Network side fault diagnosis method, system and device for PDA device in electric power Internet of Things, and medium

Abstract

The invention discloses a network side fault diagnosis method, system and device for PDA equipment in the electric power Internet of Things and a medium, and relates to the field of network management and fault diagnosis of the electric power Internet of Things, and the method comprises the steps: collecting the operation data of the PDA equipment at a self-adaptive sampling frequency, and carrying out the abnormal judgment of the equipment; performing path tracking based on an equipment abnormity judgment result, determining a specific network node where a fault occurs, and obtaining position information of a fault point; performing fault diagnosis based on the equipment abnormity judgment result and the position information of the fault point to obtain a fault diagnosis result; and generating a diagnosis report and a repair suggestion based on the fault diagnosis result. The PDA equipment fault detection timeliness, the positioning speed, the diagnosis accuracy and the maintenance efficiency in the electric power Internet of Things are improved, and the overall reliability and stability of the electric power Internet of Things are enhanced.

Description

Technical Field

[0001] This invention relates to the field of power Internet of Things (IoT) network management and fault diagnosis, and in particular to a method, system, device and medium for network-side fault diagnosis of PDA devices in power IoT. Background Technology

[0002] Partial discharge (PD) monitoring is a core technology for ensuring the reliable operation of high-voltage equipment (such as transformers, cables, and switchgear) in power systems. Significant progress has been made in PD monitoring technology in recent years. Traditional PD monitoring relies on offline testing during power outages, which is inefficient. Today, online monitoring systems are becoming mainstream, enabling real-time assessment of insulation health during normal equipment operation. For example, instruments can monitor partial discharge in hydroelectric generators through specific capacitors and provide early fault warnings. However, there are still many shortcomings in network-side fault diagnosis using PDA (Partial Discharge Analyzer) devices in the power Internet of Things (IoT). In complex power networks, especially in IoT environments with interconnected devices, accurately locating PD sources is a challenge. Existing systems struggle to quickly distinguish PD signals generated by different assets, and related reviews indicate that PD location in complex power grids still needs improvement. PD signals are susceptible to external electromagnetic interference and difficult to distinguish from noise. Differentiating different types of PD (such as internal discharge, surface discharge, and corona discharge) requires advanced signal processing techniques. While methods exist to address the problem of noise obscuring PD signals, further optimization is needed.

[0003] Many PD monitoring systems only provide raw data and lack advanced analytical tools to predict faults or provide maintenance recommendations, limiting their predictive maintenance capabilities in the power Internet of Things (IoT). Some articles have pointed out limitations in data interpretation within existing systems. Advanced PD monitoring equipment (such as online monitoring systems) is expensive, limiting its application in small and medium-sized utility companies; cost is a major obstacle to the widespread adoption of PD monitoring systems. Differences in measurement and data interpretation methods among PD monitoring systems from different manufacturers lead to inconsistent results; therefore, standardized PD testing methods are essential to ensure data comparability across systems and regions. Integrating PD monitoring into large-scale power IoT requires robust communication infrastructure and data management systems. Existing systems may face performance bottlenecks when handling geographically dispersed or large-scale networks; for example, traditional data acquisition and monitoring systems perform poorly in wide area networks. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is: how to overcome the shortcomings of existing PDA device network-side fault diagnosis in the power Internet of Things in terms of fault location, signal classification and noise interference processing, data processing capabilities, cost, standardization methods, and large-scale Internet of Things integration and scalability.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for network-side fault diagnosis of PDA devices in a power Internet of Things, comprising: Collect PDA device operation data using an adaptive sampling frequency and determine device anomalies; Based on the equipment anomaly determination results, path tracing is performed to identify the specific network node where the fault occurred and obtain the location information of the fault point; Fault diagnosis is performed based on the equipment anomaly determination results and the location information of the fault point to obtain the fault diagnosis results; A diagnostic report and repair suggestions are generated based on the fault diagnosis results.

