Power line communication control method and system for subway DC power supply system

Abstract

The present invention relates to the field of data processing technology, and discloses a power line communication control method and system for a subway DC power supply system, which is used to improve the efficiency and accuracy of the power line communication control of the subway DC power supply system. The method includes: performing orthogonal frequency division multiplexing modulation on the voltage signal of the DC power supply line to obtain a high-frequency carrier signal containing operation data; performing power line noise suppression on the high-frequency carrier signal through an adaptive notch filter to obtain a low-noise data signal; performing time synchronization and alignment on the low-noise data signal to obtain a synchronous multi-source data stream; performing feature extraction and dimension reduction on the synchronous multi-source data stream through a distributed compressed sensing algorithm to obtain a state feature vector; performing reasoning analysis on the state feature vector through a fuzzy neural network to obtain a control decision; broadcasting and executing the control decision on the DC power supply line through an improved time division multiple access protocol to obtain a real-time control instruction set.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a power line communication control method and system for a subway direct current power supply system. Background Art

[0002] The DC power supply system of subway is an important part of urban rail transit, providing traction power for trains. Traditional DC power supply systems of subways usually adopt fixed parameter control strategies, which are difficult to adapt to complex and changing operating environments. With the development of smart grid technology, power line communication (PLC) technology has gradually been applied to the monitoring and management of power supply systems. PLC technology uses existing power lines as a communication medium, without the need for additional wiring, and can achieve the integration of power supply and communication.

[0003] However, the existing PLC technology in the subway DC power supply system still has some shortcomings. First, the power line noise interference is serious, affecting the communication quality and reliability. Second, the synchronization and alignment problems of multi-source data have not been effectively solved, resulting in inaccurate data analysis results. Third, the traditional feature extraction and dimensionality reduction methods have high computational complexity and are difficult to meet the needs of real-time control. Finally, the existing control decision-making methods lack adaptive capabilities and are difficult to cope with dynamic changes in system status. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a power line communication control method and system for a subway DC power supply system, which are used to improve the efficiency and accuracy of the power line communication control of the subway DC power supply system.

[0005] The present invention provides a power line communication control method for a subway DC power supply system, comprising: performing orthogonal frequency division multiplexing modulation on a voltage signal of a DC power supply line to obtain a high-frequency carrier signal containing operation data;

[0006] Performing power line noise suppression on the high frequency carrier signal by using an adaptive notch filter to obtain a low noise data signal;

[0007] Performing time synchronization and alignment on the low-noise data signal to obtain a synchronous multi-source data stream;

[0008] Performing feature extraction and dimension reduction on the synchronous multi-source data stream by using a distributed compressed sensing algorithm to obtain a state feature vector;

[0009] Performing reasoning and analysis on the state feature vector through a fuzzy neural network to obtain a control decision;

[0010] The control decision is broadcast and executed on the DC power supply line through an improved time division multiple access protocol to obtain a real-time control instruction set.

[0011] The present invention also provides a power line communication control system for a subway DC power supply system, comprising:

[0012] A modulation module, used for performing orthogonal frequency division multiplexing modulation on the voltage signal of the DC power supply line to obtain a high-frequency carrier signal containing operation data;

[0013] A suppression module, used for suppressing power line noise of the high-frequency carrier signal through an adaptive notch filter to obtain a low-noise data signal;

[0014] An alignment module, used for performing time synchronization and alignment on the low-noise data signal to obtain a synchronous multi-source data stream;

[0015] A dimension reduction module, used for performing feature extraction and dimension reduction on the synchronous multi-source data stream by using a distributed compressed sensing algorithm to obtain a state feature vector;

[0016] An analysis module, used for performing reasoning analysis on the state feature vector through a fuzzy neural network to obtain a control decision;

[0017] The execution module is used to broadcast and execute the control decision on the DC power supply line through an improved time division multiple access protocol to obtain a real-time control instruction set.

[0018] In the technical solution provided by the present invention, by orthogonal frequency division multiplexing modulation of the voltage signal of the DC power supply line, efficient data transmission is achieved, the existing power supply line is effectively used as a communication medium, no additional communication line is required, and the installation and maintenance costs of the system are greatly reduced. Orthogonal frequency division multiplexing technology also improves the spectrum utilization rate and enhances the anti-interference ability of the system, so that the communication can remain stable and reliable in a complex power line environment. Secondly, an adaptive notch filter is used to suppress power line noise on the high-frequency carrier signal, which significantly improves the signal quality, effectively overcomes the common noise interference problem in power line communication, and lays a good foundation for subsequent data processing. Time synchronization and alignment of low-noise data signals solves the time inconsistency of multi-source data, ensures that data from different devices can accurately correspond, and improves the overall coordination of the system. The distributed compressed sensing algorithm is used to extract features and reduce the dimension of the synchronous multi-source data stream, which greatly reduces the amount of data, improves data processing efficiency, and retains key information, providing high-quality input for real-time control decisions. The application of fuzzy neural network realizes the intelligent reasoning analysis of state feature vectors, combining the interpretability of fuzzy logic and the learning ability of neural network, making the control decision more accurate and adaptable, and able to cope with complex and changeable power supply system conditions. Finally, the improved time division multiple access protocol is used to broadcast and execute control decisions, achieving efficient resource allocation and priority management, and ensuring the timely transmission and execution of key control instructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 The present invention is a flowchart of a power line communication control method for a subway DC power supply system according to an embodiment of the present invention.

[0021] Figure 2 The figure is a schematic diagram of a power line communication control system of a subway DC power supply system in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0024] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0025] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , Figure 1 FIG. 1 is a flow chart of a power line communication control method for a subway DC power supply system according to an embodiment of the present invention. Figure 1 As shown, the following steps are included:

[0026] S101, performing orthogonal frequency division multiplexing modulation on a voltage signal of a DC power supply line to obtain a high-frequency carrier signal containing operation data;

[0027] S102, suppressing power line noise of the high-frequency carrier signal by using an adaptive notch filter to obtain a low-noise data signal;

[0028] S103, performing time synchronization and alignment on the low-noise data signal to obtain a synchronous multi-source data stream;

[0029] S104, extracting features and reducing dimensions of the synchronous multi-source data streams by using a distributed compressed sensing algorithm to obtain a state feature vector;

[0030] S105, performing reasoning analysis on the state feature vector through a fuzzy neural network to obtain a control decision;

[0031] S106. Broadcasting and executing the control decision on the DC power supply line through an improved time division multiple access protocol to obtain a real-time control instruction set.

