A method and device for data processing of power distribution ring main unit

Abstract

This application provides a data processing method and apparatus for distribution ring main units (RNBs), applied in the field of power system data processing technology. By collecting operational data from the RNBs and identifying their operating modes, it enables the selection of a suitable data processing model based on the operating mode. The selected model is then applied to process the data, outputting equipment health status assessment results and anomaly detection results. The model's parameters are adjusted based on the assessment results, determining whether the data processing model's performance meets actual conditions or preset standards. Therefore, this application solves the technical problem of decreased accuracy in complex dynamic environments using existing methods. It offers the advantages of improving data processing accuracy, reducing data processing error rates, and dynamically adjusting parameters to adapt to different operating modes of the RNBs, thereby improving assessment reliability.

Description

Technical Field

[0001] This application relates to the field of power system data processing technology, and in particular to a data processing method and apparatus for a distribution ring network cabinet. Background Technology

[0002] As a key piece of equipment in power distribution networks, CGS switchgear undertakes the functions of power distribution, control, and protection. Its internal structure is complex, containing main circuits (such as circuit breakers, disconnectors, grounding switches, busbars, and contacts) and auxiliary circuits (such as operating mechanisms, control loops, sensors, and protective relays). To achieve real-time monitoring of equipment operating status and timely detection of anomalies, CGS switchgear is typically equipped with various types of sensors to continuously collect large amounts of operating data. The collected data is transmitted to local or remote data processing units, where it is analyzed using preset algorithms and models to assess the health status of the equipment, predict potential faults, and identify current abnormal operating conditions.

[0003] However, during long-term operation, the internal components of CGS switchgear inevitably undergo a dynamic evolution process. In the initial stages of operation, new components require a break-in period, and their physical characteristics and operational performance may not yet be fully stable. This can lead to fluctuations in initial operating data or discrepancies between the initial data and the data distribution under long-term stable operation. As operating time accumulates, equipment components experience natural wear, aging, and deterioration, causing the characteristics of equipment operating data to drift slowly over time. Status assessment benchmarks or trained models built based on "normal" data collected during the initial operation phase or at a specific point in time may not accurately reflect the true operating characteristics of the equipment under long-term operation or aging conditions. This can lead to biased assessment results, increasing the risk of false alarms (mistaking normal aging characteristics for abnormalities) or missed alarms (failing to detect true deterioration trends in a timely manner).

[0004] Furthermore, CCGS switchgear does not always operate under a single, stable load mode. The equipment experiences different load levels (light load, heavy load) and operating frequencies. More importantly, the equipment periodically enters atypical operating modes, whose data characteristics differ significantly from those of normal load power supply modes. In atypical modes, the main circuit current and voltage waveforms may be completely different (e.g., fault current), the control signal sequence and timing of auxiliary circuits may differ from normal operation (e.g., specific logic for standby switching), and the sequence and time intervals of switching operations may also differ from remote / local operation under normal operation. If the data processing method cannot effectively identify and distinguish these different operating modes, conflating data from atypical modes with normal operation data, models trained based on normal modes may misjudge normal data characteristics in these modes as abnormal, leading to numerous false alarms and interfering with the judgment of operators.

[0005] In summary, the operational data characteristics of CGS switchgear face challenges as the equipment evolves throughout its lifecycle (initial break-in, long-term aging) and its operating modes switch (normal, atypical modes). Existing equipment status assessment and anomaly identification methods that rely on fixed data processing logic or models are ill-suited to these complex and ever-changing factors. This leads to a decrease in the accuracy of assessments throughout the equipment's lifecycle and under various operating modes, and an increase in false alarm or missed alarm rates, failing to meet the requirements of modern power distribution networks for equipment reliability and intelligent operation and maintenance.

[0006] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0007] In view of the shortcomings of the prior art, this application provides a data processing method and apparatus for distribution ring main units, which has the beneficial effects of improving data processing accuracy, reducing data processing error rate, and dynamically adjusting parameters to adapt to different operating modes of distribution ring main units, thereby improving the reliability of evaluation.

[0008] A first aspect includes a data processing method for a power distribution ring main unit, the method comprising:

[0009] S1: Obtain the operating data of the power distribution ring network cabinet;

[0010] S2: Analyze the operating data to identify the current operating mode of the power distribution ring network cabinet;

[0011] S3: Based on the identified current operating mode, select the corresponding data processing model from the preset mode corresponding model library. The data processing model includes an equipment status assessment model and an anomaly detection model.

[0012] S4: Analyze the operating data according to the equipment status assessment model and the anomaly detection model to obtain the equipment status assessment results and anomaly detection results;

[0013] S5: Determine the performance of the data processing model based on the equipment status assessment results and the anomaly detection results, obtain the determination results, and adjust the parameters of the data processing model based on the determination results.

[0014] This application proposes a data processing method for distribution ring main units (RMS). By collecting various data during the operation of the RMS and analyzing the collected data, the method identifies the operating mode. This allows for the selection of a suitable data processing model based on the operating mode. The selected model is then applied to process the data, outputting equipment health status assessment results and anomaly detection results. By comparing the output equipment assessment results and anomaly detection results with the output of the data processing model, the method determines whether the model's performance meets the actual situation or preset standards. If the model's performance is unsatisfactory, its parameters are adjusted based on the assessment results. This feedback adjustment mechanism enables the model to adaptively learn and adapt to the drift and changes in equipment operating data over time, maintaining the effectiveness of the data processing model. Therefore, this application solves the technical problem of decreased accuracy in existing methods under complex dynamic environments. It has the beneficial effects of improving data processing accuracy, reducing data processing error rate, and dynamically adjusting parameters to adapt to different operating modes of the RMS, thereby improving assessment reliability.

[0015] Furthermore, in step S1, the power distribution ring network cabinet is equipped with at least a main circuit and an auxiliary circuit, and the main circuit and the auxiliary circuit are respectively equipped with multiple sensors for collecting analog signal data including current, voltage, temperature and switch status.

[0016] Step S1 includes:

[0017] S11: Convert the analog signal data into digital signal data, and preprocess the digital signal data to unify the operating data of different dimensions into the [0,1] interval;

[0018] S12: Perform data verification on the preprocessed digital signal data;

[0019] S13: For the digital signal data that fails verification, initiate an automatic retransmission request until the digital signal data is successfully verified or the maximum number of retransmissions is reached; if the verification still fails after reaching the maximum number of retransmissions, mark the digital signal data as unreliable data, record the error type and occurrence time, and use the successfully verified digital signal data as the running data.