[0007] As a preferred solution for network-side fault diagnosis of PDA devices in the power Internet of Things, the following is provided: The step of collecting PDA device operating data at an adaptive sampling frequency and determining device anomalies includes: PDA device operation data is collected in an adaptive manner, and the sampling frequency is dynamically adjusted according to the device status. Based on the collected PDA device operation data, the statistical characteristics of partial discharge level are calculated using a sliding window algorithm, and spatiotemporal feature vectors are constructed by combining skewness and kurtosis.

[0008] As a preferred solution for network-side fault diagnosis of PDA devices in the power Internet of Things, the following is provided: The step of collecting PDA device operating data at an adaptive sampling frequency and determining device anomalies also includes: Anomaly detection is performed on preprocessed data by combining LSTM network and random forest algorithm. LSTM network determines whether the device is in an abnormal state based on prediction error, and random forest algorithm generates anomaly score through classification confidence.

[0009] As a preferred solution for network-side fault diagnosis of PDA devices in the power Internet of Things, the following is provided: The step of collecting PDA device operating data at an adaptive sampling frequency and determining device anomalies also includes: Based on survival analysis, the constructed spatiotemporal feature vectors are used to drive the survival function to predict the conditional probability of equipment failure before the prediction time window.

[0010] The beneficial effects of this preferred technical solution are: by predicting the conditional probability of equipment failure before the prediction time window through survival analysis, early warning of equipment failure can be given, which facilitates timely maintenance measures, reduces losses caused by equipment failure, and improves the reliability and stability of the power Internet of Things system.

[0011] As a preferred solution for network-side fault diagnosis of PDA devices in the power Internet of Things, the following is provided: The process of tracing the path based on the device anomaly determination result, identifying the specific network node where the fault occurred, and obtaining the location information of the fault point includes: The optimized parallel routing tracing algorithm is used for path tracing. The parallel routing tracing algorithm uses Internet Control Message Protocol and User Datagram Protocol probes, combined with the topology query operation of the software-defined network controller and the analysis results of the parallel algorithm, to perform fault location analysis, determine the specific network node where the fault occurred, and generate the location information of the fault point after the fault location is completed.

[0012] The beneficial effects of this preferred technical solution are as follows: the optimized parallel routing tracing algorithm, combined with two protocol probes, can perform path tracing more efficiently; combined with software-defined network controller query topology and parallel algorithm analysis, it can quickly and accurately locate the specific network node where the fault occurs, improve the efficiency and accuracy of fault location, and provide strong support for subsequent fault diagnosis and repair.

[0013] As a preferred solution for network-side fault diagnosis of PDA devices in the power Internet of Things, the following is provided: The fault diagnosis based on the equipment anomaly determination result and the location information of the fault point yields the following results: Construct a semantic network containing device entities and the relationships between them, and use SPARQL 1.1 to perform association queries on the knowledge graph to associate device alarms and network events.

[0014] As a preferred solution for network-side fault diagnosis of PDA devices in the power Internet of Things, the following is provided: The fault diagnosis based on the equipment anomaly determination result and the location information of the fault point also includes: When multiple candidate causes for a failure are detected, the confidence level of each candidate cause is evaluated based on historical data and the correlation between events. Based on the confidence level evaluation results, network failures and equipment failures are distinguished, and the root cause of the failure is determined.

[0015] The beneficial effects of this preferred technical solution are: by evaluating the confidence level of multiple candidate causes, the root cause of the fault can be determined more scientifically and reasonably, accurately distinguishing between network faults and equipment faults, providing an accurate basis for subsequent development of targeted repair solutions, and improving the efficiency and effectiveness of fault handling.

