Evaluation method and device for communication power supply system

Abstract

The invention discloses an evaluation method and device for a communication power supply system, and relates to the technical field of communication, and the method comprises the steps: collecting dynamic operation data, static configuration parameters and historical maintenance records in real time through a multi-source heterogeneous data collection module; constructing an equipment-level, system-level and subsystem-level three-layer risk assessment architecture; calculating a subjective weight by adopting an improved fuzzy analytic hierarchy process; deploying an LSTM-GAN hybrid model to generate equipment degradation simulation data; establishing a deep residual convolutional network; constructing a dynamic risk propagation matrix; establishing a virtual mirror image system based on a digital twinning technology; generating a three-dimensional visual risk thermodynamic diagram; outputting a customized maintenance scheme; and updating the equipment health archive through the block chain smart contract. Therefore, multi-dimensional indexes such as the voltage ripple, the temperature rise rate and the equipment aging curve are integrated through a multi-source data fusion framework, information islands are eliminated, and therefore the dynamic information stability under the complex working condition is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to an evaluation method and device for a communication power supply system. Background Art

[0002] With the continuous development of the electric power communication network, the number of communication equipment has continued to increase, which has put a great test on the power supply capacity of the communication power supply system. In addition, as high-power consumption equipment is put into operation, the power consumption of the equipment changes dramatically, which has a great impact on the safe operation of the communication power supply system.

[0003] However, traditional communication power supply assessment methods are mostly based on single parameter threshold judgment or static model analysis, which is difficult to adapt to the dynamic risk quantification needs under complex working conditions. Summary of the Invention

[0004] In view of the above defects, the purpose of the present invention is to provide a method and device for evaluating a communication power supply system, aiming to solve the problem that the existing technology is difficult to adapt to dynamic risk quantification under complex working conditions.

[0005] In order to solve the above technical problems, the technical solution of the present invention is: A method for evaluating a communication power supply system, comprising: S1. Real-time acquisition of dynamic operation data, static configuration parameters, and historical maintenance records through a multi-source heterogeneous data acquisition module. The dynamic operation data includes voltage fluctuation waveforms, current harmonic spectrum temperature gradient distribution; the static configuration parameters include equipment model code, design life index, and topology connection relationship; S2. Construct a three-tiered risk assessment framework at the device, system, and subsystem levels. The device level is used to assess the efficiency decay rate of rectifier modules, the system level is used to assess the robustness of network topology, and the subsystem level is used to assess the stability of power supply buses. S3. Use the improved fuzzy analytic hierarchy process to calculate subjective weights, combine it with the entropy weight method to calculate objective weights, and achieve dynamic weight fusion through Bayesian reasoning; S4. Deploy an LSTM-GAN hybrid model to generate simulated device degradation data. The generator uses a bidirectional gated recurrent unit, and the discriminator integrates a self-attention mechanism. S5. Build a deep residual convolutional network. The first branch processes the time-domain characteristics of the electrical signal, the second branch analyzes the thermodynamic spatial distribution, and the third branch extracts the frequency-domain characteristics of the vibration signal. S6. Construct a dynamic risk propagation matrix to quantify the Markov transition probability of the impact of single-point failures on the system level; S7. Establish a virtual mirror system based on digital twin technology to achieve online parameter calibration of risk assessment models; S8. Generate a 3D visual risk heat map, marking red warning areas where the failure probability exceeds the threshold. S9. Output customized maintenance plans and automatically generate work orders including spare parts lists, operation sequences, and safety procedures. S10. Update the device health profile through blockchain smart contracts and record an unalterable assessment log.

[0006] Among them, in S1, the data collected by the multi-source heterogeneous data acquisition module also includes: a wide-band current sensor to obtain 0-100kHz spectrum data with a sampling rate of no less than 200kS / s; a distributed fiber optic temperature measurement system with a spatial resolution of 1cm and a temperature accuracy of ±0.5°C; a three-axis MEMS vibration sensor with a frequency range of 5Hz-10kHz and a dynamic range of ≥80dB.

[0007] The improved fuzzy analytic hierarchy process described in S3 includes: a. Establish a triangular fuzzy judgment matrix ,in , which represents the fuzzy importance of index i to index j, satisfying ,and ; b. Calculate the fuzzy weight of each indicator: ; c. Defuzzify using α-cut and calculate the weights using the cutoff interval when α = 0.8.

[0008] Among them, in S3, the entropy weight method calculation includes: constructing a decision matrix for m samples and n indicators ; Calculate the normalized matrix ; Calculate information entropy in ; Determine objective weights .

[0009] Among them, in S4, the LSTM-GAN model includes: the generator G adopts a bidirectional GRU structure, the number of hidden layer units is set to 128, and the input noise vector With conditional tags ; The discriminator D contains 5 causal convolution layers, and the convolution kernel width increases exponentially.