[0032] It should be noted that the voltage signal of the DC power supply line is first modulated by orthogonal frequency division multiplexing. It starts with sampling and preprocessing the voltage signal, and then converts the signal to the frequency domain through Fourier transform. In the frequency domain, the signal is assigned to different subcarriers, each of which is orthogonally modulated. Subsequently, the signal is converted back to the time domain through inverse Fourier transform, and a cyclic prefix is ​​added to prevent inter-symbol interference. Finally, after digital-to-analog conversion, power amplification and bandpass filtering, the high-frequency carrier signal is injected into the DC power supply line. Next, the adaptive notch filter suppresses the power line noise of the high-frequency carrier signal. This step first analyzes the spectral characteristics of the signal and identifies the noise frequency. Then, the initial notch parameters are set according to the identified noise frequency, and these parameters are continuously adjusted through an adaptive algorithm. The notch filter filters the signal multiple times, and the residual noise is evaluated after each filtering, and the notch parameters are optimized accordingly. Finally, the quality of the filtered signal is evaluated to ensure that a low-noise data signal is obtained.

[0033] Time synchronization and alignment of low-noise data signals is the next key step. This process first extracts the signal’s timestamp, estimates the clock bias, and corrects the timestamp. The corrected data is then divided into time windows and interpolated within each window to ensure that the data points are evenly spaced. After that, the data is phase aligned and delay compensated, and finally the information from multiple data sources is fused and outliers are removed to obtain a synchronized multi-source data stream. The synchronized multi-source data stream is feature extracted and dimension reduced using a distributed compressed sensing algorithm. This process first sparsely represents the data and then obtains a compressed measurement matrix through random projection. The signal is reconstructed using an orthogonal matching pursuit algorithm and multi-scale coefficients are obtained through wavelet transform. These coefficients are thresholded and inversely transformed to obtain a de-noised signal. Subsequently, the feature vector reflecting the system state is finally obtained through methods such as principal component analysis and nonlinear feature extraction.

[0034] Fuzzy neural networks are used to perform reasoning analysis on state feature vectors. First, the feature vectors are converted into fuzzy input variables and their membership values ​​are calculated. Then, matching and weight calculation are performed according to the preset fuzzy rules to obtain a preliminary control decision. This decision is forward propagated through the neural network, the error is calculated and back-propagated, and the network parameters are continuously adjusted to finally obtain the optimized control decision. Finally, the control decision is broadcast and executed on the DC power supply line through the improved time division multiple access protocol. This process includes prioritization of control decisions, time slot allocation, channel coding, spread spectrum processing, carrier modulation and power control. The processed signal is injected into the power supply line through the power line coupler, demodulated and decoded at the receiving end, and finally a real-time control instruction set is obtained.

[0035] For example, in the DC power supply system of a subway line, the control center needs to adjust the output power of multiple substations to balance the load. The control center first collects the voltage signal, and after orthogonal frequency division multiplexing modulation, obtains a high-frequency carrier signal containing the current load information of each substation. This signal is processed by an adaptive notch filter to effectively remove the harmonic interference generated by power electronic equipment. After time synchronization and alignment, a multi-source data stream containing real-time data of all substations is formed. The distributed compressed sensing algorithm extracts key features from this data, such as the load rate, voltage level, and temperature of each substation. The fuzzy neural network analyzes these features and generates control decisions for adjusting the output power of each substation. Finally, this decision is broadcast on the power supply line through an improved time division multiple access protocol. After receiving the instruction, each substation adjusts its output power accordingly, thereby achieving dynamic load balance.

[0036] By executing the above steps, the voltage signal of the DC power supply line is modulated by orthogonal frequency division multiplexing, which realizes efficient data transmission, effectively utilizes the existing power supply line as a communication medium, and does not require additional communication lines, which greatly reduces the installation and maintenance costs of the system. Orthogonal frequency division multiplexing technology also improves spectrum utilization and enhances the system's anti-interference ability, so that communication can remain stable and reliable in a complex power line environment. Secondly, an adaptive notch filter is used to suppress power line noise on high-frequency carrier signals, which significantly improves signal quality, effectively overcomes the common noise interference problem in power line communication, and lays a good foundation for subsequent data processing. Time synchronization and alignment of low-noise data signals solves the time inconsistency of multi-source data, ensures that data from different devices can accurately correspond, and improves the overall coordination of the system. The distributed compressed sensing algorithm is used to extract features and reduce the dimension of the synchronous multi-source data stream, which greatly reduces the amount of data, improves data processing efficiency, and retains key information, providing high-quality input for real-time control decisions. The application of fuzzy neural network realizes the intelligent reasoning analysis of state feature vectors, combining the interpretability of fuzzy logic and the learning ability of neural network, making the control decision more accurate and adaptable, and able to cope with complex and changeable power supply system conditions. Finally, the improved time division multiple access protocol is used to broadcast and execute control decisions, achieving efficient resource allocation and priority management, and ensuring the timely transmission and execution of key control instructions.

[0037] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0038] (1) Sampling the voltage signal of the DC power supply line to obtain raw voltage data, and preprocessing the raw voltage data to obtain preprocessed voltage data;

[0039] (2) Performing Fourier transform on the preprocessed voltage data to obtain a frequency domain signal, and performing subcarrier allocation on the frequency domain signal to obtain a subcarrier signal set;

[0040] (3) performing orthogonal modulation on the subcarrier signal set to obtain an orthogonal modulated signal, and performing inverse Fourier transform on the orthogonal modulated signal to obtain a time domain composite signal;

[0041] (4) Adding a cyclic prefix to the time domain composite signal to obtain an OFDM symbol with the prefix, and performing digital-to-analog conversion on the OFDM symbol with the prefix to obtain an analog OFDM signal;

[0042] (5) Amplifying the power of the analog OFDM signal to obtain a high-power OFDM signal, and performing bandpass filtering on the high-power OFDM signal to obtain a filtered high-frequency carrier signal;

[0043] (6) The filtered high-frequency carrier signal is injected into the DC power supply line through a coupling circuit to obtain a high-frequency carrier signal containing operating data.