[0020] The data processing method for distribution ring main units proposed in this application ensures that the operational data input to subsequent processing stages is reliable data that has been converted, standardized in dimensions, and verified. This significantly improves the robustness and accuracy of the entire data processing method and solves the negative impact of raw data quality issues on subsequent analysis.

[0021] Furthermore, step S2 includes:

[0022] S21: Extract the current, voltage, and temperature from the operating data to construct a multidimensional time series;

[0023] S22: Using the sliding window method, the multidimensional time series is divided into multiple time windows, and for each time window, the statistical characteristics of current, voltage, and temperature are calculated;

[0024] S23: Input the statistical features into a pre-trained classification model, which is built based on the support vector machine algorithm, to identify the current operating mode of the power distribution ring network cabinet and output the probability distribution of the operating mode;

[0025] S24: Based on the probability distribution, select the operating mode with the highest probability as the current operating mode of the power distribution ring network cabinet.

[0026] This application proposes a data processing method for distribution ring main units. This method utilizes the statistical characteristics of operating parameters and combines them with a pre-trained classification model to identify different operating modes of distribution ring main units, including distinguishing between normal load modes and atypical modes. This provides mode information for subsequent steps to select data processing models and avoids misjudgments caused by inaccurate mode recognition.

[0027] Furthermore, step S23 includes:

[0028] S231: Extract the statistical feature vectors of current and voltage, as well as the maximum and minimum values ​​of temperature, from the statistical features;

[0029] S232: Input the extracted current and voltage statistical feature vectors and the maximum and minimum values ​​of temperature into the classification model;

[0030] S233: Using the classification model, with the statistical feature vector and the maximum and minimum values ​​of temperature as inputs, calculate the probability distribution characterizing the current operating mode of the power distribution ring main unit.

[0031] This application proposes a data processing method for distribution ring main units. By optimizing the input features of the classification model, the accuracy of pattern recognition is improved. Compared with using all statistical features, the combined input of current and voltage statistical feature vectors and the maximum and minimum values ​​of temperature can more effectively capture key discriminative information of different operating modes, reduce the influence of redundant or interfering features, and thus enable the classification model to more accurately identify the current operating mode, providing a more reliable foundation for subsequent pattern-based data processing.

[0032] Furthermore, the preset mode-corresponding model library in step S3 contains at least corresponding operating modes and data processing models, with each operating mode corresponding to one or more data processing models.

[0033] Step S3 includes:

[0034] S31: Based on the identified current operating mode, select one or more of the corresponding data processing models from the preset mode corresponding model library;

[0035] S32: When the current operating mode corresponds to multiple data processing models, evaluate the electromagnetic interference intensity in the operating environment of the power distribution ring network cabinet, and adjust the selection confidence level of each data processing model according to the electromagnetic interference intensity.

[0036] S33: Obtain the accuracy of each of the data processing models in the historical data processing process, and adjust the selection weight corresponding to each of the data processing models according to the accuracy;

[0037] S34: Multiply the selection confidence level by the selection weight to obtain the selection probability of each data processing model;

[0038] S35: Select the data processing model with the highest selection probability as the data processing model corresponding to the current operating mode.

[0039] Furthermore, step S4 includes:

[0040] S41: Using the equipment status assessment model, with current, voltage, and temperature as inputs, assess the health status parameters of the power distribution ring main unit;

[0041] S42: Using the aforementioned anomaly detection model, with the degree of sudden change and fluctuation frequency of current and voltage as input, detect whether there are any abnormal situations that do not conform to the historical normal operation mode, and obtain the anomaly detection result;

[0042] S43: Calculate the overall health score of the power distribution ring network cabinet based on the health status parameters, correct the overall health score based on the anomaly detection results, and obtain the equipment status assessment result.

[0043] Furthermore, step S5 includes:

[0044] S51: Based on the collected equipment status assessment results and anomaly detection results, if the anomaly detection results show an anomaly and at least one health status parameter in the equipment status assessment results exceeds a preset threshold, it is judged as a true positive; otherwise, it is a false positive. If the anomaly detection results show no anomaly and all health status parameters in the equipment status assessment results do not exceed the preset threshold, it is judged as a true negative; otherwise, it is a false negative. Based on the number of true positives, false positives, true negatives, and false negatives, the precision, recall, and F1 score are calculated as comprehensive performance indicators of the data processing model.

[0045] S52: Determine whether the comprehensive performance index meets the preset threshold. If not, determine the adjustment direction and magnitude of the parameters of the data processing model based on the precision, recall and F1 score.

[0046] S53: Based on the adjustment direction and magnitude, the parameters of the data processing model are adjusted using an adaptive learning rate optimization algorithm.

[0047] Furthermore, step S52 includes:

[0048] S521: If the accuracy is lower than a preset threshold, the anomaly threshold of the anomaly detection model is increased, and the increase in the anomaly threshold is set according to the degree to which the accuracy is lower than the threshold.

[0049] S522: If the recall rate is lower than a preset threshold, the anomaly threshold of the anomaly detection model is reduced, and the reduction range of the anomaly threshold is set according to the degree to which the recall rate is lower than the threshold.

[0050] S523: If the F1 value is lower than the preset threshold, the weights of the corresponding parameters in the equipment condition assessment model are adjusted according to the degree of insulation aging, contact wear, and heat dissipation performance degradation in the abnormal detection results, and the weight adjustment range is determined.

[0051] Furthermore, step S53 includes:

[0052] S531: Calculate the adjusted learning rate based on the adjustment direction and the adjustment magnitude. The calculation formula is: Adjusted learning rate = Current learning rate + Adjustment direction * Adjustment magnitude;

[0053] S532: The Adam optimization algorithm is used, with an adjusted learning rate, and momentum estimation and adaptive moment estimation are combined to update the parameters of the data processing model to accelerate convergence and improve model performance.

[0054] Secondly, a power distribution ring main unit data processing device is applied to the steps of any of the above-described methods, the device comprising:

[0055] Acquisition module: Acquires the operating data of the power distribution ring network cabinet;

[0056] First analysis module: Analyzes the operating data and identifies the current operating mode of the power distribution ring network cabinet;

[0057] Selection module: Based on the identified current operating mode, select the corresponding data processing model from the preset mode corresponding model library. The data processing model includes the equipment status assessment model and the anomaly detection model.