[0016] Secondly, the present invention provides a network-side fault diagnosis system for PDA devices in the power Internet of Things, comprising: The device anomaly detection module is used to collect the operating data of the PDA device at an adaptive sampling frequency and perform device anomaly detection. The path tracing module is used to perform path tracing based on the device anomaly determination results, identify the specific network node where the fault occurred, and obtain the location information of the fault point; The fault diagnosis module is used to perform fault diagnosis based on the equipment anomaly determination results and the location information of the fault point, and obtain the fault diagnosis results. The diagnostic report and repair suggestion generation module is used to generate diagnostic reports and repair suggestions based on the fault diagnosis results.

[0017] Thirdly, the present invention provides an electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the one or more programs are executed by the one or more processors, the one or more processors implement the network-side fault diagnosis method for PDA devices in the power Internet of Things as described in this invention.

[0018] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the network-side fault diagnosis method for PDA devices in the power Internet of Things.

[0019] The beneficial effects of this invention are as follows: The adaptive sampling frequency used in this invention is dynamically adjusted based on the partial discharge level of the PDA device, reducing unnecessary data acquisition, lowering the computational resources and storage space required for data storage and processing, and preventing the system from being overburdened by processing large amounts of irrelevant data. Conversely, when the device malfunctions and the partial discharge level reaches or exceeds a threshold, the maximum sampling frequency is used. High-frequency sampling can promptly capture subtle changes in the device's state, providing a rich and accurate data foundation for subsequent anomaly detection. Sufficient data volume makes anomaly detection more accurate, improves the timeliness of fault detection, and reduces the possibility of overlooking potential faults. In the anomaly detection process, an LSTM network and a random forest algorithm are combined. The LSTM network, with its powerful processing capabilities for time-series data, can capture long-term dependencies in the device's operating data and determine whether the device is abnormal based on the prediction error threshold. The random forest algorithm generates anomaly scores through classification confidence, evaluating the device state from different perspectives. The combination of the two compensates for the limitations of a single algorithm and improves the reliability of anomaly detection. Accurate anomaly detection provides clear direction for subsequent fault investigation and handling, enabling the system to respond promptly in the early stages of a fault and prevent further escalation. Path tracing is performed based on a Software-Defined Networking (SDN) controller and a parallelized traceroute algorithm. The SDN controller can directly acquire network topology information, avoiding the delays of traditional hop-by-hop probing and accelerating fault location. Parallelized probing further improves efficiency, and the use of Internet Control Message Protocol (ICMP) and User Datagram Protocol (UDP) probes adapts to the response characteristics of different network devices, ensuring accurate and rapid fault location in various network environments. Rapid fault node location reduces the time cost of fault investigation, allowing maintenance personnel to quickly reach the fault site and mitigate the impact of the fault on the operation of the power IoT. The fault diagnosis phase employs a knowledge graph-based reasoning mechanism. The knowledge graph constructs a semantic network containing device entities and relationships. By querying related device alarms and network events using SPARQL, it is possible to accurately distinguish between network faults and device faults and deeply analyze the root causes of the faults. When faced with multiple possible causes of failure, the most probable causal relationship is determined by calculating normalized probabilities, thus improving diagnostic accuracy. Accurate fault diagnosis results provide crucial information for developing effective repair plans, avoiding blind repairs and improving repair efficiency and success rates. Diagnostic reports and repair recommendations are generated based on the fault diagnosis results. These recommendations, based on historical data and expert knowledge, use recommendation models to match the best repair actions for different fault characteristics. These recommendations are highly targeted and actionable, helping maintenance personnel quickly develop repair plans and shorten equipment repair time.Meanwhile, the system also supports human-machine collaboration, allowing engineers to provide feedback and continuously optimize repair suggestions, enabling the system to provide more effective solutions when facing various complex faults and improve the overall reliability and stability of the power Internet of Things. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is an overall flowchart of the network-side fault diagnosis method for PDA devices in the power Internet of Things provided by the present invention. Detailed Implementation

[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0023] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for network-side fault diagnosis of PDA devices in a power Internet of Things, including: S1: Collect the operating data of the PDA device at an adaptive sampling frequency and determine device anomalies; S2: Based on the equipment anomaly determination result, perform path tracing, identify the specific network node where the fault occurred, and obtain the location information of the fault point; S3: Based on the equipment anomaly determination results and the location information of the fault point, perform fault diagnosis to obtain the fault diagnosis results; S4: Generate diagnostic reports and repair suggestions based on fault diagnosis results.

