Method, device and equipment for evaluating health condition of distributed power distribution network fittings under sky-ground monitoring system and storage medium

Abstract

The application discloses a kind of distributed power distribution network hardware health condition evaluation method, device and equipment under sky-ground monitoring system and storage medium;The method includes the multi-modal hardware data collected by sky-ground monitoring system is preprocessed and mapped to the same dimension, constructs input data space matrix;Based on feature correlation mode rule and probability fuzzy logic modeling strategy Fuzzy Importance Analysis Model is constructed, generate fuzzy importance diagnostic index, based on fuzzy importance diagnostic index optimization deep factor decomposition machine model obtains hardware risk assessment model, outputs risk score and normalizes, generates hardware health condition risk heat map.Due to the present application effectively fuses multi-modal data, realizes model optimization by fuzzy importance analysis, to accurately identify the low-frequency high-loss hidden danger of hardware, realize the change of hardware of distributed power distribution network from post-determination and passive repair to prior prediction and evaluation, effectively depict the prospective health of distribution hardware and risk prediction.

Description

Technical Field

[0001] This invention relates to the field of distributed power distribution network technology, and in particular to a method, apparatus, equipment and storage medium for assessing the health status of distributed power distribution network fittings under a sky-ground monitoring system. Background Technology

[0002] In distributed power distribution networks, fittings perform fundamental functions such as connection, fixation, and support, directly affecting the quality of conductor electrical contacts, mechanical integrity, insulation margin, and heat dissipation capacity. Due to the combined effects of complex suburban environments, power flow reversals caused by distributed power grid connection, and load fluctuations, fittings are prone to latent degradation such as contact deterioration, corrosion, and loosening during long-term service. Under high temperature and humidity or peak load conditions, this degradation can amplify into overheating and discharge risks, leading to power outages and equipment damage. Currently, common health assessments fall into two main categories: manual inspections rely on on-site observation and auxiliary testing, offering high precision but low coverage efficiency, high cost, and personnel safety risks; and vehicle-based inspections utilize mobile vehicles such as drones for remote inspections, offering superior efficiency and safety compared to manual methods. However, constraints such as low-altitude flight restrictions, ground obstructions, range limitations, and scheduling make it difficult to achieve high-frequency, full-area coverage of the large and dispersed number of fittings. These methods are more focused on "discovering existing or ongoing problems," and are insufficient for early identification and trend tracking of latent degradation.

[0003] Current methods for assessing the health of electrical fittings have the following shortcomings: First, manual pole climbing or ground-based observation is hampered by wide-area dispersion, terrain obstruction, and safety constraints, making it difficult to maintain high-frequency coverage. Evidence collection is scattered and relies on experience. Detection equipment signals are mostly collected from single points, with inconsistent clock calibers, and fixed thresholds are prone to false alarms and missed alarms under varying operating conditions and noise. Second, vehicle-based inspections rely on mobile vehicles such as drones to acquire images, which are limited by flight restrictions, endurance, weather, and nighttime visibility. Mountainous and forested areas are prone to obstruction. Image recognition is not robust enough to small-scale defects and weak textures, and there is a lack of unified registration between images and electrical parameters, making it difficult to form calibrable cross-modal evidence and feed it forward to differentiated operation and maintenance. Third, existing methods for assessing electrical fittings are mostly reactive and depend on post-event judgment, lacking the ability to proactively characterize and prevent issues related to distribution fittings. Summary of the Invention

[0004] The main objective of this invention is to provide a method, device, equipment, and storage medium for assessing the health status of distributed distribution network fittings under a sky-ground monitoring system. This aims to solve the technical problems of existing technologies for assessing the health status of distribution network fittings in wide-area distributed distribution network scenarios, which are difficult to effectively integrate multimodal information, have difficulty identifying low-frequency high-loss hidden dangers, and lack the ability to proactively characterize and prevent distribution fittings.

[0005] To achieve the above objectives, this invention provides a method for assessing the health status of distributed distribution network fittings under a sky-ground monitoring system, the method comprising the following steps:

[0006] The multimodal hardware data collected by the sky-ground monitoring system are preprocessed and mapped to the same dimensional space to construct an input data space matrix. The sky-ground monitoring system includes a space-based module, an air-based module, and a ground-based module. The input data space matrix includes multiple evaluation records. Each evaluation record includes a feature set and an evaluation result. The feature set contains discrete features and continuous features. Both the discrete features and the continuous features are composed of multiple factors.

[0007] A fuzzy importance analysis model is constructed based on feature-related pattern rules and probabilistic fuzzy logic modeling strategies. The continuous and discrete features in the input data space matrix are processed by the fuzzy importance analysis model to generate fuzzy importance diagnostic indicators, which include significance coefficient, confidence coefficient and consistency coefficient.

[0008] The deep factor decomposition machine model is optimized based on the aforementioned fuzzy importance diagnostic index to generate a hardware risk assessment model.

[0009] The hardware risk assessment model outputs the risk score corresponding to each assessment record in the input data space matrix, and normalizes the risk score. Based on the normalized risk score, a hardware health status risk heat map is generated.

[0010] Optionally, the fuzzy importance analysis model is constructed based on feature-related pattern rules and probabilistic fuzzy logic modeling strategies, and the continuous and discrete features in the input data space matrix are processed by the fuzzy importance analysis model to generate fuzzy importance diagnostic indicators, including:

[0011] Based on the feature-related pattern rule and probabilistic fuzzy logic modeling strategies, a feature-related pattern rule recognition module and a probabilistic fuzzy logic modeling module are constructed and combined to form a fuzzy importance analysis model.

[0012] The discrete features in the input data space matrix are input into the feature-related pattern rule recognition module to generate a feature-related rule set.

[0013] The continuous features in the input data space matrix are input into the probabilistic fuzzy logic modeling module to generate continuous risk scores;

[0014] Based on the feature-related rule set and the continuous risk score, calculation functions for the significance coefficient, confidence coefficient, and consistency coefficient are constructed respectively.

[0015] The fuzzy importance diagnostic index is generated by performing calculations on the feature-related rule set and the continuous risk score using the calculation function.

[0016] Optionally, the step of inputting discrete features from the input data space matrix into the feature-related pattern rule recognition module to generate a feature-related rule set includes:

[0017] Obtain the set of transaction numbers corresponding to the discrete features in the input data space matrix;

[0018] Calculate the intersection of the set of transaction IDs corresponding to the preceding feature and the subsequent evaluation result to obtain the preceding-subsequent candidate pair;

[0019] The features of the preceding term in the preceding-following candidate pair are expanded vertically to obtain multiple preceding combinations;

[0020] Based on antimonotonicity, the multiple antecedent combinations are pruned to retain stable co-occurring multiple antecedent-consequence candidate pairs, generating a feature-related rule set.

[0021] Optionally, the step of inputting continuous features from the input data space matrix into the probabilistic fuzzy logic modeling module to generate continuous risk scores includes:

[0022] Obtain the feature probability distribution curves of continuous features in the input data space matrix, and set preset boundary values ​​to divide multiple input regions;

[0023] Multiple overlapping trapezoidal input membership functions are adaptively generated based on the input region;

[0024] A hierarchical structure is used to perform pairwise rule-decoupled calculations on the continuous features, and risk evidence is aggregated layer by layer.

[0025] The centroid method is used to deblur the aggregated risk evidence and generate a continuous risk score.

[0026] Optionally, the step of calculating the fuzzy importance diagnostic index by operating the calculation function on the feature-related rule set and the continuous risk score includes:

[0027] The feature-related rule set is expanded to include rule pairs containing fuzzy sets corresponding to the set of preceding feature factors and fuzzy sets corresponding to the set of subsequent result-end target factors;

[0028] When the sum of participation corresponding to the feature fuzzy set exceeds the preset participation threshold, the significance coefficient is calculated, referring to the following formula:

[0029]

[0030]

[0031] in, Indicates the significance coefficient. Let S represent the i-th evaluation record in the input data space matrix. Represents the set of the preceding characteristic factors. Let represent the set of transaction IDs corresponding to the evaluation records that satisfy the i-th feature-related rule. and These represent the total number of records in the input data space matrix S and the total number of occurrences in the set of preceding feature factors, respectively. Indicates evaluation record Chinese characteristics Factors, This represents the fuzzy membership degree of the j-th feature in the set of preceding feature factors A corresponding to the i-th evaluation record. This represents the fuzzy set associated with the j-th antecedent feature in the i-th evaluation record. The level of ambiguity risk Indicates the feature risk adjustment weight. This indicates the preset participation threshold. This represents the fuzzy set corresponding to the set of the preceding feature factors;

[0032] The conjunction is obtained by combining the set of feature factors of the preceding term and the set of target factors of the result term of the following term.

