Power communication line monitoring method and system based on optical fiber multi-source data fusion

Abstract

The application relates to the technical field of optical fiber communication monitoring, and discloses a power communication line monitoring method and system based on optical fiber multi-source data fusion. The method comprises the following steps: calculating the weight of each data source through a confidence evaluation neural network according to the historical stability and the error code rate variation of each data source; calculating a communication service quality correction factor and correcting to obtain a line state fusion characteristic value according to the error code rate and the optical signal-to-noise ratio after weighting and fusing each weight and normalized data characteristics; when the weight ratio of optical signal data weight and communication performance data weight exceeds a set threshold, extracting the optical fiber disturbance parameters and the optical power sequence of adjacent monitoring positions to calculate a compensation disturbance value through optical power weighted least square fitting; and calculating a theoretical optical power variation according to the compensation disturbance value and the optical fiber loss coefficient and taking the deviation between the theoretical optical power variation and the measured value as a compensation effectiveness criterion. The application improves the accuracy and reliability of optical fiber communication line monitoring.

Description

Technical Field

[0001] This application relates to the field of optical fiber communication monitoring technology, and in particular to a method and system for monitoring power communication lines using optical fiber multi-source data fusion. Background Technology

[0002] Power communication lines, as critical infrastructure for information transmission in power systems, carry important communication services such as power dispatching, protection control, and production management. Existing fiber optic communication line monitoring methods primarily employ optical time domain reflectometry (OTDR) to monitor optical power attenuation and breakpoint locations. Simultaneously, they collect communication performance parameters such as bit error rate and optical signal-to-noise ratio through network management systems. Some solutions also incorporate distributed fiber optic sensing technology to collect physical parameters such as temperature and strain, or integrate power grid operation data and meteorological environmental data to construct a multi-source monitoring system. Multi-source data fusion methods, by comprehensively analyzing monitoring information from different sources, can more comprehensively assess the line's operating status. Current technologies typically employ fixed-weight weighted averaging or rule-based fusion strategies for data fusion, while some solutions utilize neural networks or Bayesian methods.

[0003] However, existing methods for monitoring power communication lines using multi-source fiber optic data fusion have significant shortcomings. First, fixed-weight or rule-based fusion strategies cannot dynamically adjust fusion parameters based on the real-time reliability of each data source. When fiber optic sensors experience data anomalies due to electromagnetic interference, fiber micro-bending, or localized faults, the system still includes the abnormal data in the fusion calculation with fixed weights, leading to an increased false alarm rate. Second, the data fusion process in existing methods is decoupled from the service quality of the fiber optic communication system. The fusion algorithm only focuses on the statistical characteristics of the monitoring data itself, failing to use parameters that directly reflect communication quality, such as bit error rate and optical signal-to-noise ratio, as constraints in the fusion process. This results in monitoring results that may not match the actual operating status of the communication service, leading to misjudgments such as abnormal physical parameters but normal communication service, or normal physical parameters but degraded communication quality. Summary of the Invention

[0004] This application provides a method and system for monitoring power communication lines using fiber optic multi-source data fusion. By establishing a dynamic weighted fusion model based on optical transmission performance constraints and combining it with a spatial gradient compensation mechanism based on fiber optic transmission characteristics, this method solves the problems of fixed data fusion weights, disconnect between monitoring results and communication quality, and loss of positioning accuracy caused by weight adjustments in existing technologies, thereby improving the accuracy and reliability of fiber optic communication line monitoring.

[0005] In a first aspect, this application provides a method for monitoring power communication lines using fiber optic multi-source data fusion, the method comprising:

[0006] Step S1: Collect backscattered light signals from power communication optical cables, bit error rate and optical signal-to-noise ratio of optical transmission lines, power grid load data and meteorological environment data, perform time synchronization processing on each data source, and generate a multi-source monitoring data matrix.

[0007] Step S2: Based on the historical stability of each data source in the multi-source monitoring data matrix and the change in bit error rate of the optical transmission line, calculate the weights of optical fiber signal data, communication performance data, power grid data, and meteorological data using a confidence evaluation neural network.

[0008] Step S3: The weights of the optical fiber signal data, communication performance data, power grid data, and meteorological data are weighted and fused with the corresponding normalized data features. The communication service quality correction factor is calculated based on the bit error rate and optical signal-to-noise ratio of the optical transmission line. The weighted fusion result is then corrected to obtain the line state fusion feature value.

[0009] Step S4: When the ratio of the weight of the fiber optic signal data to the weight of the communication performance data at the monitoring location exceeds a set threshold, extract the fiber optic disturbance parameter sequence and optical power sequence of the adjacent monitoring locations before and after, calculate the compensation disturbance value of the monitoring location by least square fitting with optical power weighting, calculate the theoretical optical power change based on the compensation disturbance value and the fiber temperature loss coefficient and strain loss coefficient, and use the deviation between the theoretical optical power change and the measured optical power change as the criterion for compensation effectiveness.

[0010] Secondly, this application provides a power communication line monitoring system based on fiber optic multi-source data fusion, the power communication line monitoring system based on fiber optic multi-source data fusion comprising:

[0011] The synchronization module is used to collect backscattered light signals from power communication optical cables, bit error rate and optical signal-to-noise ratio of optical transmission lines, power grid load data and meteorological environmental data, and to perform time synchronization processing on each data source to generate a multi-source monitoring data matrix.

[0012] The calculation module is used to calculate the weights of optical fiber signal data, communication performance data, power grid data, and meteorological data by using a confidence evaluation neural network based on the historical stability of each data source in the multi-source monitoring data matrix and the change in bit error rate of the optical transmission line.

[0013] The weighting module is used to perform weighted fusion of the optical fiber signal data weight, communication performance data weight, power grid data weight, and meteorological data weight with the corresponding normalized data features, calculate the communication service quality correction factor based on the bit error rate and optical signal-to-noise ratio of the optical transmission line, and correct the weighted fusion result to obtain the line state fusion feature value.

[0014] The monitoring module is used to extract the fiber optic disturbance parameter sequence and optical power sequence of adjacent monitoring locations when the ratio of the weight of the fiber optic signal data to the weight of the communication performance data at the monitoring location exceeds a set threshold. It then calculates the compensation disturbance value of the monitoring location by least squares fitting with optical power weighting, calculates the theoretical optical power change based on the compensation disturbance value and the fiber temperature loss coefficient and strain loss coefficient, and uses the deviation between the theoretical optical power change and the measured optical power change as the criterion for compensation effectiveness.

