An LSTM power IoT ring main unit optimization method, system, equipment and medium

Abstract

The application discloses an LSTM power internet of things looped netowrk cabinet optimization method, system, device and medium, belongs to the technical field of power internet of things management and data analysis, including real-time collection of communication state and network environment data, generating a multi-dimensional feature matrix through dynamic standard score standardization and derivative feature calculation; dynamically adjusting the LSTM forgetting gate weight based on the bandwidth fluctuation index, outputting the device performance score; constructing a dynamic scoring algorithm and triggering differentiated maintenance instructions in conjunction with multi-threshold scene strategies; screening high-bias data using a closed-loop feedback mechanism, strengthening new scene training through a dynamic loss function, and realizing minute-level hot updating of the model; The application solves the problems of static threshold misjudgment, poor environmental adaptability of the prediction model, and single decision mechanism, significantly improves feature quality, equipment evaluation accuracy, and resource allocation rationality, and strengthens the intelligent operation and maintenance capability of the system.

Description

Technical Field

[0001] This invention relates to the field of power Internet of Things (IoT) management and data analysis technology, specifically to an LSTM power IoT ring network cabinet optimization method, system, equipment, and medium. Background Technology

[0002] The operation and maintenance management of power IoT ring main units faces severe challenges in dynamic network environments. Existing technologies generally rely on static models or single indicators to formulate maintenance strategies, such as triggering alarms based on fixed latency thresholds or using offline models trained solely on historical data for fault prediction. These methods have significant limitations: they cannot respond to real-time network fluctuations, leading to delayed maintenance of high-load equipment and deterioration of communication quality; they lack the ability to analyze the synergistic impact of communication status and network environment, resulting in severe resource imbalances in multi-device concurrent scenarios. More critically, existing solutions do not establish a long-term dependency model between network fluctuations and equipment performance. When communication status remains abnormal, the system cannot predict performance degradation trends, leading to frequent equipment failures. Although some technologies have attempted to introduce time-series models to optimize data transmission, these are not linked to maintenance strategies, and high-priority equipment still fails due to insufficient response during sudden network congestion. These defects in static mechanisms, lack of time-series correlation, and insufficient data fusion directly result in low operation and maintenance efficiency, soaring costs, and decreased system stability.

[0003] The fundamental problem with traditional methods lies in their lack of closed-loop optimization capabilities and weak dynamic adaptability. Static standardization mechanisms experience a significant increase in noise misjudgment rates during sudden traffic spikes, while offline training modes prevent models from absorbing real-time feedback data. For example, network topology changes require prolonged manual intervention, during which the prediction function completely fails. Simultaneously, the fixed weight mechanism of existing time-series models struggles to perceive the nonlinear impact of network fluctuations on device performance, leading to increased prediction bias. The reliance on single threshold rules at the decision-making level is particularly problematic, failing to quantify dynamic characteristic changes or construct a multi-factor collaborative evaluation system, resulting in excessive or insufficient maintenance instructions. These deficiencies are amplified in complex network environments: when bandwidth utilization is extremely high and topology changes drastically, existing systems, lacking real-time optimization mechanisms, cannot quickly adjust model parameters, ultimately leading to systemic operational failures. Therefore, a dynamic maintenance solution integrating real-time fluctuation perception, adaptive time-series modeling, and closed-loop feedback is urgently needed. 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 to build an intelligent decision-making model by integrating real-time communication status (such as latency and packet loss rate) with network environment data (such as bandwidth fluctuations and topology changes), aiming to solve problems such as lagging maintenance strategies and inefficient resource allocation of ring main units in complex network environments, and promote the intelligent transformation of power distribution networks.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: an LSTM power IoT ring network cabinet optimization method, comprising:

[0007] Real-time acquisition of communication status data and network environment data of ring network cabinets, real-time calculation of data within the sliding window, determination of outliers through dynamic standard score standardization and historical data interpolation correction, calculation of dynamic delay change rate and bandwidth fluctuation index, and generation of multi-dimensional feature matrix.

[0008] The multidimensional feature matrix is ​​input into the long short-term memory network, and the forget gate weights are adjusted in real time based on the bandwidth fluctuation index. The output device performance score and key influencing factors are then used.