[0024] It should be noted that steps S1-S4 achieve a complete process for handling PDA device network-side faults, from data collection, anomaly detection, fault location, precise diagnosis to the generation of repair suggestions. By fully utilizing advanced technologies such as adaptive sampling, machine learning algorithms, and software-defined networking, network-side faults can be quickly and accurately identified and located, the root causes of faults can be comprehensively analyzed, and targeted repair suggestions can be provided. This method significantly enhances the availability of PDA devices, improves fault response speed and maintenance efficiency, provides strong support for the stable operation of the power Internet of Things (IoT), and thus helps improve the reliability and security of power equipment, promoting the efficient development of the power IoT.

[0025] Example 2, refer to Figure 1 As an embodiment of the present invention, based on the previous embodiment, a method for network-side fault diagnosis of PDA devices in a power Internet of Things is provided, including: In this embodiment, step S1 above, which involves collecting the operating data of the PDA device at an adaptive sampling frequency and determining device anomalies, includes: The system collects operational data from the PDA device at an adaptive sampling frequency, including heartbeat signals, connection status, signal strength, and metadata for partial discharge measurements (such as pulse count, amplitude, and phase angle).

[0026] It should be noted that adaptive sampling dynamically adjusts the frequency according to the device status to ensure timely data capture in abnormal situations, while optimizing resource utilization.

[0027] Specifically, the sampling frequency is adjusted based on the partial discharge level. and predefined thresholds , is represented as: in, This is the minimum sampling frequency under normal conditions (e.g., once per hour). This is the maximum sampling frequency under abnormal conditions (e.g., once every 5 minutes).

[0028] In time period Total number of internal samplings for: In time period Total amount of data collected internally for: in, The sampling interval is... For time Sampling frequency (unit: ), or This represents the number of bytes per sample.

[0029] The system maintains low-frequency sampling under normal conditions to reduce network load, and when it detects... The system automatically triggers a high-frequency mode, ensuring complete capture of transient process data through a frequency step response. This mechanism can improve the data resolution of abnormal events by 12 times while maintaining an 83% reduction in average resource consumption.

[0030] For example, a PDA device monitors the insulation status of a transformer. During normal operation ( The heartbeat signal is sent once per hour and the average pulse count is summarized. If the partial discharge level suddenly increases ( The sampling frequency was increased to once every 5 minutes to capture detailed data to support subsequent diagnosis.

[0031] After collecting the operating data of the PDA device, the device data is preprocessed in real time using a sliding window algorithm (window size w) to calculate the partial discharge level. Statistical characteristics, including the mean ,variance And by combining parameters such as skewness and kurtosis, a 72-dimensional spatiotemporal feature vector is constructed, which is expressed as: Where w is the number of samples within the window, and the window duration (minutes) is the time range covered by the window.

[0032] For equally spaced sampling, the window time length is calculated as follows: For variable interval sampling, the window time length is calculated as follows: .

[0033] Default recommendation: Under normal conditions, the sampling interval and the number of samples in the window are as follows: In abnormal states: The window duration can be adjusted according to the device's importance level or maintenance strategy. Configured within a minute range.

[0034] Window selection rules: When a target window is given a time length (denoted as...) ) and sampling interval Time (equal interval scenario), take: In engineering practice, the recommended window duration for normal mode is [length missing]. minutes (e.g.) minute, The recommended window duration for abnormal mode is... minutes (e.g.) minute, ), and can Select from the minute set based on equipment level and operation and maintenance strategy.