[0033] The confidence coefficient is calculated based on the conjunct set and the feature-related rule set, referring to the following formula:

[0034]

[0035] in, Represents the confidence coefficient. The conjunctive set is obtained by combining the set of feature factors A of the preceding term with the set of target factors B of the following term. This represents the i-th evaluation result in the conjunction set. This represents the membership degree of the evaluation result in the conjunction set corresponding to the i-th evaluation record;

[0036] A consistency coefficient is generated by calculating the expectations of the fuzzy sets corresponding to the set of feature factors of the preceding term, the fuzzy sets corresponding to the set of target factors of the following term, and the fuzzy sets corresponding to the conjunctive set, as shown in the following formula:

[0037]

[0038] in, Represents the consistency coefficient. Let represent the mathematical expectation of the fuzzy set corresponding to the conjunctive set. This represents the mathematical expectation of the set of preceding feature factors and their corresponding fuzzy sets. This represents the mathematical expectation of the target factor set B and its corresponding fuzzy set D at the result end of the subsequent term. Represents the set of the preceding characteristic factors. This represents the fuzzy set corresponding to the set of the preceding feature factors. This represents the set of target factors at the end of the subsequent result. This represents the fuzzy set corresponding to the target factor set of the subsequent result;

[0039] By integrating the significance coefficient, confidence coefficient, and consistency coefficient, a fuzzy importance diagnostic index is obtained.

[0040] Optionally, the step of optimizing the deep factor decomposition machine model based on the fuzzy importance diagnostic index to generate a hardware risk assessment model includes:

[0041] A deep factorization machine model is constructed, which includes an input layer, a factorization machine layer, and a deep neural network layer.

[0042] The significance coefficients in the fuzzy importance diagnostic index are incorporated as prior coefficients into the embedding layer and the factor decomposition machine layer;

[0043] The confidence coefficient and consistency coefficient in the fuzzy importance diagnostic index are combined to generate a weighted coefficient;

[0044] The prior coefficients and the weighting coefficients are multiplied by the eigenvalues ​​of the first-order terms of the factorization machine layer.

[0045] The prior coefficients are used to modulate the embedding vector of the input features to obtain the modulated embedding vector. Then, the inner product of the weighting coefficients and the modulated embedding vector is multiplied by the feature value corresponding to the second-order interaction term.

[0046] The first-order and second-order interaction terms after computation are summed with the bias term to generate the optimized factorization machine layer output;

[0047] The prior coefficients and the weighting coefficients are applied to the embedding features of the deep neural network layer;

[0048] By integrating and optimizing the input layer, factorization machine layer, and deep neural network layer, a hardware risk assessment model is generated.

[0049] Optionally, the mathematical expression of the hardware risk assessment model is:

[0050]

[0051] in, This represents the risk score output by the model. It is the sigmoid activation function. This represents the output features of the last layer in a deep neural network. This represents the output of the factorization machine layer;

[0052] The mathematical expression for each layer of the deep neural network is:

[0053]

[0054] in, This represents the weight matrix of the l-th layer. This represents the bias vector of the l-th layer. Represents the ReLU activation function. This represents the output feature of the l-th layer. Indicates the first The output features of the layer;

[0055] The mathematical expression for the factorization engine layer is:

[0056]

[0057]

[0058] in, and These represent the i-th and j-th input features, respectively. This represents the weighting coefficient corresponding to the i-th preceding feature. These represent the prior coefficients, which are used during the encoding process to gate and modulate the embedded representation. This represents the weighting coefficient corresponding to the interaction of the i-th and j-th preceding features. Indicates the bias term. and This represents the embedding vector of the i-th and j-th preceding features. This represents the vector inner product operation, used to characterize the degree of second-order interaction between features. It represents the mapping function for embedding operations from high-dimensional sparse features to low-dimensional dense features.

[0059] Furthermore, to achieve the above objectives, the present invention also proposes a distributed distribution network fittings health status assessment device under a sky-ground monitoring system. The device is configured to implement the steps of the distributed distribution network fittings health status assessment method under a sky-ground monitoring system as described above. The distributed distribution network fittings health status assessment device under a sky-ground monitoring system includes:

[0060] The data processing module preprocesses the multimodal hardware data collected by the sky-ground monitoring system and maps it to the same dimensional space to construct an input data space matrix. The sky-ground monitoring system includes a space-based module, an air-based module, and a ground-based module. The input data space matrix includes multiple evaluation records. Each evaluation record includes a feature set and an evaluation result. The feature set contains discrete features and continuous features, and both the discrete features and the continuous features are composed of multiple factors.

[0061] The dual-path feature processing module is used to construct a fuzzy importance analysis model based on feature-related pattern rules and probabilistic fuzzy logic modeling strategies, and to process the continuous and discrete features in the input data space matrix through the fuzzy importance analysis model to generate fuzzy importance diagnostic indicators, which include significance coefficient, confidence coefficient and consistency coefficient.

[0062] The model optimization module is used to optimize the deep factor decomposition machine model based on the fuzzy importance diagnostic index to generate a hardware risk assessment model.

[0063] The risk assessment module is used to output the risk score corresponding to each assessment record in the input data space matrix through the hardware risk assessment model, and to normalize the risk score, and generate a hardware health status risk heat map based on the normalized risk score.

[0064] Furthermore, to achieve the above objectives, this application also proposes a distributed distribution network fittings health status assessment device under a sky-ground monitoring system. The device includes: a memory, a processor, and a distributed distribution network fittings health status assessment program stored in the memory. The processor is used to run the distributed distribution network fittings health status assessment program, and the computer program is configured to implement the steps of the distributed distribution network fittings health status assessment method under the sky-ground monitoring system described above.

[0065] In addition, to achieve the above objectives, this application also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the distributed distribution network hardware health status assessment method under the above-described sky-ground monitoring system.

[0066] This invention aggregates multi-source data from an integrated sky-ground monitoring architecture, establishes a unified spatiotemporal reference and quality control mechanism, constructs an input data spatial matrix, employs a dual-path feature identification mechanism to process discrete and continuous features, generates three types of fuzzy importance diagnostic criteria, integrates these criteria into an improved deep factor decomposition machine model, predicts the health risk score of hardware fittings, normalizes the risk score, generates a risk heatmap, and delineates risk regions. Because this invention effectively integrates multimodal data and optimizes the model through fuzzy importance analysis, it accurately identifies low-frequency, high-damage hidden dangers in hardware fittings, achieving [a more precise and accurate assessment]. The shift from post-event judgment and passive maintenance to pre-event prediction and assessment of distributed power distribution network fittings effectively enables forward-looking health characterization and risk prediction of power distribution fittings. It effectively mines the characteristic value of multimodal fitting data, fully leverages the data advantages of each monitoring module, improves the accuracy and efficiency of fitting risk assessment, simplifies the assessment process, reduces the cost and difficulty of manual assessment, and visually presents the risk distribution through heat maps. This provides a scientific and reliable basis for the operation and maintenance of fittings, effectively avoids the risk of fitting failure, ensures the safe and stable operation of fittings, and comprehensively improves the level of intelligence in fitting health management. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 This is a schematic diagram of the structure of a distributed power distribution network hardware health status assessment device under the sky-ground monitoring system of the hardware operating environment involved in the embodiments of the present invention;

[0069] Figure 2 This is a flowchart illustrating the first embodiment of the distributed distribution network hardware health status assessment method under the sky-ground monitoring system of the present invention.

[0070] Figure 3 This is a schematic diagram of the input data space matrix in one embodiment of the distributed distribution network hardware health status assessment method under the sky-ground monitoring system of the present invention;

[0071] Figure 4 This is a schematic diagram of the structure of the fittings risk assessment model in one embodiment of the distributed distribution network fittings health status assessment method under the sky-ground monitoring system of the present invention.

[0072] Figure 5 This is a risk heat map for assessing the health status of distributed distribution network fittings in an embodiment of the distributed distribution network fittings health status assessment method under the sky-ground monitoring system of the present invention.

[0073] Figure 6 This is a flowchart illustrating the second embodiment of the distributed distribution network hardware health status assessment method under the sky-ground monitoring system of the present invention.

[0074] Figure 7 This is a schematic diagram of the probabilistic fuzzy logic modeling module of the distributed distribution network hardware health status assessment method under the sky-ground monitoring system of the present invention.

[0075] Figure 8(a) is a flowchart of Step 1 in the probabilistic fuzzy risk solution process in an embodiment of the distributed distribution network hardware health status assessment method under the sky-ground monitoring system of the present invention.

[0076] Figure 8(b) is a flowchart of Step 2 in the probabilistic fuzzy risk solution process in one embodiment of the distributed distribution network hardware health status assessment method under the sky-ground monitoring system of the present invention.

[0077] Figure 9 This is a flowchart illustrating the third embodiment of the distributed distribution network hardware health status assessment method under the sky-ground monitoring system of the present invention.

[0078] Figure 10 This is a structural block diagram of the first embodiment of the distributed power distribution network hardware health status assessment device under the sky-ground monitoring system of the present invention.

[0079] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings.

[0080] The reference numerals are as follows: Processor 1001, Communication Bus 1002, User Interface 1003, Network Interface 1004, Memory 1005. Detailed Implementation

[0081] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0082] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a distributed power distribution network hardware health status assessment device under the sky-ground monitoring system of the hardware operating environment involved in the embodiments of the present invention.