[0009] A dynamic scoring algorithm is constructed by adjusting the dynamic latency change rate, bandwidth fluctuation index, and packet loss rate based on real-time network status, and a multi-threshold linkage strategy is used to define latency and bandwidth overload scenarios to trigger differentiated maintenance instructions.

[0010] Collect equipment status feedback data after the execution of maintenance instructions, filter data whose prediction deviation exceeds a set threshold, use a dynamic loss function to strengthen the weight of new scenario data, train the model, and update the model version when the accuracy of the validation set exceeds a set threshold and support rollback.

[0011] As a preferred embodiment of the LSTM power IoT ring network cabinet optimization method described in this invention, the real-time calculation of data within the sliding window includes setting a sliding window of fixed duration, capturing communication status data and network environment data within the window in real time; calculating the mean and standard deviation of the data within the sliding window, calculating the standard score using the dynamic Z-score formula, and determining an outlier if the absolute value of the standard score exceeds a preset outlier threshold; and correcting it through historical data interpolation.

[0012] As a preferred embodiment of the LSTM power IoT ring network cabinet optimization method described in this invention, the multidimensional feature matrix includes a mechanism that transforms the original communication data into a high-precision multidimensional feature matrix through dynamic sliding window Z-score standardization and historical data interpolation correction.

[0013] The raw communication data includes real-time latency, packet loss rate, bandwidth, and topology status.

[0014] The high-precision multidimensional feature matrix includes latency, packet loss rate, bandwidth, topology status, dynamic latency change rate, and bandwidth fluctuation index.

[0015] As a preferred embodiment of the LSTM power IoT ring network cabinet optimization method described in this invention, the output device performance score and key influencing factors include dynamically adjusting the LSTM gate weights based on the bandwidth fluctuation index, constructing an adaptive time-series fusion model, output device performance score, and key influencing factors.

[0016] As a preferred embodiment of the LSTM power IoT ring network cabinet optimization method described in this invention, the construction of the dynamic scoring algorithm includes calculating a comprehensive score by comprehensively considering the dynamic delay change rate, bandwidth fluctuation index, and current packet loss rate through dynamic weight coefficients.

[0017] As a preferred embodiment of the LSTM power IoT ring network cabinet optimization method described in this invention, the multi-threshold linkage strategy includes triggering a high-priority maintenance command and performing link switching when the delay exceeds a first threshold and the bandwidth utilization exceeds a second threshold; and triggering a communication path priority dynamic adjustment command when a topology change is detected and bandwidth allocation is uneven.

[0018] As a preferred embodiment of the LSTM power IoT ring network cabinet optimization method described in this invention, the dynamic loss function includes: fusing historical and new loss values, weighting and summing the historical data loss values ​​and the new data loss values; and adjusting the weights according to the scenario, selecting new data with prediction deviations exceeding a set threshold to participate in the calculation.

[0019] This invention provides an LSTM power IoT ring network cabinet optimization system. It generates a multi-dimensional feature matrix that integrates communication status and network environment in real time through a dynamic feature construction module; dynamically adjusts the LSTM forget gate weights based on bandwidth fluctuation index using a time-series adaptive prediction module to output device performance scores; generates differentiated maintenance instructions by combining dynamic weighted scoring and a scenario-based rule base through a multi-threshold decision module; and achieves incremental learning-driven model hot updates at the minute level using a closed-loop optimization module. This addresses the technical problems of static models being unable to adapt to network fluctuations, lagging maintenance strategies, and the lack of a closed-loop mechanism.

[0020] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an LSTM power Internet of Things ring network cabinet optimization system, comprising:

[0021] The dynamic feature construction module collects real-time communication status data and network environment data of the ring network cabinet, calculates data within the sliding window in real time, judges outliers by standardization through dynamic standard scores and performs historical data interpolation correction, calculates dynamic delay change rate and bandwidth fluctuation index, and generates a multi-dimensional feature matrix.

[0022] The time-adaptive prediction module inputs a multi-dimensional feature matrix into the long short-term memory network, adjusts the forget gate weights in real time based on the bandwidth fluctuation index, and outputs device performance scores and key influencing factors.

[0023] The multi-threshold decision module adjusts the dynamic latency change rate, bandwidth fluctuation index, and packet loss rate according to the real-time network status to build a dynamic scoring algorithm, and uses a multi-threshold linkage strategy to define latency and bandwidth overload scenarios to trigger differentiated maintenance instructions.