[0058] The second analysis module analyzes the operating data based on the equipment status assessment model and the anomaly detection model to obtain the equipment status assessment results and anomaly detection results.

[0059] Adjustment module: Based on the equipment status assessment results and the anomaly detection results, the module determines the performance of the data processing model, obtains the determination result, and adjusts the parameters of the data processing model based on the determination result.

[0060] Beneficial Effects: This application proposes a data processing method and apparatus for distribution ring main units. By collecting various data during the operation of the distribution ring main unit and analyzing the collected operational data, the operating mode is identified. This allows for the selection of a suitable data processing model based on the operating mode, application of the selected model for data processing, and output of equipment health status assessment results and anomaly detection results. By comparing the output equipment assessment results and anomaly detection results with the output of the data processing model, it is determined whether the performance of the data processing model meets the actual situation or preset standards. If the performance of the data processing model is poor, the model parameters are adjusted according to the assessment results. This feedback adjustment mechanism enables the model to adaptively learn and adapt to the drift and changes in equipment operating data over time, maintaining the effectiveness of the data processing model. Therefore, this application solves the technical problem of decreased accuracy in existing methods under complex dynamic environments. It has the beneficial effects of improving data processing accuracy, reducing data processing error rate, and dynamically adjusting parameters to adapt to different operating modes of the distribution ring main unit, thereby improving the reliability of the assessment. Attached Figure Description

[0061] Figure 1 This is a flowchart of a data processing method for a power distribution ring main unit proposed in this application.

[0062] Figure 2 This is a structural diagram of a power distribution ring main unit data processing device proposed in this application.

[0063] Figure 3 This is a structural block diagram of a power distribution ring main unit data processing method proposed in this application.

[0064] Labeling Explanation: 201, Acquisition Module; 202, First Analysis Module; 203, Selection Module; 204, Second Analysis Module; 205, Adjustment Module. Detailed Implementation

[0065] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and marked in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0066] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0067] Please refer to Figure 1 , Figure 3 A method for processing data from a power distribution ring main unit, the method comprising:

[0068] S1: Obtain the operating data of the power distribution ring network cabinet;

[0069] S2: Analyze operational data to identify the current operating mode of the power distribution ring network cabinet;

[0070] S3: Based on the identified current operating mode, select the corresponding data processing model from the preset model library corresponding to the mode. The data processing models include the equipment status assessment model and the anomaly detection model.

[0071] S4: Analyze the operating data based on the equipment condition assessment model and the anomaly detection model to obtain the equipment condition assessment results and anomaly detection results;

[0072] S5: Determine the performance of the data processing model based on the equipment status assessment results and anomaly detection results, and adjust the parameters of the data processing model according to the judgment results.

[0073] This method aims to address the issue of decreased accuracy in existing fixed models caused by the evolution of power distribution ring main unit (GN) operation data and changes in operating modes throughout the equipment's lifecycle. Specifically, in step S1, the operation data is typically collected by sensors installed on the main and auxiliary circuits, including analog signal data such as current, voltage, temperature, and switch status. The acquired data is preprocessed and verified before being used as the operation data.

[0074] In step S2, features are extracted from the operational data, such as statistical features of current, voltage, and temperature, and then input into a pre-trained classification model. The classification model outputs a probability distribution of the operational modes, and selects the mode with the highest probability as the current operational mode.

[0075] In step S3, the model library stores data processing models optimized or trained for different operating modes. Each operating mode can correspond to one or more data processing models. The selection process can consider the electromagnetic interference intensity of the operating environment and the accuracy of the model in historical data processing to determine the model selection probability, and select the model with the highest probability.

[0076] In step S4, the equipment status assessment model can use inputs such as current, voltage, and temperature to evaluate the health status parameters of the distribution ring main unit, while the anomaly detection model can use inputs such as the degree of sudden changes and fluctuation frequency of current and voltage to detect abnormal conditions. Based on the assessment and detection results, the overall health score of the distribution ring main unit can be calculated.

[0077] In step S5, performance evaluation can be performed by comparing the model output with the actual situation or a preset standard, and calculating comprehensive performance indicators such as precision, recall, and F1 score. If the comprehensive performance indicators do not meet the preset threshold, the direction and magnitude of parameter adjustment are determined based on the indicators, and an adaptive learning rate optimization algorithm is used to adjust the model parameters.

[0078] By combining operation mode recognition and performance-based model adaptive adjustment, this method can effectively address the challenges posed by changes in the operation data of distribution ring main units over time and under operating conditions, thereby improving the accuracy and reliability of data processing.

[0079] The purpose of the equipment condition assessment model is to evaluate the health status parameters of the distribution ring main unit based on the input current, voltage, and temperature data. It is assumed that the health status parameters are a multi-dimensional vector representing the equipment's health status in different aspects. Let the input data be:

[0080] I(t): Current data, a sequence of changes over time t.

[0081] V(t): Voltage data, a sequence over time t.

[0082] T(t): Temperature data, a sequence of changes over time t.

[0083] The equipment condition assessment model can be represented as a function H = f(I(t), V(t), T(t)), where H is a health state parameter vector representing the health state of the equipment. Assuming it is a 3-dimensional vector, i.e., H = [H1, H2, H3], where H1 represents the insulation performance health state, H2 represents the temperature stability health state, and H3 represents the electrical performance health state. Specifically, the equipment condition assessment model is a complex function, typically implemented through machine learning; here, it can be described in a simplified form: in,

[0084] u I u V u T These are the average values ​​of current, voltage, and temperature, respectively, σ I σ V σ T , respectively, are the standard deviations of current, voltage, and temperature, W is the weight matrix, representing the influence of input features on health state parameters, and b is the bias vector.

[0085] The purpose of an anomaly detection model is to detect the presence of abnormalities based on the degree of abrupt changes and the frequency of fluctuations in the input current and voltage. The output is the anomaly detection result, typically a binary variable (abnormal or normal).

[0086] ΔI(t) represents the degree of abrupt change in current, which can be defined as the rate of change or difference of current.

[0087] ΔV(t) represents the degree of voltage abrupt change, which can be defined as the rate of change or difference of voltage.

[0088] f I (t) is the fluctuation frequency of the current, which can be obtained through Fourier transform or wavelet transform.

[0089] f V (t) represents the voltage fluctuation frequency, which can be obtained through Fourier transform or wavelet transform.

[0090] Anomaly detection models can be represented as a function A = g(ΔI(t), ΔV(t), f I (t),f V (t)), where A is the anomaly detection result, which is usually a binary variable (0 indicates normal and 1 indicates abnormal).