[0083] like Figure 1As shown, the distributed power distribution network hardware health assessment device under this sky-ground monitoring system may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen and an input unit such as a keyboard; the user interface 1003 may also include standard wired and wireless interfaces. The network interface 1004 may optionally include standard wired and wireless interfaces (such as Wireless-Fidelity (Wi-Fi) interfaces). The memory 1005 may be high-speed random access memory (RAM) or stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0084] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the health status assessment equipment for distributed distribution network fittings under the sky-ground monitoring system. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0085] like Figure 1 As shown, the memory 1005, which is a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a distributed power distribution network hardware health status assessment program.

[0086] exist Figure 1 In the distributed distribution network hardware health status assessment device under the sky-ground monitoring system shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and memory 1005 in the distributed distribution network hardware health status assessment device under the sky-ground monitoring system of the present invention can be set in the distributed distribution network hardware health status assessment device under the sky-ground monitoring system. The distributed distribution network hardware health status assessment device under the sky-ground monitoring system calls the distributed distribution network hardware health status assessment program stored in the memory 1005 through the processor 1001 and executes the distributed distribution network hardware health status assessment method under the sky-ground monitoring system provided in the embodiment of the present invention.

[0087] This invention provides a method for assessing the health status of distributed distribution network fittings under a sky-ground monitoring system, referring to... Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the distributed power distribution network hardware health status assessment method under the sky-ground monitoring system of the present invention.

[0088] In this embodiment, the method for assessing the health status of distributed distribution network fittings under the sky-ground monitoring system includes the following steps:

[0089] Step S10: Preprocess the multimodal hardware data collected by the sky-ground monitoring system and map it to the same dimensional space to construct the input data space matrix.

[0090] It should be understood that the executing entity of this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or a terminal electronic device capable of performing the above functions. The following description uses a distributed distribution network hardware health status assessment device (hereinafter referred to as the assessment device) under a sky-ground monitoring system as an example to illustrate this embodiment and the following embodiments.

[0091] It should be noted that the sky-ground monitoring system can be composed of space-based modules, air-based modules, and ground-based modules working together to form a comprehensive monitoring system for collecting hardware-related data in all directions, multiple dimensions, and all weather conditions. It can achieve full-scenario data collection for hardware, from large-scale coverage to close-range precise monitoring.

[0092] The sky-ground monitoring system includes a space-based module, an air-based module, and a ground-based module. The input data spatial matrix includes multiple evaluation records. Each evaluation record includes a feature set and an evaluation result. The feature set contains discrete features and continuous features, and both the discrete features and the continuous features are composed of multiple factors.

[0093] It should be noted that the input data space matrix refers to the matrix structure formed by organizing the preprocessed multimodal hardware data and mapping it to the same dimension according to the evaluation records. It is the core input for subsequent model training and analysis. The matrix structure clearly presents the correspondence between features and evaluation results.

[0094] Discrete features can be features in a feature set whose values ​​are discontinuous and in the form of discrete categories. They are composed of multiple categorical influencing factors, cannot be represented by continuous numerical values, and can only be classified into specific categories.

[0095] Continuous features can be features in a feature set that take continuously changing numerical values. They consist of multiple numerical influencing factors and can be used to quantify a certain operating state or health parameter of the hardware through specific numerical values.

[0096] Understandably, the evaluation equipment receives multimodal fitting data collected by the space-based, air-based, and ground-based modules; performs temporal alignment and gridded mapping on the multimodal fitting data to identify outliers and fill in missing segments; normalizes and scales the filled multimodal fitting data to obtain a standardized feature stream; integrates the standardized feature stream with the operation and maintenance records to divide it into discrete and continuous features; and constructs an input data space matrix based on the integrated discrete and continuous features and the evaluation results.

[0097] In practice, the evaluation equipment gathers operational and environmental data such as satellite remote sensing, UAV imagery and ground sensing under the integrated sky-ground monitoring architecture, establishes a unified spatiotemporal reference and quality control mechanism, and forms a learnable mapping space as subsequent modeling corpus.

[0098] In some embodiments, based on a hierarchical collaborative perception model of "wide-area-close-range-local," the space layer utilizes satellite remote sensing to periodically acquire regional area observations, forming a "scene prior layer" for hardware risk assessment; the air layer relies on mobile carriers such as UAVs to perform close-range imaging and online detection of hardware, quickly verifying abnormal areas indicated by satellite, forming a "detailed evidence layer"; the ground layer deploys low-power sensors at key nodes of hardware to collect parameters such as temperature rise, humidity, and vibration, unifying the time reference at the edge-end side and performing primary quality control, constructing a "causal indicator layer of electrical participation in the environment." Addressing issues such as sampling rhythm differences, missing measurements, and abnormal fluctuations in multi-source heterogeneous data, temporal alignment and gridded mapping are performed, and anomaly identification and missing segment filling are conducted; normalization and scale alignment are implemented based on dimensional and distribution differences, outputting a high-quality feature stream that can be directly used for modeling, achieving an integrated closed loop of "acquisition-alignment-cleaning-standardization."

[0099] This integrates multimodal observations, including satellite remote sensing, UAV imagery, ground sensing, and operation and maintenance records, to construct an input data mapping space. Let the record set be... Each element Represents an evaluation record number; a set of features , Indicates a certain feature It consists of multiple factors; let the target factor set be defined. Each element This represents the final evaluation result corresponding to an evaluation record. To analyze discrete and continuous features separately, Divided into two parts: Represents the sum of elements in discrete features This represents the elements in the continuous feature. The final result is a matrix S as shown in the figure, represented as follows: Figure 3 As shown, Figure 3 This is a schematic diagram of the input data space matrix.

[0100] Understandably, this embodiment effectively removes noise and outliers from the collected data, repairs missing data, cleans up redundant information, and improves data quality through multimodal data preprocessing, thereby avoiding interference from invalid data with subsequent model analysis and evaluation results. Mapping multimodal data to the same dimensional space eliminates the differences in dimensions and dimensional heterogeneity between different modal data, achieving homogeneous integration of multi-source data and breaking down barriers between data from different monitoring modules. Constructing an input data space matrix organizes scattered multimodal data into a standardized and structured model input format, clearly presenting the correspondence between features and evaluation results, providing high-quality and standardized data support for subsequent feature importance analysis and model construction, and laying the foundation for the entire technical solution.

[0101] Step S20: Construct a fuzzy importance analysis model based on feature-related pattern rules and probabilistic fuzzy logic modeling strategy, and process the continuous and discrete features in the input data space matrix through the fuzzy importance analysis model to generate fuzzy importance diagnostic indicators.

[0102] It should be noted that the fuzzy importance diagnostic index can be a set of indicators generated by the fuzzy importance analysis model, used to quantify the importance of features. The fuzzy importance diagnostic index includes significance coefficient, confidence coefficient, and consistency coefficient.

[0103] The significance coefficient is used to measure the degree of significance of a feature on the risk assessment results of hardware. The higher the coefficient, the stronger the influence of the feature.

[0104] The confidence coefficient is used to measure the credibility of the relationship between a feature and the evaluation result. A higher or lower coefficient corresponds to a higher reliability of the relationship between the feature and the evaluation result.

[0105] The consistency coefficient is used to measure the consistency of the influence of the same feature in different evaluation records. The higher the coefficient, the more stable the influence pattern of the feature.

[0106] It should be noted that the feature-related pattern rules can be mined from the input data space matrix and are a set of rules used to describe the relationship between features and evaluation results, as well as between different features. They can reflect the influence and correlation strength of various features on the hardware risk assessment results.

[0107] Probabilistic fuzzy logic modeling strategy can be a modeling approach that integrates probabilistic statistical methods and fuzzy logic theory. Its core purpose is to deal with the uncertainty and fuzzy information in multimodal data. By quantifying fuzzy information into probabilistic variables, it can achieve accurate analysis and modeling of fuzzy features, taking into account the fuzziness and randomness of the data.

[0108] The fuzzy importance analysis model can be built based on feature-related pattern rules and probabilistic fuzzy logic modeling strategies. It is a core model used to analyze the importance of various features in the input data space matrix. It can achieve unified processing of discrete and continuous features and quantify the influence of various features on hardware risk assessment.

[0109] Understandably, this embodiment employs a dual-path characterization and probabilistic fuzzy logic modeling (PFLM) approach—Feature-correlated pattern rule identification (FCPRI) and probabilistic fuzzy logic modeling (PFLM)—to form a multi-dimensional importance metric for low-probability, high-damage hazards. The "FCPRI+PFLM" dual-path feature identification mechanism processes discrete and continuous features separately, and incorporates the resulting fuzzy risks into the importance metric to generate three fuzzy importance diagnostic criteria. This allows low-probability, high-damage features to be evaluated alongside high-damage features, rather than being discarded as irrelevant features.