[0044] Specifically, the voltage signal of the DC power supply line is first sampled. The sampling process uses a high-speed analog-to-digital converter to capture the instantaneous value of the voltage signal at a sufficiently high sampling rate to obtain the original voltage data. These raw data are then preprocessed, including removing DC bias, eliminating high-frequency noise and normalization, and finally obtaining the preprocessed voltage data. Next, the preprocessed voltage data is subjected to a fast Fourier transform (FFT) to convert the time domain signal into a frequency domain signal. The FFT algorithm can efficiently calculate the discrete Fourier transform and convert the sampling points in the time domain into frequency components in the frequency domain. After obtaining the frequency domain signal, subcarrier allocation is performed. During the subcarrier allocation process, subcarriers of different frequencies are allocated to different data streams according to the channel conditions and data transmission requirements to form a subcarrier signal set.

[0045] The subcarrier signal set is then orthogonally modulated. Orthogonal modulation uses quadrature amplitude modulation (QAM) technology to map digital data to constellation points on the complex plane. Each subcarrier is modulated independently and carries part of the data information. The orthogonal modulated signal is then converted back to the time domain through an inverse fast Fourier transform (IFFT) to obtain a time domain composite signal. The IFFT process realizes the synthesis of multiple orthogonal subcarriers to form a continuous time domain signal. In order to prevent inter-symbol interference, a cyclic prefix is ​​added to the time domain composite signal. The cyclic prefix is ​​a small segment at the end of the time domain signal copied to the beginning of the signal to form an OFDM (Orthogonal Frequency Division Multiplexing) symbol with a prefix. This structure can effectively counteract the effects of multipath propagation. The prefixed OFDM symbol is then converted to a continuous analog OFDM signal through a digital-to-analog converter.

[0046] The analog OFDM signal is amplified by a power amplifier to ensure that the signal has sufficient strength to be transmitted on the power line. The amplified high-power OFDM signal is processed by a bandpass filter to filter out out-of-band noise and harmonics to obtain a filtered high-frequency carrier signal. The center frequency and bandwidth of the bandpass filter are designed according to the frequency band requirements of power line communication, usually in the megahertz range. Finally, the filtered high-frequency carrier signal is injected into the DC power supply line through a specially designed coupling circuit. The coupling circuit includes a high-pass filter and an isolation transformer, which can separate the high-frequency communication signal from the low-frequency power signal, while ensuring the electrical isolation of the communication equipment from the high-voltage power system. The injected high-frequency carrier signal coexists with the DC power signal in the power supply line to form a high-frequency carrier signal containing operating data.

[0047] For example, in a 1500V DC power supply system of a subway line, it is necessary to transmit operating data between the substation and the train. First, the DC voltage is sampled at the substation end, the sampling rate is set to 1MHz, and 1ms of data is collected to obtain 1000 sampling points. These data are preprocessed to remove 50Hz power frequency interference and high-frequency noise. The preprocessed data is converted to the frequency domain through a 1024-point FFT to obtain 512 effective frequency points. In the frequency band of 2-30MHz, 100 subcarriers are selected for data transmission. Each subcarrier uses 16QAM modulation and can carry 4 bits of data. After IFFT and adding a 200μs cyclic prefix, a complete OFDM symbol is formed with a duration of 1.2ms. The symbol is converted to analog and digital, amplified to 100mW, and passed through a bandpass filter with a center frequency of 15MHz and a bandwidth of 28MHz. Finally, the signal is injected into the DC power supply line through a coupling circuit, realizing the transmission of high-frequency carrier signals containing operating data such as train position, speed, and current consumption.

[0048] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0049] (1) Perform spectrum analysis on the high-frequency carrier signal to obtain the signal spectrum characteristics, and perform noise frequency identification on the signal spectrum characteristics to obtain a noise frequency set;

[0050] (2) Setting the notch center frequency of the noise frequency set to obtain initial notch parameters, and adaptively adjusting the initial notch parameters to obtain dynamic notch parameters;

[0051] (3) performing notch filtering on the high-frequency carrier signal according to the dynamic notch parameter to obtain a preliminary filtered signal, and performing residual noise evaluation on the preliminary filtered signal to obtain a noise evaluation result;

[0052] (4) Iteratively optimize the dynamic notch parameters according to the noise evaluation results to obtain the optimized notch parameters, and perform secondary notch filtering on the high-frequency carrier signal according to the optimized notch parameters to obtain a deep filtered signal;

[0053] (5) Calculate the signal-to-noise ratio of the deep filtering signal to obtain a signal-to-noise ratio index, and evaluate the quality of the deep filtering signal based on the signal-to-noise ratio index to obtain a filtering quality result;

[0054] (6) Post-process the deep filtered signal according to the filtering quality result to obtain a low-noise data signal.

[0055] Specifically, in the power line communication control method of the subway DC power supply system, noise suppression of the high-frequency carrier signal is a key step to ensure the communication quality. First, the high-frequency carrier signal is subjected to spectrum analysis, and the time domain signal is converted into frequency domain representation using fast Fourier transform (FFT) to obtain the signal spectrum characteristics. The spectrum analysis result can be expressed as:

[0056]

[0057] in, is the spectrum of the signal, is the time domain signal, is the frequency, is time, j is an imaginary unit. Next, the noise frequency is identified based on the signal spectrum characteristics. By setting the amplitude threshold and frequency range, the frequency points that exceed the threshold and do not belong to the valid signal frequency band are identified, and these frequency points are defined as the noise frequency set. The notch center frequency is set for the noise frequency set to obtain the initial notch parameters. Each noise frequency Corresponding to a notch filter, its transfer function can be expressed as:

[0058]

[0059] in, is the normalized angular frequency, is the sampling frequency, is the pole radius of the notch filter, which determines the bandwidth of the notch filter, and z is the initial notch parameter. The initial notch parameter is adaptively adjusted to obtain the dynamic notch parameter. The minimum mean square error (LMS) algorithm is used to continuously update the notch filter coefficient to minimize the output error. The update formula is:

[0060]

[0061] in, It is The filter coefficients are The value of the moment, It is The filter coefficients are The value of the moment, is the step size factor, controlling the speed of the update, yes The estimated error of time, The input signal is The value of the moment, is the index of the filter coefficient, is the current time index. Perform notch filtering on the high-frequency carrier signal according to the dynamic notch parameters to obtain a preliminary filtered signal. Perform residual noise evaluation on the preliminary filtered signal, calculate the power spectrum density of the signal, identify the remaining noise components, and obtain the noise evaluation result.