[0024] The closed-loop optimization module collects equipment status feedback data after the execution of maintenance instructions, filters data whose prediction deviation exceeds a set threshold, uses a dynamic loss function to strengthen the weight of new scenario data, trains the model, and updates the model version when the accuracy of the validation set exceeds a set threshold and supports rollback.

[0025] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the LSTM power Internet of Things ring network cabinet optimization method.

[0026] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the LSTM power Internet of Things ring network cabinet optimization method.

[0027] The beneficial effects of this invention are as follows: This invention effectively eliminates the problem of misjudging sudden network fluctuations by static thresholds through a dynamic data processing mechanism, significantly improving feature quality; it adopts network-aware time-series modeling technology, enabling the prediction model to adapt to complex environmental changes, greatly enhancing the accuracy of equipment performance evaluation in sudden scenarios such as topology reconfiguration; it achieves dynamic optimization of maintenance strategies based on a multi-dimensional collaborative decision-making mechanism, completely changing the traditional single-indicator decision-making mode and significantly improving the rationality of resource allocation for high-load equipment; and through a closed-loop feedback-driven model optimization system, it achieves the ability to respond to network mutation events at the minute level, comprehensively strengthening the intelligent operation and maintenance level and long-term operational stability of the power Internet of Things system. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 The above is a flowchart of an LSTM power Internet of Things ring network cabinet optimization method provided in one embodiment of the present invention.

[0030] Figure 2 This diagram illustrates a practical application example of an LSTM power IoT ring network cabinet optimization method provided in one embodiment of the present invention. Detailed Implementation

[0031] 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.

[0032] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides an LSTM power IoT ring network cabinet optimization method, including:

[0033] S10: Real-time acquisition of ring main unit communication status data and network environment data; real-time calculation of data within the sliding window; standardization using dynamic standard scores to identify outliers and perform historical data interpolation correction; calculation of dynamic delay change rate and bandwidth fluctuation index; generation of a multi-dimensional feature matrix; by calculating the delay difference within adjacent time windows, combined with the time interval, the dynamic delay change rate is defined as follows: Dynamic delay change rate = ,in, This is the delay value at the current moment. This is the delay value from the previous moment. The time interval is defined as follows: based on the bandwidth data within the sliding window, the standard deviation of bandwidth within each time interval is calculated and defined as the bandwidth fluctuation index. The bandwidth fluctuation index = ,in, The standard deviation of bandwidth, This represents the average bandwidth.

[0034] It should be noted that by using dynamic sliding window Z-score (dynamic standard score standardization) standardization and historical data interpolation correction mechanism, the original communication data is transformed into a high-precision 6-dimensional feature matrix, that is, a high-precision multi-dimensional feature matrix, which solves the problem of noise misjudgment caused by static threshold and standardization parameters in traditional methods and reduces the feature extraction error rate.

[0035] Furthermore, the mean of data within a real-time calculation window (e.g., 5 minutes) is standardized using dynamic Z-score. with standard deviation , is represented as:

[0036] ,

[0037] in, This refers to raw data (such as real-time collected values ​​of latency, packet loss rate, bandwidth, etc.). This represents the mean of the data within a sliding window (5 minutes). The standard deviation of the data within the sliding window. These are standardized values ​​used for outlier detection.

[0038] If the device's delay data is within a 5-minute window Calculated , ,but ;because It is not considered an outlier; when Values ​​identified as outliers are removed in real time and corrected using historical data interpolation. This represents a standardized value in the dynamic Z-score standardization calculation.

[0039] Furthermore, feature extraction uncovers multidimensional dynamic features, such as calculating the dynamic latency change rate (reflecting the fluctuation trend of latency) and the network bandwidth fluctuation index (quantifying the stability of bandwidth).

[0040] Specifically, the dynamic delay change rate is obtained by calculating the delay change trend of adjacent time windows, and is expressed as:

[0041] ,

[0042] in, This is the delay value for the current time window (e.g., 1 minute). This is the delay value of the previous time window. The time window interval (e.g., 1 minute). For dynamic delay rate of change; if , ,but This reflects an upward trend in delays;

[0043] Quantizing bandwidth stability, using standardized calculations, is expressed as:

[0044] ,

[0045] in, For the first Bandwidth value for each time window, The average bandwidth within the sliding window. This represents the number of data points within the window; the output is a standardized feature matrix containing 6 dimensions (latency, packet loss rate, bandwidth, topology status, ...). , This provides high-quality input for subsequent model building;

[0046] If the bandwidth within the window is ,but Calculated The intensity of bandwidth fluctuations is quantified. If a device experiences a sudden increase in latency data during a certain period, the cleaning process will remove the abnormal value and correct it through historical data interpolation to ensure the accuracy of feature extraction.

