Distribution network operation and maintenance management method and system based on multi-dimensional intelligent perception

Abstract

The invention relates to the technical field of distribution network operation and maintenance management, and discloses a distribution network operation and maintenance management method and system based on multi-dimensional intelligent perception, and the method comprises the steps: constructing a distributed cooperative sampling perception architecture, collecting distribution network operation and maintenance data through multiple types of sensors, and carrying out the space-time alignment compensation of the distribution network operation and maintenance data through combining with a digital twin space-time coordinate system; establishing an electromagnetic interference self-adaptive filtering architecture to filter interference data, performing equipment characteristic decoupling of equipment type adaptation, performing decoding at a cloud side after encoding, and reconstructing an equipment core operation state; constructing a three-dimensional topological twin body of the distribution network, establishing a health assessment model to predict a degradation track, calculating a fault risk entropy, and substituting the fault risk entropy into a corrected life prediction result; constructing an operation and maintenance decision space in combination with the fault risk entropy and the distribution network three-dimensional topological twin state, and solving an optimal decision; and injecting the decision into the distribution network three-dimensional topology twinborn simulation verification security constraint, evaluating the fault risk entropy change, and feeding back and updating the health evaluation model and the decision engine.

Description

Distribution network operation and maintenance management methods and systems based on multi-dimensional intelligent sensing Technical Field

[0001] This invention relates to the field of distribution network operation and maintenance management technology, specifically to a distribution network operation and maintenance management method and system based on multi-dimensional intelligent perception. Background Technology

[0002] In the current technology, most intelligent operation and maintenance systems for power distribution networks are based on SCADA systems, combined with IoT sensors to achieve basic data collection, and use cloud-based centralized processing for fault diagnosis and operation and maintenance scheduling; some systems have introduced rudimentary technologies such as digital twins and edge computing, attempting to improve response speed through virtual mapping and local data processing, but have not yet formed a full-link collaborative system; in terms of data fusion, simple time-series alignment and filtering are mostly used, relying on traditional machine learning models for fault prediction;

[0003] Existing fusion algorithms fail to consider the unique load fluctuations and electromagnetic interference of distribution networks, resulting in high deviation rates in fusion results. The volatility of distributed power sources leads to frequent changes in power flow in distribution networks, causing traditional fault location algorithms to fail in multi-power source scenarios. Equipment failures caused by extreme weather require real-time linkage between meteorological warnings and equipment status, but existing patents often process meteorological and equipment data independently. Self-healing decisions for damaged network structures lack dynamic reconstruction and resource optimization capabilities. Distribution network operation and maintenance need to balance maintenance costs and power supply reliability, but existing patents lack big data analysis and resource optimization algorithms based on real-time status. The installation of high-voltage side sensors requires power outages, and low-voltage side Bluetooth Mesh solutions cannot be directly applied to high-voltage scenarios. Existing decision-making focuses primarily on reducing risk or controlling costs, neglecting power flow changes after the integration of distributed power sources and lacking pre-execution verification mechanisms.

[0004] Therefore, there is a need to provide a distribution network operation and maintenance management method and system based on multi-dimensional intelligent perception. Summary of the Invention

[0005] The purpose of this invention is to provide a distribution network operation and maintenance management method and system based on multi-dimensional intelligent sensing. To solve the above-mentioned problems in the prior art, this invention achieves this through the following technical solution:

[0006] The first part, the intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception provided in this invention, specifically includes the following steps:

[0007] Step 1: Construct a distributed collaborative sampling and sensing architecture, collect distribution network operation and maintenance data through multiple types of sensors, and combine the digital twin spatiotemporal coordinate system to perform spatiotemporal alignment compensation on the distribution network operation and maintenance data;

[0008] Step 2: Combine the aligned and compensated distribution network operation and maintenance data to establish an electromagnetic interference adaptive filtering architecture to filter interference data, and decouple the equipment features for equipment type adaptation. After encoding, decode and reconstruct the core operating status of the equipment in the cloud.

[0009] Step 3: Construct a three-dimensional topological twin of the distribution network, map equipment features to the virtual equipment model, combine dynamic transfer learning to establish a health assessment model to predict degradation trajectory, calculate fault risk entropy, and substitute it into the corrected lifetime prediction results;

[0010] Step 4: Construct an operation and maintenance decision space by combining the fault risk entropy and the state of the distribution network 3D topology twin, and solve for the optimal decision; inject the decision into the distribution network 3D topology twin simulation to verify the safety constraints, evaluate the changes in fault risk entropy, and update the health assessment model and decision engine.