[0015] The technical solution provided in this application generates a multi-source monitoring data matrix by collecting backscattered light signals from power communication optical cables, bit error rate and optical signal-to-noise ratio of optical transmission lines, power grid load data, and meteorological environmental data, and performing time synchronization processing. This constructs a complete monitoring data system covering optical fiber physical layer disturbances, communication layer performance, service layer correlations, and environmental layer parameters, overcoming the problems of information partiality and timing misalignment caused by existing technologies relying on only a single data source or inconsistent time bases for multi-source data. Based on the historical stability of each data source in the multi-source monitoring data matrix and the change in bit error rate of the optical transmission line, a confidence assessment neural network calculates dynamic weights, realizing adaptive adjustment of data source weights according to their real-time reliability. When optical fiber sensing data is abnormal due to interference, its weight is automatically reduced while the weights of communication performance data and other data sources are increased, avoiding the problem of abnormal data sources continuously polluting the fusion results in fixed-weight fusion methods. At the same time, the confidence assessment neural network uses the optical transmission performance degradation degree as a key input for weight calculation, making the data source confidence assessment not only based on statistical characteristics but also deeply correlated with communication quality, ensuring the consistency between monitoring logic and optical fiber communication service status. By weighting and fusing dynamic weights with normalized data features, and calculating a communication service quality correction factor based on the bit error rate and optical signal-to-noise ratio of the optical transmission line, the fusion result is corrected to obtain the fused characteristic value of the line status. This design treats communication service quality as a physical constraint of the fusion process rather than a simple data input. When the bit error rate increases or the optical signal-to-noise ratio decreases, the correction factor automatically adjusts the amplitude of the fused characteristic value, so that even if the basic anomaly calculated by multi-source data fusion is low, the system can still respond to communication quality degradation in a timely manner. This solves the problem of missed reports caused by the disconnect between monitoring data evaluation and the actual state of communication services in existing technologies, and truly realizes a monitoring mechanism with the core objective of ensuring the transmission quality of optical fiber communication systems. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of an embodiment of the power communication line monitoring method based on fiber optic multi-source data fusion in this application.

[0018] Figure 2 This is a schematic diagram illustrating the dynamic change of the weights of each data source with the bit error rate of the optical transmission line in the embodiments of this application;

[0019] Figure 3 This is a schematic diagram illustrating the comparison and verification of the theoretical optical power change and the measured optical power change in the embodiments of this application. Detailed Implementation

[0020] This application provides a method and system for monitoring power communication lines using fiber optic multi-source data fusion. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0021] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the power communication line monitoring method based on fiber optic multi-source data fusion in this application includes:

[0022] Step S1: Collect backscattered light signals from power communication optical cables, bit error rate and optical signal-to-noise ratio of optical transmission lines, power grid load data and meteorological environment data, perform time synchronization processing on each data source, and generate a multi-source monitoring data matrix.

[0023] The backscattered light signal acquisition of the power communication optical cable is achieved through a phase-sensitive optical time-domain reflectometer. This device injects a laser pulse with a pulse width of 100 nanoseconds into the optical cable. When the laser propagates in the optical fiber, it encounters fiber molecules and causes Rayleigh backscattering. The scattered light carries the phase information of the local position of the fiber, and the phase change reflects the physical disturbances such as temperature, strain, and vibration experienced by the fiber. The bit error rate and optical signal-to-noise ratio of the optical transmission line are acquired through the monitoring port of the optical transmission network management system. The bit error rate is the ratio of the number of erroneous bits during transmission to the total number of transmitted bits, and the optical signal-to-noise ratio is the ratio of optical signal power to noise power. Both directly reflect the communication quality. Grid load data, including the load current and bus voltage of the substation, are acquired through the communication interface of the power dispatching system. The load current reflects the power demand, and the voltage reflects the power supply quality. These data are related to the operating status of the communication line because grid faults may affect the power supply to the communication equipment. Meteorological environmental data, including temperature, humidity, and wind speed, is acquired through the data interface of meteorological monitoring stations. Changes in ambient temperature cause thermal expansion and contraction of optical cables, strong winds exert mechanical stress on overhead optical cables, and humidity affects the aging rate of the optical cable sheath. Time synchronization processing employs a network time protocol. This protocol calibrates the timestamps of various data sources to a unified time base through round-trip delay measurement and clock offset calculation between the client and the time server, controlling time accuracy to the millisecond level. After synchronization, data from different sampling rates are resampled to the same period using an interpolation algorithm. The interpolation algorithm calculates the estimated value of the intermediate time based on the time distance weighted by the data values ​​at adjacent times. The multi-source monitoring data matrix organizes the synchronized data into a two-dimensional matrix structure according to time and spatial location. Rows in the matrix correspond to different monitoring locations, and columns correspond to different data types and times. Each element in the matrix stores the monitoring value of a specific location, specific time, and specific type.

[0024] Step S2: Based on the historical stability of each data source in the multi-source monitoring data matrix and the change in bit error rate of the optical transmission line, calculate the weights of optical fiber signal data, communication performance data, power grid data, and meteorological data through a confidence evaluation neural network.

[0025] Specifically, historical stability is calculated by extracting historical samples from each data source within a past time window (300 seconds) from the multi-source monitoring data matrix. Statistical variance is calculated for these historical samples; variance measures the degree of data fluctuation, with smaller variance indicating more stable and reliable data. The variance is normalized, and its reciprocal is taken to obtain a stability index. A larger stability index value indicates a more reliable data source. The change in bit error rate (BER) of the optical transmission line is obtained by calculating the difference between the current BER and the historical normal BER. An increase in BER indicates deterioration in optical transmission quality. The change in BER serves as an important basis for judging the reliability of the data source. When the BER increases significantly, even if the statistical characteristics of the fiber optic disturbance data are normal, its weight should be reduced because deterioration in optical transmission quality indicates a problem with the fiber itself or the transmission path. The confidence assessment neural network consists of three layers: an input layer, a hidden layer, and an output layer. The input layer receives feature vectors from each data source, the hidden layer contains 128 neurons and uses the ReLU activation function for nonlinear transformation, and the output layer generates weight values ​​for the four data sources. The neural network is trained using historical monitoring data and manually labeled fault types as training samples. The network parameters are adjusted through backpropagation to ensure that the weights output by the network accurately reflect the reliability of each data source under different scenarios. The weights for fiber optic signal data correspond to the reliability of fiber phase changes; the weights for communication performance data correspond to the reliability of bit error rate and optical signal-to-noise ratio; the weights for power grid data correspond to the reliability of load current and voltage; and the weights for meteorological data correspond to the reliability of environmental parameters. These four weights are normalized using a softmax function to ensure their sum equals 1. This normalization ensures that the weight distribution is inversely proportional; an increase in the weight of one data source inevitably leads to a decrease in the weights of other data sources.

[0026] Step S3: Weight the fiber optic signal data weight, communication performance data weight, power grid data weight, and meteorological data weight with the corresponding normalized data features, calculate the communication service quality correction factor based on the bit error rate and optical signal-to-noise ratio of the optical transmission line, and correct the weighted fusion result to obtain the line state fusion feature value.

[0027] The generation of normalized data features maps raw data of different dimensions to a unified numerical range. Temperature variation is normalized by subtracting the minimum value from the current value and then dividing by the range. Strain variation and vibration frequency use the same normalization method. Bit error rate (BER) is normalized after logarithmic transformation due to its large numerical range. Optical power and optical signal-to-noise ratio (SNR) are linearly normalized according to their respective normal ranges. Grid load current is represented by the ratio of the current value to the rated value. Meteorological temperature is normalized according to a reasonable range of ambient temperature. Weighted fusion multiplies the weights of fiber optic signal data with the normalized temperature, strain, and vibration features and sums them. It also multiplies the weights of communication performance data with the normalized BER, optical power, and optical SNR features and sums them. Grid data weights are multiplied with the normalized load features and sums them. Meteorological data weights are multiplied with the normalized temperature features and sums them. The sum of these four weighted results yields the basic fusion value, which comprehensively reflects the assessment of the line status by each data source. The communication service quality correction factor is calculated based on the current state of the bit error rate (BER) and optical signal-to-noise ratio (SNR). The correction factor decreases when the BER exceeds the normal threshold and also decreases when the SNR falls below the normal threshold. The formula for calculating the correction factor is 1 minus the portion of the BER exceeding the standard, the portion of optical power deviation, and the portion of insufficient SNR. These three parts are then multiplied by their respective weighting coefficients and summed. The base fusion value is multiplied by the communication service quality correction factor to obtain the line status fusion characteristic value. This multiplication operation amplifies the final fusion characteristic value even if the base fusion value is low when communication quality deteriorates. This design ensures that the monitoring system prioritizes communication service quality, triggering timely alarms when BER or SNR anomalies occur. A higher line status fusion characteristic value indicates a higher degree of line anomaly.