[0047] S20: Input the multidimensional feature matrix into the long short-term memory network, adjust the forget gate weights in real time based on the bandwidth fluctuation index, and output the device performance score and key influencing factors.

[0048] It should be noted that a multi-dimensional LSTM temporal fusion model is constructed using the Long Short-Term Memory (LSTM) algorithm, based on dynamically adjusting the LSTM gating weights according to the bandwidth fluctuation index (input: the 6-dimensional feature matrix output by S10). This model includes a feature fusion input layer, a dynamically weighted LSTM layer, and an output layer; the output layer contains device performance scores and key influencing factors (output: (Score and sensitivity label) to address the problem that existing LSTM models cannot integrate multi-dimensional data from the network environment and increase prediction accuracy in scenarios of topological changes; among them, the key influencing factor is the sensitivity label, such as latency surge sensitivity.

[0049] Furthermore, the feature fusion input layer has the following input dimensions: the input dimension is a preprocessed 6-dimensional feature matrix (from S10); the data format is a sliding window data within a time window (e.g., 30 minutes);

[0050] Dynamic weighted LSTM layer: In the LSTM gating mechanism, the forget gate Input gate and output gate The calculation is expressed as:

[0051] ,

[0052] in, The state was hidden in the previous moment. Given the current input feature vector, , , Here are the weight matrices for the forget gate, input gate, and output gate. , , For bias terms, for Activation function , , These represent the Forget Gate, Input Gate, and Output Gate, respectively.

[0053] Output layer: Output device performance rating ( (Points) and key influencing factors (such as "delayed burst sensitivity").

[0054] The device performance score is a score between 0 and 100 calculated using an LSTM model based on multiple network characteristics (such as latency, packet loss rate, bandwidth, etc.), representing the overall performance of the device. The calculation steps are as follows:

[0055] (1) Input data: LSTM model receive delay, packet loss rate, bandwidth, topology and other data are used as input;

[0056] (2) LSTM learning: The LSTM model learns the relationships between data and automatically identifies the impact of each feature on device performance;

[0057] (3) Scoring Calculation: The equipment performance score is calculated using the following weighted formula:

[0058] ,

[0059] in, , , , These are weight coefficients automatically learned by the LSTM model, representing the impact of each feature on device performance.

[0060] Key influencing factors (such as latency surge sensitivity) are used to represent the sensitivity of devices to network changes; the calculation method is as follows:

[0061] (1) Input data: Similar to device performance scoring, the data input to the LSTM model includes network features such as latency, bandwidth, and packet loss rate, but special attention is paid to how these features affect device performance when the network fluctuates;

[0062] (2) LSTM learning: The LSTM model learns how the device responds to different network changes, especially sudden increases in latency and bandwidth fluctuations, by training on historical data. LSTM can identify which network changes will cause a sharp drop in device performance and map these changes to sensitivity labels.

[0063] (3) Sensitivity label calculation: For example, latency surge sensitivity is determined by the change in the computing device's response to latency surges; the calculation formula is as follows:

[0064] ,

[0065] in, It is the change in the device response time. It represents the increase in latency. When latency increases significantly, the device's response time usually changes as well; this ratio indicates the device's sensitivity to changes in latency.

[0066] (4) Output sensitivity labels: The LSTM model outputs sensitivity labels based on the changes in the device's response time, which are usually divided into multiple levels. For example, high sensitivity is defined as the device reacts strongly to sudden increases in latency and its performance drops significantly; medium sensitivity is defined as the device reacts to latency changes to a certain extent, but does not affect normal operation; low sensitivity is defined as the device can adapt well to latency fluctuations and its performance does not change much.

[0067] Furthermore, according to the bandwidth fluctuation index ( )Adjustment Weights, expressed as:

[0068] ,

[0069] in, Adjustment coefficient (default) ), control the intensity of the impact of bandwidth fluctuations on memory, This represents the current bandwidth fluctuation index. For the original forget gate weights, The adjusted forget gate weights represent the degree to which the LSTM forget gate retains historical information after bandwidth fluctuation exponential adjustment; when hour, Strengthen the memory of characteristics during periods of high volatility.