[0091] An anomaly detection model can be a classifier (such as logistic regression, support vector machine, neural network, etc.), which can be described in a simplified form here:

[0092] Where s is the activation function, such as the sigmoid function, used to map the output to the interval [0,1]. W A It is a weight vector, representing the influence of input features on the anomaly detection result, b A It is a bias term.

[0093] In some specific implementations, the data processing system of the distribution ring main unit executes this method periodically. For example, every ten minutes, the system acquires the latest operating data. The system analyzes this data and identifies the current operating mode as "normal load." Based on the "normal load" mode, the system selects a pre-trained equipment status assessment model A and anomaly detection model B for the normal load mode from the model library. Model A uses current, voltage, and temperature data to assess the insulation status, contact wear, and heat dissipation performance of the equipment, obtaining health status parameters. Model B uses the rate of change of current and voltage data to detect whether there are sudden anomalies. Based on the outputs of Model A and Model B, the system determines the overall health score of the equipment and whether there are any anomalies. At the same time, the system records the output results of Model A and Model B. In subsequent processing cycles, the system compares the model output results with the actual situation (e.g., through manual inspection or feedback confirmation from higher-level systems) and calculates the precision, recall, and F1 score of Model A and Model B.

[0094] If the calculated F1 score is lower than the preset threshold of 0.8, the system determines that the model performance has deteriorated. Based on the degree of the F1 score decline and the specific false positive / false negative situation, the system determines that model parameters need to be adjusted. For example, if a high false positive rate leads to low precision, the system will increase the anomaly threshold of anomaly detection model B; if a high false negative rate leads to low recall, the system will decrease the anomaly threshold of anomaly detection model B. The system uses an adaptive learning rate optimization algorithm to update the internal parameters of model A and model B, such as the weights of the neural network or the kernel function parameters of the support vector machine, according to the determined adjustment direction and magnitude, to improve the accuracy of the model in subsequent data processing.

[0095] Furthermore, in step S1, the power distribution ring network cabinet is equipped with at least a main circuit and an auxiliary circuit. Multiple sensors are installed on the main circuit and the auxiliary circuit respectively to collect analog signal data including current, voltage, temperature and switch status.

[0096] Step S1 includes:

[0097] S11: Convert analog signal data into digital signal data and preprocess the digital signal data to unify the operating data of different dimensions into the [0,1] interval;

[0098] S12: Perform data verification on the preprocessed digital signal data;

[0099] S13: For digital signal data that fails verification, initiate an automatic retransmission request until the digital signal data is successfully verified or the maximum number of retransmissions is reached; if the verification still fails after reaching the maximum number of retransmissions, mark the digital signal data as unreliable data, record the error type and occurrence time, and use the successfully verified digital signal data as the running data.

[0100] The main circuit of the distribution ring main unit is equipped with sensors. The main circuit includes components such as circuit breakers, disconnect switches, grounding switches, busbars, and contacts. The sensors can be installed near these components or integrated into them to collect operating parameters directly related to power transmission. Furthermore, sensors are also installed on the auxiliary circuits. The auxiliary circuits include operating mechanisms, control circuits, sensors, and protective relays. The sensors can be installed on these auxiliary components to collect parameters related to equipment control and status indication.

[0101] Specifically, sensors used to acquire current can be implemented using current transformers or Hall effect sensors; sensors used to acquire voltage can be implemented using voltage transformers or voltage sensors; sensors used to acquire temperature can be implemented using resistance temperature detectors (RTDs) or thermocouples; and sensors used to acquire switch states can be implemented using limit switches or proximity switches. These sensors acquire analog signal data, the values ​​of which change continuously with the physical quantity.

[0102] In step S11, the conversion of analog signal data into digital signal data is typically achieved using an analog-to-digital converter (ADC). The digital signal data undergoes preprocessing to unify the different dimensions of the operating data into the 0,1 range. This can be achieved using the minimum-maximum normalization (MMR) method, which linearly scales the original data to a specified range. For example, assuming the analog signal range acquired by the voltage sensor is 0V to 1000V, and the analog signal range acquired by the temperature sensor is 0℃ to 100℃, in step S11, the 500V analog voltage signal is converted into a digital signal by the ADC, and then converted to (500-0) / (1000-0) = 0.5 using the MMR formula (x-min) / (max-min). The 50℃ analog temperature signal is converted into a digital signal and then normalized to (50-0) / (100-0) = 0.5.

[0103] In step S12, the preprocessed digital signal data undergoes data verification. This can be achieved using techniques such as Cyclic Redundancy Check (CRC) to calculate the check value of the data block and compare it with the received check value. Specifically, during the transmission of digital signal data, the data to be transmitted is first divided into several data blocks of equal or unequal size. The size of the data blocks can be determined based on the specific application scenario and verification algorithm. Then, for each data block, a check value is calculated using the Cyclic Redundancy Check (CRC) algorithm. The calculated check value is then appended to the end of the data block to form a complete data block containing both data and verification information. Finally, this complete data block is transmitted. The receiving end receives and reads the complete data block, comparing its check value with the received check value. If they match, the verification is successful.

[0104] In step S13, an automatic retransmission request is initiated for digital signal data that fails verification. When data verification fails, the data receiver sends a retransmission request to the data sender, requesting the data to be retransmitted. This process repeats until data verification succeeds or the preset maximum number of retransmissions is reached. This can be implemented using a communication protocol based on an acknowledgment / retransmission mechanism, such as TCP or a custom application layer protocol. If verification still fails after reaching the maximum number of retransmissions, the digital signal data is marked as unreliable, and the error type and occurrence time are recorded. For example, a flag bit can be added to the data record to indicate that the data is unreliable, and the type of verification failure (such as CRC error, range error) and the timestamp of occurrence can be recorded. Successfully verified digital signal data is used as running data; this data is considered accurate and reliable and can be used in subsequent data processing steps.

[0105] Only data frames that pass verification are used to extract operational data (normalized current, voltage, and temperature) and stored as reliable operational data in the database for use by subsequent operational mode recognition, equipment status assessment, and anomaly detection algorithms.

[0106] Furthermore, step S2 includes:

[0107] S21: Extract current, voltage, and temperature from the operating data to construct a multidimensional time series;

[0108] S22: The sliding window method is used to divide the multidimensional time series into multiple time windows, and the statistical characteristics of current, voltage and temperature are calculated for each time window.