[0110] It should be understood that this embodiment, by constructing a fuzzy importance analysis model and combining feature-related pattern rules and probabilistic fuzzy logic modeling strategies, can effectively solve the problems of coexistence of discrete and continuous features and the existence of fuzziness and uncertainty in multimodal data, avoiding the limitations of a single analysis method that cannot take into account various features. By processing features through the model and generating fuzzy importance diagnostic indicators, the influence of various features on hardware risk assessment can be quantified, core features that play a key role in the assessment results can be accurately identified, irrelevant or secondary features can be eliminated, reducing computational redundancy in subsequent models, and providing clear direction and quantitative basis for the optimization of deep factorization machine models, thereby improving the pertinence and effectiveness of subsequent model optimization.

[0111] In practical implementation, generating fuzzy importance diagnostic indicators may include: feature-related pattern rule extraction, implementation of probabilistic fuzzy logic modeling strategies, construction of fuzzy importance analysis models, and feature processing and diagnostic indicator generation, specifically including:

[0112] The association rule mining algorithm is used to explore the relationship between features and evaluation results, as well as between different features, and to extract statistically significant feature-related pattern rules to clarify the potential impact of various features on the hardware evaluation results.

[0113] By integrating probabilistic statistical methods and fuzzy logic theory, and introducing core concepts such as fuzzy sets and membership functions, the uncertain and fuzzy feature information in the input data (such as the category fuzziness of discrete features and the boundary fuzziness of continuous features) is transformed into quantifiable logical variables, taking into account both the fuzziness and randomness of feature information.

[0114] Based on the extracted feature-related pattern rules, a probabilistic fuzzy logic modeling strategy is incorporated to build a model framework, define fuzzy inference rules, determine the core parameters of the model, and construct a fuzzy importance analysis model that can simultaneously handle discrete and continuous features.

[0115] The discrete features in the input data space matrix are fuzzified, and the continuous features are normalized and then fuzzified. These are then substituted into the fuzzy importance analysis model. Through the core process of fuzzy inference and probability calculation of the model, the importance of various features is quantitatively analyzed, and finally, a fuzzy importance diagnostic index containing significance coefficient, confidence coefficient, and consistency coefficient is generated.

[0116] Step S30: Optimize the deep factor decomposition machine model based on the fuzzy importance diagnostic index to generate a hardware risk assessment model.

[0117] It should be noted that the Deep Factorization Machine (DeepFM) model can be a feature interaction modeling framework that combines Factorization Machine (FM) and Deep Neural Network (DNN). Its basic idea is to simultaneously model low-order feature interactions and high-order nonlinear combinations within the same structure: on the one hand, it retains FM's ability to model the inner product of second-order feature interactions, used to characterize explicit and structurally clear interaction relationships; on the other hand, it automatically learns higher-order and more complex feature combination patterns through embedding layers and multi-layer fully connected networks.

[0118] It is understood that this embodiment uses a priori weighted deep representation learning network based on fuzzy importance diagnostic index to achieve end-to-end modeling and prediction of the health status of distribution fittings, outputting risk scores and early warnings at the equipment and line segment levels, and realizing the transformation of distributed distribution network fittings from ex-post judgment and passive maintenance to ex-ante prediction and assessment.

[0119] In its implementation, the evaluation device integrates fuzzy importance diagnostic criteria with the DeepFM structure, making the model more interpretable and calibrable under prior constraints. The significance coefficients from the fuzzy importance diagnostic indicators are incorporated as prior coefficients into the model's embedding and factorization layers. Confidence and consistency coefficients are combined as prior weights and incorporated into the FM and DNN layers of the DeepFM structure. By exposing the saliency of features and guiding the synergistic reinforcement of low-order explicit interactions and high-order nonlinear representations, the ability to identify and distinguish highly damaging features is improved, achieving better discrimination accuracy in small-sample scenarios. Its structural diagram is shown below. Figure 4 As shown, Figure 4 This is a schematic diagram of the structure of the hardware risk assessment model obtained after optimizing the deep factor decomposition machine model.

[0120] It should be noted that this embodiment optimizes the deep factor decomposition machine model by using a fuzzy importance diagnostic index, which allows the model to focus more on the core features that have a key impact on hardware risk assessment. This solves the problems of unreasonable feature weight allocation and core features being masked by secondary features in traditional deep factor decomposition machine models, and improves the model's ability to capture feature interaction relationships. The optimized hardware risk assessment model can better adapt to the characteristics of multimodal hardware data, improve the model's assessment accuracy and generalization ability, reduce model overfitting, and ensure that the model can stably and reliably output hardware risk assessment results, providing accurate model support for subsequent hardware risk score calculation and health status visualization.

[0121] Step S40: Output the risk score corresponding to each assessment record in the input data space matrix through the hardware risk assessment model, normalize the risk score, and generate a hardware health status risk heat map based on the normalized risk score.

[0122] It should be noted that the risk score can be output by the hardware risk assessment model and is used to quantify the degree of health risk of a single assessment record. The value is positively correlated with the degree of health risk of the hardware; the higher the value, the higher the health risk of the hardware.

[0123] A hardware health status risk heat map is a visual chart used to intuitively present the spatial distribution of hardware health risks. Different color gradients correspond to different normalized risk scores, clearly showing the differences in health risk levels of different regions and hardware, making it easy to quickly identify high-risk targets.

[0124] In its implementation, the hardware risk assessment model outputs a uniform risk score for each assessment record. This corresponds to a risk intensity scale of "impossible to occur → inevitable to occur," serving as the basic variable for spatial mapping. After normalizing all risk scores, a risk heatmap is generated to more intuitively display the distribution of health status of different hardware in different areas. In the risk heatmap, the color intensity reflects the spatial distribution changes in the health status of hardware, and high-risk, medium-risk, and low-risk areas are determined based on the risk distribution (red areas represent high risk, yellow areas represent medium risk, and green areas represent low risk). This helps personnel quickly locate key areas and supports inspection route planning and maintenance priority ranking, such as... Figure 5 As shown, Figure 5 This is a heat map for assessing the health status of distributed distribution network fittings in one embodiment.

[0125] Understandably, this embodiment outputs risk scores corresponding to each assessment record through a hardware risk assessment model, enabling a quantitative assessment of hardware health risks. It transforms the abstract health status of hardware into quantifiable and comparable values, improving the objectivity and accuracy of risk assessment and avoiding the subjectivity and limitations of manual assessment. Normalizing the risk scores eliminates score differences between different assessment records, achieving a unified standard for all hardware risk levels. This facilitates unified comparison, sorting, and grading of the health status of all hardware by staff. Generating a hardware health status risk heatmap transforms the quantified risk scores into intuitive visualizations, clearly presenting the spatial distribution patterns of hardware risks. This allows staff to quickly identify high-risk areas and high-risk hardware, improving the targeting and efficiency of hardware maintenance and providing intuitive and reliable support for maintenance decisions.

[0126] This embodiment aggregates multi-source data from an integrated sky-ground monitoring architecture, establishes a unified spatiotemporal reference and quality control mechanism, constructs an input data spatial matrix, employs a dual-path feature identification mechanism to process discrete and continuous features, generates three types of fuzzy importance diagnostic criteria, integrates these criteria into an improved deep factor decomposition machine model, predicts the health risk score of hardware fittings, normalizes the risk score, generates a risk heatmap, and delineates risk regions. Because this embodiment effectively integrates multimodal data and optimizes the model through fuzzy importance analysis, it accurately identifies low-frequency, high-damage hidden dangers in hardware fittings, achieving… The shift from post-event assessment and passive maintenance to pre-event prediction and evaluation of distributed distribution network fittings effectively enables forward-looking health characterization and risk prediction of distribution fittings. It also effectively mines the characteristic value of multimodal fitting data, fully leverages the data advantages of each monitoring module, improves the accuracy and efficiency of fitting risk assessment, simplifies the assessment process, reduces the cost and difficulty of manual assessment, and visually presents the risk distribution through heat maps. This provides a scientific and reliable basis for the operation and maintenance of fittings, effectively avoids the risk of fitting failure, ensures the safe and stable operation of fittings, and comprehensively improves the level of intelligence in fitting health management.

[0127] refer to Figure 6 , Figure 6 This is a flowchart illustrating the second embodiment of the distributed power distribution network hardware health status assessment method under the sky-ground monitoring system of the present invention.

[0128] Based on the first embodiment described above, in this embodiment, step S20 further includes:

[0129] Step S201: Construct a feature-related pattern rule recognition module and a probabilistic fuzzy logic modeling module based on feature-related pattern rules and probabilistic fuzzy logic modeling strategies, and combine them to form a fuzzy importance analysis model.

[0130] It should be noted that the feature-related pattern rule recognition module is a functional module that uses feature-related pattern rules as its core basis and adopts association rule mining algorithms (such as the Apriori algorithm) to process discrete features, mine the correlation between discrete features and evaluation results, and generate feature-related rule sets.

[0131] The probabilistic fuzzy logic modeling module is based on the probabilistic fuzzy logic modeling strategy, integrating fuzzy set theory and probabilistic statistical methods. It is a functional module specifically designed to handle continuous features, resolve the fuzziness and uncertainty of continuous features, and generate continuous risk scores.