[0062] According to the noise evaluation results, the dynamic notch parameters are iteratively optimized. The center frequency, bandwidth and depth of the notch filter are adjusted to further suppress the residual noise. The optimized notch parameters are used for secondary notch filtering to obtain a deep filtered signal. The signal-to-noise ratio of the deep filtered signal is calculated to obtain the signal-to-noise ratio index. The signal-to-noise ratio calculation formula is:

[0063]

[0064] in, is the signal power, is the noise power. The quality of the deep filtering signal is evaluated according to the signal-to-noise ratio index to obtain the filtering quality result. Finally, the deep filtering signal is post-processed according to the filtering quality result, including amplitude normalization, phase correction, etc., to obtain a low-noise data signal.

[0065] For example, in a subway DC power supply system, the frequency range of the high-frequency carrier signal is 1-30MHz. Through FFT analysis, it is found that there is obvious narrowband noise at 5MHz, 15MHz and 25MHz, with amplitudes of -40dB, -35dB and -45dB respectively. The noise recognition threshold is set to -50dB, and the noise frequency set N = {5MHz, 15MHz, 25MHz} is obtained. A notch filter is designed for each noise frequency, and the initial pole radius r is set to 0.95. The LMS algorithm is used for adaptive adjustment, and the step factor μ is set to 0.01. After 1000 iterations, the bandwidth of the notch filter at 5MHz is adjusted from the initial 100kHz to 50kHz, and the suppression depth is increased from 10dB to 20dB. After two notch filters, the signal-to-noise ratio is increased from the initial 5dB to 15dB. In the final low-noise data signal, the useful signals at 2MHz and 28MHz that were originally masked by noise were restored, with amplitudes of -55dB and -58dB respectively, providing a reliable basis for subsequent data demodulation.

[0066] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0067] (1) Extracting the timestamp of the low-noise data signal to obtain the original timestamp sequence, and estimating the clock deviation of the original timestamp sequence to obtain the clock deviation value;

[0068] (2) Correct the original timestamp sequence according to the clock deviation value to obtain a corrected timestamp sequence, and divide the corrected timestamp sequence into time windows to obtain a time window set;

[0069] (3) interpolating the data in the time window set to obtain equally spaced sampling data, and performing phase alignment on the equally spaced sampling data to obtain phase aligned data;

[0070] (4) performing delay estimation on the phase-aligned data to obtain a delay estimation value, and performing delay compensation on the phase-aligned data according to the delay estimation value to obtain compensated data;

[0071] (5) Perform multi-source fusion on the compensated data to obtain a fused data stream, and perform data quality assessment on the fused data stream to obtain data quality indicators;

[0072] (6) Detect and remove outliers on the fused data stream based on data quality indicators to obtain a synchronized multi-source data stream.

[0073] Specifically, in the power line communication control method of the subway DC power supply system, time synchronization and alignment of low-noise data signals are key steps to ensure the accuracy of multi-source data. First, the low-noise data signal is timestamped by identifying the time stamp in the data packet or using the system clock when the data is received to obtain the original timestamp sequence. Subsequently, the clock deviation of the original timestamp sequence is estimated, and the linear regression method is used to analyze the change trend of the timestamp sequence to calculate the deviation value between the local clock and the reference clock. According to the obtained clock deviation value, the original timestamp sequence is corrected. The correction process is achieved by subtracting the accumulated clock deviation from each timestamp to obtain a corrected timestamp sequence. Next, the corrected timestamp sequence is divided into time windows, and the continuous time series is divided into fixed-length or overlapping time periods to form a time window set. This division method is helpful for subsequent data processing and analysis.

[0074] The data in the time window set is interpolated to obtain equally spaced sampled data. Interpolation uses methods such as linear interpolation or spline interpolation to add new data points between the original data points so that the entire data sequence is evenly distributed in time. After obtaining the equally spaced sampled data, a phase alignment operation is performed. Phase alignment converts the signal to the frequency domain through Fourier transform, adjusts the phase of each frequency component, and then converts it back to the time domain to obtain phase-aligned data. The phase-aligned data is subjected to delay estimation. This step uses the cross-correlation method to calculate the time offset between different data sources. Based on the obtained delay estimate, the phase-aligned data is delayed compensated, and the data from different sources are accurately aligned in time to obtain compensated data.

[0075] Next, the compensated data is fused from multiple sources. Multi-source fusion uses algorithms such as weighted averaging or Kalman filtering to combine information from different sensors or data sources to form a more complete and accurate data stream. The fusion process takes into account the reliability and accuracy of each data source and assigns different weights to obtain a fused data stream. Subsequently, the fused data stream is evaluated for data quality, and indicators such as signal-to-noise ratio, data integrity and consistency are calculated to obtain data quality indicators. Finally, the fused data stream is detected and removed for outliers based on the data quality indicators. Outlier detection uses statistical methods or machine learning algorithms to identify data points that are significantly different from normal data patterns. These outliers are removed from the data stream by setting thresholds or using clustering algorithms, and ultimately a synchronized multi-source data stream is obtained.

[0076] For example, in the DC power supply system of a subway line, three substations transmit data to the control center at the same time. Each substation collects 100 data points per second, including voltage, current and temperature information. Through timestamp extraction, it is found that the clocks of the three substations have deviations of +0.5ms / s, -0.3ms / s and +0.1ms / s respectively. After clock deviation correction, the data is divided into time windows of 10 seconds. In each time window, the data is resampled to 200 points per second through cubic spline interpolation. During the phase alignment process, it is found that the data of the three substations have delays of 5ms, 8ms and 3ms. The three sets of data are accurately aligned through delay compensation. When multi-source fusion is performed, weights of 0.4, 0.35 and 0.25 are assigned according to the historical reliability of each substation. Data quality assessment shows that the signal-to-noise ratio of the fused data stream is improved by 3dB. The outlier detection uses the 3σ principle to identify 5 abnormal points and remove them. The resulting synchronized multi-source data stream contains 200 high-quality data points per second.

[0077] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0078] (1) Sparsely represent the synchronous multi-source data stream to obtain a sparse signal matrix, and randomly project the sparse signal matrix to obtain a compressed measurement matrix;

[0079] (2) Perform orthogonal matching pursuit reconstruction on the compressed measurement matrix to obtain a reconstructed signal, and perform wavelet transform on the reconstructed signal to obtain multi-scale coefficients;

[0080] (3) Perform threshold processing on the multi-scale coefficients to obtain the key coefficient set, and perform inverse wavelet transform on the key coefficient set to obtain the noise reduction signal;

[0081] (4) Performing principal component analysis on the denoised signal to obtain a principal component set, and sorting the principal component set by feature importance to obtain a sorted feature set;

[0082] (5) Perform nonlinear feature extraction on the sorted feature set to obtain a nonlinear feature set, and perform feature fusion on the nonlinear feature set to obtain a fused feature vector;

[0083] (6) Perform feature selection on the fused feature vector to obtain the state feature vector.