[0028] Step S4: When the ratio of the weight of the fiber optic signal data to the weight of the communication performance data at the monitoring location exceeds the set threshold, extract the fiber optic disturbance parameter sequence and optical power sequence of the adjacent monitoring locations before and after. Calculate the compensation disturbance value of the monitoring location through least squares fitting with optical power weighting. Calculate the theoretical optical power change based on the compensation disturbance value, the fiber temperature loss coefficient, and the strain loss coefficient. Use the deviation between the theoretical optical power change and the measured optical power change as the criterion for compensation effectiveness.

[0029] The ratio of the weight of fiber optic signal data to the weight of communication performance data at a monitoring location reflects the reliability of the data source at that location. When the ratio exceeds a set threshold, such as 2.5, it indicates that the weight of fiber optic disturbance data is significantly lower than that of communication performance data. In this case, the fiber optic sensing function may be malfunctioning due to a local fault, and the monitoring system's over-reliance on communication performance data leads to a decrease in positioning accuracy. The fiber optic disturbance parameter sequence for adjacent monitoring locations includes the temperature change, strain change, and vibration frequency of five adjacent monitoring locations. The optical power sequence includes the corresponding optical power measurements at these locations. These sequences are extracted to determine the reliability of data from adjacent locations and to perform gradient fitting calculations. The optical power weighted least squares fitting converts the optical power deviation into a weighting factor. The larger the optical power deviation from the reference value, the worse the transmission quality of the measurement point and the lower its data reliability. The weighting factor is calculated using an exponential function, giving high weights to measurement points with normal optical power and low weights to measurement points with abnormal optical power. When constructing the objective function using the weighted least squares method, the squared residuals of each measurement point are multiplied by the corresponding weighting factor. By minimizing the weighted objective function, the polynomial fitting coefficients are solved. The resulting quadratic polynomial model can describe the variation trend of the disturbance parameter along the spatial position of the optical cable. Substituting the coordinates of the abnormal monitoring position into the fitted polynomial, the compensation disturbance value is calculated. The theoretical optical power change is calculated based on the temperature and strain changes in the compensation disturbance value. Temperature changes cause changes in the refractive index of the optical fiber, which in turn leads to changes in transmission loss. Strain changes also affect the refractive index and transmission loss. Multiplying the temperature loss coefficient by the temperature change and then by the monitoring point spacing yields the optical power loss caused by temperature. Multiplying the strain loss coefficient by the strain change and then by the monitoring point spacing yields the optical power loss caused by strain. Adding the two losses gives the theoretical optical power change. The measured optical power change is directly read from the multi-source monitoring data matrix as the difference between the measured optical power and the reference optical power at the monitoring location. If the deviation between the theoretical optical power change and the measured optical power change is less than a preset threshold, such as 1 dB, the compensation is considered effective. If the deviation is too large, it indicates that there is a defect in the optical fiber itself at the location, such as a precursor to fiber breakage. When the compensation is effective, the weight correction coefficient corresponding to the compensation disturbance value is weighted and summed with the average weight of the optical fiber signal data of the adjacent normal measurement points. The corrected weight enables the abnormal monitoring location to regain a certain optical fiber sensing and positioning capability, and the positioning accuracy is restored from the kilometer level to the hundred-meter level.

[0030] In one specific embodiment, step S1 includes:

[0031] Pulsed laser light is injected into the power communication optical cable using a phase-sensitive optical time-domain reflectometer. The backscattered light signal is received and the phase change is demodulated. The spatial resolution is calculated based on the fiber refractive index and the speed of light, and the phase change at each monitoring location is extracted.

[0032] The optical power, bit error rate and optical signal-to-noise ratio of each repeater section are collected through the network management interface of the optical transmission line; the load current and voltage of the substation are collected through the power dispatching system interface; and the temperature, humidity and wind speed along the optical cable path are collected through the meteorological data interface.

[0033] The clock protocol is used to align the timestamps of phase change, optical power, bit error rate, optical signal-to-noise ratio, load current, voltage, temperature, humidity and wind speed to a unified time base;

[0034] For load current, voltage, temperature, humidity and wind speed with low sampling rates, a cubic spline interpolation algorithm is used to resample to a sampling period consistent with the phase change, optical power, bit error rate and optical signal-to-noise ratio to generate a multi-source monitoring data matrix.

[0035] Specifically, when the pulsed laser injected by the phase-sensitive optical time-domain reflectometer propagates in the optical fiber, the light wave interacts with the fiber molecules to produce Rayleigh backscattering. The phase of the scattered light carries information about the strain, temperature, and vibration of the local location of the fiber. Phase demodulation involves receiving the returned scattered light through a photodetector, converting it into an electrical signal, and then performing coherent demodulation to extract the phase change. Spatial resolution is calculated based on the pulse width and the speed of light in the optical fiber. The speed of light in vacuum is equal to the refractive index of the fiber. The round-trip time of the pulse is multiplied by the speed of light and then divided by 2 to obtain the fiber distance. A pulse width of 100 nanoseconds corresponds to a spatial resolution of approximately 10 meters. The network management interface of the optical transmission line follows standard communication protocols to access the performance monitoring module of the optical transmission equipment, reading the optical power, bit error rate statistics, and optical signal-to-noise ratio calculation results of each repeater segment. The power dispatching system interface queries real-time data of the substation's load current and bus voltage based on industrial communication standards. The meteorological data interface connects to the meteorological monitoring station to obtain temperature, humidity, and wind speed sensor readings.

[0036] The clock protocol measures clock deviation through bidirectional message interaction between the client and the time server. The client sends a timestamp request and records the sending time; the server receives the request, records the receiving time, and returns a response. The client calculates the clock offset and network latency based on the four timestamps and adjusts its local clock accordingly. Timestamp alignment across data sources calibrates all acquisition devices to the same reference time base. A cubic spline interpolation algorithm resamples low-sampling-rate data. This algorithm constructs a cubic polynomial function between adjacent sampling points. The polynomial coefficients are solved using boundary and continuity conditions. The boundary conditions set the second derivative at the first and last points to zero, while the continuity condition requires that the function value, first derivative, and second derivative of adjacent polynomial segments be equal at the connection points. The interpolation results are calculated by substituting the polynomial values ​​for each segment into the target time. After resampling, the sampling period of each data source is uniformly set to 100 milliseconds. The generated multi-source monitoring data matrix stores different monitoring locations by row and different data types and times by column.

[0037] In one specific embodiment, step S2 includes:

[0038] The optical fiber phase change, optical power, bit error rate, optical signal-to-noise ratio, load current and meteorological temperature in the multi-source monitoring data matrix are used as input features and input to the input layer of a three-layer feedforward neural network.

[0039] The input features are nonlinearly transformed by 128 neurons in the hidden layer. The historical standard deviation of each data source within the time window is calculated. The historical stability index is calculated based on the historical standard deviation. The optical transmission performance degradation degree is calculated based on the deviation between optical power and reference optical power, the ratio of bit error rate to reference bit error rate, and the deviation between optical signal-to-noise ratio and reference optical signal-to-noise ratio. The abnormal deviation degree is obtained by weighted summation of the historical stability index and the optical transmission performance degradation degree.