[0011] Specifically, the spatiotemporal alignment compensation method is as follows:

[0012] By combining the digital twin spatiotemporal coordinate system and using the micron-level time scale of the synchronous phasor measurement device (PMU) as a reference, a global spatiotemporal coordinate system for the distribution network is established; automatic alignment of distribution network operation and maintenance data is achieved through a distributed collaborative sampling protocol, and the spatiotemporal alignment compensation time of the data is analyzed.

[0013] Based on the obtained data spatiotemporal alignment compensation time, the timestamp of the original data is corrected, and the original timestamp is added to the data spatiotemporal alignment compensation time to achieve time synchronization of electrical quantities, mechanical quantities and environmental quantities.

[0014] Specifically, the method for filtering interference data by the device is as follows:

[0015] Data on the power frequency magnetic field and high-frequency interference field strength around the equipment are collected and used as input for the interference suppression algorithm.

[0016] A two-stage processing mechanism combining physical filtering and neural compensation is employed to eliminate interference signals using the formula:

[0017]

[0018] Analysis yields the filtered interference data at time t. ,in, This is the physical filter function. The data collected by the sensor at time t. This is the original data. For the preset compensation coefficient, For Fourier transform, For the interference transfer function The square of the modulus, This is the interference transfer function used for training the neural network. As a regularization factor;

[0019] Specifically, the method for decoupling the device features is as follows:

[0020] After interference removal, feature decoupling is performed on the data to adapt to different device types, transmitting only key features, as shown in the following formula: The compressed feature vectors output by the edge nodes are obtained through analysis. ,in, Select a feature matrix for device type. For element-wise multiplication, It is the transpose of the wavelet envelope basis matrix. This is the original data matrix after removing interference;

[0021] The edge nodes perform Huffman coding compression on the decoupled feature vectors; the vectors are then transmitted to the cloud via dual 5G and fiber optic links. The cloud receives and decodes the vectors, and reconstructs the core operating status of the device by combining the compressed feature vectors.

[0022] Specifically, the method for predicting the degradation trajectory is as follows:

[0023] Construct a three-dimensional topological twin of the distribution network, map compressed feature vectors to virtual device models, and realize real-time linkage between physical devices and virtual models;

[0024] Embedded electrical field, temperature field and mechanical field coupling module to simulate the equipment operation status under distributed power access and extreme weather conditions. The coupling equation is based on the original method to ensure that the virtual model is consistent with the operating characteristics of the physical equipment.

[0025] To address the heterogeneity of different device models, a dynamic transfer learning framework is adopted to achieve cross-device adaptation of the health assessment model, and the health assessment value is obtained by analyzing the transfer learning loss function.

[0026] Specifically, the method for correcting the lifetime prediction results is as follows:

[0027] By combining the characteristics after migration and the cumulative degradation effect, the equipment failure risk entropy is calculated and used as a quantitative indicator of health status to analyze and obtain the failure risk entropy.

[0028] Substituting the failure risk entropy into the remaining life formula, the life prediction results are corrected to ensure the linkage between health assessment and life management, and the remaining health value is obtained through analysis.

[0029] Specifically, the method for solving the optimal decision is as follows:

[0030] By combining equipment failure risk entropy with the state of the twin model, an operation and maintenance decision space is constructed, which covers the following operation and maintenance solutions: in, For the k-th operation and maintenance decision, For operation and maintenance decision indexing, For maintenance time window, For the m-th maintenance team, For the nth fault isolation scheme, For the maintenance team index, Index of fault isolation schemes;

[0031] With the objectives of minimizing failure risk and minimizing total operation and maintenance cost, and taking into account equipment safety constraints, the optimal decision is solved.

[0032] Specifically, the method for verifying security constraints is as follows:

[0033] By combining a non-dominated sorting genetic algorithm and introducing distributed power source output constraints, the Pareto optimal decision set is found by optimizing the particle search range.