[0084] Specifically, in the power line communication control method of the subway DC power supply system, feature extraction and dimensionality reduction of the synchronous multi-source data stream are key steps to improve data processing efficiency. First, the synchronous multi-source data stream is sparsely represented, and the data is converted to the frequency domain using discrete cosine transform (DCT) to obtain the sparse signal matrix X. The sparse representation can be expressed as follows:

[0085]

[0086] in, is the DCT basis matrix, is a sparse coefficient matrix. Next, the sparse signal matrix is ​​randomly projected to obtain the compressed measurement matrix Y. The random projection process can be expressed as:

[0087]

[0088] in, is a random projection matrix, usually a Gaussian random matrix or a Bernoulli random matrix.

[0089] The compressed measurement matrix is ​​reconstructed by orthogonal matching pursuit (OMP) to obtain the reconstructed signal. The OMP algorithm reconstructs the original signal step by step by iteratively selecting the most relevant basis vectors. The reconstructed signal is then wavelet transformed to obtain multi-scale coefficients. Wavelet transform uses discrete wavelet transform (DWT) to capture the characteristics of the signal at different scales. The multi-scale coefficients are thresholded and the coefficients greater than the set threshold are retained to obtain the key coefficient set. Thresholding can effectively remove noise and unimportant details. The key coefficient set is inversely wavelet transformed to obtain a denoised signal. This step restores the time domain signal while retaining important feature information.

[0090] Next, perform principal component analysis (PCA) on the denoised signal to obtain a set of principal components. PCA projects the original data onto a new orthogonal basis by calculating the eigenvalues ​​and eigenvectors of the covariance matrix. The principal components are sorted according to the size of the explained variance to obtain a sorted feature set. Nonlinear feature extraction is performed on the sorted feature set using the kernel principal component analysis (KPCA) method. KPCA maps the data to a high-dimensional feature space through a kernel function, and then performs linear PCA in this space to capture the nonlinear characteristics of the data. This step results in a nonlinear feature set.

[0091] Finally, the nonlinear feature set is fused using an autoencoder. The autoencoder compresses the input features into a low-dimensional representation through nonlinear transformation, and then reconstructs the original features. The fused feature vector is subjected to feature selection and the recursive feature elimination (RFE) algorithm is used to finally obtain the state feature vector.

[0092] For example, in a subway DC power supply system, the synchronous multi-source data stream collected from three substations contains voltage, current and temperature data. Each variable is sampled 100 times per second for 10 minutes, forming a 180,000×9 data matrix. First, the data matrix is ​​transformed by DCT to obtain a sparse signal matrix, and it is found that 95% of the energy is concentrated in the first 20% of the coefficients. Then, a 100×180,000 random Gaussian matrix is ​​used for random projection to compress the data into a 100×9 compressed measurement matrix.

[0093] The signal was reconstructed by the OMP algorithm, and the sparsity was set to 3600 (i.e., 2% of the original data volume). The reconstructed signal was obtained after 50 iterations. The reconstructed signal was subjected to a 5-layer db4 wavelet transform to obtain multi-scale coefficients. The threshold was set to the 95% quantile of the absolute value of the wavelet coefficients, and soft threshold processing was performed to obtain the key coefficient set. After the inverse wavelet transform, the signal-to-noise ratio was increased from the original 15dB to 22dB. PCA was performed on the denoised signal, and the principal components with a cumulative explained variance rate of 99% were selected, and the 9 original variables were reduced to 5 principal components. KPCA with the RBF kernel function was used to extract nonlinear features, and the kernel parameter σ was set to 0.1, and 10 nonlinear features were extracted. A 3-layer stacked autoencoder was used for feature fusion, with the number of hidden layer nodes being 8, 4, and 8, respectively, to obtain a 4-dimensional fusion feature vector. Finally, the three most important features were selected by the RFE algorithm to form the final state feature vector. This process compresses the original 180,000×9 data matrix into a 3-dimensional state feature vector, greatly reducing the computational complexity of subsequent processing while retaining the most critical information in the data.

[0094] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0095] (1) Fuzzify the state feature vector to obtain fuzzy input variables, and calculate the membership function of the fuzzy input variables to obtain a membership value set;

[0096] (2) Perform fuzzy rule matching on the membership value set to obtain the activated rule set, and perform weight calculation on the activated rule set to obtain the rule weight vector;

[0097] (3) Perform fuzzy reasoning on the rule weight vector to obtain a fuzzy output set, and then defuzzify the fuzzy output set to obtain a preliminary control decision;

[0098] (4) Perform fuzzy neural network forward propagation on the preliminary control decision to obtain the output layer activation value, and perform error calculation on the output layer activation value to obtain the error signal;

[0099] (5) Back-propagating the error signal to obtain the weight adjustment amount, and updating the parameters of the fuzzy neural network according to the weight adjustment amount to obtain the optimized neural network;

[0100] (6) Fine-tune the preliminary control decision through the optimized neural network to obtain the control decision.

[0101] Specifically, each input variable is mapped to a predefined fuzzy set, such as "low", "medium", and "high". Subsequently, the membership function is calculated for the fuzzy input variables to obtain a set of membership values. The membership function can be a triangular, trapezoidal, or Gaussian function, etc., which is used to quantify the degree to which each input variable belongs to each fuzzy set. Next, fuzzy rule matching is performed on the membership value set. Fuzzy rules are expert knowledge expressed in the form of "if-then", such as "if the voltage is low and the current is high, then increase the power supply". By comparing the membership value of the input variable with the antecedent part of the rule, it is determined which rules are activated, thereby obtaining an activated rule set. The activated rule set is weighted to obtain a rule weight vector. The weight calculation is usually based on the membership value of the input variable, reflecting the importance of each rule in the current situation.