[0010] Among them, in S5, the deep residual convolutional network includes: the electrical signal branch adopts a void convolution structure to capture multi-scale time domain features; the thermodynamic branch deploys a 3D convolution kernel to extract the spatial gradient of the temperature field; the vibration signal branch applies wavelet packet transform, with the decomposition layer number reaching 8 layers, and extracts 32 frequency band energy features.

[0011] Among them, in S6, constructing a dynamic risk propagation matrix includes: defining a system state vector; constructing a state transfer matrix; and calculating a system risk index.

[0012] Among them, in S7, the virtual mirror system includes: physical entity layer, virtual model layer, service application layer, and data connection layer; the virtual model layer runs a high-fidelity simulation model and updates the time step ; The high-fidelity simulation model includes rectifier module loss calculation, battery degradation model, and cable aging model.

[0013] Among them, in S8, the three-dimensional visualization risk heat map includes: three-dimensional coordinate system, color coding and interactive functions.

[0014] Among them, an evaluation device for a communication power supply system includes: a multi-source data acquisition unit, an edge computing unit, a risk assessment engine, a visualization terminal, and a self-maintenance actuator; the multi-source data acquisition unit includes a Rogowski coil structure, a scorer circuit, and a self-calibration module; the edge computing unit includes a real-time signal processor, a data buffer component, and a communication interface.

[0015] After adopting the above technical solution, the beneficial effects of the present invention are: First, a multi-source data fusion architecture integrates multi-dimensional indicators such as voltage ripple, temperature rise rate, and equipment aging curves, eliminating information silos and ensuring dynamic information stability under complex operating conditions. Second, a dynamic weight fusion matrix adjusts evaluation parameters in real time to accommodate nonlinear aging characteristics such as changes in battery internal resistance. Third, an LSTM-GAN model generates an extended dataset and uses a generative adversarial network to synthesize rare fault samples (such as transient overvoltage waveforms), improving model recognition accuracy. DETAILED DESCRIPTION

[0016] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] A method for evaluating a communication power supply system, comprising: S1. Real-time acquisition of dynamic operation data, static configuration parameters, and historical maintenance records through a multi-source heterogeneous data acquisition module. The dynamic operation data includes voltage fluctuation waveforms, current harmonic spectrum temperature gradient distribution; the static configuration parameters include equipment model code, design life index, and topology connection relationship; S2. Construct a three-tiered risk assessment framework at the device, system, and subsystem levels. The device level is used to assess the efficiency decay rate of rectifier modules, the system level is used to assess the robustness of network topology, and the subsystem level is used to assess the stability of power supply buses. S3. Use the improved fuzzy analytic hierarchy process to calculate subjective weights, combine it with the entropy weight method to calculate objective weights, and achieve dynamic weight fusion through Bayesian reasoning; S4. Deploy an LSTM-GAN hybrid model to generate simulated device degradation data. The generator uses a bidirectional gated recurrent unit, and the discriminator integrates a self-attention mechanism. S5. Build a deep residual convolutional network. The first branch processes the time-domain characteristics of the electrical signal, the second branch analyzes the thermodynamic spatial distribution, and the third branch extracts the frequency-domain characteristics of the vibration signal. S6. Construct a dynamic risk propagation matrix to quantify the Markov transition probability of the impact of single-point failures on the system level; S7. Establish a virtual mirror system based on digital twin technology to achieve online parameter calibration of risk assessment models; S8. Generate a 3D visual risk heat map, marking red warning areas where the failure probability exceeds the threshold. S9. Output customized maintenance plans and automatically generate work orders including spare parts lists, operation sequences, and safety procedures. S10. Update the device health profile through blockchain smart contracts and record an unalterable assessment log.

[0018] Through the multi-source data fusion architecture, multi-dimensional indicators such as voltage ripple, temperature rise rate, equipment aging curve, etc. are integrated to eliminate information islands, thereby ensuring the stability of dynamic information under complex working conditions.

[0019] Through the dynamic weight fusion matrix, the evaluation parameters are adjusted in real time to adapt to nonlinear aging characteristics such as the resistance change inside the battery.

[0020] The LSTM-GAN model generates an extended data set and synthesizes rare fault samples (such as instantaneous overvoltage waveforms) through a generative adversarial network, thereby improving the model's recognition accuracy.

[0021] The dynamic risk assessment matrix supports three levels of warning (RPN ≥ 120 triggers a red alert), the fault warning time is shortened from 24 hours to 72 hours, and the incidence of major accidents is reduced by 60%.

[0022] In S1, the data collected by the multi-source heterogeneous data acquisition module also includes: a wide-band current sensor that obtains 0-100kHz spectrum data with a sampling rate of no less than 200kS / s; a distributed fiber optic temperature measurement system with a spatial resolution of 1cm and a temperature accuracy of ±0.5°C; and a three-axis MEMS vibration sensor with a frequency range of 5Hz-10kHz and a dynamic range of ≥80dB.