[0040] Historical stability indicators and abnormal deviations are input into the output layer, and the weights of fiber optic signal data, communication performance data, power grid data, and meteorological data are calculated using the softmax normalization function.

[0041] The weight matrix and bias vector of a three-layer feedforward neural network are trained using the backpropagation algorithm and historical labeled fault data. The cross-entropy between the predicted fault type and the actual fault type is used as the loss function for iterative optimization.

[0042] Specifically, the input layer of the three-layer feedforward neural network receives feature vectors extracted from the multi-source monitoring data matrix. Fiber optic phase changes include temperature and strain changes calculated using phase-temperature and phase-strain conversion coefficients. Optical power, bit error rate, and optical signal-to-noise ratio are directly extracted from optical transmission line monitoring data. Load current is extracted from power grid data, and meteorological temperature is extracted from meteorological data. The number of neurons in the input layer equals the total dimension of all features. The 128 neurons in the hidden layer are fully connected to the input layer. The calculation for each neuron involves element-wise multiplication of the input feature vector with its corresponding weight vector, summing the results, and adding the bias value. The result is processed using the ReLU activation function, which sets negative values ​​to zero and retains positive values ​​to achieve a non-linear transformation. The historical standard deviation is calculated by looking back 300 seconds from the current time, extracting sample sequences from each data source within that time period from the multi-source monitoring data matrix, calculating the sum of the squares of the differences between the values ​​in the sample sequence and the mean, dividing by the number of samples, and then taking the square root to obtain the standard deviation. The historical stability index normalizes the reciprocal of the standard deviation to a range of 0 to 1. Data sources with smaller standard deviations have a stability index close to 1, while those with larger standard deviations have a stability index close to 0. The calculation of optical transmission performance degradation includes three components: the optical power deviation component (current optical power minus reference optical power, absolute value divided by reference optical power); the bit error rate ratio component (current bit error rate divided by reference bit error rate, logarithm); and the optical signal-to-noise ratio (SNR) deviation component (reference SNR minus current SNR, divided by reference SNR). These three components are multiplied by weighting coefficients of 0.3, 0.4, and 0.3 respectively, and then summed to obtain the optical transmission performance degradation degree. The anomaly deviation is calculated by multiplying the historical stability index by a coefficient of 0.6 and subtracting the optical transmission performance degradation multiplied by a coefficient of 0.4. This design ensures that even if the data statistics remain stable, anomaly deviation increases when optical transmission performance deteriorates, thus reducing the weight of that data source.

[0043] The output layer has 4 neurons, corresponding to the weights of the four data sources. The input value of each neuron in the output layer is the product of the hidden layer output vector and the output layer weight matrix, plus the output layer bias vector. The softmax normalization function calculates the natural exponent for each of the four input values ​​and divides it by the sum of the four exponent values. The four normalized values ​​are the weights for fiber optic signal data, communication performance data, power grid data, and meteorological data. All four weights are positive and their sum is always equal to 1. The backpropagation algorithm uses historical labeled fault data as training samples. Each sample contains a multi-source monitoring data matrix at the time of the fault and the corresponding real fault type label. The label is encoded as a one-hot vector. The network forward propagation calculates the predicted fault type probability distribution. The cross-entropy loss function calculates the cross-entropy between the real label vector and the predicted probability vector. Specifically, it is calculated by multiplying each dimension value of the real label by the corresponding dimension value of the predicted probability, taking the logarithm, summing the results, and then taking the negative value. The smaller the loss function value, the more accurate the prediction. Backpropagation adjusts the weight matrix and bias vector based on the gradient of the network parameters according to the loss function. The gradient is passed from the output layer to the input layer layer by layer through the chain rule. The weight is updated by subtracting the learning rate and multiplying the gradient from the current weight using the gradient descent method. The iterative optimization process repeats the forward propagation to calculate the loss and the backpropagation to update the parameters until the loss function converges or the preset number of iterations is reached.

[0044] Figure 2 This is a schematic diagram illustrating the dynamic change of the weights of each data source with the bit error rate of the optical transmission line in an embodiment of this application. For example... Figure 2 As shown, the horizontal axis represents the logarithmic scale of the bit error rate (BER), and the vertical axis represents the normalized weight values ​​of each data source. The graph includes four curves: fiber optic signal data weight, communication performance data weight, power grid data weight, and meteorological data weight. As the BER gradually increases from 10⁻¹² to 10⁻... 9 At that time, the weight of fiber optic signal data decreased from 0.70 to 0.20, while the weight of communication performance data increased from 0.15 to 0.60. The two curves showed a similar trend at a bit error rate of approximately 10⁻¹. 0 The weight reversal point occurs at this location, where the weight of the fiber optic signal data and the weight of the communication performance data are equal, both at 0.45. The red dashed line in the figure marks the bit error rate alarm threshold of 10⁻⁻⁶. 9 When the bit error rate exceeds this threshold, the system prioritizes communication performance data for fault determination. The blue dashed line marks the weight threshold of 0.3. When the weight of a data source is lower than this threshold, it is determined to be in a low-reliability state. The weights of power grid data and meteorological data remain relatively stable throughout the bit error rate variation, remaining within the ranges of 0.10 to 0.15 and 0.05 to 0.10, respectively, indicating that these two types of data sources have a relatively small weight proportion as auxiliary data in the fusion process.

[0045] In one specific embodiment, step S3 includes:

[0046] The temperature change, strain change, and vibration frequency converted from the fiber phase change in the multi-source monitoring data matrix are normalized respectively. The bit error rate, optical power, and optical signal-to-noise ratio are normalized respectively. The grid load current and meteorological temperature are mapped to each monitoring location through inverse distance weighted interpolation and then normalized.

[0047] The weighted values ​​of the fiber optic signal data are summed with the normalized temperature change, strain change, and vibration frequency; the weighted values ​​of the communication performance data are summed with the normalized bit error rate, optical power, and optical signal-to-noise ratio; the weighted values ​​of the power grid data are summed with the normalized load current; and the weighted values ​​of the meteorological data are summed with the normalized meteorological temperature. The four weighted values ​​are then summed to obtain the basic fusion value.

[0048] The communication service quality correction factor is calculated based on the difference between the bit error rate and the reference bit error rate, the ratio of optical power to the reference optical power, and the difference between the optical signal-to-noise ratio and the reference optical signal-to-noise ratio.

[0049] Multiplying the base fusion value by the communication service quality correction factor yields the line status fusion characteristic value.

[0050] Specifically, the fiber phase change is converted into temperature change, strain change, and vibration frequency using phase-temperature conversion coefficient, phase-strain conversion coefficient, and spectral analysis, respectively. Temperature change is normalized by subtracting a set minimum temperature change of -20 degrees Celsius from the current value and then dividing by the temperature change range of 60 degrees Celsius. Strain change is normalized by dividing the current value by the maximum strain change of 1000 microstrains. Vibration frequency is normalized by dividing the current value by the maximum vibration frequency of 80 Hz. Due to the large numerical range, the bit error rate (BER) is normalized using a logarithmic transformation. The BER is normalized by taking the commonly used logarithm, subtracting the minimum logarithm, and then dividing by the logarithmic range. Optical power is normalized by subtracting the reference optical power from the current value and then dividing by the allowable deviation range. Optical signal-to-noise ratio (SNR) is normalized by subtracting a minimum SNR of 10 dB from the current value and then dividing by the SNR range of 20 dB. The spatial locations of the power grid load current and meteorological temperature do not correspond to the fiber optic monitoring locations. Inverse distance weighted interpolation calculates the Euclidean distance from the monitoring location to each power grid station and meteorological station, uses the inverse square of the distance as a weighting factor, multiplies the data of each station by the corresponding weighting factor, sums them, and then divides by the sum of the weighting factors to obtain the interpolation result of the monitoring location. The load current is normalized by dividing the interpolation result by the rated current, and the meteorological temperature is normalized by subtracting -20 degrees Celsius from the interpolation result and then dividing by 70 degrees Celsius.