[0034] Each operational decision in the Pareto optimal decision set is injected into a digital twin to simulate the execution process;

[0035] Real-time monitoring of overvoltage, current, and equipment temperature in the simulation to determine whether the requirements are met. ,in, For divergence operators, For the electrical quantity field in digital twin simulation. Let Chebyshev norm be the divergence of electrical quantities in the simulation. Preset safety threshold;

[0036] Filter out the decisions that meet the constraints, sort them according to the priority of minimum risk entropy and minimum consumption cost, and output the final execution decision;

[0037] Specifically, the method for assessing the change in fault risk entropy is as follows:

[0038] After the operation is completed, the edge node re-collects the device parameters and calculates the change in fault risk entropy before and after the operation and maintenance. If the change in fault risk entropy is less than the preset entropy change threshold, the operation and maintenance effect is deemed to be qualified; otherwise, the operation and maintenance effect is deemed to be unqualified.

[0039] The evaluation results and operation and maintenance data are fed back to the health assessment model and decision engine to update the sample library for transfer learning and the parameters for decision optimization.

[0040] The second part, the intelligent operation and maintenance management system for distribution networks based on multi-dimensional perception provided in this embodiment of the invention, specifically includes the following units:

[0041] Sampling and sensing unit: Construct a distributed collaborative sampling and sensing architecture, collect distribution network operation and maintenance data through multiple types of sensors, and combine the digital twin spatiotemporal coordinate system to perform spatiotemporal alignment compensation on the distribution network operation and maintenance data;

[0042] Decoupling and Reconstruction Unit: Combining the aligned and compensated distribution network operation and maintenance data, an electromagnetic interference adaptive filtering architecture is established to filter interference data, and equipment features are decoupled to adapt to equipment types. After encoding, the core operating status of the equipment is decoded and reconstructed in the cloud.

[0043] Assessment and prediction unit: Construct a three-dimensional topological twin of the distribution network, map equipment features to a virtual equipment model, combine dynamic transfer learning to establish a health assessment model to predict degradation trajectory, calculate fault risk entropy, and substitute it into the corrected lifetime prediction results;

[0044] Verification Feedback Unit: Constructs an operation and maintenance decision space by combining fault risk entropy with the state of the distribution network's three-dimensional topology twin, and solves for the optimal decision; injects the decision into the simulation of the distribution network's three-dimensional topology twin to verify the safety constraints, evaluates the changes in fault risk entropy, and provides feedback to update the health assessment model and decision engine.

[0045] The beneficial effects of this invention are:

[0046] 1. Construct a distributed sensing architecture that coordinates multiple types of sensors. Combine the micron-level timescale of the PMU with the topological gradient matrix, and achieve spatiotemporal alignment compensation of multi-dimensional distribution network operation and maintenance data through a dedicated formula to solve the problem of data time consistency across devices. Design a two-level anti-interference mechanism that combines physical filtering and ResNet18 neural compensation, dynamically adjust the weight coefficients, and simultaneously match feature decoupling adapted to device type and 5G / fiber dual-link transmission to balance data fidelity and transmission efficiency. Create a digital twin with embedded electrical / temperature / mechanical field coupling modules. Combine a dynamic transfer learning framework based on Frobenius norm and KL divergence to solve the limitations of static twins and the generalization problem of heterogeneous device models. Propose a fault risk entropy model that integrates LSTM degradation prediction and economic adjustment coefficient to replace fixed threshold alarms. Solve the Pareto optimality of the operation and maintenance decision space through a non-dominated sorting genetic algorithm. Combine digital twin simulation to verify the safety of decision-making.

[0047] 2. In terms of spatiotemporal synchronization, the device topology gradient matrix, physical spacing, and signal propagation speed are combined to break through the traditional correction method that relies solely on communication delay and improve time synchronization accuracy. In terms of interference suppression, physical filtering and neural network compensation are integrated to dynamically adjust the regularization factor. In terms of model adaptation, dynamic transfer learning is applied to the health assessment of heterogeneous devices in the distribution network, and KL divergence is used to avoid model overfitting. In terms of operation and maintenance decision-making, digital twins are upgraded from display tools to decision verification platforms, and the balance between economic efficiency and safety is quantified by combining fault risk entropy. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 is a flowchart of the steps of the intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception provided in Embodiment 1 of the present invention;

