Power optical fiber on-line state monitoring and early warning method and system

Abstract

The invention relates to an electric power optical fiber online state monitoring and early warning method and system. The method comprises the following steps that a non-intrusive monitoring module is adopted to collect optical power data of an optical fiber; high-frequency noise is filtered out through multi-scale wavelet denoising preprocessing in combination with an improved SureShrink algorithm; inputting the de-noised data into a simulated annealing-long short-term memory SA-LSTM prediction model of simulated annealing optimization, and capturing key time sequence features by using a time attention mechanism; and dynamically calculating a threshold value based on a sliding window, and when a predicted value exceeds a limit, triggering degradation early warning and automatically switching an optical fiber path. Signal interference is avoided through non-contact monitoring, the data quality is improved in combination with wavelet denoising, the prediction precision is improved by optimizing hyper-parameters of a long short-term memory (LSTM) model through a simulated annealing SA algorithm, a dynamic threshold value adapts to environmental fluctuation, and real-time monitoring and intelligent maintenance of the health state of the optical fiber are achieved.

Description

Technical Field

[0001] This application relates to the field of optical communication technologies, and more specifically, to a method and system for on-line status monitoring and early warning of power optical fibers. Background Art

[0002] With the development of modern power systems, optical fiber communication technologies are increasingly widely used in power transmission networks. As an important carrier for power transmission, the stability and security of power optical fibers are crucial for the reliable operation of power systems. During long-term use, power optical fibers may deteriorate due to various factors such as environmental factors, mechanical stress, and chemical corrosion. Such deterioration not only affects the efficiency of power transmission but may also trigger power accidents, causing serious economic losses and social impacts. Therefore, it is particularly important to timely monitor and evaluate the health status of optical fibers and conduct effective early warning of optical fiber status.

[0003] Currently, methods for detecting the status of optical fibers mainly include optical fiber testing, manual inspection, etc. These traditional methods have deficiencies such as long response times and reliance on manual judgment, and cannot meet the requirements of modern power systems for real-time monitoring and rapid response. Therefore, developing a power optical fiber status early warning system based on advanced data analysis methods, which can achieve real-time monitoring and analysis of the status of optical fibers, can not only improve the safety and reliability of power systems, but also effectively reduce operation and maintenance costs and extend the service life of optical fibers. The research and application of this system have important theoretical value and broad market prospects, and can provide strong support for the intelligent development of the power industry.

[0004] The prior art, such as the Chinese patent application with the publication number "CN118890092A", discloses an optical power monitoring system and method applied to optical fibers. The system includes an optical power status identification module, an abnormal signal detection module, a status conversion analysis module, a network adjustment instruction module, an implementation effect tracking module, and an optical power monitoring optimization module. In this invention, by analyzing the stable stage of optical power data in the optical fiber network and setting status boundaries, the accuracy and response speed of fault detection are improved. After setting the status boundaries, threshold division is used to identify and mark abnormal signals, reducing false positives and missed detections. The path of optical power change is analyzed, the status conversion mode is identified and the trend is tracked, network management is optimized and the ability to prevent faults is improved, the light source power and optical fiber path are adjusted to optimize the transmission efficiency, the maintenance cost is reduced, the stability after adjustment is monitored in real time, rapid feedback is provided, and the monitoring parameters are adjusted according to the effectiveness test to enhance the adaptability of the monitoring strategy, ensuring the long-term stable and efficient operation of the network.

[0005] The problems existing in the above-mentioned prior art are that only anomaly detection is carried out through the classification of historical data states, relying on fixed state boundaries and K-means clustering, which cannot adapt to dynamic environmental factors such as optical fiber aging and temperature drift, resulting in a high false alarm rate and possible lag in early warning; the clustering algorithm depends on manual setting of distance indices and weight coefficients, and parameter adjustment takes a long time and is prone to falling into local optima. Summary of the Invention

[0006] To solve the above technical problems, the present invention proposes a method and system for on-line status monitoring and early warning of power optical fibers.