[0070] S30: Based on the real-time network status, adjust the dynamic latency change rate, bandwidth fluctuation index, and packet loss rate to construct a dynamic scoring algorithm, and use a multi-threshold linkage strategy to define latency surge and bandwidth overload scenarios to trigger differentiated maintenance instructions.

[0071] It should be noted that, based on the dynamic scoring algorithm (input: performance score, bandwidth utilization, and topology status), differentiated maintenance instructions are generated (output: JSON format instructions). The resource waste caused by single-indicator decision-making is solved by a multi-threshold linkage strategy, thereby improving resource allocation efficiency.

[0072] Furthermore, a dynamic scoring algorithm and a multi-threshold linkage strategy are designed;

[0073] Specifically, the dynamic scoring algorithm is expressed as follows:

[0074] ,

[0075] in, For dynamic scoring results, The dynamic rate of change of delay. This is the bandwidth fluctuation index. The current packet loss rate, , , The dynamic weighting coefficients (initial values ​​are respectively) , , When the topology changes (e.g., when a new device node is added); Increase by 20% (i.e.) ), enhancing latency sensitivity; when bandwidth utilization > 85%, Increase by 30% (i.e.) ), amplifying the impact of bandwidth fluctuations;

[0076] Multi-threshold linkage strategies include: Scenario 1: Sudden increase in latency And bandwidth utilization Scenario 1: Triggering link switching + prioritizing maintenance of high-load devices; Scenario 2: Topology changes lead to uneven bandwidth distribution, triggering dynamic adjustment of communication path priority (such as prioritizing the bandwidth of core nodes).

[0077] Furthermore, the command generation outputs commands in JSON format, for example:

[0078] {

[0079] Device ID: R001

[0080] "Maintenance Action": "Switch to Backup Link"

[0081] Priority: 1,

[0082] Triggering conditions: Latency = 62ms, Bandwidth utilization = 93%

[0083] }

[0084] S40: Collect equipment status feedback data after the execution of maintenance instructions, filter data whose prediction deviation exceeds the set threshold, use dynamic loss function to strengthen the weight of new scenario data, train the model, and update the model version when the accuracy of the validation set continuously improves beyond the set threshold and supports rollback.

[0085] It should be noted that incremental learning is driven by real-time feedback data (input: device status after command execution), enabling minute-level hot updates of the model (output: new version model and parameters), solving the problem of lag in traditional offline training and reducing prediction errors in sudden scenarios.

[0086] Furthermore, an incremental learning-driven closed-loop feedback mechanism is designed, including incremental training and model version management;

[0087] Incremental training includes prioritizing data with high scoring errors (such as prediction bias). ), using the loss function to increment the random forest Training, denoted as:

[0088] ,

[0089] in, The loss value is the historical data. The loss value for newly added data, such as the dynamic weight coefficient during topology reconstruction. In typical scenarios, dynamic weighting coefficients If new features are added after topology reconstruction) data, The model will prioritize learning new data features, thus improving its adaptation speed. ;

[0090] Model version management includes retaining historical model versions (such as...) , Supports fast rollback; version update trigger condition is an improvement in validation set accuracy after 3 consecutive iterations. If the new version The accuracy on the validation set decreased. Then it will automatically roll back to This ensures system stability.

[0091] Example 2, refer to Figure 2 This invention provides an LSTM power IoT ring network cabinet optimization method. To verify the beneficial effects of this invention, scientific demonstration is carried out through experiments.

[0092] A large-scale power Internet of Things (IoT) system includes multiple ring main units (RNBs) for distributing power to different device nodes. Each device node in the system transmits information such as latency, packet loss rate, bandwidth, and topology changes via a wireless network. The RNBs need to monitor this data in real time and adjust maintenance strategies according to changes in the network environment to ensure stable operation of the equipment.