[0051] The weights of the fiber optic signal data are multiplied by the normalized temperature change, strain change, and vibration frequency to obtain three products. These three products are then weighted and summed in proportions of 0.4, 0.3, and 0.3 to obtain the weighted contribution value of the fiber optic signal. The weights of the communication performance data are multiplied by the normalized bit error rate, optical power, and optical signal-to-noise ratio to obtain three products. These three products are then weighted and summed in proportions of 0.3, 0.5, and 0.2 to obtain the weighted contribution value of the communication performance. The weights of the power grid data are directly multiplied by the normalized load current to obtain the weighted contribution value of the power grid. The weights of the meteorological data are directly multiplied by the normalized meteorological temperature to obtain the weighted contribution value of the meteorological data. The four weighted contribution values ​​are summed to obtain the basic fusion value. The calculation of the communication service quality correction factor first involves calculating the difference between the bit error rate (BER) and the reference BER. A positive difference indicates that the BER exceeds the standard. The excess portion is normalized and used as a penalty term. Next, the ratio of optical power to reference optical power is calculated. A ratio deviating from 1 indicates abnormal optical power. The absolute value of the deviation is normalized and used as an optical power penalty term. Then, the difference between the optical signal-to-noise ratio (SNR) and the reference SNR is calculated. A negative difference indicates insufficient SNR. The insufficient portion is normalized and used as an SNR penalty term. These three penalty terms are multiplied by weighting coefficients and summed to obtain the total penalty. The communication service quality correction factor equals 1 minus the total penalty. The basic fusion value is multiplied by the communication service quality correction factor. When communication quality is normal, the correction factor is close to 1, and the basic fusion value remains essentially unchanged. When the BER increases or the SNR decreases, the correction factor is less than 1, amplifying the fusion characteristic value and making it easier to trigger the alarm threshold. The line status fusion characteristic value comprehensively reflects the degree of line anomaly in multi-source data evaluation and communication quality constraints.

[0052] In one specific embodiment, step S4 includes:

[0053] Calculate the ratio of the fiber optic signal data weight to the communication performance data weight at the monitoring location. When the ratio exceeds a set threshold, the fiber optic sensing function at the monitoring location is determined to be abnormal.

[0054] Extract the temperature change sequence, strain change sequence, and corresponding optical power sequence of the fiber phase change conversion from five consecutive adjacent monitoring positions in front of and five consecutive adjacent monitoring positions behind the monitoring position.

[0055] Calculate the weighted average of the fiber optic signal data from the five adjacent monitoring positions in front and the five adjacent monitoring positions behind. When both the forward average and the backward average are higher than the reliability threshold, the data from the adjacent monitoring positions are determined to be reliable, and gradient compensation calculation is performed.

[0056] Specifically, the ratio of the fiber optic signal data weight to the communication performance data weight is obtained by dividing the former by the latter. A ratio greater than 1 indicates a high weight for the fiber optic signal data, while a ratio less than 1 indicates that the communication performance data weight dominates. When the ratio exceeds a set threshold of 2.5, it indicates that the monitoring location is overly reliant on communication performance data while the weight of fiber optic disturbance data is too low. This situation is usually caused by local fiber optic sensor failure, increased fiber microbending loss, or electromagnetic interference leading to a decrease in fiber optic phase signal quality. After determining that the fiber optic sensing function is abnormal, the spatial gradient compensation program is initiated. When extracting data from adjacent monitoring locations, the five preceding locations refer to the monitoring points corresponding to the current abnormal monitoring location number minus 1 to minus 5, and the five following locations refer to the monitoring points corresponding to the current abnormal monitoring location number plus 1 to plus 5. The fiber optic phase change of these ten locations is read from the multi-source monitoring data matrix. The phase change is converted into a temperature change using a phase-temperature conversion coefficient to form a preceding temperature sequence and a following temperature sequence. The phase change is converted into a strain change using a phase-strain conversion coefficient to form a preceding strain sequence and a following strain sequence. At the same time, the optical power measurement values ​​of these ten locations are extracted to form a preceding optical power sequence and a following optical power sequence.

[0057] The average weighted value of the fiber optic signal data from the five adjacent monitoring locations ahead is calculated by adding the weights of the five locations and dividing by 5. The average weighted value of the fiber optic signal data from the five adjacent monitoring locations behind is calculated using the same method. The reliability threshold is set to 0.6. When both the forward and backward average values ​​are greater than or equal to 0.6, it indicates that the fiber optic sensors on both sides of the abnormal monitoring location are working normally, and the data has the reliability for gradient fitting. In this case, the disturbance parameters of adjacent locations can reflect the true physical state along the fiber optic line, satisfying the spatial continuity assumption. If the forward or backward average values ​​are lower than the reliability threshold, it indicates that the abnormal area is large or that there are also sensor anomalies at adjacent locations. In this case, gradient compensation is not performed, and fault determination is directly based on communication performance data and external data sources to avoid error accumulation caused by compensation based on unreliable data.

[0058] In one specific embodiment, performing gradient compensation calculation includes:

[0059] The optical power weighting factor is calculated based on the deviation between the optical power of the five adjacent monitoring positions and the reference optical power. The optical power weighting factor is used as a weighting coefficient to perform weighted least squares fitting on the temperature change sequence and position coordinates of the five adjacent monitoring positions to construct a forward quadratic polynomial model and calculate the forward predicted temperature change of the monitoring position.

[0060] The optical power weighting factor is calculated based on the deviation between the optical power of the five adjacent monitoring positions and the reference optical power. The optical power weighting factor is used as a weighting coefficient to perform weighted least squares fitting on the temperature change sequence and position coordinates of the five adjacent monitoring positions to construct a backward quadratic polynomial model and calculate the backward predicted temperature change of the monitoring position.

[0061] The forward confidence level is calculated based on the variance of the fitting residuals of the forward quadratic polynomial model, and the backward confidence level is calculated based on the variance of the fitting residuals of the backward quadratic polynomial model. The forward and backward predicted temperature changes are then weighted and averaged according to the forward and backward confidence levels to obtain the compensation disturbance value of the monitoring location.

[0062] Specifically, when calculating the optical power weighting factor for the five adjacent monitoring positions ahead, the measured optical power at each position is first subtracted from the reference optical power by -3dBm to obtain the deviation. The absolute value of the deviation is divided by 3dBm, and the negative number is used as the input of the exponential function. The natural exponential value is then calculated to obtain the optical power weighting factor. The smaller the optical power deviation, the closer the weighting factor is to 1; the larger the optical power deviation, the closer the weighting factor is to 0. This design ensures that measurement points with good optical transmission quality have a greater weight in the fitting process. A weighted least squares fitting method is used to construct the objective function, which is the sum of the squared fitting residuals of each measurement point multiplied by the corresponding optical power weighting factor. The fitting residual is the difference between the measured temperature change and the value calculated by the quadratic polynomial model. The quadratic polynomial model is expressed as: the temperature change equals the coefficient 'a' multiplied by the position coordinate minus the square of the anomaly point's position coordinate, plus the coefficient 'b' multiplied by the position coordinate minus the anomaly point's position coordinate, plus a constant term 'c'. By taking the partial derivatives of the objective function with respect to the coefficients 'a', 'b', and 'c' respectively and setting the partial derivatives to zero, a system of three linear equations is obtained. Solving the system of equations yields the fitting coefficients. The coordinates of the anomaly monitoring location are then substituted into the forward quadratic polynomial model to calculate the forward predicted temperature change. The same method is used to calculate the optical power weighting factor and construct a weighted least squares objective function for the five adjacent monitoring locations. The coefficients of the backward quadratic polynomial model are then fitted, and the coordinates of the anomaly monitoring location are substituted into the backward model to calculate the backward predicted temperature change.