[0109] S23: Input the statistical features into the pre-trained classification model. The classification model is built based on the support vector machine algorithm and is used to identify the current operating mode of the power distribution ring network cabinet and output the probability distribution of the operating mode.

[0110] S24: Based on the probability distribution, select the operating mode with the highest probability as the current operating mode of the power distribution ring network cabinet.

[0111] In step S21, current, voltage, and temperature are key parameters reflecting the operating status of the distribution ring main unit. These data can be collected by sensors installed on the main and auxiliary circuits of the distribution ring main unit. These parameters are then integrated over time to form a data sequence containing multiple variables (current, voltage, and temperature) changing over time.

[0112] In step S22, the time window and sliding step size of the sliding window are determined according to the actual application. For example, the sliding window size can be set to 60 seconds and the sliding step size to 10 seconds. For the data within each time window, statistical characteristics of current, voltage, and temperature are calculated. These statistical characteristics may include, but are not limited to, mean, variance, standard deviation, maximum value, and minimum value. These statistical characteristics can summarize the distribution and variation patterns of the data within the window, transforming the original time-series data into a more compact feature representation.

[0113] Step S23 inputs the calculated statistical features into a pre-trained classification model. This classification model is built based on the Support Vector Machine (SVM) algorithm. SVM is a supervised learning model that classifies data by constructing a hyperplane in a high-dimensional space. The model is pre-trained on a large amount of labeled historical operating data to learn the statistical feature patterns corresponding to different operating modes (such as normal load, light load, heavy load, switching operation, fault, etc.). The model receives the statistical feature vector of the current time window as input and outputs the probability distribution of the data in that time window belonging to each preset operating mode.

[0114] Step S24 selects the mode with the highest probability as the current operating mode based on the operating mode probability distribution output in step S23. The probability distribution reflects the model's confidence in the current data belonging to various modes. Selecting the mode with the highest probability means adopting the operating state that the model considers most likely as the final identification result. For example, for the normal operating mode, a device status assessment model focusing on trend analysis and health scoring might be selected, while for the switch operation mode, an anomaly detection model focusing on transient waveform analysis and operation sequence verification might be selected. This method of dynamically selecting the model based on the mode allows the entire data processing system to better adapt to the data characteristics of the distribution ring main unit under different states, improving the accuracy of device status assessment and anomaly detection, avoiding misjudging normal data under atypical modes as anomalies, thereby reducing false alarms and improving the overall system reliability.

[0115] Furthermore, step S23 includes:

[0116] S231: Extract statistical feature vectors of current and voltage, as well as the maximum and minimum values ​​of temperature from statistical features;

[0117] S232: Input the extracted statistical feature vectors of current and voltage, and the maximum and minimum values ​​of temperature into the classification model;

[0118] S233: Using a classification model, with statistical feature vectors and the maximum and minimum values ​​of temperature as inputs, the probability distribution characterizing the current operating mode of the distribution ring main unit is calculated.

[0119] In step S231, the statistical feature vectors of current and voltage refer to a numerical sequence formed by combining multiple statistical features of current or voltage. Specifically, this can be implemented using an array or list data structure. For example, the current statistical feature vector can include the average value, root mean square value, maximum value, and variance of the current, while the voltage statistical feature vector can include the average value, root mean square value, minimum value, and variance of the voltage. The dimension of these vectors depends on the number of statistical features they contain.

[0120] The maximum and minimum temperatures refer to the highest and lowest temperatures recorded within a specific time window. This can be achieved by directly selecting the corresponding values ​​from a set of temperature statistical features.

[0121] In step S232, the extracted current and voltage statistical feature vectors, as well as the maximum and minimum values ​​of temperature, are converted into input vectors. The format of the input vectors is consistent with the input format used during the training of the classification model. In step S234, the distance from the sample point (i.e., the input vector, representing the operating state of the distribution ring main unit) to each category hyperplane is calculated and converted into a probability value. The hyperplane for each category is the decision boundary learned by the classification model to distinguish different operating modes. The purpose of calculating the distance is to determine which side of the hyperplane the sample point is located on, thereby determining which category the sample point belongs to, and thus obtaining different operating modes. This facilitates the subsequent selection of the operating mode with the highest probability as the current operating mode.

[0122] Furthermore, the preset mode-corresponding model library in step S3 contains at least corresponding operating modes and data processing models, with each operating mode corresponding to one or more data processing models.

[0123] Step S3 includes:

[0124] S31: Based on the identified current operating mode, select one or more corresponding data processing models from the preset model library corresponding to the mode;

[0125] S32: When the current operating mode corresponds to multiple data processing models, evaluate the electromagnetic interference intensity in the operating environment of the power distribution ring network cabinet, and adjust the selection confidence level of each data processing model according to the electromagnetic interference intensity.

[0126] S33: Obtain the accuracy of each data processing model in the historical data processing process, and adjust the selection weight of each data processing model according to the accuracy.

[0127] S34: Multiply the selection confidence score by the selection weight to obtain the selection probability for each data processing model;

[0128] S35: Select the data processing model with the highest probability as the data processing model corresponding to the current operating mode.

[0129] In this method, after identifying the current operating mode of the distribution ring main unit, the first step is to search for the data processing model associated with that mode from a preset model library. The preset model library stores the mapping relationship between different operating modes and their corresponding data processing models; one operating mode can be associated with one or more data processing models. When an operating mode corresponds to only one data processing model, that model is directly selected.

[0130] When a single operating mode corresponds to multiple data processing models, the system initiates a further evaluation process. This evaluation process considers external environmental factors, specifically assessing the electromagnetic interference (EMI) intensity in the operating environment of the distribution ring main unit. EMI intensity can affect the performance of certain data processing models; for example, it may have a significant impact on models that rely on high-frequency signal analysis. Based on the assessed EMI intensity, the system adjusts the selection confidence level for each alternative model, assigning higher confidence levels to models whose performance is expected to be less affected under the current electromagnetic environment, and lower confidence levels to those whose performance is expected to be more affected.

[0131] Simultaneously, the system also references the actual performance of each candidate model in historical data processing tasks to obtain its historical accuracy. Historical accuracy reflects the model's effectiveness in processing similar data in the past. Based on historical accuracy, the system adjusts the selection weight of each model, with models having higher historical accuracy receiving higher weights. Finally, the selection confidence level adjusted based on environmental factors is multiplied by the selection weight adjusted based on historical performance to calculate the overall selection probability of each candidate model.