[0050] Figure 2 is a schematic diagram of the structure of the intelligent operation and maintenance management system for distribution networks based on multi-dimensional perception provided in Embodiment 2 of the present invention. Detailed Implementation

[0051] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0052] Example 1

[0053] As shown in Figure 1, the intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception provided in this embodiment of the invention specifically includes the following steps:

[0054] Step 1: Construct a distributed collaborative sampling and sensing architecture, collect distribution network operation and maintenance data through multiple types of sensors, and combine the digital twin spatiotemporal coordinate system to perform spatiotemporal alignment compensation on the distribution network operation and maintenance data;

[0055] It should be explained that distribution network operation and maintenance data includes: electrical quantities, mechanical vibration, distribution network wiring topology, environmental parameters, and electromagnetic interference;

[0056] In a specific embodiment, electrical quantities of high-voltage equipment are collected. Wideband CT / PT is deployed on the high and low voltage sides of the ring main unit and transformer to capture high-frequency transient signals of lightning strikes and arc flashovers. The capture frequency range is preset to [0Hz, 500kHz]. At the same time, steady-state parameters of three-phase voltage, current and power factor are collected. The sampling interval is preset to 10μs.

[0057] The system dynamically senses the mechanical state of the equipment. Anti-magnetic interference MEMS vibration sensors are installed on the circuit breaker operating mechanism and the transformer body. The system adopts an impact response adaptive sampling mode. When there is no abnormal vibration, the sampling interval is preset to 100ms. If the vibration amplitude is detected to exceed twice the gravitational acceleration, the system automatically switches to the preset 1ms sampling interval and records the vibration frequency and phase information simultaneously.

[0058] To monitor the distribution network environment and partial discharge, a temperature and humidity-partial discharge composite sensor is deployed at switchgear and cable joints. The temperature and humidity-partial discharge composite sensor integrates infrared temperature measurement and UHF detection, and collects the ambient temperature and humidity once every 30 seconds. The sampling rate of the partial discharge signal is preset to 1MHz, and the amplitude, frequency and phase distribution of the partial discharge pulse are recorded.

[0059] The distribution network wiring topology is obtained through the GIS system. The distribution network wiring topology includes: physical spacing and connection relationship. Micro weather stations are deployed in the distribution radio area to collect environmental data according to the preset collection cycle. The environmental data includes, but is not limited to: wind speed, rainfall and lightning strikes.

[0060] It should be noted that GIS system refers to Geographic Information System, not gas-insulated switchgear. It is a technical tool that integrates geospatial data and equipment attribute data to realize distribution network topology visualization, spatial analysis and operation and maintenance decision support.

[0061] By combining the digital twin spatiotemporal coordinate system and using the micron-level timescale of the synchronous phasor measurement unit (PMU) as a reference, a global spatiotemporal coordinate system for the distribution network is established. Automatic alignment of distribution network operation and maintenance data is achieved through a distributed collaborative sampling protocol, using the following formula:

[0062]

[0063] Analysis yields data spatiotemporal alignment compensation time This refers to the time difference that needs to be adjusted, among which... The equipment topology gradient matrix, generated based on the power distribution network wiring diagram, reflects the electrical connections and physical location relationships between equipment. This is the physical distance vector between sensors, i.e., the straight-line distance between two data acquisition nodes, which is directly obtained from the GIS system. The speed of signal propagation includes the speed of electrical signal propagation and the speed of mechanical vibration signal propagation. The inherent delay of the sensor is obtained through factory calibration experiments.

[0064] Based on the obtained data spatiotemporal alignment compensation time, the timestamp of the original data is corrected, and the original timestamp is added to the data spatiotemporal alignment compensation time to achieve time synchronization of electrical quantities, mechanical quantities and environmental quantities.

[0065] The aligned distribution network operation and maintenance data is uniformly stored in the device-level spatiotemporal database, with a time resolution uniformly preset to 1ms, supporting edge computing calls;

[0066] Step 2: Based on the aligned and compensated distribution network operation and maintenance data, an electromagnetic interference adaptive filtering architecture is proposed to simultaneously solve the problems of data distortion caused by electromagnetic interference and excessive smoothing of useful signals by traditional filtering. In addition, the characteristic decoupling compression transmission technology is adapted according to the equipment type to reduce the amount of wireless transmission data.