[0007] The technical solution of the present invention is as follows:

[0008] The present invention proposes a method for on-line status monitoring and early warning of power optical fibers, including the following steps:

[0009] Step S1, collecting optical power data of the power optical fiber by using a non-invasive status monitoring module;

[0010] Step S2, performing multi-scale wavelet denoising preprocessing on the collected optical power data;

[0011] Step S3, taking the preprocessed optical power data as the input of a trained simulated annealing-long short-term memory SA-LSTM prediction model to predict the optical power;

[0012] Step S4, when the predicted value of the optical power exceeds the dynamic threshold, sending out an optical fiber deterioration early warning and automatically switching the optical fiber path.

[0013] As a preferred embodiment, the specific steps of performing multi-scale wavelet denoising preprocessing on the collected optical power data are as follows:

[0014] Performing normalization processing on the collected optical power data, specifically:

[0015]

[0016] where: is the optical power data after normalization processing; x i is the original optical power data; x max 、x min are respectively the maximum and minimum values in the original optical power data;

[0017] Performing wavelet decomposition on the normalized optical power data by using an improved SureShrink algorithm to obtain the coefficients d of each layer of wavelet decomposition j,k , and performing threshold denoising, specifically:

[0018]

[0019] where the dynamic threshold λj The adjustment formula is as follows:

[0020]

[0021] Where: is the wavelet coefficient after denoising; d j,k is the k-th wavelet coefficient of the j-th layer; λ j is the dynamic threshold of the j-th layer; σ j is the noise standard deviation of the j-th layer; N j is the number of data in the j-th layer;

[0022] Reconstruct the wavelet coefficients after denoising to obtain the optical power data after denoising, specifically:

[0023]

[0024] Where: is the optical power data after denoising; is the wavelet basis function; τ is the reconstruction time point.

[0025] As a preferred embodiment, the simulated annealing-long short-term memory SA-LSTM prediction model optimizes the number of iterations, the number of hidden units, and the learning rate in the long short-term memory LSTM model through the simulated annealing algorithm SA. The specific steps are as follows:

[0026] Define the parameter search space and the energy function: Set the value ranges of the parameters to be optimized, namely the number of iterations, the number of hidden units, and the learning rate. The energy function is the mean square error of the validation set of the long short-term memory LSTM model, specifically:

[0027]

[0028] Where: E(S) is the energy function; N val is the number of samples in the validation set; y n is the actual optical power; is the optical power predicted by the LSTM model;

[0029] Initialize the simulated annealing parameters: Set the initial temperature, the termination temperature, the cooling coefficient, the maximum number of iterations, and the neighborhood perturbation step size;

[0030] Generate the initial solution and calculate the energy: Generate a random parameter combination and train the LSTM model to calculate the initial energy;

[0031] Iteratively optimize until the termination condition is met and output the optimal parameter combination, where:

[0032] The formula for generating the neighborhood solution is:

[0033] S ′ =(e ′ ,h′ , η ′ );

[0034] Wherein:

[0035] e ′ = Clip(e current + Δe, e min , e max );

[0036] h ′ = Clip(h current + Δh, h min , h max );

[0037] η ′ = Clip(η current × (1 + Δη), η min , η max );

[0038] In the formula: S ′ is the neighborhood solution; e ′ , h ′ and η ′ are respectively the number of iterations of the parameter combination of the neighborhood solution, the number of hidden units, and the learning rate; Clip is to limit the parameters of the neighborhood solution within the upper and lower limit intervals; e current , h current and η current are respectively the number of iterations, the number of hidden units, and the learning rate in the iterative state; Δe, Δh, and Δη are respectively the perturbation parameters for generating the neighborhood solution of the number of iterations, the number of hidden units, and the learning rate; e min , e max , h min , h max , η min , η max are respectively the value ranges of the number of iterations, the number of hidden units, and the learning rate;

[0039] The energy difference calculation formula is:

[0040] ΔE = E(S ′ ) - E(S);

[0041] In the formula: ΔE is the energy difference between the new solution and the current solution; if ΔE < 0, accept the neighborhood solution S ′ , otherwise, accept the neighborhood solution S with probability ′ , T is the current temperature;

[0042] The cooling strategy during the iteration process is:

[0043] T t+1 = α · T t ;

[0044] where: α is the temperature reduction coefficient; T t is the temperature at the t-th iteration;

[0045] The global optimal solution is:

[0046] S best = argmin(E(S));

[0047] where: S best is the global optimal solution; when the maximum number of iterations is reached or the temperature reaches the termination temperature, the global optimal solution is output.