[0132] It is understandable that this embodiment adopts a modular design to construct two core modules, clearly defining the functional boundaries of each module. This enables the separate processing and collaborative optimization of discrete and continuous features, avoiding the limitation that a single module cannot take into account both types of features. By configuring reasonable algorithms and parameters, the rule recognition module can efficiently mine the correlation between discrete features, and the modeling module can accurately handle the fuzziness of continuous features. The fuzzy importance analysis model formed by the combination of the two has the core capability to simultaneously process multiple types of features and quantify the importance of features.

[0133] Step S202: Input the discrete features in the input data space matrix into the feature-related pattern rule recognition module to generate a feature-related rule set.

[0134] It should be noted that the feature-related rule set is output by the feature-related pattern rule recognition module. It is a structured collection of discrete feature association rules, presented in the form of "feature combination → evaluation result". It includes core information such as rule support and confidence, and can reflect the influence of discrete features on the hardware risk assessment results.

[0135] In some embodiments, the evaluation device can extract all discrete features (such as hardware defect type, installation environment category, material type, etc.) and corresponding evaluation results from the input data space matrix, encode the discrete features (using one-hot encoding to convert categorical features into numerical features), and remove redundant features after encoding; input the processed discrete feature data into the feature-related pattern rule recognition module, and use the built-in Apriori algorithm of the module to mine association rules that meet the minimum support and minimum confidence requirements, deduplicate and filter the mined rules, and remove invalid rules without actual physical meaning (such as feature association rules unrelated to hardware health), and finally generate a structured feature-related rule set. The rule set is stored in the form of "feature combination → evaluation result", containing core information such as rule ID, feature combination, support, and confidence.

[0136] Furthermore, in order to effectively mine discrete features, step S202 above may include:

[0137] Step S2021: Obtain the set of transaction numbers corresponding to the discrete features in the input data space matrix;

[0138] Step S2022: Calculate the intersection of the set of transaction numbers corresponding to the preceding feature and the subsequent evaluation result to obtain the preceding-subsequent candidate pair;

[0139] Step S2023: Vertically expand the features of the preceding term in the preceding-following term candidate pair to obtain multiple preceding term combinations;

[0140] Step S2024: Prune the multi-antecedent combination based on anti-monotonicity, retain stable co-occurring multi-antecedent-consequence candidate pairs, and generate a feature-related rule set.

[0141] In practical implementation, focusing on discrete characteristics, we take... for For any subset of , define a feature-related pattern rule as follows: (and ), representing the current item feature group When it appears, the subsequent item is marked It is bound to happen. A scan of the dataset monitored by the air-space-ground monitoring system is performed, and Eclat is selected as the core method of FCPRI. For any level item... A set of transaction IDs that record its occurrence. The preceding term is obtained according to the following formula. Direct mapping of risk:

[0142]

[0143] We obtain candidate pairs of "antecedent-consequence". Similarly, we continue to vertically expand the antecedent to form multiple antecedent combinations:

[0144]

[0145] By leveraging anti-monotonicity (if any item set is infrequent, then any subset thereof is also infrequent), pruning is performed to retain only the stable co-occurring "multiple antecedent-consequence" candidates, thereby generating a rule set that is significantly relevant to and interpretable in relation to risk outcomes.

[0146] Step S203: Input the continuous features in the input data space matrix into the probabilistic fuzzy logic modeling module to generate a continuous risk score.

[0147] It should be noted that the continuous risk score is output by the probabilistic fuzzy logic modeling module and is used to quantify the degree of risk of the hardware corresponding to the continuous feature. The score range is usually [0,1]. The score is positively correlated with the degree of risk of the hardware corresponding to the continuous feature and can reflect the degree of influence of the continuous feature on the health status of the hardware.

[0148] In some embodiments, the evaluation device can extract all continuous features (such as hardware operating temperature, vibration frequency, corrosion rate, etc.) from the input data space matrix, perform Min-Max normalization on the continuous features, map them to the [0,1] interval, and eliminate dimensional differences; input the normalized continuous features into the probabilistic fuzzy logic modeling module, and use the built-in triangular membership function of the module to fuzzify the continuous features into fuzzy sets (such as three fuzzy subsets: low temperature, medium temperature, and high temperature), calculate the membership degree of each continuous feature to each fuzzy subset; combine the Bayesian probability method to calculate the risk probability corresponding to each fuzzy subset, and perform a weighted product operation on the membership degree and risk probability. The weighting coefficient is set according to the actual influence weight of the feature (such as the weight of temperature feature is set to 0.3, and the weight of vibration frequency is set to 0.25), and finally output the continuous risk score corresponding to each continuous feature. The score range is mapped to [0,1], and the higher the score, the higher the risk of the hardware corresponding to the continuous feature.

[0149] Furthermore, in order to accurately mine continuous features, step S203 above may include:

[0150] Step S2031: Obtain the feature probability distribution curve of continuous features in the input data space matrix, and set preset boundary values ​​to divide multiple input regions;

[0151] Step S2032: Adaptively generate multiple overlapping trapezoidal input membership functions based on the input region;

[0152] Step S2033: Perform pairwise rule decoupling calculations on the continuous features using a hierarchical structure, and aggregate risk evidence layer by layer;

[0153] Step S2034: Defuzzify the aggregated risk evidence using the centroid method to generate a continuous risk score.

[0154] In its implementation, for continuous features, the system first sets boundary values ​​based on the Feature Probability Distribution Curve (FPDC) of each feature and divides the input region into four segments: Normal (N), Slight (M), Moderate (G), and Severe (S). The probability of occurrence in each segment is calculated, and four overlapping trapezoidal input membership functions are adaptively generated accordingly, avoiding subjective uncertainty introduced by manual hard thresholding. To balance efficiency and scalability, a hierarchical Mamdani structure is adopted, decoupling the calculation of pairwise feature rules from the combination rules of "feature + previous layer output," aggregating risk evidence layer by layer. Logical AND operations take the minimum membership degree to ensure convergence and interpretability. The output end sets four triangular membership functions corresponding to four risk levels: Low (L), Low-Medium (ML), Medium-High (MH), and High (H), with weights configured according to historical statistics to reflect the relative risk dimensions from mild to severe. For a single record, the membership degree of each level of risk is first calculated based on the input membership degree and the rule base, then weighted aggregation is performed, and finally, the centroid method is used for defuzzification to obtain a calculable continuous risk score and classification result. For example... Figure 7 Figure 8(a) and Figure 8(b) are flowcharts of the probabilistic fuzzy logic modeling module. Figure 8(a) is a flowchart of Step 1, and Figure 8(b) is a flowchart of Step 2. The surface temperature of the fittings is 14.3°C, the humidity is 55%, the vibration frequency is 26.5Hz, and the final probabilistic fuzzy risk is 0.809.

[0155] Step S204: Based on the feature-related rule set and the continuous risk score, construct calculation functions for the significance coefficient, confidence coefficient and consistency coefficient, respectively.

[0156] It should be noted that the calculation function can take the core parameters (support, confidence) of the feature-related rule set and the statistical characteristics (mean, standard deviation, variance) of the continuous risk score as inputs, and output the corresponding coefficient values. It is the core tool for generating fuzzy importance diagnostic indicators.

[0157] Step S205: Calculate the feature-related rule set and the continuous risk score using the calculation function to generate the fuzzy importance diagnostic index.

[0158] In some embodiments, the evaluation device can input the feature-related rule set and continuous risk scores into the constructed calculation function; using the significance coefficient calculation function, it calculates the rule support and mean risk score for each feature one by one to obtain the significance coefficient of each feature; using the confidence coefficient calculation function, it calculates the rule confidence and standard deviation of risk score for each feature one by one to obtain the confidence coefficient of each feature; using the consistency coefficient calculation function, it calculates the rule support variance and risk score variance for each feature one by one to obtain the consistency coefficient of each feature; the three coefficients are grouped by feature and integrated to form a complete fuzzy importance diagnostic index, which is stored in matrix form, with each row corresponding to a feature and each column corresponding to a coefficient.

[0159] Furthermore, in order to achieve multi-dimensional quantification of feature importance, step S205 above may include:

[0160] Step S2051: Expand the feature-related rule set into rule pairs that include the fuzzy set corresponding to the previous feature factor set and the fuzzy set corresponding to the subsequent result target factor set;

[0161] Step S2052: When the total participation of the feature fuzzy set exceeds the preset participation threshold, calculate the significance coefficient;

[0162] Step S2053: Combine the set of feature factors of the preceding term and the set of target factors of the result term to obtain the combined set;

[0163] Step S2054: Calculate the confidence coefficient based on the conjunctive set and the feature-related rule set;

[0164] Step S2055: Generate a consistency coefficient by calculating the expectations of the fuzzy set corresponding to the set of feature factors of the preceding term, the fuzzy set corresponding to the set of target factors of the following term, and the fuzzy set corresponding to the conjunctive set;

[0165] Step S2056: Integrate the significance coefficient, confidence coefficient, and consistency coefficient to obtain the fuzzy importance diagnostic index.