[0132] The selection probability comprehensively reflects the model's applicability in the current environment and its historical performance. The system selects the data processing model with the highest selection probability as the data processing model used in the current operating mode. By incorporating environmental factors and historical performance as the basis for model selection, this method can dynamically select a data processing model that is more suitable for the current situation.

[0133] Furthermore, step S4 includes:

[0134] S41: Using the equipment condition assessment model, with current, voltage, and temperature as inputs, assess the health status parameters of the power distribution ring main unit.

[0135] S42: Using an anomaly detection model, with the degree of sudden change and fluctuation frequency of current and voltage as input, detect whether there are any abnormal situations that do not conform to the historical normal operation mode, and obtain the anomaly detection results;

[0136] S43: Calculate the overall health score of the power distribution ring main unit based on the health status parameters, correct the overall health score based on the abnormal detection results, and obtain the equipment status assessment result.

[0137] Step S41 specifies that the specific inputs to the equipment condition assessment model are current, voltage, and temperature data, and clarifies that its output is the health status parameters of the distribution ring main unit. This provides the data types of input and the assessment objectives for equipment condition assessment. Therefore, the health status of key components of the distribution ring main unit, such as insulation, contacts, and heat dissipation, can be quantitatively assessed.

[0138] Step S42 specifies that the specific inputs to the anomaly detection model are the degree of abrupt changes in current and voltage and the frequency of fluctuations, and clarifies its function is to detect abnormal situations and output anomaly detection results. This provides the input data characteristics and detection targets for anomaly detection. Therefore, dynamic abnormal events such as transient faults and abnormal oscillations can be identified.

[0139] Step S43 calculates an overall health score based on the health status parameters obtained in S41, which is a comprehensive assessment. Then, the anomaly detection results obtained in S42 are used to correct this overall health score. This correction mechanism indicates that the anomaly detection results can serve as a calibration or confirmation of the health parameter-based assessment results, thereby obtaining the final equipment condition assessment result. Thus, by combining the long-term degradation trend of the equipment with instantaneous abnormal events, the reliability of the assessment results is improved.

[0140] Furthermore, step S5 includes:

[0141] S51: Based on the collected equipment status assessment results and anomaly detection results, if the anomaly detection results show an anomaly and at least one health status parameter in the equipment status assessment results exceeds a preset threshold, it is judged as a true positive; otherwise, it is a false positive. If the anomaly detection results show no anomaly and all health status parameters in the equipment status assessment results do not exceed the preset threshold, it is judged as a true negative; otherwise, it is a false negative. Based on the number of true positives, false positives, true negatives, and false negatives, the precision, recall, and F1 score are calculated as comprehensive performance indicators of the data processing model.

[0142] S52: Determine whether the overall performance index meets the preset threshold. If not, determine the adjustment direction and magnitude of the data processing model parameters based on precision, recall and F1 score.

[0143] S53: Based on the adjustment direction and magnitude, an adaptive learning rate optimization algorithm is used to adjust the parameters of the data processing model.

[0144] Based on the collected equipment status assessment results and anomaly detection results, each processing result is classified into true positive, false positive, true negative, or false negative through logical judgment. A true positive indicates that the model correctly identified the anomaly, and the equipment's health status does indeed have a problem. A false positive indicates that the model falsely reported an anomaly, even though the equipment's health status is normal. A true negative indicates that the model correctly judged that the equipment has no anomalies and its health status is good. A false negative indicates that the model failed to detect any actual anomalies or health problems.

[0145] Based on the cumulative number of these classifications, precision (number of true positives / number of all predicted anomalies), recall (number of true positives / number of all actual anomalies), and F1 score (harmonic average of precision and recall) are calculated. These metrics quantify the model's predictive accuracy, coverage, and overall performance.

[0146] Furthermore, the calculated precision, recall, and F1 score are compared with preset performance thresholds. If any metric fails to meet the threshold, the model parameters that need adjustment (e.g., the anomaly threshold for an anomaly detection model or the weights of parameters in an equipment condition assessment model) and the direction (increase or decrease) and magnitude of the adjustment are determined based on the specific metric that failed to meet the threshold and its severity. This establishes a direct correlation between model performance and parameter adjustment strategies.

[0147] Furthermore, step S52 includes:

[0148] S521: If the accuracy is lower than the preset threshold, increase the anomaly threshold of the anomaly detection model, and set the increase of the anomaly threshold according to the degree to which the accuracy is lower than the threshold.

[0149] S522: If the recall rate is lower than the preset threshold, the anomaly threshold of the anomaly detection model is reduced, and the reduction range of the anomaly threshold is set according to the degree to which the recall rate is lower than the threshold.

[0150] S523: If the F1 value is lower than the preset threshold, the weights of the corresponding parameters in the equipment condition assessment model are adjusted according to the degree of insulation aging, contact wear, and heat dissipation performance degradation in the abnormal detection results, and the weight adjustment range is determined.

[0151] When the overall performance indicators of the data processing model do not meet the preset requirements, the parameters are adjusted according to the specific performance.

[0152] Specifically, when the accuracy rate is lower than the preset threshold, it indicates that the model is misclassifying normal situations as abnormal in many cases. In this case, increasing the abnormality threshold of the anomaly detection model can improve the strictness of the model's anomaly judgment and thus reduce false positives.

[0153] The magnitude of the increase in the anomaly threshold is related to the degree to which precision deviates from the threshold. When the recall rate is lower than the preset threshold, it indicates that the model has failed to detect a large number of actual anomalies. In this case, reducing the anomaly threshold of the anomaly detection model can reduce the strictness of the model's anomaly judgment, thereby reducing false negatives. The magnitude of the decrease in the anomaly threshold is related to the degree to which recall deviates from the threshold.

[0154] When the F1 score is below a preset threshold, it indicates that the overall performance of the model needs improvement. In this case, based on the specific anomaly types identified in the anomaly detection results, such as insulation aging, contact wear, or decreased heat dissipation, the weights of the parameters related to these anomaly types in the equipment condition assessment model are adjusted. This makes the assessment results more reflective of the impact of these specific anomalies on the equipment condition. The adjustment range of the weights is determined based on the degree of anomaly detected.

[0155] Specifically, let P be the precision rate, P threshold R is the preset precision threshold, and R is the recall rate. threshold F1 is the preset recall threshold. threshold The preset F1 value threshold.

[0156] When P < P threshold When this happens, the model needs to be adjusted to reduce false alarms. The adjustment direction is positive (increasing the threshold), and the adjustment magnitude can be determined based on how much the accuracy falls below the threshold.