[0048] As a preferred embodiment, a time attention module is added before the LSTM layer of the simulated annealing-long short-term memory SA-LSTM prediction model, and the weight calculation formula of the time attention module is:

[0049]

[0050] where: w t is the importance weight of the hidden state at the t-th time step for the current prediction; v T is the transpose of the trainable parameter vector v; W h is the trainable weight matrix; h t is the hidden state of the LSTM at the t-th time step; b h is the trainable bias term; h r is the hidden state of the LSTM at the r-th time step; tanh is the hyperbolic tangent activation function; R is the total number of time steps of the input sequence.

[0051] As a preferred embodiment, the dynamic threshold calculates the mean and standard deviation of historical data based on a sliding window and dynamically updates the threshold, and the calculation formula is:

[0052] θ dynamic = μ window + β·σ window ;

[0053] where: θ dynamic is the dynamic threshold; μ window is the mean optical power within the sliding window; β is the safety factor; σ window is the standard deviation of the optical power within the sliding window.

[0054] On the other hand, the present invention also provides an on-line state monitoring and early warning system for power optical fibers, including:

[0055] A non-intrusive state monitoring module for collecting optical power data of power optical fibers;

[0056] Data preprocessing module, which performs multi-scale wavelet denoising preprocessing on the collected optical power data;

[0057] Prediction module, which takes the preprocessed optical power data as the input of the trained simulated annealing-long short-term memory SA-LSTM prediction model to predict the optical power;

[0058] Early warning module, when the predicted value of the optical power exceeds the dynamic threshold, it issues a warning of optical fiber deterioration and automatically switches the optical fiber path.

[0059] As a preferred implementation, the data preprocessing module performs multi-scale wavelet denoising preprocessing on the collected optical power data, and the specific steps are as follows:

[0060] Perform normalization processing on the collected optical power data, specifically:

[0061]

[0062] Where: is the optical power data after normalization processing; x i is the original optical power data; x max 、x min are the maximum and minimum values in the original optical power data respectively;

[0063] Use the improved SureShrink algorithm to perform wavelet decomposition on the normalized optical power data to obtain the coefficients d of each layer of wavelet decomposition j,k , and perform threshold denoising, specifically:

[0064]

[0065] Among them, the adjustment formula of the dynamic threshold λ j is:

[0066]

[0067] Where: is the wavelet coefficient after denoising; d j,k is the k-th wavelet coefficient of the j-th layer; λ j is the dynamic threshold of the j-th layer; σ j is the noise standard deviation of the j-th layer; N j is the number of data in the j-th layer;

[0068] Reconstruct the wavelet coefficients after denoising to obtain the optical power data after denoising, specifically:

[0069]

[0070] Where: is the optical power data after denoising; is a wavelet basis function; τ is the reconstruction time point.

[0071] As a preferred embodiment, the dynamic threshold in the early warning module calculates the mean and standard deviation of historical data based on a sliding window, and dynamically updates the threshold. The calculation formula is:

[0072] θ dynamic = μ window + β·σ window ;

[0073] In the formula: θ dynamic is the dynamic threshold; μ window is the mean optical power within the sliding window; β is the safety factor; σ window is the standard deviation of the optical power within the sliding window.

[0074] On the other hand, the present invention also provides an electronic device, on which a computer program is stored. When the computer program is executed by a processor, it implements a method for online status monitoring and early warning of power optical fibers as described in any embodiment of the present invention.

[0075] On the other hand, the present invention also provides a computer-readable medium for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement a method for online status monitoring and early warning of power optical fibers as described in any embodiment of the present invention.