[0166] In practical implementation, considering that some hardware failures have the characteristics of low probability and high damage, and are easily regarded as insignificant features and discarded, three fuzzy importance diagnostic criteria are constructed by combining fuzzy risk level and score: significance coefficient SC, confidence coefficient CC, and consistency coefficient CnC. Firstly, the feature-related pattern rule is further extended to... ,in and Respectively represent and and Corresponding fuzzy sets. If the feature-fuzzy set pair has sufficient historical data involved, and the sum of these involvements exceeds an involvement threshold, a risk pattern can be identified. The SC expression is:

[0167]

[0168] In the formula, and They represent the total number of records in S and The total number of occurrences in express One of the features Factors, This represents the base number of records that simultaneously meet all internal standards. Representing fuzzy sets The fuzzy membership degree, if the feature exhibits low probability and high damage, is expressed as follows:

[0169]

[0170] In the formula, for middle The value, This indicates that the support coefficient is included in the threshold. and Indicates and Related fuzzy sets The level of ambiguity risk This indicates the weight in the fuzzy rule table corresponding to the fuzzy risk value.

[0171] The previous feature set and the feature set of the latter term Combined ,in , , , The CC expression is as follows:

[0172]

[0173] In the formula, for Relevant characteristic factors in Representing fuzzy sets Membership degree. If the feature exhibits low probability - high damage, and Its expression is as follows:

[0174]

[0175]

[0176] In the formula, express exist The value in This indicates that the confidence coefficient is involved in the threshold. One of the final results indicating the health status of hardware. and Representing fuzzy sets The level of ambiguity risk and This indicates the corresponding weight.

[0177] To better evaluate and The degree of compactness, CnC expression is as follows:

[0178]

[0179] in, Represents the consistency coefficient. Let represent the mathematical expectation of the fuzzy set corresponding to the conjunctive set. This represents the mathematical expectation of the set of preceding feature factors and their corresponding fuzzy sets. This represents the mathematical expectation of the target factor set B and its corresponding fuzzy set D at the result end of the subsequent term. Represents the set of the preceding characteristic factors. This represents the fuzzy set corresponding to the set of the preceding feature factors. This represents the set of target factors at the end of the subsequent result. This represents the fuzzy set corresponding to the target factor set of the subsequent result;

[0180] The solution can be obtained using the following formula:

[0181]

[0182]

[0183]

[0184]

[0185]

[0186] In the formula, , , They represent , , For the fuzzy membership degree, if the feature exhibits low probability and high damage, the expressions for the three are as follows:

[0187]

[0188]

[0189]

[0190] In the formula, This indicates that the correlation coefficient is included in the threshold. and Representing fuzzy sets The level of ambiguity risk This indicates the corresponding weight.

[0191] This embodiment achieves separate processing and collaborative analysis of discrete and continuous features, transforming the correlation patterns and fuzzy information of features into quantifiable diagnostic indicators. Overall, it effectively solves the problems of coexistence of two types of features and the existence of fuzziness and uncertainty in multimodal hardware data, accurately mines the core influence patterns of features, quantifies the importance of features, and the generated fuzzy importance diagnostic indicators can provide high-quality support for subsequent model optimization. At the same time, it improves the accuracy and reliability of the feature analysis link in the entire technical solution, laying a solid feature analysis foundation for hardware risk assessment.

[0192] refer to Figure 9 , Figure 9 This is a flowchart illustrating the third embodiment of the distributed distribution network hardware health status assessment method under the sky-ground monitoring system of the present invention.

[0193] Based on the above embodiments, in this embodiment, step S30 further includes:

[0194] Step S301: Construct a deep factorization machine model, which includes an input layer, a factorization machine layer, and a deep neural network layer;

[0195] Step S302: Incorporate the significance coefficients in the fuzzy importance diagnostic index as prior coefficients into the embedding layer and the factor decomposition machine layer;

[0196] Step S303: Combine the confidence coefficient and consistency coefficient in the fuzzy importance diagnostic index to generate a weighted coefficient;

[0197] Step S304: Multiply the prior coefficients and the weighting coefficients with the eigenvalues ​​of the first-order terms of the factorization machine layer;

[0198] Step S305: The prior coefficients are used to modulate the embedding vector of the input features to obtain the modulated embedding vector. Then, the inner product of the weighting coefficients and the modulated embedding vector is multiplied by the feature value corresponding to the second-order interaction term.

[0199] Step S306: Summing the calculated first-order and second-order interaction terms with the bias term to generate the optimized factorization machine layer output;

[0200] Step S307: Apply the prior coefficients and the weighting coefficients to the embedding features of the deep neural network layer;

[0201] Step S308: Integrate the optimized input layer, factorization machine layer, and deep neural network layer to generate a hardware risk assessment model.

[0202] It should be noted that the Deep Factorization Machine (DFM) model includes: an input data layer, a sparse feature layer, an embedding layer, a prior coefficient layer, a Factorization Machine (FM) layer, a Deep Neural Network (DNN) layer, a weighted coefficient layer, and an output layer (Sigmoid layer). The embedding layer is located above the sparse feature layer, and its output serves as the input to both the FM and DNN layers. The embedding layer transforms high-dimensional sparse features into low-dimensional dense embedding vectors, effectively capturing the relationships between features, mitigating the curse of dimensionality, and improving model training efficiency and prediction performance.

[0203] The input layer is the first layer of the deep factorization machine model. Its core function is to receive input data (significant coefficient vectors in this step), complete data format conversion and preliminary feature mapping, and transform the input data into a feature format that is suitable for subsequent layers. It is the foundation for model data transmission.

[0204] The factorization machine layer is one of the core layers of the deep factorization machine model. It is divided into first-order term computation units and second-order interaction term computation units. The first-order term is used to capture the linear influence of a single feature, while the second-order interaction term is used to capture the interaction influence between any two features. It is the key to the model capturing the linear relationship and low-order interaction relationship of features.

[0205] The deep neural network layer is one of the core layers of the deep factorization machine model. It consists of multiple perceptrons and its core function is to capture the high-order nonlinear interaction relationships between features, making up for the inability of the factorization machine layer to capture high-order interactions and improving the model's fitting and prediction capabilities.

[0206] It should be noted that the prior coefficient is a coefficient mapped from the significance coefficient in the fuzzy importance diagnostic index; the weighted coefficient is a coefficient generated by weighting the confidence coefficient and consistency coefficient in the fuzzy importance diagnostic index. It is used to perform weighted optimization on the features of each layer of the model, reflecting the credibility of the features and the influence of consistency on the model output. It is the core intermediate parameter for model optimization.

[0207] The eigenvalues ​​of the first-order terms are the outputs of the first-order term calculation units of the factorization machine layer. They correspond to the degree of linear influence of a single feature on the model output, with higher and lower values ​​corresponding to the strength of the linear influence of the feature.

[0208] The inner product of the embedding vectors of the second-order interaction term is an intermediate result in the calculation of the second-order interaction term in the factorization machine layer. First, the features are mapped to low-dimensional embedding vectors, and then the inner product of any two embedding vectors is calculated to measure the interaction strength between the two features. The larger the inner product, the more significant the feature interaction.

[0209] Embedded features are low-dimensional feature vectors output by the embedding layer of a deep neural network. They are a further compression and mapping of the input features, which reduces the computational complexity of the model while retaining the core information of the features, and facilitates the subsequent capture of higher-order interaction relationships.

[0210] Understandably, the collected data contains both discrete and continuous attributes, and some of these attributes exhibit low-probability, high-damage characteristics. Directly inputting the raw features into the model may fail to adequately represent these heterogeneous features and may not fully capture the feature patterns associated with low-probability, high-damage characteristics. Therefore, this section embeds a fuzzy importance diagnostic metric into the model to obtain more reliable evaluation results in real-world, complex data scenarios.

[0211] When multi-source raw observations are directly input into the model, the discrimination signal of low-probability-high-damage modes may be overwhelmed by the dominant mode. The model first maps the calculated SC to prior coefficients to explicitly strengthen the significance of risk, thereby improving the model's sensitivity to low-probability-high-damage scenarios and enhancing its overall discrimination ability.

[0212] This is then introduced into the encoding layer (Embedding layer) to map each high-dimensional sparse feature to a low-dimensional dense feature, as shown in the following expression:

[0213]

[0214] All embedded vectors are concatenated to form an embedded feature sequence, which also serves as a shared input sequence for both the FM and DNN branches, thus realizing an integrated input path of "prior-embedding-sharing".

[0215] To determine the degree to which the importance of a particular feature is amplified, we define a weighting coefficient. Its expression is as follows:

[0216]

[0217] in and These represent the learning coefficients obtained from the CC and CnC mappings, respectively. Parameters and Controlling the relative emphasis on confidence priors and consistent priors to satisfy... ,and , .