[0023] In S3, the improved fuzzy analytic hierarchy process includes: a. Establish a triangular fuzzy judgment matrix ,in , which represents the fuzzy importance of index i to index j, satisfying ,and ; b. Calculate the fuzzy weight of each indicator: ; c. Defuzzify using α-cut and calculate the weights using the cutoff interval when α = 0.8.

[0024] Among them, in S3, the entropy weight method calculation includes: constructing a decision matrix for m samples and n indicators ; Calculate the normalized matrix ; Calculate information entropy in ; Determine objective weights .

[0025] In S4, the LSTM-GAN model includes: the generator G adopts a bidirectional GRU structure, the number of hidden layer units is set to 128, and the input noise vector With conditional tags The discriminator D consists of five causal convolutional layers, with kernel widths increasing exponentially (3, 7, 15, 31, 63). The Wasserstein distance is used to optimize the objective function: ,in, The gradient penalty term constrains the Lipschitz constant. By constraining the discriminator gradient modulus, the instability problem of traditional GAN ​​training is solved.

[0026] In S5, the deep residual convolutional network includes: the electrical signal branch adopts a void convolution structure, with the expansion coefficient progressively increasing from [1, 2, 4, 8] to capture multi-scale time domain features; the thermodynamic branch deploys a 3D convolution kernel with a size of 5×5×5 to extract the spatial gradient of the temperature field; the vibration signal branch applies wavelet packet transform with 8 decomposition layers to extract 32 frequency band energy features.

[0027] In S6, a dynamic risk propagation matrix is ​​constructed, including: defining the system state vector ,in, Identify the reliability function of device i; Constructing the state transition matrix ,in, ; Calculating the system risk index , where C(i) identifies the set of nodes associated with device i.

[0028] In S7, the virtual mirror system includes: a physical entity layer for deploying edge computing nodes, performing data preprocessing and feature extraction; a virtual model layer for running high-fidelity simulation models and updating time steps. The service application layer provides a RESTful API interface to support risk assessment service calls. The data connection layer uses the OPC UA protocol to achieve physical-virtual layer data synchronization.

[0029] The high-fidelity simulation model includes rectifier module loss calculation: ; Among them, k=0.8 is the empirical coefficient.

[0030] Battery degradation model: capacity decay rate , where α=2.3×10 -6 , β=0.056, γ=1.2; Cable Aging Model: Insulation Resistance , k=3.2×10 -4 .

[0031] In S8, the three-dimensional visual risk heat map includes: a three-dimensional coordinate system, with the X-axis representing the physical location of the equipment, the Y-axis representing the risk level, and the Za-axis showing the fault propagation path; color coding, red (R>0.8), orange (0.6≤R<0.6), and green (R<0.4); interactive functions, supporting gesture control to rotate the viewing angle, and calling up the life prediction curve of motor equipment.

[0032] An evaluation device for a communication power supply system includes: a multi-source data acquisition unit, an edge computing unit, a risk assessment engine, a visualization terminal, and a self-maintenance actuator; the multi-source data acquisition unit includes a Rogowski coil structure, a scorer circuit, and a self-calibration module; the edge computing unit includes a real-time signal processor, a data buffer component, and a communication interface.

[0033] The multi-source data acquisition unit includes an integrated broadband current sensor, an infrared thermal imager, and an acoustic radio frequency detection module. The integrated broadband current sensor includes a Rogowski coil structure, a scorer circuit, and a self-calibration module; an edge computing unit is used for data preprocessing; a risk assessment engine is used to run LSTM-GAN; a visualization terminal is used to display various information and provide an interactive module; and a self-maintenance actuator includes an intelligent circuit breaker, a programmable load box, and a robotic inspection unit.

[0034] During operation, sensor zero drift detection is performed daily, with an allowable deviation of ≤±0.5%FS; model accuracy verification is performed weekly, requiring F1-score ≥0.92; and equipment health status reports are uploaded to the supervision platform every month.

[0035] In summary, this solution offers the following advantages: First, a multi-source data fusion architecture integrates multi-dimensional indicators such as voltage ripple, temperature rise rate, and equipment aging curves, eliminating information silos and ensuring dynamic information stability under complex operating conditions. Second, a dynamic weighted fusion matrix allows for real-time adjustment of evaluation parameters to accommodate nonlinear aging characteristics such as changes in battery internal resistance. Third, the LSTM-GAN model generates an extended dataset and uses a generative adversarial network to synthesize rare fault samples (such as transient overvoltage waveforms), improving model recognition accuracy.

[0036] The present invention is not limited to the above-mentioned specific implementation methods. Various changes made by ordinary technicians in this field based on the above-mentioned concept without creative work are all within the scope of protection of the present invention.