[0063] The variance of the fitting residuals of the forward quadratic polynomial model is obtained by multiplying the squares of the fitting residuals of the five forward measurement points by the optical power weighting factor, summing the results, and then dividing by 5. A smaller variance indicates a better fit to the data. The forward confidence score uses a natural exponential function to map the negative value of the fitting residual variance to a value between 0 and 1; a small residual variance results in a confidence score close to 1, while a large residual variance results in a confidence score close to 0. The variance of the fitting residuals and the backward confidence score of the backward quadratic polynomial model are calculated using the same method. The forward predicted temperature change is multiplied by the forward confidence score, and the backward predicted temperature change is multiplied by the backward confidence score. The sum of these two products is divided by the sum of the forward and backward confidence scores to obtain a weighted average. This weighted average assigns higher weights to the prediction directions with better fitting quality. The resulting value serves as the compensation for temperature changes at the anomaly monitoring location. The compensation value for strain changes is calculated using the same gradient fitting process. The compensation values ​​for temperature changes and strain changes together constitute the compensation disturbance value for the monitoring location.

[0064] In one specific embodiment, the deviation between the theoretical change in optical power and the measured change in optical power is used as a criterion for the effectiveness of compensation, including:

[0065] The temperature-induced refractive index change is calculated based on the temperature change in the compensation disturbance value and the fiber temperature refractive index coefficient; the strain-induced refractive index change is calculated based on the strain change in the compensation disturbance value and the fiber strain refractive index coefficient.

[0066] The optical power loss caused by temperature is calculated based on the change in refractive index caused by temperature, the change in temperature, and the optical fiber temperature loss coefficient. The optical power loss caused by strain is calculated based on the change in refractive index caused by strain, the change in strain, and the optical fiber strain loss coefficient. The optical power loss caused by temperature and the optical power loss caused by strain are added together to obtain the theoretical change in optical power.

[0067] The measured optical power at the monitoring location is extracted from the multi-source monitoring data matrix, and the difference between the measured optical power and the reference optical power is calculated to obtain the change in measured optical power.

[0068] The absolute deviation between the theoretical optical power change and the measured optical power change is calculated. When the absolute deviation is less than the preset deviation threshold, the compensation is deemed effective. The weighted correction coefficient of the compensation disturbance value is weighted and summed with the average weight of the fiber optic signal data at adjacent monitoring locations to obtain the corrected fiber optic signal data weight at the monitoring location.

[0069] Specifically, the temperature-induced refractive index change is calculated by multiplying the temperature change in the compensation disturbance value by the fiber temperature refractive index coefficient of 8.6 multiplied by 10 to the power of -6 per degree Celsius, and then by the fiber effective refractive index of 1.468. The strain-induced refractive index change is calculated by multiplying the strain change in the compensation disturbance value by the fiber strain refractive index coefficient of 0.22, and then by the fiber effective refractive index of 1.468. The refractive index change reflects the degree of influence of temperature and strain on the optical fiber light transmission characteristics. The optical power loss caused by temperature is obtained by multiplying the optical fiber temperature loss coefficient of 0.0005 dB per kilometer per degree Celsius by the temperature change, then by the monitoring point spacing of 10.2 meters, and finally by 1000 meters to convert the units. The optical power loss caused by strain is obtained by multiplying the optical fiber strain loss coefficient of 0.0000015 dB per kilometer per microstrain by the strain change, then by the monitoring point spacing of 10.2 meters, and finally by 1000 meters to convert the units. The sum of the two optical power losses gives the theoretical optical power change, which is the expected change in optical power calculated based on the physical characteristics of the optical fiber and the compensation disturbance value.

[0070] The measured optical power value at the current moment is read from the spatial coordinate index of the abnormal monitoring location in the multi-source monitoring data matrix. The change in measured optical power is obtained by subtracting the reference optical power -3dBm from the measured optical power. The measured value reflects the deviation of the optical power actually measured by the optical transmission system. The absolute deviation is obtained by subtracting the theoretical optical power change from the measured optical power change. The preset deviation threshold is set to 1 dB. When the absolute deviation is less than 1 dB, it indicates that the theoretical optical power change calculated by the compensation disturbance value matches the measured value, verifying the accuracy of the compensation disturbance value and determining that the compensation is effective. The weight correction factor for the compensation disturbance value is set to 0.5. The average weight of the fiber optic signal data of the five adjacent monitoring positions in front and behind has been calculated in the previous step. The smaller of the two average values ​​is taken as the representative weight of the adjacent monitoring position. The weight correction factor of 0.5 is multiplied by the original weight of the fiber optic signal data of the abnormal monitoring position, plus the weight correction factor of 0.5 multiplied by the representative weight of the adjacent position. The weighted sum is used as the corrected weight of the fiber optic signal data. The corrected weight enables the abnormal monitoring position to partially restore the fiber optic sensing and positioning capability. The weight increase enables the fiber optic disturbance data to play a role again in the subsequent fusion calculation. The positioning accuracy is restored from the kilometer level to the hundred-meter level. If the absolute deviation is greater than or equal to 1 dB, the weight is not corrected and the fiber optic body fault warning is triggered.

[0071] Figure 3 This is a schematic diagram illustrating the comparison and verification of the theoretical optical power change and the measured optical power change in the embodiments of this application. Figure 3As shown, the horizontal axis represents the numbers of the 20 test monitoring points, and the vertical axis represents the change in optical power in decibels (dB). The graph includes both theoretical and measured optical power change bar charts, with the red dashed line marking the 1dB deviation threshold. Among the 20 test points, the measured optical power changes at test points 8 and 16 were 1.2dB and 1.5dB, respectively, significantly higher than the theoretical calculation values ​​of 0.38dB and 0.68dB, with deviations exceeding the 1dB threshold. These points are marked with a red cross in the graph as fiber defects, indicating ineffective compensation. These two points may have inherent problems such as microcracks or joint degradation in the fiber. The deviations between the theoretical and measured values ​​for the remaining 18 test points were all less than 1dB, with relative errors between theoretical and measured values ​​ranging from 5% to 15%, verifying the accuracy of the compensation disturbance value. The effective compensation rate was 90%. The upper right corner of the figure shows that 18 out of 20 points are valid compensation points, achieving an efficiency of 100%. Subtracting the two invalid points, the proportion is 90%, indicating that the theoretical optical power change calculated based on the fiber temperature loss coefficient and strain loss coefficient can effectively verify the accuracy of gradient compensation. When the theoretical value matches the measured value, the compensation is confirmed to be effective and the fiber signal data weight is corrected. When the deviation is too large, it is identified as a fiber optic fault and a low weight is maintained to avoid introducing new positioning errors through incorrect compensation.