[0076] The present invention has the following beneficial effects:

[0077] 1. By using a clamping coupler to extract the leaked optical signal, it avoids the interference of traditional invasive detection on optical fiber transmission;

[0078] 2. By improving the multi-scale wavelet denoising algorithm, it accurately filters out environmental noise, and the signal-to-noise ratio is increased by more than 40%;

[0079] 3. The SA algorithm globally searches for the optimal parameter combination of LSTM, and the prediction error is significantly reduced compared with traditional LSTM;

[0080] 4. By combining the time attention mechanism to locate the deterioration node, it reduces the path switching delay and ensures the continuity of power communication. Description of the Drawings

[0081] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required to be used in the embodiments of the present application. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0082] Figure 1Schematic diagram of the method flow of the present invention;

[0083] Figure 2 Schematic diagram for improving the prediction effect of SA-LSTM;

[0084] Figure 3 Schematic diagram of the prediction effect of traditional LSTM. Specific implementation manners

[0085] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0086] It should be understood that the step numbers used in the text are only for convenient description and do not limit the execution order of the steps.

[0087] It should be understood that the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0088] The terms "comprising" and "including" indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their combinations.

[0089] The term "and / or" refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0090] Embodiment 1:

[0091] To make the purpose, technical solutions and advantages of the present invention clearer, the following will combine specific embodiments of the present application and refer to the attached Figure 1 , and clearly and completely describe the technical solutions of the present invention.

[0092] To solve the problems of the prior art, the present invention provides a method for on-line status monitoring and early warning of power optical fibers, including the following steps:

[0093] Step S1, collecting the optical power data of the power optical fiber by using a non-intrusive status monitoring module;

[0094] The non-invasive status monitoring module includes: an optical signal non-invasive detection unit for bending an optical fiber through a clamping coupler and extracting a leaked optical signal; a photoelectric conversion unit for converting the optical signal into an electrical signal using an avalanche diode; and a data regeneration unit for amplifying and shaping the electrical signal and then outputting it.

[0095] Step S2, perform multi-scale wavelet denoising preprocessing on the collected optical power data;

[0096] The original optical power data contains high-frequency noise, which affects the convergence of the SA-LSTM model. Multi-scale wavelet denoising preprocessing is used to filter out the high-frequency noise.

[0097] Step S201, perform normalization processing on the collected optical power data, specifically:

[0098]

[0099] In the formula: is the optical power data after normalization processing; x i is the original optical power data; x max , x min are respectively the maximum and minimum values in the original optical power data;

[0100] Step S202, perform wavelet decomposition on the normalized optical power data using an improved SureShrink algorithm to obtain the coefficients d j,k of each layer of wavelet decomposition, and perform threshold denoising, specifically:

[0101]

[0102] where the formula for adjusting the dynamic threshold λ j of the SureShrink algorithm is:

[0103]

[0104] Among them:

[0105]

[0106] In the formula: is the wavelet coefficient after denoising; d j,k is the k-th wavelet coefficient of the j-th layer; λ j is the dynamic threshold of the j-th layer; σ j is the noise standard deviation of the j-th layer; N j is the number of data in the j-th layer; MAD is the median absolute deviation of the coefficients in the highest-frequency subband;

[0107] Reconstruct the wavelet coefficients after denoising to obtain the optical power data after denoising, specifically:

[0108]

[0109] In the formula: is the optical power data after denoising; is the wavelet basis function; τ is the reconstruction time point.

[0110] Step S3, use the preprocessed optical power data as the input of the trained simulated annealing-long short-term memory SA-LSTM prediction model to predict the optical power;

[0111] The simulated annealing-long short-term memory SA-LSTM prediction model optimizes the number of iterations, the number of hidden units, and the learning rate in the long short-term memory LSTM model through the simulated annealing algorithm SA. The specific steps are as follows:

[0112] Define the parameter search space and the energy function: Set the value ranges of the parameters to be optimized, namely the number of iterations, the number of hidden units, and the learning rate. The energy function is the mean square error of the validation set of the long short-term memory LSTM model. Specifically:

[0113]

[0114] In the formula: E(S) is the energy function; N val is the number of samples in the validation set; y n is the actual optical power; is the optical power predicted by the LSTM model;

[0115] Initialize the simulated annealing parameters: Set the initial temperature, the termination temperature, the cooling coefficient, the maximum number of iterations, and the neighborhood perturbation step size;

[0116] Generate the initial solution and calculate the energy: Generate a random parameter combination and train the LSTM model to calculate the initial energy;

[0117] Iteratively optimize until the termination condition is met and output the optimal parameter combination, where:

[0118] The formula for generating the neighborhood solution is:

[0119] S ′ =(e ′ , h ′ , η ′ );

[0120] Among them:

[0121] e ′ =Clip(e current +Δe, e min , e max );

[0122] h ′ =Clip(hcurrent +Δh,h min ,h max );

[0123] η ′ =Clip(η current ×(1+Δη),η min ,η max );

[0124] Where: S ′ is the neighborhood solution; e ′ , h ′ and η ′ are the iteration times of the parameter combination, the number of hidden units, and the learning rate of the neighborhood solution respectively; Clip is to limit the parameters of the neighborhood solution within the upper and lower limit intervals; e current , h current and η current are the iteration times, the number of hidden units, and the learning rate in the iterative state respectively; Δe, Δh, and Δη are the perturbation parameters for generating the neighborhood solution of the iteration times, the number of hidden units, and the learning rate respectively; e min , e max , h min , h max , η min , η max are the value ranges of the iteration times, the number of hidden units, and the learning rate respectively;

[0125] The energy difference calculation formula is:

[0126] ΔE = E(S ′ ) - E(S);

[0127] Where: ΔE is the energy difference between the new solution and the current solution; if ΔE < 0, accept the neighborhood solution S ′ , otherwise, accept the neighborhood solution S with a probability of ′ , T is the current temperature;

[0128] The cooling strategy during iteration is:

[0129] T t+1 = α·T t ;

[0130] Where: α is the cooling coefficient; T t is the temperature at the t-th iteration;

[0131] The global optimal solution is:

[0132] S best = argmin(E(S));

[0133] Where: S bestis the global optimal solution; when the maximum number of iterations is reached or the temperature reaches the termination temperature, the global optimal solution is output.

[0134] A time attention module is added before the LSTM layer of the simulated annealing-long short-term memory SA-LSTM prediction model to solve the problem of insufficient ability of traditional LSTM to capture key features of long sequences and strengthen the feature expression of key detection points. The weight calculation formula of the time attention module is:

[0135]

[0136] In the formula: w t is the importance weight of the hidden state at the t-th time step for the current prediction; v T is the transpose of the trainable parameter vector v; W h is the trainable weight matrix; h t is the hidden state of the LSTM at the t-th time step; b h is the trainable bias term; h r is the hidden state of the LSTM at the r-th time step; tanh is the hyperbolic tangent activation function; R is the total number of time steps of the input sequence.

[0137] In this embodiment: after optimizing the parameters of the long short-term memory model through simulated annealing, the number of iterations, the number of hidden units, and the learning rate are respectively selected as 289, 51, and 0.02. And after adding the time attention module, the prediction effect of the model is as Figure 2 shown; the prediction results of the LSTM without adding the time attention module and without parameter optimization are as Figure 3 shown. By comparing Figure 2 Figure 3 , it can be seen that the improved SA-LSTM prediction model can effectively improve the performance in the optical power prediction task and significantly reduce the prediction error.

[0138] Step S4, when the optical power prediction value exceeds the dynamic threshold, an optical fiber deterioration warning is issued and the optical fiber path is automatically switched.

[0139] The fixed threshold cannot adapt to the baseline drift caused by environmental changes. The dynamic threshold calculates the mean and standard deviation of historical data based on a sliding window and dynamically updates the threshold. The calculation formula is:

[0140] θ dynamic = μ window + β·σ window ;

[0141] In the formula: θ dynamic is the dynamic threshold; μ window is the mean optical power within the sliding window; β is the safety factor, and its value range is 2.0 - 3.0; σ window is the standard deviation of the optical power within the sliding window.

[0142] When the predicted value exceeds the dynamic threshold, an early warning of optical fiber deterioration is carried out, and the automatic switching of the optical fiber path is started when the limit is exceeded three times in a row.

[0143] Embodiment 2:

[0144] This embodiment provides an on-line state monitoring and early warning system for power optical fibers, including:

[0145] A non-invasive state monitoring module for collecting the optical power data of the power optical fiber;

[0146] A data preprocessing module for performing multi-scale wavelet denoising preprocessing on the collected optical power data;

[0147] A prediction module that takes the preprocessed optical power data as the input of a trained simulated annealing-long short-term memory SA-LSTM prediction model to predict the optical power;

[0148] An early warning module that issues an early warning of optical fiber deterioration and automatically switches the optical fiber path when the optical power prediction value exceeds the dynamic threshold.