[0102] Fuzzy inference operations are performed on the rule weight vector to obtain a fuzzy output set. Common fuzzy inference methods include Mamdani reasoning and Sugeno reasoning. In this process, the fuzzy set of each output variable is calculated based on the activation rules and weights. Subsequently, the fuzzy output set is defuzzified to convert the fuzzy output into a clear numerical value to obtain a preliminary control decision. Defuzzification methods include the centroid method, the maximum average method, etc. After obtaining the preliminary control decision, it is input into the fuzzy neural network for forward propagation. The fuzzy neural network combines the interpretability of fuzzy logic with the learning ability of the neural network. During the forward propagation process, the input signal passes through the fuzzification layer, the rule layer, the normalization layer, and the defuzzification layer, and finally the output layer activation value is obtained. The output layer activation value is compared with the expected output to calculate the error signal.

[0103] Back propagation is performed based on the error signal to calculate the gradient of each layer of the network and obtain the weight adjustment. This process uses gradient descent or its variant algorithm, such as the Adam optimizer. The parameters of the fuzzy neural network are updated according to the weight adjustment, including the membership function parameters, rule weights, etc., to obtain the optimized neural network. Finally, the preliminary control decision is fine-tuned through the optimized neural network to obtain the final control decision. This fine-tuning process takes into account the new knowledge learned by the network and can refine and optimize the preliminary decision.

[0104] For example, in a subway DC power supply system, the control center needs to adjust the power supply strategy based on the voltage, current and temperature data of the substation. The state feature vector contains these three variables. After fuzzification, each variable is divided into three fuzzy sets: "low", "medium" and "high". Assuming that the current voltage is 1450V, the corresponding membership value is {low: 0.2, medium: 0.7, high: 0.1}. Similarly, the membership values ​​of current and temperature are calculated. The fuzzy rule base contains 27 rules (3×3×3), such as "if the voltage is low and the current is high and the temperature is medium, increase the power supply." Through rule matching, it is found that 5 rules are activated, and the calculated rule weight vector is [0.3, 0.2, 0.25, 0.15, 0.1]. After fuzzy reasoning and defuzzification, a preliminary control decision is obtained: increase the power supply by 5%. This preliminary decision is input into a 5-layer fuzzy neural network, which contains 10 fuzzy rule nodes. After forward propagation, the output layer activation value is 0.06, indicating that it is recommended to increase the power supply by 6%. Assuming the expected output is 0.07, the calculated error signal is 0.01. Through back propagation, the network parameters are adjusted, such as moving the center point of a membership function from 1450V to 1455V.

[0105] After 100 iterations of optimization, the average error of the fuzzy neural network was reduced from 0.015 to 0.005. Finally, the optimized network fine-tuned the preliminary control decision and obtained the final control decision: to increase the power supply by 6.2%. This decision not only took into account the preset fuzzy rules, but also learned the potential patterns in the historical data through the neural network, achieving more accurate power supply control.

[0106] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0107] (1) Prioritizing the control decisions to obtain priority-marked control decisions, and grouping and packaging the priority-marked control decisions to obtain control data packets;

[0108] (2) Allocate time slots for control data packets to obtain a time slot allocation table, and schedule the control data packets according to the time slot allocation table to obtain a scheduling sequence;

[0109] (3) Channel coding the scheduling sequence to obtain a coded data stream, and performing spread spectrum processing on the coded data stream to obtain a spread spectrum signal;

[0110] (4) Carrier modulating the spread spectrum signal to obtain a modulated signal, and power controlling the modulated signal to obtain a power-controlled signal;

[0111] (5) performing power line coupling injection on the power-controlled signal to obtain an injection signal, and performing multi-path reception on the injection signal to obtain a received signal set;

[0112] (6) Demodulate and decode the received signal set to obtain a real-time control instruction set.

[0113] It should be noted that according to the urgency and importance of control decisions, they are divided into three priorities: high, medium and low, to obtain priority-marked control decisions. For example, emergency power outage instructions are marked as high priority, load adjustment instructions are marked as medium priority, and daily status reports are marked as low priority. Subsequently, decisions of the same priority are grouped and packaged to form control data packets, each of which contains priority information, target device ID and specific control instructions.

[0114] The key to achieving orderly transmission is to allocate time slots for control data packets. The improved time division multiple access protocol is used to divide the communication cycle into multiple time slots. According to the priority and data volume, the corresponding time slots are allocated to different control data packets to generate a time slot allocation table. High-priority data packets obtain more or earlier time slots to ensure their priority transmission. The control data packets are scheduled according to the time slot allocation table to form a scheduling sequence, which determines the order in which the data packets are sent. Channel coding is performed on the scheduling sequence to enhance the reliability of communication. Forward error correction coding (FEC) technology, such as Reed-Solomon code or LDPC code, is used to add redundant information to the original data, so that the receiving end can detect and correct some transmission errors. The encoded data stream is spread spectrum processed to improve the anti-interference ability. Direct sequence spread spectrum (DSSS) technology is widely used, which replaces each data bit with a longer pseudo-random code sequence, significantly increasing the signal bandwidth and reducing the power spectrum density.

[0115] The spread spectrum signal is then carrier modulated to convert the baseband signal into a high-frequency signal suitable for transmission on the power line. Considering the characteristics of the power line channel, an orthogonal frequency division multiplexing (OFDM) modulation scheme is adopted to distribute the data to multiple orthogonal subcarriers, effectively resisting frequency selective fading and narrowband interference. The modulated signal is power controlled to dynamically adjust the transmission power according to the real-time status of the power line channel, while ensuring signal quality and minimizing interference to other devices. The power-controlled signal is injected into the DC power supply line through a specially designed coupling circuit. The coupling circuit includes a high-pass filter and an isolation transformer to separate the communication signal from the power signal while ensuring the safety of the equipment. The injected signal propagates on the power line and is received in multiple channels at each receiving point. Each receiving point is equipped with a matching coupling circuit and receiver to extract the high-frequency communication signal from the power line.

[0116] Finally, the received signal set is demodulated and decoded. The demodulation process corresponds to the modulation process at the transmitter, converting the high-frequency carrier signal back to the baseband signal. The decoding process uses the forward error correction coding capability to correct errors that may occur during transmission and restore the original control instructions. After this series of processing, a real-time control instruction set is finally obtained for execution by various devices.