[0157] Adjust direction:

[0158] Adjustment range: Amplitude P =α·(P threshold -P), where α is an adjustment coefficient used to control the magnitude of the adjustment, which is usually set based on experience.

[0159] When R < R threshold In such cases, the model needs to be adjusted to reduce false negatives. The adjustment direction is negative (reducing the threshold), and the adjustment magnitude can be determined based on how much the recall rate falls below the threshold.

[0160] Adjust direction:

[0161] Adjustment range: Amplitude R =β·(R) threshold-R), where β is an adjustment coefficient used to control the magnitude of the adjustment, which is usually set based on experience.

[0162] When F1 < F1 threshold In such cases, the model needs to be adjusted to balance precision and recall. The direction and magnitude of the adjustment can be determined based on factors such as the degree to which the F1 value falls below the threshold, the degree of insulation aging, the degree of contact wear, and the degree of degradation in heat dissipation performance.

[0163] Adjust direction:

[0164] Adjustment range: Amplitude F1 =γ·(F1) threshold -F1), where γ is an adjustment coefficient used to control the magnitude of the adjustment, which is usually set based on experience.

[0165] Furthermore, step S53 includes:

[0166] S531: Calculate the adjusted learning rate based on the adjustment direction and adjustment magnitude. The calculation formula is: Adjusted learning rate = Current learning rate + Adjustment direction * Adjustment magnitude;

[0167] S532: Employs the Adam optimization algorithm, utilizing an adjusted learning rate, and combining momentum estimation and adaptive moment estimation to update the parameters of the data processing model, thereby accelerating convergence and improving model performance.

[0168] Specifically, this technical solution provides a method for adjusting data processing model parameters based on performance evaluation results. Step S531 calculates a dynamically adjusted learning rate based on the parameter adjustment direction and magnitude determined in the performance evaluation. This learning rate is adjusted according to the model's current performance. If the model performance is poor and a significant parameter adjustment is required, the learning rate will be adjusted accordingly to allow for a larger parameter update step size; conversely, if only fine-tuning is needed, the learning rate will also be adjusted accordingly. This performance feedback-based learning rate adjustment mechanism allows the parameter update process to better adapt to the model's current state and optimization needs.

[0169] Step S532 employs the Adam optimization algorithm to perform parameter updates. The Adam algorithm is an optimization method that combines momentum and adaptive learning rates. It maintains an independent learning rate for each model parameter and adaptively adjusts these learning rates based on historical gradient information.

[0170] By applying the adjusted learning rate calculated in step S531 to the Adam algorithm, the parameter update process not only benefits from the Adam algorithm's inherent adaptability and stability but also incorporates global learning rate adjustment information based on the overall model performance. This combined approach accelerates the model's convergence towards better performance and improves the final model's performance level, thus solving the problem of how to efficiently and stably update model parameters to enhance performance.

[0171] Please refer to Figure 2 A data processing device for a power distribution ring main unit, applied in the steps of any of the above methods, the device comprising:

[0172] Module 201: Acquires operating data of the power distribution ring network cabinet;

[0173] First analysis module 202: Analyzes operational data and identifies the current operating mode of the power distribution ring network cabinet;

[0174] Selection module 203: Based on the identified current operating mode, select the corresponding data processing model from the preset mode corresponding model library. The data processing models include equipment status assessment model and anomaly detection model.

[0175] Second analysis module 204: Analyzes the operating data based on the equipment status assessment model and the anomaly detection model to obtain the equipment status assessment results and anomaly detection results;

[0176] Adjustment module 205: Determines the performance of the data processing model based on the equipment status assessment results and anomaly detection results, and adjusts the parameters of the data processing model based on the judgment results.

[0177] The acquisition module 201 receives data streams from sensors in the power distribution ring main unit, converts analog signals into digital signals, performs preprocessing and verification to ensure data availability. Furthermore,

[0178] The first analysis module 202 receives the verified digital signal data, extracts key features such as current, voltage, and temperature statistics, and uses a pre-trained classification model to determine the current operating mode of the device.

[0179] Therefore, the selection module 203 searches and selects from a model library containing various data processing models based on the operating mode information output by the first analysis module 202. Specifically, when one operating mode corresponds to multiple models, the selection module 203 can further consider environmental factors and historical performance to determine the final model to be used.

[0180] The second analysis module 204 receives the data processing model determined by the selection module 203 and the operating data provided by the acquisition module 201, executes specific evaluation and detection algorithms, and outputs the health status score of the device and a judgment on whether there is an anomaly.

[0181] The adjustment module 205 receives the output results of the second analysis module 204, calculates the performance indicators of the model by comparing the evaluation and detection results with the actual situation or preset standards, and calculates the adjustment direction and magnitude of the model parameters based on these indicators, such as adjusting the anomaly detection threshold or model weights.

[0182] Specifically, this device automates the processing of operational data from power distribution ring main units through inter-module collaboration. The acquisition module 201 first collects and processes raw sensor data, transforming it into standardized and reliable operational data. This data is then sent to the first analysis module 202, which accurately identifies the current operating mode of the equipment, such as normal load, light load, heavy load, or a specific operating mode, using feature extraction and classification algorithms. The identified operating mode information is passed to the selection module 203, which selects the most suitable equipment status assessment model and anomaly detection model from a model library based on preset rules or strategies. The selected model is then invoked by the second analysis module 204 to perform in-depth analysis of the operational data, assessing the equipment's health status parameters and detecting any anomalies. Finally, the adjustment module 205 receives the assessment and detection results and evaluates the performance of the current model, such as calculating precision and recall. If the performance of the data processing model does not meet the requirements, the adjustment module 205 will calculate the adjustment amount of the model parameters (such as thresholds and weights) based on the deviation of the performance indicators, and update the model parameters using optimization algorithms, thereby forming a closed loop of data processing, result feedback and model optimization. This enables the data processing model to continuously adapt to changes in equipment operating status and the influence of environmental factors, improve the accuracy of assessment and detection, reduce false alarms and false negatives, and solve the problem that existing methods are difficult to adapt to the evolution of equipment life cycle and the switching of operating modes.