[0149] Embodiment 3:

[0150] This embodiment provides an electronic device with a computer program stored thereon, and when the computer program is executed by a processor, it implements a method for on-line state monitoring and early warning of power optical fibers as described in any embodiment of the present invention.

[0151] Embodiment 4:

[0152] This embodiment provides a computer-readable medium for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement a method for on-line state monitoring and early warning of power optical fibers as described in any embodiment of the present invention.

[0153] In the embodiments of the present application, "at least one" means one or more, and "a plurality" means two or more. "And / or" describes the association relationship of associated objects and indicates that three relationships may exist. For example, A and / or B may represent the situation where A exists alone, A and B exist simultaneously, or B exists alone. Where A and B may be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one of the following" and its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, and c may represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c may be single or multiple.

[0154] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of both. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application.

[0155] Those skilled in the art can clearly understand that for the sake of convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0156] In several embodiments provided by this application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM for short), random access memories (RAM for short), magnetic disks, or optical discs that can store program codes.

[0157] The above are only the embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A method for monitoring and early warning of online status of power optical fiber, characterized in that: The following steps are involved: Step S1, using a non-intrusive status monitoring module to collect optical power data of the power optical fiber; Step S2, performing multi-scale wavelet denoising preprocessing on the collected optical power data; Step S3, using the preprocessed optical power data as input to the trained simulated annealing-long short-term memory SA-LSTM prediction model to predict the optical power; Step S4: When the predicted optical power value exceeds the dynamic threshold, an optical fiber degradation warning is issued and the optical fiber path is automatically switched.

2. The method for monitoring and early warning of online status of electric optical fiber according to claim 1, characterized in that: The multi-scale wavelet denoising preprocessing of the collected optical power data is performed in the following specific steps: The collected optical power data is normalized as follows: Where: is the normalized optical power data; x i is the original optical power data; x max 、x min are the maximum and minimum values ​​in the original optical power data respectively; The improved SureShrink algorithm is used to perform wavelet decomposition on the normalized optical power data to obtain the wavelet decomposition coefficients d j,k , and perform threshold denoising, specifically: The dynamic threshold λ j The adjustment formula is: Where: is the wavelet coefficient after denoising; d j,k is the kth wavelet coefficient of the jth layer; j is the dynamic threshold of the jth layer; σ j is the noise standard deviation of the jth layer; N j is the number of data in the jth layer; The denoised wavelet coefficients are reconstructed to obtain the denoised optical power data, specifically: Where: is the denoised optical power data; is the wavelet basis function; τ is the reconstruction time point.

3. The method for online status monitoring and early warning of power optical fiber according to claim 1, characterized in that: The simulated annealing-long short-term memory SA-LSTM prediction model optimizes the number of iterations, the number of hidden units and the learning rate in the long short-term memory LSTM model through the simulated annealing algorithm SA. The specific steps are: Define the parameter search space and energy function: set the number of iterations of the parameters to be optimized, the number of hidden units, and the range of the learning rate. The energy function is the mean square error of the validation set of the long short-term memory LSTM model, specifically: Where: E(S) is the energy function; N val is the number of samples in the validation set; y n is the actual optical power; Predict optical power for the LSTM model; Initialize simulated annealing parameters: set initial temperature, end temperature, cooling coefficient, maximum number of iterations and neighborhood perturbation step size; Generate initial solution and calculate energy: Generate random parameter combinations and train the LSTM model to calculate initial energy; Iterate the optimization until the termination condition is met and output the optimal parameter combination, where: The formula for generating the neighborhood solution is: S′=(e′,h′,η′); in: e′=Clip(e current +Δe,e min ,And max ); h′=Clip(h current +Δh,h min ,h max ); η′=Clip(η current ×(1+Dη),η min ,or max ); Where: S′ is the neighborhood solution; e′, h′ and η′ are the number of iterations of the parameter combination of the neighborhood solution, the number of hidden units and the learning rate respectively; Clip is to limit the parameters of the neighborhood solution to the upper and lower limits; e current 、h current and η current are the number of iterations, number of hidden units and learning rate in the iteration state; Δe, Δh and Δη are the perturbation parameters for the number of iterations, number of hidden units and learning rate to generate neighborhood solutions respectively; e min 、e max 、h min 、h max , η min , η max are the ranges of the number of iterations, the number of hidden units, and the learning rate, respectively; The energy difference calculation formula is: ΔE=E(S′)-E(S); Where: ΔE is the energy difference between the new solution and the current solution; if ΔE<0, accept the neighborhood solution S ′ , otherwise, with probability Accept the neighborhood solution S ′ , T is the current temperature; The cooling strategy during the iteration process is: T t+1 =α·T t ; Where: α is the temperature drop coefficient; T t is the temperature at the tth iteration; The global optimal solution is: S best =argmin(E(S)); Where: S best is the global optimal solution; when the maximum number of iterations is reached or the temperature reaches the termination temperature, the global optimal solution is output.