[0218] The FM layer not only possesses the functionality of a linear regression model but also introduces a second-order combination term, enabling the learning of interactions between features. In the FM layer, the first-order term maintains linear interpretability, and the introduced prior weights can be used to characterize the marginal contribution of a single feature to risk. The second-order term characterizes the explicit interactions between paired features, and prior weights are introduced for the "stable co-occurrence-robust interaction" risk pattern to amplify effective coupling and suppress accidental co-occurrence. Its expression is shown below:

[0219]

[0220] in, and These represent the i-th and j-th input features, respectively. This represents the weighting coefficient corresponding to the i-th preceding feature. These represent the prior coefficients, which are used during the encoding process to gate and modulate the embedded representation. This represents the weighting coefficient corresponding to the interaction of the i-th and j-th preceding features. Indicates the bias term. and This represents the embedding vector of the i-th and j-th preceding features. This represents the vector inner product operation, used to characterize the degree of second-order interaction between features. It represents the mapping function for embedding operations from high-dimensional sparse features to low-dimensional dense features.

[0221] DNN layers are used to process feature combinations of order three and above, performing high-order feature crossing on the input to enhance the interaction between features. First, the embedded features are modified as follows:

[0222]

[0223] in, This represents the weighting coefficient corresponding to the i-th preceding feature. This represents the weighting coefficient corresponding to the interaction of the i-th and j-th preceding features. Indicates the bias term. and This represents the embedding vector of the i-th and j-th preceding features. This represents the vector inner product operation, used to characterize the degree of second-order interaction between features. It represents the mapping function for embedding operations from high-dimensional sparse features to low-dimensional dense features.

[0224] The concatenated embedding vectors are then used as the final input to the DNN and substituted into the following formula to calculate the output of each layer of the DNN up to the last layer. The expression for each layer is as follows:

[0225]

[0226] in, This represents the weight matrix of the l-th layer. This represents the bias vector of the l-th layer. This represents the output feature of the l-th layer. Indicates the first The output features of the layer The ReLU activation function is expressed as follows:

[0227]

[0228] The final layer output uses linear readout. The DNN output is then fused with the FM output, and the final risk probability is obtained through Sigmoid. The final prediction expression is as follows:

[0229]

[0230] in, This represents the risk score output by the model. It is the sigmoid activation function. This represents the output features of the last layer in a deep neural network. This represents the output of the factorization machine layer.

[0231] This embodiment deeply integrates multi-dimensional information of fuzzy importance diagnostic indicators into each layer of the model, optimizing the entire process from model input, intermediate calculation, to output. Overall, it strengthens the influence of core features, weakens the interference of secondary features, improves the model's accuracy in capturing various relationships between features, reduces model overfitting and prediction bias, and makes the generated hardware risk assessment model more suitable for the characteristics of multimodal hardware data. This provides more accurate and stable model support for subsequent risk score output, further improving the reliability of the risk assessment of the entire technical solution.

[0232] Furthermore, this embodiment of the invention also proposes a computer-readable storage medium storing a distributed distribution network hardware health status assessment program. When the distributed distribution network hardware health status assessment program is executed by a processor, it implements the steps of the distributed distribution network hardware health status assessment method under the sky-ground monitoring system described above.

[0233] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0234] The aforementioned computer-readable storage medium may be included in the distributed distribution network fittings health status assessment equipment under the sky-ground monitoring system; or it may exist independently and not be installed in the distributed distribution network fittings health status assessment equipment under the sky-ground monitoring system.

[0235] Furthermore, this invention also proposes a computer program product, including a distributed distribution network hardware health status assessment program, which, when executed by a processor, implements the steps of the distributed distribution network hardware health status assessment method under the above-described sky-ground monitoring system.

[0236] The specific implementation of the computer program product of the present invention is basically the same as the embodiments of the distributed distribution network hardware health status assessment method under the above-mentioned sky-ground monitoring system, and will not be repeated here.

[0237] Reference Figure 10 , Figure 10 This is a structural block diagram of the first embodiment of the distributed power distribution network hardware health status assessment device under the sky-ground monitoring system of the present invention.

[0238] like Figure 10 As shown in the embodiment of the present invention, the distributed distribution network hardware health status assessment device under the sky-ground monitoring system includes:

[0239] The data processing module 10 preprocesses the multimodal hardware data collected by the sky-ground monitoring system and maps it to the same dimensional space to construct an input data space matrix. The sky-ground monitoring system includes a space-based module, an air-based module, and a ground-based module. The input data space matrix includes multiple evaluation records. Each evaluation record includes a feature set and an evaluation result. The feature set contains discrete features and continuous features. Both the discrete features and the continuous features are composed of multiple factors.

[0240] The dual-path feature processing module 20 is used to construct a fuzzy importance analysis model based on feature-related pattern rules and probabilistic fuzzy logic modeling strategies, and to process the continuous and discrete features in the input data space matrix through the fuzzy importance analysis model to generate fuzzy importance diagnostic indicators, which include significance coefficient, confidence coefficient and consistency coefficient.

[0241] Model optimization module 30 is used to optimize the deep factor decomposition machine model based on the fuzzy importance diagnostic index to generate a hardware risk assessment model.

[0242] The risk assessment module 40 is used to output the risk score corresponding to each assessment record in the input data space matrix through the hardware risk assessment model, and to normalize the risk score, and generate a hardware health status risk heat map based on the normalized risk score.

[0243] This embodiment aggregates multi-source data from an integrated sky-ground monitoring architecture, establishes a unified spatiotemporal reference and quality control mechanism, constructs an input data spatial matrix, employs a dual-path feature identification mechanism to process discrete and continuous features, generates three types of fuzzy importance diagnostic criteria, integrates these criteria into an improved deep factor decomposition machine model, predicts the health risk score of hardware fittings, normalizes the risk score, generates a risk heatmap, and delineates risk regions. Because this embodiment effectively integrates multimodal data and optimizes the model through fuzzy importance analysis, it accurately identifies low-frequency, high-damage hidden dangers in hardware fittings, achieving… The shift from post-event assessment and passive maintenance to pre-event prediction and evaluation of distributed distribution network fittings effectively enables forward-looking health characterization and risk prediction of distribution fittings. It also effectively mines the characteristic value of multimodal fitting data, fully leverages the data advantages of each monitoring module, improves the accuracy and efficiency of fitting risk assessment, simplifies the assessment process, reduces the cost and difficulty of manual assessment, and visually presents the risk distribution through heat maps. This provides a scientific and reliable basis for the operation and maintenance of fittings, effectively avoids the risk of fitting failure, ensures the safe and stable operation of fittings, and comprehensively improves the level of intelligence in fitting health management.

[0244] The distributed distribution network fittings health status assessment device under the sky-ground monitoring system provided in this application adopts the distributed distribution network fittings health status assessment method under the sky-ground monitoring system in the above embodiments, and can solve the technical problem of distributed distribution network fittings health status assessment. Compared with the prior art, the beneficial effects of the distributed distribution network fittings health status assessment device under the sky-ground monitoring system provided in this application are the same as the beneficial effects of the distributed distribution network fittings health status assessment method under the sky-ground monitoring system provided in the above embodiments, and other technical features in the distributed distribution network fittings health status assessment device under the sky-ground monitoring system are the same as the features disclosed in the methods of the above embodiments, and will not be repeated here.

[0245] It should be understood that the above are merely illustrative examples and do not constitute any limitation on the technical solutions of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any restrictions on this.

[0246] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0247] In addition, for technical details not described in detail in this embodiment, please refer to the method for assessing the health status of distributed distribution network fittings under the sky-ground monitoring system provided in any embodiment of the present invention, which will not be repeated here.

[0248] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0249] It should be noted that the user information (including but not limited to user device information, user personal information, user location information, user behavior information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0250] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0251] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0252] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for assessing the health status of distributed distribution network fittings under a sky-ground monitoring system, characterized in that, The method for assessing the health status of distributed distribution network fittings under the aforementioned sky-ground monitoring system includes: The multimodal hardware data collected by the sky-ground monitoring system are preprocessed and mapped to the same dimensional space to construct an input data space matrix. The sky-ground monitoring system includes a space-based module, an air-based module, and a ground-based module. The input data space matrix includes multiple evaluation records. Each evaluation record includes a feature set and an evaluation result. The feature set contains discrete features and continuous features. Both the discrete features and the continuous features are composed of multiple factors. A fuzzy importance analysis model is constructed based on feature-related pattern rules and probabilistic fuzzy logic modeling strategies. The continuous and discrete features in the input data space matrix are processed by the fuzzy importance analysis model to generate fuzzy importance diagnostic indicators, which include significance coefficient, confidence coefficient and consistency coefficient. The deep factor decomposition machine model is optimized based on the aforementioned fuzzy importance diagnostic index to generate a hardware risk assessment model. The hardware risk assessment model outputs the risk score corresponding to each assessment record in the input data space matrix, and normalizes the risk score. Based on the normalized risk score, a hardware health status risk heat map is generated. The optimization of the deep factor decomposition machine model based on the fuzzy importance diagnostic index to generate a hardware risk assessment model includes: A deep factorization machine model is constructed, which includes an input layer, a factorization machine layer, and a deep neural network layer. The significance coefficients in the fuzzy importance diagnostic index are incorporated as prior coefficients into the embedding layer and the factor decomposition machine layer; The confidence coefficient and consistency coefficient in the fuzzy importance diagnostic index are combined to generate a weighted coefficient; The prior coefficients and the weighting coefficients are multiplied by the eigenvalues ​​of the first-order terms of the factorization machine layer. The prior coefficients are used to modulate the embedding vector of the input features to obtain the modulated embedding vector. Then, the inner product of the weighting coefficients and the modulated embedding vector is multiplied by the feature value corresponding to the second-order interaction term. The first-order and second-order interaction terms after computation are summed with the bias term to generate the optimized factorization machine layer output; The prior coefficients and the weighting coefficients are applied to the embedding features of the deep neural network layer; By integrating and optimizing the input layer, factorization machine layer, and deep neural network layer, a hardware risk assessment model is generated.