[0072] The above describes the power communication line monitoring method based on fiber optic multi-source data fusion in the embodiments of this application. The following describes the power communication line monitoring system based on fiber optic multi-source data fusion in the embodiments of this application. One embodiment of the power communication line monitoring system based on fiber optic multi-source data fusion in the embodiments of this application includes:

[0073] The synchronization module is used to collect backscattered light signals from power communication optical cables, bit error rate and optical signal-to-noise ratio of optical transmission lines, power grid load data and meteorological environmental data, and to perform time synchronization processing on each data source to generate a multi-source monitoring data matrix.

[0074] The calculation module is used to calculate the weights of optical fiber signal data, communication performance data, power grid data, and meteorological data by using a confidence evaluation neural network based on the historical stability of each data source in the multi-source monitoring data matrix and the change in bit error rate of the optical transmission line.

[0075] The weighting module is used to perform weighted fusion of the optical fiber signal data weight, communication performance data weight, power grid data weight, and meteorological data weight with the corresponding normalized data features, calculate the communication service quality correction factor based on the bit error rate and optical signal-to-noise ratio of the optical transmission line, and correct the weighted fusion result to obtain the line state fusion feature value.

[0076] The monitoring module is used to extract the fiber optic disturbance parameter sequence and optical power sequence of adjacent monitoring locations when the ratio of the weight of the fiber optic signal data to the weight of the communication performance data at the monitoring location exceeds a set threshold. It then calculates the compensation disturbance value of the monitoring location by least squares fitting with optical power weighting, calculates the theoretical optical power change based on the compensation disturbance value and the fiber temperature loss coefficient and strain loss coefficient, and uses the deviation between the theoretical optical power change and the measured optical power change as the criterion for compensation effectiveness.

[0077] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring power communication lines using fiber optic multi-source data fusion, characterized in that, The method includes: Step S1: Collect backscattered light signals from power communication optical cables, bit error rate and optical signal-to-noise ratio of optical transmission lines, power grid load data and meteorological environment data, perform time synchronization processing on each data source, and generate a multi-source monitoring data matrix. Step S2: Based on the historical stability of each data source in the multi-source monitoring data matrix and the change in bit error rate of the optical transmission line, calculate the weights of optical fiber signal data, communication performance data, power grid data, and meteorological data using a confidence evaluation neural network. Step S3: Weight the fiber optic signal data weights, communication performance data weights, power grid data weights, and meteorological data weights with the corresponding normalized data features. Calculate the communication service quality correction factor based on the bit error rate and optical signal-to-noise ratio of the optical transmission line. Correct the weighted fusion result to obtain the line state fusion feature value. This includes: normalizing the temperature change, strain change, and vibration frequency converted from fiber phase changes in the multi-source monitoring data matrix; normalizing the bit error rate, optical power, and optical signal-to-noise ratio; mapping the power grid load current and meteorological temperature to each monitoring location using inverse distance weighted interpolation and normalizing them; and... The signal data weights are weighted and summed with normalized temperature change, strain change, and vibration frequency. The communication performance data weights are weighted and summed with normalized bit error rate, optical power, and optical signal-to-noise ratio. The power grid data weights are weighted and summed with normalized load current. The meteorological data weights are weighted and summed with normalized meteorological temperature. The four weighted results are summed to obtain the basic fusion value. The communication service quality correction factor is calculated based on the difference between the bit error rate and the reference bit error rate, the ratio of the optical power to the reference optical power, and the difference between the optical signal-to-noise ratio and the reference optical signal-to-noise ratio. The basic fusion value is multiplied by the communication service quality correction factor to obtain the line state fusion characteristic value. Step S4: When the ratio of the weight of the fiber optic signal data to the weight of the communication performance data at the monitoring location exceeds a set threshold, extract the fiber optic disturbance parameter sequence and optical power sequence of the adjacent monitoring locations before and after, calculate the compensation disturbance value of the monitoring location by least square fitting with optical power weighting, calculate the theoretical optical power change based on the compensation disturbance value and the fiber temperature loss coefficient and strain loss coefficient, and use the deviation between the theoretical optical power change and the measured optical power change as the criterion for compensation effectiveness.

2. The method for monitoring power communication lines using fiber optic multi-source data fusion according to claim 1, characterized in that, Step S1 includes: Pulsed laser light is injected into the power communication optical cable using a phase-sensitive optical time-domain reflectometer. The backscattered light signal is received and the phase change is demodulated. The spatial resolution is calculated based on the fiber refractive index and the speed of light, and the phase change at each monitoring location is extracted. The optical power, bit error rate and optical signal-to-noise ratio of each repeater segment are collected through the network management interface of the optical transmission line; the load current and voltage of the substation are collected through the power dispatching system interface; and the temperature, humidity and wind speed along the optical cable path are collected through the meteorological data interface. The timestamps of the phase change, optical power, bit error rate, optical signal-to-noise ratio, load current, voltage, temperature, humidity and wind speed are aligned to a unified time base using a clock protocol; The load current, voltage, temperature, humidity, and wind speed with low sampling rates are resampled using a cubic spline interpolation algorithm to a sampling period consistent with the phase change, optical power, bit error rate, and optical signal-to-noise ratio, thereby generating the multi-source monitoring data matrix.

3. The method for monitoring power communication lines using fiber optic multi-source data fusion according to claim 1, characterized in that, Step S2 includes: The optical fiber phase change, optical power, bit error rate, optical signal-to-noise ratio, load current, and meteorological temperature in the multi-source monitoring data matrix are used as input features and input to the input layer of a three-layer feedforward neural network. The input features are nonlinearly transformed using 128 neurons in the hidden layer. The historical standard deviation of each data source within the time window is calculated. The historical stability index is calculated based on the historical standard deviation. The optical transmission performance degradation degree is calculated based on the deviation between the optical power and the reference optical power, the ratio of the bit error rate to the reference bit error rate, and the deviation between the optical signal-to-noise ratio and the reference optical signal-to-noise ratio. The anomaly deviation degree is obtained by weighted summation of the historical stability index and the optical transmission performance degradation degree. The historical stability index and the abnormal deviation are input to the output layer, and the weights of the optical fiber signal data, communication performance data, power grid data, and meteorological data are calculated using the softmax normalization function. The weight matrix and bias vector of the three-layer feedforward neural network are trained using the backpropagation algorithm and historical labeled fault data. The cross-entropy between the predicted fault type and the actual fault type is used as the loss function for iterative optimization.

4. The method for monitoring power communication lines using fiber optic multi-source data fusion according to claim 1, characterized in that, Step S4 includes: Calculate the ratio of the fiber optic signal data weight to the communication performance data weight at the monitoring location. When the ratio exceeds a set threshold, determine that the fiber optic sensing function at the monitoring location is abnormal. Extract the temperature change sequence, strain change sequence, and corresponding optical power sequence of the fiber phase change conversion from five consecutive adjacent monitoring positions in front of and five consecutive adjacent monitoring positions behind the monitoring position. Calculate the average weight of the fiber optic signal data from the five adjacent monitoring positions in front and the five adjacent monitoring positions behind. When both the average value in front and the average value behind are higher than the reliability threshold, the data from the adjacent monitoring positions are determined to be reliable, and gradient compensation calculation is performed.

