

[0080] (3) Performance analysis of game theory-driven distributed power control: In a typical multi-node co-channel environment, a distributed network system with 6 nodes is constructed. The transmit power of each node is randomly initialized, and the power strategy is iteratively solved based on a non-cooperative game model. Based on the comparison between centralized optimal power control and traditional fixed power methods, the convergence, total energy consumption and SINR level of the proposed algorithm are analyzed.

[0081] like Figure 7 As shown, the game-theoretic control strategy achieves rapid convergence within 20 iterations, with the power of each node tending to be balanced and maintaining stability. As shown in Table 4, the proposed algorithm achieves an optimization of approximately 23.5% in terms of total power consumption control, while effectively improving the average SINR, verifying its adaptability and efficiency in complex interference scenarios.

[0082] Table 4. Comparative Analysis of Power Control Performance in Multi-Node Co-channel Interference Scenarios

[0083]

[0084] To evaluate the effectiveness of the proposed "circular polarization + CNN–LSTM" joint anti-interference strategy in a typical deep well environment, this application designed a set of system comparison experiments on the MATLAB communication link simulation platform, covering two test scenarios: static channel (fixed dust concentration and support type) and dynamic channel (interference intensity and environmental parameter changes).

[0085] In static channel testing, this invention compared the convergence performance and recovery accuracy of three equalizers—traditional LMS, MMSE, and the CNN-LSTM proposed in this application—under an environment of SNR=10dB and impulse noise duty cycle of 10%. Figure 8 As shown, the NMSE of the LMS and MMSE equalizers converges slowly, tending to 0.035 and 0.024 after 50 epochs, respectively, while the CNN-LSTM model converges to 0.014 after about 20 epochs, ultimately reducing the NMSE by 42.5% compared to MMSE and 60.1% compared to LMS. Table 5 shows that the comparison between the final NMSE and the static average BER of the three models further verifies the advantage of CNN-LSTM in static deep well channels: its BER is controlled at 4.7 × 10⁻⁶. -3 The following methods are significantly superior to traditional methods.

[0086] Table 5. Comparison of final NMSE and BER performance of the three equalizers under static test conditions.

[0087]

[0088] In dynamic channel testing, this invention simulates a sudden scenario where dust concentration increases abruptly from 50 g / m³ to 300 g / m³, and interference intensity changes abruptly from 85 dBμV / m to 95 dBμV / m. For example... Figure 9 As shown, when the frame number exceeds 50, the BER of the traditional equalizer increases significantly and is almost impossible to recover in subsequent frames. In contrast, the CNN-LSTM equalizer completes adaptive adjustment in only 3 frame updates, keeping the BER fluctuation stable within 0.01, demonstrating excellent real-time interference suppression capability and environmental adaptability.

[0089] In summary, the experiments fully verified the robustness and generalization ability of the intelligent anti-interference architecture in the variable deep well environment, especially under the conditions of coupled noise sources, nonlinear interference and time-varying channels, it has significant performance advantages.

[0090] The above description is merely one embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, or improvements made within the technical solution and principle of the present invention, especially the application of principles and design methods, such as the design principles and specific effective methods of the present invention, should be included within the protection scope of the present invention.

Claims

1. A smart anti-interference method based on multi-physics coupling and polarization diversity, characterized in that, Includes the following steps: Shielding attenuation factor calculated based on eddy current shielding effect of metal support materials Total attenuation factor calculated using Mie scattering theory based on dust concentration Construct a multi-physics coupled path loss model ,in, Let d0 represent the free-space reference loss, d0 = 1 m, λ = 0.125 m @ 2.4 GHz, and n represent the path loss exponent. Indicates the shadow fading component. To shield the attenuation factor, This is the total attenuation factor; The design incorporates a hierarchical intelligent anti-interference architecture. The first layer employs direct sequence spread spectrum technology for narrowband interference mitigation; the second layer uses an adaptive threshold denoising method based on discrete wavelet transform to suppress impulse noise; and the third layer constructs a distributed power control model based on non-cooperative game theory, iteratively optimizing the transmit power of each node to maximize the system utility function. The utility function of the i-th node is defined as follows: ,in, Let be the transmit power of node i. Its corresponding channel gain, For noise power, This indicates interference from other nodes. Energy consumption weighting coefficient; By employing polarization diversity and deep learning channel equalization strategies, firstly, a circularly polarized antenna is used as the front-end receiving device to suppress multipath effects by utilizing its polarization orthogonality of reflected and scattered waves. Then, a dynamic channel equalizer based on a CNN-LSTM architecture is constructed. Through the modeling and symbol recovery of dynamic channel state information by the CNN-LSTM network, dynamic channel equalization is achieved.

2. The intelligent anti-interference method for multi-physics coupling and polarization diversity according to claim 1, characterized in that, The processing gain of the direct sequence spread spectrum technique is no less than 20 dB, the code length of the Gold code is 63, and the maximum cross-correlation value is no greater than 7.

3. The intelligent anti-interference method for multi-physics coupling and polarization diversity according to claim 1, characterized in that, The adaptive threshold denoising method employs a hybrid soft and hard thresholding strategy, with the threshold dynamically determined based on the median absolute deviation.

4. The intelligent anti-interference method for multi-physics coupling and polarization diversity according to claim 1, characterized in that, The utility function of the distributed power control model considers channel gain, noise power, interference power, and energy consumption weight, and iteratively solves for the Nash equilibrium point using a gradient projection algorithm.

5. The intelligent anti-interference method for multi-physics coupling and polarization diversity according to claim 1, characterized in that, The CNN-LSTM dynamic equalizer includes a one-dimensional convolutional layer, a max pooling layer, an LSTM layer, and a fully connected layer.

6. The intelligent anti-interference method for multi-physics coupling and polarization diversity according to claim 1, characterized in that, The input to the CNN-LSTM dynamic equalizer is a time-domain signal sequence of length 256, and the output is the equalized symbol estimate.

7. A wireless communication system for implementing the method as described in any one of claims 1 to 6, characterized in that, include: The transmitting module is used to generate and transmit DSSS-modulated signals; The receiving module includes a circularly polarized antenna and a CNN-LSTM dynamic equalizer; The processing module is used to perform wavelet denoising, power control, and channel equalization operations.

8. The wireless communication system according to claim 7, characterized in that, The system is suitable for deep well safety monitoring, emergency communication, or intelligent mine wireless communication scenarios.

Images