4. The method for monitoring and early warning of online status of electric optical fiber according to claim 1, characterized in that: A time attention module is added before the LSTM layer of the simulated annealing-long short-term memory SA-LSTM prediction model. The weight calculation formula of the time attention module is: Where: w t is the importance weight of the hidden state at the tth time step to the current prediction; v T is the transpose of the trainable parameter vector v; W h is the trainable weight matrix; h t is the hidden state of LSTM at the tth time step; b h is a trainable bias term; h r is the hidden state of LSTM at the rth time step; tanh is the hyperbolic tangent activation function; R is the total time step length of the input sequence.

5. The method for monitoring and early warning of online status of electric optical fiber according to claim 1, characterized in that: The dynamic threshold is calculated based on the sliding window to calculate the mean and standard deviation of historical data, and the threshold is dynamically updated. The calculation formula is: i dynamic =μ window +b·s window ; Where: θ dynamic is the dynamic threshold; μ window is the mean optical power in the sliding window; β is the safety factor; σ window is the standard deviation of optical power in the sliding window.

6. A power optical fiber online status monitoring and early warning system, characterized in that: include: Non-intrusive status monitoring module for collecting optical power data of power optical fiber; The data preprocessing module performs multi-scale wavelet denoising preprocessing on the collected optical power data; The prediction module uses the preprocessed optical power data as the input of the trained simulated annealing-long short-term memory SA-LSTM prediction model to predict the optical power; The early warning module issues an early warning of fiber degradation and automatically switches the fiber path when the predicted optical power value exceeds the dynamic threshold.

7. The power optical fiber online status monitoring and early warning system according to claim 6, characterized in that: The data preprocessing module performs multi-scale wavelet denoising preprocessing on the collected optical power data. The specific steps are as follows: The collected optical power data is normalized as follows: Where: is the normalized optical power data; x i is the original optical power data; x max 、x min are the maximum and minimum values ​​in the original optical power data respectively; The improved SureShrink algorithm is used to perform wavelet decomposition on the normalized optical power data to obtain the wavelet decomposition coefficients d j,k , and perform threshold denoising, specifically: The dynamic threshold λ j The adjustment formula is: Where: is the wavelet coefficient after denoising; d j,k is the kth wavelet coefficient of the jth layer; j is the dynamic threshold of the jth layer; σ j is the noise standard deviation of the jth layer; N j is the number of data in the jth layer; The denoised wavelet coefficients are reconstructed to obtain the denoised optical power data, specifically: Where: is the denoised optical power data; is the wavelet basis function; τ is the reconstruction time point.

8. The power optical fiber online status monitoring and early warning system according to claim 6, characterized in that: The dynamic threshold in the early warning module calculates the mean and standard deviation of historical data based on the sliding window, and dynamically updates the threshold. The calculation formula is: i dynamic =μ window +b·s window ; Where: θ dtnamic is the dynamic threshold; μ window is the mean optical power in the sliding window; β is the safety factor; σ window is the standard deviation of optical power in the sliding window.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, an electric fiber online status monitoring and early warning method as described in any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, an electric optical fiber online status monitoring and early warning method as claimed in any one of claims 1 to 5 is implemented.

Images