[0093] like Figure 2 As shown, the dynamic maintenance and optimization process of this invention under sudden network fluctuation scenarios is illustrated, including:

[0094] Network fluctuations and data acquisition: The power IoT system where the ring main unit is located experienced network fluctuations, specifically manifested as a sudden delay from... Rise to bandwidth from Rising sharply to Furthermore, due to the addition of 5 device nodes, the network topology has changed; at this point, traditional static threshold methods will struggle to make accurate maintenance decisions because they typically rely on fixed rules (e.g., latency exceeding a certain threshold). The system cannot adapt to sudden network fluctuations or topology changes, resulting in excessive or insufficient maintenance.

[0095] Data preprocessing and feature extraction: The ring main unit maintenance system adopts an intelligent decision-making method based on LSTM and real-time network fluctuations. When a network fluctuation event occurs, the system first performs data preprocessing, calculates the average value and standard deviation of latency and bandwidth in real time using dynamic Z-score normalization technology, and judges whether the latency fluctuation exceeds the normal fluctuation range based on a sliding window method. For example, when the latency suddenly increases, the system will correct it by interpolating historical data to avoid misjudging the sudden increase in latency as an abnormal situation.

[0096] LSTM Temporal Modeling and Dynamic Weight Adjustment: The system fuses multi-dimensional data using an LSTM model to learn the temporal relationships between latency, bandwidth, packet loss rate, and topology changes. The LSTM network can capture the long-term dependence of latency and bandwidth fluctuations on device performance and make performance predictions accordingly. Figure 2 As shown, when the bandwidth fluctuation index exceeds the threshold, the system dynamically adjusts the forget gate weights (e.g., the weights are increased by 3 times) to improve the prediction accuracy of the model in complex network environments.

[0097] Dynamic scoring and maintenance instruction generation: When making equipment maintenance decisions, the system comprehensively evaluates the impact of multiple factors through a dynamic scoring model. The system not only considers sudden increases in latency, but also calculates the fluctuations in bandwidth, packet loss rate, and the impact of topology changes in real time to generate a comprehensive performance score. For example, when latency suddenly increases and bandwidth utilization is close to saturation, the system will prioritize high-load devices and generate a JSON instruction of "switching links + high-priority maintenance".

[0098] Execution maintenance and closed-loop feedback optimization: After executing maintenance commands, the system collects equipment status feedback data in real time; if the prediction deviation exceeds... If this happens, the incremental learning mechanism is triggered to update the model parameters. For example, if the addition of a new device node causes a change in the topology, the system can quickly adapt to the network environment through incremental learning, thereby improving the accuracy of subsequent decisions.

[0099] Example 3 is an embodiment of the present invention, which provides an LSTM power IoT ring network cabinet optimization system, including:

[0100] The dynamic feature construction module collects real-time communication status data and network environment data of the ring network cabinet, calculates data within the sliding window in real time, judges outliers by standardization through dynamic standard scores and performs historical data interpolation correction, calculates dynamic delay change rate and bandwidth fluctuation index, and generates a multi-dimensional feature matrix.

[0101] The time-adaptive prediction module inputs a multi-dimensional feature matrix into the long short-term memory network, adjusts the forget gate weights in real time based on the bandwidth fluctuation index, and outputs device performance scores and key influencing factors.

[0102] The multi-threshold decision module adjusts the dynamic latency change rate, bandwidth fluctuation index, and packet loss rate according to the real-time network status to build a dynamic scoring algorithm, and uses a multi-threshold linkage strategy to define latency and bandwidth overload scenarios to trigger differentiated maintenance instructions.

[0103] The closed-loop optimization module collects equipment status feedback data after the execution of maintenance instructions, filters data whose prediction deviation exceeds a set threshold, uses a dynamic loss function to strengthen the weight of new scenario data, trains the model, and updates the model version and supports rollback when the accuracy of the validation set continuously improves beyond the set threshold.

[0104] This embodiment also provides an electronic device applicable to an LSTM power IoT ring main unit optimization method, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the LSTM power IoT ring main unit optimization method proposed in the above embodiment.

[0105] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements an LSTM power IoT ring network cabinet optimization method as proposed in the above embodiment.

[0106] The storage medium proposed in this embodiment and the method for optimizing an LSTM power IoT ring network cabinet 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.