[0035] Secondly, anomaly detection is performed by combining LSTM networks with the random forest algorithm: LSTM is based on prediction error threshold An abnormal state is indicated as follows: in, This is the actual value. For predicted values, The model is obtained by one-step prediction using an LSTM with the feature sequences of the most recent m time steps as input (m=6 for abnormal mode or m=3 for normal mode by default). The model is trained with historical normal data, and the loss function can be MSE or MAE. The threshold ϵ is taken as the 99th percentile of the training residuals or is calibrated by the validation set.

[0036] Random forests generate outlier scores based on classification confidence. ,when If the value is ≥τ, it is considered abnormal; otherwise, it is considered normal. ,default It can be calibrated in conjunction with a validation set.

[0037] Finally, based on survival analysis to predict future failure risks, the conditional probability of the equipment failing before time t is calculated and expressed as: Where T is a non-negative random variable (in minutes) representing the time from the current moment until the device fails. The survival function is driven by the feature vector X. For the prediction time window; The default time is 120 minutes, but you can also adjust the time based on device importance or maintenance strategy. Configured in minutes.

[0038] For example, a PDA device suddenly stops sending heartbeat signals and exhibits an abnormally high level of partial discharge. The system extracts features using a sliding window, detects time-series biases using LSTM, confirms the anomaly using random forest, and predicts that the device has an 80% risk of going offline within the next 120 minutes.

[0039] In another possible implementation, a Support Vector Machine (SVM) algorithm can be introduced in conjunction with LSTM and Random Forest when determining device anomalies. SVM can classify device data and, through training, obtain an optimal hyperplane that separates normal and abnormal data.

[0040] In another possible implementation, a fuzzy logic system can be used to fuzzify the device data. For example, data such as partial discharge level and signal strength can be divided into different fuzzy sets (e.g., "low", "medium", "high"), and anomaly determination can be made according to fuzzy rules.

[0041] In another possible implementation, a Bayesian network model can be established by combining historical fault data of the equipment. The probability of equipment malfunction can then be calculated using known equipment status information.

[0042] In this embodiment, step S2 above, based on the device anomaly determination result, performs path tracing to identify the specific network node where the fault occurred and obtains the location information of the fault point, including: Based on a software-defined network (SDN) controller and a parallel traceroute algorithm, this system enables real-time mapping of end-to-end network paths and precise fault location.

[0043] Total fault location time in the fault location model The time required for SDN query and algorithm analysis is determined by both factors, and can be expressed as follows: in, The time required for the SDN controller to query the topology. For analysis time. The time complexity of traditional traceroute is O(n log n). , The time is significantly reduced by optimizing the hop count through SDN.

[0044] The network topology is obtained directly through the SDN controller, avoiding the latency of traditional traceroute hop-by-hop probing. The optimized algorithm supports parallel probing, using Internet Control Message Protocol (ICMP) and User Datagram Protocol (UDP) probes to address the response characteristics of different network devices. After location is complete, the system generates the fault location information.

[0045] For example, if a PDA device stops reporting data, the system obtains the topology through the SDN controller, uses optimized traceroute to discover a router port failure within 200 milliseconds, and locates the specific switch.

[0046] In another possible implementation, when performing path tracing, a depth-first search (DFS) or breadth-first search (BFS) algorithm based on the network topology can be used to search the network nodes layer by layer, starting from the PDA device, to find possible faulty paths.

[0047] In another possible implementation, network traffic analysis tools can be used to analyze the direction of data packets and changes in traffic within the network. If the traffic of a node suddenly increases or decreases abnormally, that node may be malfunctioning.

[0048] In another possible implementation, distributed sensor nodes can be deployed to monitor the status of each node in the network in real time. When an anomaly is detected, the faulty node can be quickly located through communication between the sensor nodes.