[0067] In a specific embodiment, the electromagnetic field strength is monitored in a strong electromagnetic environment, and the power frequency magnetic field and high frequency interference field strength data around the equipment are collected in real time as the input basis for the interference suppression algorithm.

[0068] A two-stage processing mechanism combining physical filtering and neural compensation is employed to eliminate interference signals using a multi-dimensional sensing-based intelligent operation and maintenance management method for distribution networks. This is achieved through the following formula:

[0069]

[0070] Analysis yields the filtered interference data at time t. ,in, This is the physical filter function. The data collected by the sensor at time t. This is the original data. The preset compensation coefficient is set, with larger values ​​for higher field strengths, to balance the weights of physical filtering and neural compensation. For Fourier transform, For the interference transfer function The square of the modulus, This is the interference transfer function trained on the neural network, using the ResNet18 architecture. The input consists of historical interference data and corresponding clean data, and the output is the frequency domain features of the interference. The regularization factor is dynamically adjusted based on the real-time monitored field strength, and its calculation formula is as follows: , To ensure real-time electromagnetic field strength and avoid calculation overflow due to an excessively small denominator;

[0071] After interference removal, feature decoupling is performed on the data to adapt to different device types, transmitting only key features, as shown in the following formula:

[0072]

[0073] The compressed feature vectors output by the edge nodes are obtained through analysis. ,in, A feature selection matrix is ​​used for equipment type-related features, where 1 indicates features to be retained and 0 indicates features to be removed. For example, switchgear focuses on temperature-current phase coupling features and mechanical vibration peak features, while transformers focus on oil chromatographic composition features and partial discharge frequency features. For element-wise multiplication, This is the transpose of the wavelet envelope basis matrix, used to decompose the time-frequency domain features of the data. This is the original data matrix after removing interference;

[0074] Edge nodes perform Huffman coding compression on the decoupled feature vectors; the vectors are then transmitted to the cloud via dual 5G and fiber optic links. Real-time features are transmitted using 5G's Ultra-Reliable Low-Latency Communication (URLLC), while non-real-time features are transmitted via fiber optics. The cloud receives and decodes the vectors, and then reconstructs the core operating status of the device by combining the compressed feature vectors.

[0075] Step 3: Establish a digital twin with multi-field coupling and dynamic transfer learning, and combine feature transfer algorithms to solve the problem of poor model generalization caused by the heterogeneity of different equipment models. At the same time, combine multi-field coupling simulation to solve the problem that static twins cannot reflect changes in operating conditions. Design a fault risk entropy health assessment model to replace traditional fixed threshold alarms, and simultaneously solve the problems of ignoring the cumulative degradation effect of equipment and lack of economic quantitative trade-offs in operation and maintenance decisions.

[0076] In a specific embodiment, a three-dimensional topological twin of the power distribution network is constructed by combining GIS and BIM technologies, and the compressed feature vectors are mapped to the virtual device model to achieve real-time linkage between physical devices and virtual models;

[0077] It should be noted that BIM technology stands for Building Information Modeling, which extends from the traditional construction field to the digital modeling of the entire life cycle of power equipment. Its core is to realize the full-process visual management and simulation analysis of power distribution network equipment from design and installation to operation and maintenance and scrapping by constructing a digital carrier of the equipment's three-dimensional geometric model and full attribute data.

[0078] Embedded electrical field, temperature field and mechanical field coupling module to simulate the equipment operation status under distributed power access and extreme weather conditions. The coupling equation is based on the original method to ensure that the virtual model is consistent with the operating characteristics of the physical equipment.

[0079] To address the heterogeneity of different device models, a dynamic transfer learning framework is employed to achieve cross-device adaptation of the health assessment model, using the following formula:

[0080]

[0081] Health assessment values ​​are obtained by analyzing the transfer learning loss function. ,in, The feature matrix of the target device, For feature mapping function, The feature matrix of the source device, These are the parameters of the feature mapping function. The Frobenius norm is used to measure the similarity between two matrices. The adaptive weights are related to the number of fitted samples; the more fitted samples, the smaller the adaptive weights. The KL divergence between the source device state distribution and the target device state distribution is calculated to avoid overfitting the model to the source device. For the distribution of source device status, The target device state distribution;

[0082] Combining the characteristics after migration and the cumulative degradation effect, the equipment failure risk entropy is calculated as a quantitative indicator of health status, using the following formula:

[0083]

[0084] Analysis yields fault risk entropy ,in, For time steps, To assess the start time, The degradation trajectory matrix is ​​predicted by an LSTM network. The input is the historical feature vector, and the output is the degradation trend of key device parameters. For time integration variables, The preset economic adjustment coefficient, This is a fault influencing factor, and it is positively correlated with the user density and load level of the area where the equipment is located.