2. The method for assessing the health status of distributed distribution network fittings under the sky-ground monitoring system as described in claim 1, characterized in that, The fuzzy importance analysis model is constructed based on feature-related pattern rules and probabilistic fuzzy logic modeling strategy. This model is then used to process continuous and discrete features in the input data space matrix to generate fuzzy importance diagnostic indicators, including: Based on the feature-related pattern rule and probabilistic fuzzy logic modeling strategies, a feature-related pattern rule recognition module and a probabilistic fuzzy logic modeling module are constructed and combined to form a fuzzy importance analysis model. The discrete features in the input data space matrix are input into the feature-related pattern rule recognition module to generate a feature-related rule set. The continuous features in the input data space matrix are input into the probabilistic fuzzy logic modeling module to generate continuous risk scores; Based on the feature-related rule set and the continuous risk score, calculation functions for the significance coefficient, confidence coefficient, and consistency coefficient are constructed respectively. The fuzzy importance diagnostic index is generated by performing calculations on the feature-related rule set and the continuous risk score using the calculation function.

3. The method for assessing the health status of distributed distribution network fittings under the sky-ground monitoring system as described in claim 2, characterized in that, The step of inputting discrete features from the input data space matrix into the feature-related pattern rule recognition module to generate a feature-related rule set includes: Obtain the set of transaction numbers corresponding to the discrete features in the input data space matrix; Calculate the intersection of the set of transaction IDs corresponding to the preceding feature and the subsequent evaluation result to obtain the preceding-subsequent candidate pair; The features of the preceding term in the preceding-following candidate pair are expanded vertically to obtain multiple preceding combinations; Based on antimonotonicity, the multiple antecedent combinations are pruned to retain stable co-occurring multiple antecedent-consequence candidate pairs, generating a feature-related rule set.

4. The method for assessing the health status of distributed distribution network fittings under the sky-ground monitoring system as described in claim 2, characterized in that, The step of inputting continuous features from the input data space matrix into the probabilistic fuzzy logic modeling module to generate continuous risk scores includes: Obtain the feature probability distribution curves of continuous features in the input data space matrix, and set preset boundary values ​​to divide multiple input regions; Multiple overlapping trapezoidal input membership functions are adaptively generated based on the input region; A hierarchical structure is used to perform pairwise rule-decoupled calculations on the continuous features, and risk evidence is aggregated layer by layer. The centroid method is used to deblur the aggregated risk evidence and generate a continuous risk score.

5. The method for assessing the health status of distributed distribution network fittings under the sky-ground monitoring system as described in claim 2, characterized in that, The step of calculating the fuzzy importance diagnostic index by performing calculations on the feature-related rule set and the continuous risk score using the calculation function includes: The feature-related rule set is expanded to include rule pairs containing fuzzy sets corresponding to the set of preceding feature factors and fuzzy sets corresponding to the set of subsequent result-end target factors; When the sum of participation corresponding to the feature fuzzy set exceeds the preset participation threshold, the significance coefficient is calculated, referring to the following formula: in, Indicates the significance coefficient. Let S represent the i-th evaluation record in the input data space matrix. Represents the set of the preceding characteristic factors. Let represent the set of transaction IDs corresponding to the evaluation records that satisfy the i-th feature-related rule. and These represent the total number of records in the input data space matrix S and the total number of occurrences in the set of preceding feature factors, respectively. Indicates evaluation record Chinese characteristics Factors, This represents the fuzzy membership degree of the j-th feature in the set of preceding feature factors A corresponding to the i-th evaluation record. This represents the fuzzy set associated with the j-th antecedent feature in the i-th evaluation record. The level of ambiguity risk Indicates the feature risk adjustment weight. This indicates the preset participation threshold. This represents the fuzzy set corresponding to the set of the preceding feature factors; The conjunction is obtained by combining the set of feature factors of the preceding term and the set of target factors of the result term of the following term. The confidence coefficient is calculated based on the conjunct set and the feature-related rule set, referring to the following formula: in, Represents the confidence coefficient. The conjunctive set is obtained by combining the set of feature factors A of the preceding term with the set of target factors B of the following term. This represents the i-th evaluation result in the conjunction set. This represents the membership degree of the evaluation result in the conjunction set corresponding to the i-th evaluation record; A consistency coefficient is generated by calculating the expectations of the fuzzy sets corresponding to the set of feature factors of the preceding term, the fuzzy sets corresponding to the set of target factors of the following term, and the fuzzy sets corresponding to the conjunctive set, as shown in the following formula: in, Represents the consistency coefficient. Let represent the mathematical expectation of the fuzzy set corresponding to the conjunctive set. This represents the mathematical expectation of the set of preceding feature factors and their corresponding fuzzy sets. This represents the mathematical expectation of the target factor set B and its corresponding fuzzy set D at the result end of the subsequent term. Represents the set of the preceding characteristic factors. This represents the fuzzy set corresponding to the set of the preceding feature factors. This represents the set of target factors at the end of the subsequent result. This represents the fuzzy set corresponding to the target factor set of the subsequent result; By integrating the significance coefficient, confidence coefficient, and consistency coefficient, a fuzzy importance diagnostic index is obtained.

6. The method for assessing the health status of distributed distribution network fittings under the sky-ground monitoring system as described in claim 1, characterized in that, The mathematical expression for the hardware risk assessment model is: in, This represents the risk score output by the model. It is the sigmoid activation function. This represents the output features of the last layer in a deep neural network. This represents the output of the factorization machine layer; The mathematical expression for each layer of the deep neural network is: in, This represents the weight matrix of the l-th layer. This represents the bias vector of the l-th layer. Represents the ReLU activation function. This represents the output feature of the l-th layer. Indicates the first The output features of the layer; The mathematical expression for the factorization engine layer is: in, and These represent the i-th and j-th input features, respectively. This represents the weighting coefficient corresponding to the i-th preceding feature. These represent the prior coefficients, which are used during the encoding process to gate and modulate the embedded representation. This represents the weighting coefficient corresponding to the interaction of the i-th and j-th preceding features. Indicates the bias term. and This represents the embedding vector of the i-th and j-th preceding features. This represents the vector inner product operation, used to characterize the degree of second-order interaction between features. It represents the mapping function for embedding operations from high-dimensional sparse features to low-dimensional dense features.

7. A device for assessing the health status of distributed distribution network fittings under a sky-ground monitoring system, characterized in that, The device is configured to implement the distributed distribution network fittings health status assessment method under the sky-ground monitoring system as described in any one of claims 1 to 6, and the device includes: The data processing module preprocesses the multimodal hardware data collected by the sky-ground monitoring system and maps it to the same dimensional space to construct an input data space matrix. The sky-ground monitoring system includes a space-based module, an air-based module, and a ground-based module. The input data space matrix includes multiple evaluation records. Each evaluation record includes a feature set and an evaluation result. The feature set contains discrete features and continuous features, and both the discrete features and the continuous features are composed of multiple factors. The dual-path feature processing module is used to construct a fuzzy importance analysis model based on feature-related pattern rules and probabilistic fuzzy logic modeling strategies, and to process the continuous and discrete features in the input data space matrix through the fuzzy importance analysis model to generate fuzzy importance diagnostic indicators, which include significance coefficient, confidence coefficient and consistency coefficient. The model optimization module is used to optimize the deep factor decomposition machine model based on the fuzzy importance diagnostic index to generate a hardware risk assessment model. The risk assessment module is used to output the risk score corresponding to each assessment record in the input data space matrix through the hardware risk assessment model, and to normalize the risk score, and generate a hardware health status risk heat map based on the normalized risk score.

8. A distributed distribution network fittings health status assessment device under a sky-ground monitoring system, characterized in that, The distributed distribution network hardware health status assessment device under the sky-ground monitoring system includes: a memory, a processor, and a distributed distribution network hardware health status assessment program stored in the memory. The processor is used to run the distributed distribution network hardware health status assessment program, which is configured to implement the distributed distribution network hardware health status assessment method under the sky-ground monitoring system as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a distributed distribution network fittings health status assessment program, which, when executed by a processor, implements the distributed distribution network fittings health status assessment method under the sky-ground monitoring system as described in any one of claims 1 to 6.

Images