[0049] In this embodiment, the fault diagnosis performed in step S3 above based on the equipment anomaly determination result and the location information of the fault point, and the resulting fault diagnosis result includes: By employing a knowledge graph-based reasoning mechanism, a semantic network containing device entities (such as PDAs and switches) and relationships (such as "connected to" and "caused") is constructed to achieve accurate differentiation between network faults and device faults and root cause analysis.

[0050] Specifically, knowledge graph association query: using SPARQL (SPARQL Protocol and RDF QueryLanguage) to query related device alarms and network events to locate potential causal relationships.

[0051] Multi-cause confidence assessment: When multiple candidate causes are detected, normalized probabilities are calculated based on historical data and the correlation with the event. , is represented as: in, It consists of the following weighting terms: .

[0052] in, Historical co-occurrence frequency (0–1, normalized by the number of occurrences). Score the time proximity of real-time events (0–1, decaying with time difference). The topological connectivity density is 0–1, normalized by path length / degree centrality.

[0053] Weighting: ,default , , Its value is determined by maximizing diagnostic accuracy or PR-AUC (area under the Precision-Recall curve) on the validation set through grid search or Bayesian optimization using historical labeled data. Alternatively, it can be set a priori based on scenario experience and then fine-tuned.

[0054] For example, the PDA device disconnects after reporting a high partial discharge level. The system uses a knowledge graph to discover that recent BGP changes have caused network instability, with a confidence level of 90%, and determines it to be a network fault rather than a device problem.

[0055] In this embodiment, the step S4 above, which generates a diagnostic report and repair suggestions based on the fault diagnosis results, includes: Based on historical data and expert knowledge, diagnostic reports and remediation suggestions are generated. During this process, an interactive dashboard supports human-machine collaboration, and engineers can provide feedback to improve the system.

[0056] For example, it is recommended to replace the faulty router and check the transformer for elevated discharge levels.

[0057] Repair suggestions are generated using a recommendation model and are represented as follows: in, To recommend the action, Represents a specific repair action. These are fault characteristics. Estimated using historical data, indicating that the known fault characteristics are... Under the conditions, take action The probability, This means finding, among all possible actions 'a', the one that makes... The action 'a' that achieves the maximum value is taken as the final recommended action.

[0058] For example, after diagnosing a network fault, the system suggests replacing the faulty router, based on PDA data, and checks if the transformer has a problem due to increased discharge levels. Engineers confirm the recommendations via a dashboard and record the repair results.

[0059] Example 3: The above is an illustrative scheme for the network-side fault diagnosis method of PDA devices in the power Internet of Things (IoT) according to this embodiment. It should be noted that the technical solution of the network-side fault diagnosis system for PDA devices in the power IoT and the technical solution of the network-side fault diagnosis method for PDA devices in the power IoT described above belong to the same concept. Details not described in detail in the technical solution of the network-side fault diagnosis system for PDA devices in the power IoT in this embodiment can be found in the description of the technical solution of the network-side fault diagnosis method for PDA devices in the power IoT described above.

[0060] This embodiment also provides a network-side fault diagnosis system for PDA devices in the power Internet of Things, including: The device anomaly detection module is used to collect the operating data of the PDA device at an adaptive sampling frequency and perform device anomaly detection. The path tracing module is used to perform path tracing based on the device anomaly determination results, identify the specific network node where the fault occurred, and obtain the location information of the fault point; The fault diagnosis module is used to perform fault diagnosis based on the equipment anomaly determination results and the location information of the fault point, and obtain the fault diagnosis results. The diagnostic report and repair suggestion generation module is used to generate diagnostic reports and repair suggestions based on the fault diagnosis results.

[0061] This embodiment also provides an electronic device applicable to network-side fault diagnosis methods for PDA devices in the power Internet of Things, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement the network-side fault diagnosis method for PDA devices in the power Internet of Things, as proposed in the above embodiments.

[0062] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the network-side fault diagnosis method for PDA devices in the power Internet of Things as proposed in the above embodiments.