[0117] For example, a subway DC power supply system needs to send control instructions to three substations. The control center generates five control decisions, including one emergency power adjustment instruction (high priority), two voltage adjustment instructions (medium priority) and two daily inspection instructions (low priority). After priority division and packaging, three control data packets are formed. Assume that the communication cycle is divided into 10 time slots, high priority data packets get 3 time slots (1-3), medium priority gets 4 time slots (4-7), and low priority gets 3 time slots (8-10). RS (255, 223) coding is used to channel encode the data, and 32 bytes of check information are added for every 223 bytes of data. The encoded data stream is spread using a Gold code with a length of 31, expanding the original 2Mbps data stream to 62Mbps. The 64-QAM-OFDM modulation scheme is adopted, using 48 subcarriers to modulate the 62Mbps data into the 5-30MHz frequency band. Power control dynamically adjusts the transmit power between 100mW and 1W according to the real-time channel conditions. The signal is injected into the 1500V DC power supply line through a coupling circuit with 20dB coupling attenuation. At the receiving end of the three substations, the signal is bandpass filtered and amplified, and the signal-to-noise ratio is about 15dB. During the demodulation process, the channel state is estimated by the pilot and the frequency selective attenuation is compensated. The RS decoder can correct up to 16 bytes of errors in every 255 bytes. In the end, the three substations received their respective control instructions. The emergency power adjustment instruction was successfully received and executed within 50ms after it was issued, the voltage adjustment instruction was completed within 100ms, and the daily inspection instruction was delivered within 200ms.

[0118] The embodiment of the present invention also provides a power line communication control system for a subway DC power supply system, such as Figure 2 As shown, the power line communication control system of the subway DC power supply system specifically includes:

[0119] The modulation module 201 is used to perform orthogonal frequency division multiplexing modulation on the voltage signal of the DC power supply line to obtain a high-frequency carrier signal containing operation data;

[0120] A suppression module 202, configured to suppress power line noise on the high frequency carrier signal through an adaptive notch filter to obtain a low-noise data signal;

[0121] An alignment module 203, used for performing time synchronization and alignment on the low-noise data signal to obtain a synchronized multi-source data stream;

[0122] A dimension reduction module 204 is used to extract features and reduce dimensions of the synchronous multi-source data stream by using a distributed compressed sensing algorithm to obtain a state feature vector;

[0123] An analysis module 205 is used to perform reasoning analysis on the state feature vector through a fuzzy neural network to obtain a control decision;

[0124] The execution module 206 is used to broadcast and execute the control decision on the DC power supply line through an improved time division multiple access protocol to obtain a real-time control instruction set.

[0125] Through the coordinated work of the above modules, the voltage signal of the DC power supply line is modulated by orthogonal frequency division multiplexing, which realizes efficient data transmission, effectively utilizes the existing power supply line as a communication medium, and does not require additional communication lines, greatly reducing the installation and maintenance costs of the system. Orthogonal frequency division multiplexing technology also improves spectrum utilization and enhances the system's anti-interference ability, so that communication can remain stable and reliable in a complex power line environment. Secondly, an adaptive notch filter is used to suppress power line noise on high-frequency carrier signals, which significantly improves signal quality, effectively overcomes the common noise interference problem in power line communication, and lays a good foundation for subsequent data processing. Time synchronization and alignment of low-noise data signals solves the time inconsistency of multi-source data, ensures that data from different devices can accurately correspond, and improves the overall coordination of the system. The distributed compressed sensing algorithm is used to extract features and reduce the dimension of the synchronous multi-source data stream, which greatly reduces the amount of data, improves data processing efficiency, and retains key information, providing high-quality input for real-time control decisions. The application of fuzzy neural network realizes the intelligent reasoning analysis of state feature vectors, combining the interpretability of fuzzy logic and the learning ability of neural network, making the control decision more accurate and adaptable, and able to cope with complex and changeable power supply system conditions. Finally, the improved time division multiple access protocol is used to broadcast and execute control decisions, achieving efficient resource allocation and priority management, and ensuring the timely transmission and execution of key control instructions.

[0126] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the embodiments, a person skilled in the art should understand that the specific implementation modes of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A power line communication control method for a subway DC power supply system, characterized in that: include: Orthogonal frequency division multiplexing modulation is performed on the voltage signal of the DC power supply line to obtain a high-frequency carrier signal containing operation data; Performing power line noise suppression on the high frequency carrier signal by using an adaptive notch filter to obtain a low-noise data signal; Performing time synchronization and alignment on the low-noise data signal to obtain a synchronous multi-source data stream; Performing feature extraction and dimension reduction on the synchronous multi-source data stream by using a distributed compressed sensing algorithm to obtain a state feature vector; Performing reasoning and analysis on the state feature vector through a fuzzy neural network to obtain a control decision; The control decision is broadcast and executed on the DC power supply line through an improved time division multiple access protocol to obtain a real-time control instruction set. The control decision is broadcast and executed on the DC power supply line through an improved time division multiple access protocol to obtain a real-time control instruction set. The steps of broadcasting and executing the control decision on the DC power supply line through an improved time division multiple access protocol to obtain a real-time control instruction set include: prioritizing the control decision to obtain a priority-tagged control decision, and grouping and packaging the priority-tagged control decision to obtain a control data packet; allocating time slots for the control data packet to obtain a time slot allocation table, and scheduling the control data packet according to the time slot allocation table to obtain a scheduling sequence; channel coding the scheduling sequence to obtain a coded data stream, and performing spread spectrum processing on the coded data stream to obtain a spread spectrum signal; carrier modulating the spread spectrum signal to obtain a modulated signal, and power controlling the modulated signal to obtain a power-controlled signal; power line coupling injection of the power-controlled signal to obtain an injection signal, and multi-channel receiving of the injection signal to obtain a received signal set; demodulating and decoding the received signal set to obtain a real-time control instruction set.

2. The power line communication control method of the subway DC power supply system according to claim 1, characterized in that: The step of performing orthogonal frequency division multiplexing modulation on the voltage signal of the DC power supply line to obtain a high-frequency carrier signal containing operation data includes: Sampling a voltage signal of a DC power supply line to obtain original voltage data, and preprocessing the original voltage data to obtain preprocessed voltage data; Performing Fourier transform on the preprocessed voltage data to obtain a frequency domain signal, and performing subcarrier allocation on the frequency domain signal to obtain a subcarrier signal set; Performing orthogonal modulation on the subcarrier signal set to obtain an orthogonal modulation signal, and performing inverse Fourier transform on the orthogonal modulation signal to obtain a time domain composite signal; Adding a cyclic prefix to the time domain composite signal to obtain an OFDM symbol with a prefix, and performing digital-to-analog conversion on the OFDM symbol with the prefix to obtain an analog OFDM signal; Amplifying the power of the analog OFDM signal to obtain a high-power OFDM signal, and performing bandpass filtering on the high-power OFDM signal to obtain a filtered high-frequency carrier signal; The filtered high-frequency carrier signal is injected into a DC power supply line through a coupling circuit to obtain a high-frequency carrier signal containing operation data.