[0085] Substituting the failure risk entropy into the remaining life formula, the life prediction results are corrected to ensure the linkage between health assessment and life management, as shown in the following formula:

[0086]

[0087] Analysis yields remaining health values ,in, The rated remaining health value of the equipment. This is the preset lifespan correction factor. For fault risk entropy, For the time step, the fault risk entropy is added as a correction term. The higher the risk entropy, the shorter the remaining health value.

[0088] Step 4: Construct an operation and maintenance decision space, solve Pareto optimal decisions, and simultaneously address the problems of traditional decision-making focusing on a single objective and the difficulty of fault location in distributed power scenarios; design a digital twin verification closed-loop mechanism to simulate and verify before decision execution, and solve the problem of execution failure caused by the mismatch between operation and maintenance instructions and actual equipment operating conditions.

[0089] In a specific embodiment, an operation and maintenance decision space is constructed by combining the equipment failure risk entropy and the twin model state, covering the following operation and maintenance solutions: in, For the k-th operation and maintenance decision, For operation and maintenance decision indexing, For maintenance time window, For the m-th maintenance team, For the nth fault isolation scheme, For the maintenance team index, Index of fault isolation schemes;

[0090] With the objectives of minimizing failure risk and minimizing total operation and maintenance cost, and taking into account equipment safety constraints, the optimal decision is solved.

[0091] By combining a non-dominated sorting genetic algorithm and introducing distributed power source output constraints, the Pareto optimal decision set is found by optimizing the particle search range.

[0092] Each operational decision in the Pareto optimal decision set is injected into a digital twin to simulate the execution process;

[0093] Real-time monitoring of overvoltage, current, and equipment temperature in the simulation to determine whether the requirements are met. ,in, For divergence operators, For the electrical quantity field in digital twin simulation. Let Chebyshev norm be the divergence of electrical quantities in the simulation. Preset safety threshold;

[0094] Filter out the decisions that meet the constraints, sort them according to the priority of minimum risk entropy and minimum consumption cost, and output the final execution decision;

[0095] For equipment with electric operation capabilities, control commands are issued through the twin platform to achieve automatic isolation of faulty sections, and load transfer is monitored in real time after isolation;

[0096] Maintenance personnel receive work tasks using mobile terminals, which display a twin model of the equipment and safety operating procedures. Work restricted areas are demarcated using electronic fences, and smart safety helmets collect personnel location data in real time. If personnel cross the boundary or the safe distance is insufficient, the terminal immediately issues an audible and visual alarm and uploads the alarm information to the cloud.

[0097] After the operation is completed, the edge node re-collects the device parameters and calculates the change in fault risk entropy before and after the operation and maintenance. If the change in fault risk entropy is less than the preset entropy change threshold, the operation and maintenance effect is deemed to be qualified; otherwise, the operation and maintenance effect is deemed to be unqualified.

[0098] The evaluation results and operation and maintenance data are fed back to the health assessment model and decision engine to update the sample library for transfer learning and the parameters for decision optimization.

[0099] Example 2

[0100] As shown in Figure 2, the intelligent operation and maintenance management system for distribution networks based on multi-dimensional perception provided in this embodiment of the invention specifically includes the following units:

[0101] Sampling and sensing unit: Construct a distributed collaborative sampling and sensing architecture, collect distribution network operation and maintenance data through multiple types of sensors, and combine the digital twin spatiotemporal coordinate system to perform spatiotemporal alignment compensation on the distribution network operation and maintenance data;

[0102] Decoupling and Reconstruction Unit: Combining the aligned and compensated distribution network operation and maintenance data, an electromagnetic interference adaptive filtering architecture is established to filter interference data, and equipment features are decoupled to adapt to equipment types. After encoding, the core operating status of the equipment is decoded and reconstructed in the cloud.