[0107] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using 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 can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0108] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not 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. An LSTM power Internet of Things ring main unit optimization method, characterized in that: comprising, Real-time acquisition of ring network cabinet communication state data and network environment data, real-time calculation of sliding window data, dynamic standard score standardization to determine abnormal values and historical data interpolation correction, calculation of dynamic delay change rate and bandwidth fluctuation index, generation of multi-dimensional feature matrix; Input multi-dimensional feature matrix into long short-term memory network, real-time adjustment of forgetting gate weight based on bandwidth fluctuation index, output device performance score and key influence factor; According to the real-time network state, adjust the dynamic delay change rate, bandwidth fluctuation index and packet loss rate to build a dynamic scoring algorithm, and use a multi-threshold linkage strategy to define delay and bandwidth overload scenarios, and trigger differentiated maintenance instructions; Collect device state feedback data after executing maintenance instructions, filter data with prediction deviation exceeding the set threshold, use dynamic loss function to strengthen the weight of new scene data, and train the model. When the accuracy of the validation set improves by more than a set threshold, update the model version and support rollback. 2.The LSTM power IOT ring main unit optimization method of claim 1, wherein: The real-time calculation of the data in the sliding window includes setting a fixed time length of the sliding window, and real-time intercepting the communication state data and network environment data in the window; calculate the mean and standard deviation of the data in the sliding window, calculate the standard score by dynamic Z-score formula, if the absolute value of the standard score exceeds the pre-set abnormal threshold, it is determined as an abnormal value; and correct it by historical data interpolation. 3.The LSTM power IOT ring main unit optimization method of claim 2, wherein: The multi-dimensional feature matrix includes converting the original communication data into a high-precision multi-dimensional feature matrix through dynamic sliding window Z-score standardization and historical data interpolation correction mechanism; Among them, the original communication data includes real-time delay, packet loss rate, bandwidth and topology state; The high-precision multi-dimensional feature matrix includes delay, packet loss rate, bandwidth, topology state, dynamic delay change rate and bandwidth fluctuation index. 4.The LSTM power IOT ring main unit optimization method of claim 3, wherein: The output device performance score and key influence factor includes dynamically adjusting the LSTM gate weight based on the bandwidth fluctuation index, building an adaptive time series fusion model, and outputting the device performance score and key influence factor.

5. The LSTM power internet of things ring main unit optimization method of claim 4, wherein: The construction of dynamic scoring algorithm includes calculating the comprehensive score through dynamic weight coefficient by comprehensively considering dynamic delay change rate, bandwidth fluctuation index and current packet loss rate.

6. The LSTM power internet of things ring main unit optimization method of claim 5, wherein: The multi-threshold linkage strategy includes triggering high-priority maintenance instructions and executing link switching when the delay exceeds the first threshold and the bandwidth utilization rate exceeds the second threshold; when topology changes and bandwidth allocation is uneven, trigger the communication path priority dynamic adjustment instruction.

7. The LSTM power internet of things ring main unit optimization method of claim 6, wherein: The dynamic loss function includes fusing historical and new loss values, weighting and summing the historical data loss value and the new data loss value; and adjusting the weight according to the scene, and selecting the new data with prediction deviation exceeding the set threshold to participate in the calculation.

8. An LSTM power IOT ring main unit optimization system, applying the LSTM power IOT ring main unit optimization method of any one of claims 1-7, characterized in that, comprising, A dynamic feature construction module acquires ring network cabinet communication state data and network environment data in real time, calculates data in a sliding window in real time, determines abnormal values through dynamic standard score standardization and performs historical data interpolation correction, calculates dynamic delay change rate and bandwidth fluctuation index, and generates a multi-dimensional feature matrix; A time series adaptive prediction module inputs the multi-dimensional feature matrix into a long short-term memory network, real-time adjusts the forgetting gate weight based on the bandwidth fluctuation index, and outputs the device performance score and key influence factor; A multi-threshold decision module adjusts the dynamic delay change rate, bandwidth fluctuation index and packet loss rate to build a dynamic scoring algorithm according to the real-time network state, and defines the delay and bandwidth overload scenarios by using a multi-threshold linkage strategy to trigger differentiated maintenance instructions; A closed-loop optimization module collects device state feedback data after the execution of the maintenance instructions, filters data with a prediction deviation exceeding a set threshold, uses a dynamic loss function to strengthen the weight of new scene data, performs model training, and updates the model version and supports rollback when the verification set accuracy rate improves by more than a set threshold. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to implement the steps of the LSTM power Internet of Things ring network cabinet optimization method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the LSTM power Internet of Things ring network cabinet optimization method in any one of claims 1 to 7.

Images