3. The power line communication control method of a subway DC power supply system according to claim 1, characterized in that: The step of suppressing power line noise of the high-frequency carrier signal by an adaptive notch filter to obtain a low-noise data signal comprises: Performing spectrum analysis on the high-frequency carrier signal to obtain signal spectrum characteristics, and performing noise frequency identification on the signal spectrum characteristics to obtain a noise frequency set; Setting a notch center frequency for the noise frequency set to obtain initial notch parameters, and adaptively adjusting the initial notch parameters to obtain dynamic notch parameters; Performing notch filtering on the high-frequency carrier signal according to the dynamic notch parameter to obtain a preliminary filtered signal, and performing residual noise evaluation on the preliminary filtered signal to obtain a noise evaluation result; Iteratively optimizing the dynamic notch parameters according to the noise evaluation results to obtain optimized notch parameters, and performing secondary notch filtering on the high-frequency carrier signal according to the optimized notch parameters to obtain a deep filtered signal; Calculating a signal-to-noise ratio of the deep filtering signal to obtain a signal-to-noise ratio index, and performing a quality assessment on the deep filtering signal according to the signal-to-noise ratio index to obtain a filtering quality result; The deep filtering signal is post-processed according to the filtering quality result to obtain a low-noise data signal.

4. The power line communication control method of a subway DC power supply system according to claim 1, characterized in that: The step of performing time synchronization and alignment on the low-noise data signal to obtain a synchronized multi-source data stream comprises: Performing timestamp extraction on the low-noise data signal to obtain an original timestamp sequence, and performing clock deviation estimation on the original timestamp sequence to obtain a clock deviation value; Correcting the original timestamp sequence according to the clock deviation value to obtain a corrected timestamp sequence, and dividing the corrected timestamp sequence into time windows to obtain a time window set; Performing interpolation processing on the data in the time window set to obtain equally spaced sampling data, and performing phase alignment on the equally spaced sampling data to obtain phase aligned data; Performing delay estimation on the phase-aligned data to obtain a delay estimation value, and performing delay compensation on the phase-aligned data according to the delay estimation value to obtain compensated data; Performing multi-source fusion on the compensated data to obtain a fused data stream, and performing data quality assessment on the fused data stream to obtain a data quality indicator; Outlier detection and elimination are performed on the fused data stream according to the data quality index to obtain a synchronous multi-source data stream.

5. The power line communication control method of a subway DC power supply system according to claim 1, characterized in that: The step of extracting features and reducing dimensions of the synchronous multi-source data streams by using a distributed compressed sensing algorithm to obtain a state feature vector comprises: Performing sparse representation on the synchronous multi-source data stream to obtain a sparse signal matrix, and performing random projection on the sparse signal matrix to obtain a compressed measurement matrix; Performing orthogonal matching pursuit reconstruction on the compressed measurement matrix to obtain a reconstructed signal, and performing wavelet transform on the reconstructed signal to obtain multi-scale coefficients; Performing threshold processing on the multi-scale coefficients to obtain a key coefficient set, and performing inverse wavelet transform on the key coefficient set to obtain a noise reduction signal; Performing principal component analysis on the noise reduction signal to obtain a principal component set, and sorting the principal component set by feature importance to obtain a sorted feature set; Performing nonlinear feature extraction on the sorted feature set to obtain a nonlinear feature set, and performing feature fusion on the nonlinear feature set to obtain a fused feature vector; Feature selection is performed on the fused feature vector to obtain a state feature vector.

6. The power line communication control method of a subway DC power supply system according to claim 1, characterized in that: The control decision steps are obtained by performing reasoning analysis on the state feature vector through a fuzzy neural network, including: Performing fuzzification processing on the state feature vector to obtain a fuzzy input variable, and performing membership function calculation on the fuzzy input variable to obtain a membership value set; Performing fuzzy rule matching on the membership value set to obtain an activated rule set, and performing weight calculation on the activated rule set to obtain a rule weight vector; Performing fuzzy reasoning operation on the rule weight vector to obtain a fuzzy output set, and defuzzifying the fuzzy output set to obtain a preliminary control decision; Performing fuzzy neural network forward propagation on the preliminary control decision to obtain an output layer activation value, and performing error calculation on the output layer activation value to obtain an error signal; Back-propagating the error signal to obtain a weight adjustment amount, and updating the parameters of the fuzzy neural network according to the weight adjustment amount to obtain an optimized neural network; The preliminary control decision is fine-tuned by the optimized neural network to obtain a control decision.

7. A power line communication control system for a subway DC power supply system, used to execute the power line communication control method for a subway DC power supply system according to any one of claims 1 to 6, characterized in that: include: A modulation module, used for performing orthogonal frequency division multiplexing modulation on the voltage signal of the DC power supply line to obtain a high-frequency carrier signal containing operation data; A suppression module, used for suppressing power line noise of the high-frequency carrier signal through an adaptive notch filter to obtain a low-noise data signal; An alignment module, used for performing time synchronization and alignment on the low-noise data signal to obtain a synchronous multi-source data stream; A dimension reduction module, used for performing feature extraction and dimension reduction on the synchronous multi-source data stream by using a distributed compressed sensing algorithm to obtain a state feature vector; An analysis module, used for performing reasoning analysis on the state feature vector through a fuzzy neural network to obtain a control decision; The execution module is used to broadcast and execute the control decision on the DC power supply line through an improved time division multiple access protocol to obtain a real-time control instruction set. The step of broadcasting and executing the control decision on the DC power supply line through an improved time division multiple access protocol to obtain a real-time control instruction set includes: prioritizing the control decision to obtain a priority-tagged control decision, and grouping and packaging the priority-tagged control decision to obtain a control data packet; allocating time slots for the control data packet to obtain a time slot allocation table, and scheduling the control data packet according to the time slot allocation table to obtain a scheduling sequence; channel coding the scheduling sequence to obtain a coded data stream, and performing spread spectrum processing on the coded data stream to obtain a spread spectrum signal; carrier modulating the spread spectrum signal to obtain a modulated signal, and power controlling the modulated signal to obtain a power-controlled signal; performing power line coupling injection on the power-controlled signal to obtain an injection signal, and performing multi-channel reception on the injection signal to obtain a received signal set; demodulating and decoding the received signal set to obtain a real-time control instruction set.

Images