[0103] Assessment and prediction unit: Construct a three-dimensional topological twin of the distribution network, map equipment features to a virtual equipment model, combine dynamic transfer learning to establish a health assessment model to predict degradation trajectory, calculate fault risk entropy, and substitute it into the corrected lifetime prediction results;

[0104] Verification Feedback Unit: Constructs an operation and maintenance decision space by combining fault risk entropy with the state of the distribution network's three-dimensional topology twin, and solves for the optimal decision; injects the decision into the simulation of the distribution network's three-dimensional topology twin to verify the safety constraints, evaluates the changes in fault risk entropy, and provides feedback to update the health assessment model and decision engine.

[0105] The above provides a detailed description of one embodiment of the present invention, but the content described is only a preferred embodiment of the present invention and should not be considered as limiting the scope of the present invention. The above formulas are all dimensionless numerical calculations, and the formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art based on actual conditions and historical experience, and can be adjusted according to actual conditions. The above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention. All equivalent changes and improvements made in accordance with the scope of the present invention should still fall within the patent coverage of the present invention.

Claims

1. A method for intelligent operation and maintenance management of distribution networks based on multi-dimensional perception, characterized in that, Includes the following steps: A distributed collaborative sampling and sensing architecture is constructed to collect distribution network operation and maintenance data through multiple types of sensors. Combined with a digital twin spatiotemporal coordinate system, spatiotemporal alignment compensation is performed on the distribution network operation and maintenance data. Based on the aligned and compensated distribution network operation and maintenance data, an electromagnetic interference adaptive filtering architecture is established to filter interference data. Equipment features are decoupled according to equipment type, encoded, decoded in the cloud, and the core operating status of the equipment is reconstructed. A three-dimensional topological twin of the distribution network is constructed, mapping equipment features to a virtual equipment model. A health assessment model is established using dynamic transfer learning to predict degradation trajectories, calculate fault risk entropy, and substitute it into the corrected lifespan prediction results. An operation and maintenance decision space is constructed by combining the fault risk entropy and the state of the three-dimensional topological twin of the distribution network, and the optimal decision is solved. The decision is injected into the simulation of the three-dimensional topological twin of the distribution network to verify safety constraints, evaluate changes in fault risk entropy, and provide feedback to update the health assessment model and decision engine.

2. The intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception as described in claim 1, characterized in that, The spatiotemporal alignment compensation method is as follows: combining the digital twin spatiotemporal coordinate system, and using the micron-level time scale of the synchronous phasor measurement device (PMU) as the reference, a global spatiotemporal coordinate system for the distribution network is established. Automatic alignment of distribution network operation and maintenance data is achieved through a distributed collaborative sampling protocol, and the spatiotemporal alignment compensation time of the data is analyzed. Based on the obtained data spatiotemporal alignment compensation time, the timestamp of the original data is corrected, and the original timestamp is added to the data spatiotemporal alignment compensation time to achieve time synchronization of electrical quantities, mechanical quantities and environmental quantities.

3. The intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception as described in claim 1, characterized in that, The method for filtering interference data in the device is as follows: Data on the power frequency magnetic field and high-frequency interference field strength around the device are collected as input to the interference suppression algorithm; a two-stage processing mechanism combining physical filtering and neural compensation is employed to eliminate interference signals using the following formula: Analysis yields the filtered interference data at time t. ,in, This is the physical filter function. The data collected by the sensor at time t. This is the original data. For the preset compensation coefficient, For Fourier transform, For the interference transfer function The square of the modulus, This is the interference transfer function used for training the neural network. This is the regularization factor.

4. The intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception according to claim 1, characterized in that, The method for device feature decoupling is as follows: After interference removal, the data is decoupled based on device type, transmitting only key features, using the following formula: The compressed feature vectors output by the edge nodes are obtained through analysis. ,in, Select a feature matrix for device type. For element-wise multiplication, It is the transpose of the wavelet envelope basis matrix. The original data matrix after interference removal; edge nodes perform Huffman coding compression on the decoupled feature vectors; the data is transmitted to the cloud via dual 5G and fiber optic links, and after being received and decoded by the cloud, the core operating status of the device is reconstructed by combining the compressed feature vectors.