[0063] The storage medium proposed in this embodiment and the network-side fault diagnosis method for PDA devices in the power Internet of Things proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0064] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for network-side fault diagnosis of PDA devices in the power Internet of Things, characterized in that, include: Collect PDA device operation data using an adaptive sampling frequency and determine device anomalies; Based on the equipment anomaly determination results, path tracing is performed to identify the specific network node where the fault occurred and obtain the location information of the fault point; Fault diagnosis is performed based on the equipment anomaly determination results and the location information of the fault point to obtain the fault diagnosis results; A diagnostic report and repair suggestions are generated based on the fault diagnosis results.

2. The method for network-side fault diagnosis of PDA devices in a power Internet of Things as described in claim 1, characterized in that, The step of collecting PDA device operating data at an adaptive sampling frequency and determining device anomalies includes: PDA device operation data is collected in an adaptive manner, and the sampling frequency is dynamically adjusted according to the device status. Based on the collected PDA device operation data, the statistical characteristics of partial discharge level are calculated using a sliding window algorithm, and spatiotemporal feature vectors are constructed by combining skewness and kurtosis.

3. The method for network-side fault diagnosis of PDA devices in a power Internet of Things as described in claim 2, characterized in that, The step of collecting PDA device operating data at an adaptive sampling frequency and determining device anomalies also includes: Anomaly detection is performed on preprocessed data by combining LSTM network and random forest algorithm. LSTM network determines whether the device is in an abnormal state based on prediction error, and random forest algorithm generates anomaly score through classification confidence.

4. The method for network-side fault diagnosis of PDA devices in a power Internet of Things as described in claim 3, characterized in that, The step of collecting PDA device operating data at an adaptive sampling frequency and determining device anomalies also includes: Based on survival analysis, the constructed spatiotemporal feature vectors are used to drive the survival function to predict the conditional probability of equipment failure before the prediction time window.

5. A method for network-side fault diagnosis of PDA devices in a power Internet of Things as described in claim 4, characterized in that, The process of tracing the path based on the device anomaly determination result, identifying the specific network node where the fault occurred, and obtaining the location information of the fault point includes: The optimized parallel routing tracing algorithm is used for path tracing. The parallel routing tracing algorithm uses Internet Control Message Protocol and User Datagram Protocol probes, combined with the topology query operation of the software-defined network controller and the analysis results of the parallel algorithm, to perform fault location analysis, determine the specific network node where the fault occurred, and generate the location information of the fault point after the fault location is completed.

6. The method for network-side fault diagnosis of PDA devices in a power Internet of Things as described in claim 5, characterized in that, The fault diagnosis based on the equipment anomaly determination result and the location information of the fault point yields the following results: Construct a semantic network containing device entities and the relationships between them, and use SPARQL 1.1 to perform association queries on the knowledge graph to associate device alarms and network events.

7. A method for network-side fault diagnosis of PDA devices in a power Internet of Things as described in claim 6, characterized in that, The fault diagnosis based on the equipment anomaly determination result and the location information of the fault point also includes: When multiple candidate causes for a failure are detected, the confidence level of each candidate cause is evaluated based on historical data and the correlation between events. Based on the confidence level evaluation results, network failures and equipment failures are distinguished, and the root cause of the failure is determined.

8. A network-side fault diagnosis system for PDA devices in a power Internet of Things, using the method described in any one of claims 1 to 7, characterized in that, include: The device anomaly detection module is used to collect the operating data of the PDA device at an adaptive sampling frequency and perform device anomaly detection. The path tracing module is used to perform path tracing based on the device anomaly determination results, identify the specific network node where the fault occurred, and obtain the location information of the fault point; The fault diagnosis module is used to perform fault diagnosis based on the equipment anomaly determination results and the location information of the fault point, and obtain the fault diagnosis results. The diagnostic report and repair suggestion generation module is used to generate diagnostic reports and repair suggestions based on the fault diagnosis results.

9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.

Images