Claims

1. A method for evaluating a communication power supply system, characterized in that: include: S1. Real-time acquisition of dynamic operation data, static configuration parameters, and historical maintenance records through a multi-source heterogeneous data acquisition module. The dynamic operation data includes voltage fluctuation waveforms, current harmonic spectrum temperature gradient distribution; the static configuration parameters include equipment model code, design life index, and topology connection relationship; S2. Construct a three-tiered risk assessment framework at the device, system, and subsystem levels. The device level is used to assess the efficiency decay rate of rectifier modules, the system level is used to assess the robustness of network topology, and the subsystem level is used to assess the stability of power supply buses. S3. Use the improved fuzzy analytic hierarchy process to calculate subjective weights, combine it with the entropy weight method to calculate objective weights, and achieve dynamic weight fusion through Bayesian reasoning; S4. Deploy an LSTM-GAN hybrid model to generate simulated device degradation data. The generator uses a bidirectional gated recurrent unit, and the discriminator integrates a self-attention mechanism. S5. Build a deep residual convolutional network. The first branch processes the time-domain characteristics of the electrical signal, the second branch analyzes the thermodynamic spatial distribution, and the third branch extracts the frequency-domain characteristics of the vibration signal. S6. Construct a dynamic risk propagation matrix to quantify the Markov transition probability of the impact of single-point failures on the system level; S7. Establish a virtual mirror system based on digital twin technology to achieve online parameter calibration of risk assessment models; S8. Generate a 3D visual risk heat map, marking red warning areas where the failure probability exceeds the threshold. S9. Output customized maintenance plans and automatically generate work orders including spare parts lists, operation sequences, and safety procedures. S10. Update the device health profile through blockchain smart contracts and record an unalterable assessment log.

2. The evaluation method of a communication power supply system according to claim 1, characterized in that: In S1, the data collected by the multi-source heterogeneous data acquisition module also includes: a wide-band current sensor that obtains 0-100kHz spectrum data with a sampling rate of no less than 200kS / s; a distributed fiber optic temperature measurement system with a spatial resolution of 1cm and a temperature accuracy of ±0.5°C; and a three-axis MEMS vibration sensor with a frequency range of 5Hz-10kHz and a dynamic range of ≥80dB.

3. The evaluation method of a communication power supply system according to claim 1, wherein: The improved fuzzy analytic hierarchy process described in S3 includes: a. Establish a triangular fuzzy judgment matrix ,in , which represents the fuzzy importance of index i to index j, satisfying ,and ; b. Calculate the fuzzy weight of each indicator: ; c. Defuzzify using α-cut and calculate the weights using the cutoff interval when α = 0.

8.

4. The evaluation method of a communication power supply system according to claim 3, characterized in that: In S3, the entropy weight method calculation includes: constructing a decision matrix for m samples and n indicators ; Calculate the normalized matrix ; Calculate information entropy in ; Determine objective weights .

5. The evaluation method for a communication power supply system according to claim 4, wherein: In S4, the LSTM-GAN model includes: the generator G adopts a bidirectional GRU structure, the number of hidden layer units is set to 128, and the input noise vector With conditional tags ; The discriminator D contains 5 causal convolution layers, and the convolution kernel width increases exponentially.

6. The evaluation method of a communication power supply system according to claim 1, wherein: In S5, the deep residual convolutional network includes: the electrical signal branch adopts a void convolution structure to capture multi-scale time domain features; the thermodynamic branch deploys a 3D convolution kernel to extract the spatial gradient of the temperature field; the vibration signal branch applies wavelet packet transform, with the decomposition layer number reaching 8 layers, and extracts 32 frequency band energy features.

7. The evaluation method for a communication power supply system according to claim 1, wherein: In S6, constructing a dynamic risk propagation matrix includes: defining a system state vector; constructing a state transfer matrix; and calculating a system risk index.

8. The evaluation method of a communication power supply system according to claim 1, wherein: In S7, the virtual mirror system includes: physical entity layer, virtual model layer, service application layer, and data connection layer; the virtual model layer runs a high-fidelity simulation model and updates the time step. ; The high-fidelity simulation model includes rectifier module loss calculation, battery degradation model, and cable aging model.

9. The evaluation method of a communication power supply system according to claim 1, wherein: In S8, the three-dimensional visual risk heat map includes: a three-dimensional coordinate system, color coding, and interactive functions.

10. An evaluation device for a communication power supply system, characterized in that: include: A multi-source data acquisition unit, an edge computing unit, a risk assessment engine, a visualization terminal, and a self-maintenance actuator; the multi-source data acquisition unit includes a Rogowski coil structure, a scorer circuit, and a self-calibration module; the edge computing unit includes a real-time signal processor, a data buffer component, and a communication interface.

Images