5. The intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception according to claim 1, characterized in that, The method for predicting degradation trajectories is as follows: a three-dimensional topological twin of the distribution network is constructed, and compressed feature vectors are mapped to virtual device models to achieve real-time linkage between physical devices and virtual models; electrical field, temperature field, and mechanical field coupling modules are embedded to simulate the operating status of equipment under distributed power access and extreme weather conditions. The coupling equations refer to the original method to ensure that the operating characteristics of the virtual model and the physical device are consistent; for the heterogeneity of different models of equipment, a dynamic transfer learning framework is adopted to achieve cross-device adaptation of the health assessment model, and the health assessment value is obtained by analyzing the transfer learning loss function.

6. The intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception according to claim 1, characterized in that, The method for correcting the lifetime prediction results is as follows: combining the characteristics after migration and the cumulative degradation effect, calculate the equipment failure risk entropy as a quantitative indicator of health status, and analyze to obtain the failure risk entropy. By substituting the failure risk entropy into the remaining life formula, the life prediction results are corrected, ensuring the linkage between health assessment and life management, and the remaining health value is obtained through analysis.

7. The intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception according to claim 1, characterized in that, The method for finding the optimal decision is as follows: A maintenance decision space is constructed by combining the equipment failure risk entropy with the twin model state, encompassing the following maintenance solutions: in, For the k-th operation and maintenance decision, For operation and maintenance decision indexing, For maintenance time window, For the m-th maintenance team, For the nth fault isolation scheme, For the maintenance team index, This provides an index for fault isolation schemes; aiming to minimize fault risk and total maintenance cost, and considering equipment safety constraints, it solves for the optimal decision.

8. The intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception according to claim 1, characterized in that, The method for verifying security constraints is as follows: combining a non-dominated sorting genetic algorithm, introducing distributed power output constraints, finding the Pareto optimal decision set by optimizing the particle search range, and injecting each operation and maintenance decision in the Pareto optimal decision set into a digital twin to simulate the execution process; Real-time monitoring of overvoltage, current, and equipment temperature in the simulation to determine whether the requirements are met. ,in, For divergence operators, For the electrical quantity field in digital twin simulation. Let Chebyshev norm be the divergence of electrical quantities in the simulation. A preset safety threshold is set; decisions that meet the constraints are selected, and then prioritized according to the minimum risk entropy and the lowest cost, and the final execution decision is output.

9. The intelligent operation and maintenance management method for distribution networks based on multi-dimensional perception according to claim 1, characterized in that, The method for assessing the change in fault risk entropy is as follows: After the operation is completed, the edge node re-collects the equipment parameters and calculates the change in fault risk entropy before and after the operation and maintenance. If the change in fault risk entropy is less than the preset entropy change threshold, the operation and maintenance effect is deemed qualified; otherwise, the operation and maintenance effect is deemed unqualified. The assessment results and operation and maintenance data are fed back to the health assessment model and decision engine to update the sample library for transfer learning and the parameters for decision optimization.

10. A distribution network intelligent operation and maintenance management system based on multi-dimensional perception, the management system being used to execute the management method described in any one of claims 1-9, characterized in that, include: The system comprises the following components: **Sampling and Sensing Unit:** Constructs a distributed collaborative sampling and sensing architecture, collecting distribution network operation and maintenance data through multiple types of sensors. Combined with a digital twin spatiotemporal coordinate system, it performs spatiotemporal alignment compensation on the distribution network operation and maintenance data. **Decoupling and Reconstruction Unit:** Based on the aligned and compensated distribution network operation and maintenance data, it establishes an adaptive electromagnetic interference filtering architecture to filter interference data. It also performs equipment feature decoupling for equipment type adaptation, encodes the data, decodes it in the cloud, and reconstructs the core operating status of the equipment. **Evaluation and Prediction Unit:** Constructs a three-dimensional topological twin of the distribution network, mapping equipment features to a virtual equipment model. It combines dynamic transfer learning to establish a health assessment model to predict degradation trajectories, calculates fault risk entropy, and substitutes it into the corrected lifespan prediction results. **Verification and Feedback Unit:** Constructs an operation and maintenance decision space by combining fault risk entropy with the state of the three-dimensional topological twin of the distribution network, solving for the optimal decision. The decision is then injected into the three-dimensional topological twin simulation to verify safety constraints, evaluate changes in fault risk entropy, and provide feedback to update the health assessment model and decision engine.

Images