5. The method for monitoring power communication lines using fiber optic multi-source data fusion according to claim 4, characterized in that, The gradient compensation calculation includes: The optical power weighting factor is calculated based on the deviation between the optical power of the five adjacent monitoring positions and the reference optical power. The optical power weighting factor is used as a weighting coefficient to perform weighted least squares fitting on the temperature change sequence and position coordinates of the five adjacent monitoring positions to construct a forward quadratic polynomial model and calculate the forward predicted temperature change of the monitoring position. The optical power weighting factor is calculated based on the deviation between the optical power of the five adjacent monitoring positions and the reference optical power. The optical power weighting factor is used as a weighting coefficient to perform weighted least squares fitting on the temperature change sequence and position coordinates of the five adjacent monitoring positions to construct a backward quadratic polynomial model and calculate the backward predicted temperature change of the monitoring position. The forward confidence level is calculated based on the variance of the fitting residuals of the forward quadratic polynomial model, and the backward confidence level is calculated based on the variance of the fitting residuals of the backward quadratic polynomial model. The forward predicted temperature change and the backward predicted temperature change are weighted and averaged according to the forward confidence level and the backward confidence level to obtain the compensation disturbance value of the monitoring location.

6. The method for monitoring power communication lines using fiber optic multi-source data fusion according to claim 5, characterized in that, The method of using the deviation between the theoretical optical power change and the measured optical power change as a criterion for compensation effectiveness includes: The temperature-induced refractive index change is calculated based on the temperature change in the compensation disturbance value and the fiber temperature refractive index coefficient; the strain-induced refractive index change is calculated based on the strain change in the compensation disturbance value and the fiber strain refractive index coefficient. The optical power loss caused by temperature is calculated based on the refractive index change caused by temperature, the temperature change, and the optical fiber temperature loss coefficient. The optical power loss caused by strain is calculated based on the refractive index change caused by strain, the strain change, and the optical fiber strain loss coefficient. The theoretical optical power change is obtained by adding the optical power loss caused by temperature and the optical power loss caused by strain. The measured optical power at the monitoring location is extracted from the multi-source monitoring data matrix, and the difference between the measured optical power and the reference optical power is calculated to obtain the change in measured optical power. The absolute deviation between the theoretical optical power change and the measured optical power change is calculated. When the absolute deviation is less than a preset deviation threshold, the compensation is deemed effective. The weighted correction coefficient of the compensation disturbance value is weighted and summed with the average weight of the fiber optic signal data at adjacent monitoring locations to obtain the corrected fiber optic signal data weight at the monitoring location.

7. A power communication line monitoring system with fiber optic multi-source data fusion, characterized in that, The power communication line monitoring method for implementing fiber optic multi-source data fusion as described in any one of claims 1-6, wherein the fiber optic multi-source data fusion power communication line monitoring system comprises: The synchronization module is used to collect backscattered light signals from power communication optical cables, bit error rate and optical signal-to-noise ratio of optical transmission lines, power grid load data and meteorological environmental data, and to perform time synchronization processing on each data source to generate a multi-source monitoring data matrix. The calculation module is used to calculate the weights of optical fiber signal data, communication performance data, power grid data, and meteorological data by using a confidence evaluation neural network based on the historical stability of each data source in the multi-source monitoring data matrix and the change in bit error rate of the optical transmission line. The weighting module is used to perform weighted fusion of the fiber optic signal data weights, communication performance data weights, power grid data weights, and meteorological data weights with the corresponding normalized data features. It calculates a communication service quality correction factor based on the bit error rate and optical signal-to-noise ratio of the optical transmission line, and corrects the weighted fusion result to obtain the line state fusion feature value. This includes: normalizing the temperature change, strain change, and vibration frequency converted from fiber phase changes in the multi-source monitoring data matrix; normalizing the bit error rate, optical power, and optical signal-to-noise ratio; mapping the power grid load current and meteorological temperature to each monitoring location through inverse distance weighted interpolation and performing normalization; and... The fiber signal data weights are weighted and summed with normalized temperature change, strain change, and vibration frequency. The communication performance data weights are weighted and summed with normalized bit error rate, optical power, and optical signal-to-noise ratio. The power grid data weights are weighted and summed with normalized load current. The meteorological data weights are weighted and summed with normalized meteorological temperature. The four weighted results are summed to obtain the basic fusion value. The communication service quality correction factor is calculated based on the difference between the bit error rate and the reference bit error rate, the ratio of the optical power to the reference optical power, and the difference between the optical signal-to-noise ratio and the reference optical signal-to-noise ratio. The basic fusion value is multiplied by the communication service quality correction factor to obtain the line state fusion characteristic value. The monitoring module is used to extract the fiber optic disturbance parameter sequence and optical power sequence of adjacent monitoring locations when the ratio of the weight of the fiber optic signal data to the weight of the communication performance data at the monitoring location exceeds a set threshold. It then calculates the compensation disturbance value of the monitoring location by least squares fitting with optical power weighting, calculates the theoretical optical power change based on the compensation disturbance value and the fiber temperature loss coefficient and strain loss coefficient, and uses the deviation between the theoretical optical power change and the measured optical power change as the criterion for compensation effectiveness.

8. The system according to claim 7, characterized in that, The system collects backscattered optical signals from power communication optical cables, bit error rate and optical signal-to-noise ratio from optical transmission lines, power grid load data, and meteorological environmental data. It performs time synchronization processing on each data source to generate a multi-source monitoring data matrix, including: Pulsed laser light is injected into the power communication optical cable using a phase-sensitive optical time-domain reflectometer. The backscattered light signal is received and the phase change is demodulated. The spatial resolution is calculated based on the fiber refractive index and the speed of light, and the phase change at each monitoring location is extracted. The optical power, bit error rate and optical signal-to-noise ratio of each repeater segment are collected through the network management interface of the optical transmission line; the load current and voltage of the substation are collected through the power dispatching system interface; and the temperature, humidity and wind speed along the optical cable path are collected through the meteorological data interface. The timestamps of the phase change, optical power, bit error rate, optical signal-to-noise ratio, load current, voltage, temperature, humidity and wind speed are aligned to a unified time base using a clock protocol; The load current, voltage, temperature, humidity, and wind speed with low sampling rates are resampled using a cubic spline interpolation algorithm to a sampling period consistent with the phase change, optical power, bit error rate, and optical signal-to-noise ratio, thereby generating the multi-source monitoring data matrix.

9. The system according to claim 8, characterized in that, Based on the historical stability of each data source in the multi-source monitoring data matrix and the change in bit error rate of the optical transmission line, the weights of fiber optic signal data, communication performance data, power grid data, and meteorological data are calculated using a confidence assessment neural network, including: The optical fiber phase change, optical power, bit error rate, optical signal-to-noise ratio, load current, and meteorological temperature in the multi-source monitoring data matrix are used as input features and input to the input layer of a three-layer feedforward neural network. The input features are nonlinearly transformed using 128 neurons in the hidden layer. The historical standard deviation of each data source within the time window is calculated. The historical stability index is calculated based on the historical standard deviation. The optical transmission performance degradation degree is calculated based on the deviation between the optical power and the reference optical power, the ratio of the bit error rate to the reference bit error rate, and the deviation between the optical signal-to-noise ratio and the reference optical signal-to-noise ratio. The anomaly deviation degree is obtained by weighted summation of the historical stability index and the optical transmission performance degradation degree. The historical stability index and the abnormal deviation are input to the output layer, and the weights of the optical fiber signal data, communication performance data, power grid data, and meteorological data are calculated using the softmax normalization function. The weight matrix and bias vector of the three-layer feedforward neural network are trained using the backpropagation algorithm and historical labeled fault data. The cross-entropy between the predicted fault type and the actual fault type is used as the loss function for iterative optimization.

Images