[0183] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0184] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A data processing method for a power distribution ring main unit, characterized in that, The method includes: S1: Obtain the operating data of the power distribution ring network cabinet; S2: Analyze the operating data to identify the current operating mode of the power distribution ring network cabinet; S3: Based on the identified current operating mode, select the corresponding data processing model from the preset mode corresponding model library. The data processing model includes an equipment status assessment model and an anomaly detection model. The preset mode-corresponding model library in step S3 contains at least the corresponding operating mode and data processing model, and each operating mode corresponds to one or more data processing models. Step S3 includes: S31: Selecting one or more of the data processing models from a preset model library based on the identified current operating mode; S32: When the current operating mode corresponds to multiple data processing models, evaluate the electromagnetic interference intensity in the operating environment of the power distribution ring network cabinet, and adjust the selection confidence level of each data processing model according to the electromagnetic interference intensity. S33: Obtain the accuracy of each of the data processing models in the historical data processing process, and adjust the selection weight corresponding to each of the data processing models according to the accuracy; S34: Multiply the selection confidence level by the selection weight to obtain the selection probability of each data processing model; S35: Select the data processing model with the highest selection probability as the data processing model corresponding to the current operating mode; S4: Analyze the operating data according to the equipment status assessment model and the anomaly detection model to obtain the equipment status assessment results and anomaly detection results; S5: Determine the performance of the data processing model based on the equipment status assessment results and the anomaly detection results, obtain the determination results, and adjust the parameters of the data processing model based on the determination results; Step S5 includes: S51: Based on the collected equipment status assessment results and anomaly detection results, if the anomaly detection results show an anomaly, and at least one health status parameter in the equipment status assessment results exceeds a preset threshold, it is determined to be a true positive; otherwise, it is a false positive. If the anomaly detection results show no anomaly, and all health status parameters in the equipment status assessment results are within acceptable limits, it is considered a true positive. If the value exceeds a preset threshold, it is judged as a true negative; otherwise, it is a false negative. Based on the number of true positives, false positives, true negatives, and false negatives, the precision, recall, and F1 score are calculated as comprehensive performance indicators of the data processing model. S52: Determine whether the comprehensive performance index meets the preset threshold. If not, determine the adjustment direction and magnitude of the parameters of the data processing model based on the precision, recall and F1 score. Step S52 includes: S521: If the accuracy is lower than a preset threshold, the anomaly threshold of the anomaly detection model is increased, and the increase in the anomaly threshold is set according to the degree to which the accuracy is lower than the threshold. S522: If the recall rate is lower than a preset threshold, the anomaly threshold of the anomaly detection model is reduced, and the reduction range of the anomaly threshold is set according to the degree to which the recall rate is lower than the threshold. S523: If the F1 value is lower than the preset threshold, the weights of the corresponding parameters in the equipment condition assessment model are adjusted according to the degree of insulation aging, the degree of contact wear, and the degree of heat dissipation performance degradation in the abnormal detection results, and the weight adjustment range is determined. S53: Based on the adjustment direction and magnitude, the parameters of the data processing model are adjusted using an adaptive learning rate optimization algorithm.

2. The data processing method for a power distribution ring main unit according to claim 1, characterized in that, In step S1, the power distribution ring network cabinet is equipped with at least a main circuit and an auxiliary circuit. The main circuit and the auxiliary circuit are respectively equipped with multiple sensors for collecting analog signal data including current, voltage, temperature and switch status. Step S1 includes: S11: Convert the analog signal data into digital signal data, and preprocess the digital signal data to unify the operating data of different dimensions into the [0,1] interval; S12: Perform data verification on the preprocessed digital signal data; S13: For the digital signal data that fails verification, initiate an automatic retransmission request until the digital signal data is successfully verified or the maximum number of retransmissions is reached; if the verification still fails after reaching the maximum number of retransmissions, mark the digital signal data as unreliable data, record the error type and occurrence time, and use the successfully verified digital signal data as the running data.

3. The data processing method for a power distribution ring main unit according to claim 1, characterized in that, Step S2 includes: S21: Extract the current, voltage, and temperature from the operating data to construct a multidimensional time series; S22: Using the sliding window method, the multidimensional time series is divided into multiple time windows, and for each time window, the statistical characteristics of current, voltage, and temperature are calculated; S23: Input the statistical features into a pre-trained classification model, which is built based on the support vector machine algorithm, to identify the current operating mode of the power distribution ring network cabinet and output the probability distribution of the operating mode; S24: Based on the probability distribution, select the operating mode with the highest probability as the current operating mode of the power distribution ring network cabinet.

4. The data processing method for a power distribution ring main unit according to claim 3, characterized in that, Step S23 includes: S231: Extract the statistical feature vectors of current and voltage, as well as the maximum and minimum values ​​of temperature, from the statistical features; S232: Input the extracted current and voltage statistical feature vectors and the maximum and minimum values ​​of temperature into the classification model; S233: Using the classification model, with the statistical feature vector and the maximum and minimum values ​​of temperature as inputs, calculate the probability distribution characterizing the current operating mode of the power distribution ring main unit.

5. The data processing method for a power distribution ring main unit according to claim 1, characterized in that, Step S4 includes: S41: Using the equipment status assessment model, with current, voltage, and temperature as inputs, assess the health status parameters of the power distribution ring main unit; S42: Using the aforementioned anomaly detection model, with the degree of sudden change and fluctuation frequency of current and voltage as input, detect whether there are any abnormal situations that do not conform to the historical normal operation mode, and obtain the anomaly detection result; S43: Calculate the overall health score of the power distribution ring network cabinet based on the health status parameters, correct the overall health score based on the anomaly detection results, and obtain the equipment status assessment result.

6. The data processing method for a power distribution ring main unit according to claim 1, characterized in that, Step S53 includes: S531: Calculate the adjusted learning rate based on the adjustment direction and the adjustment magnitude. The calculation formula is: Adjusted learning rate = Current learning rate + Adjustment direction * Adjustment magnitude; S532: The Adam optimization algorithm is used, with an adjusted learning rate, and momentum estimation and adaptive moment estimation are combined to update the parameters of the data processing model to accelerate convergence and improve model performance.

7. A data processing device for a power distribution ring main unit, characterized in that, When applied to the steps of the method according to any one of claims 1-6, the apparatus comprises: Acquisition module: Acquires the operating data of the power distribution ring network cabinet; First analysis module: Analyzes the operating data and identifies the current operating mode of the power distribution ring network cabinet; Selection module: Based on the identified current operating mode, select the corresponding data processing model from the preset mode corresponding model library. The data processing model includes the equipment status assessment model and the anomaly detection model. The second analysis module analyzes the operating data based on the equipment status assessment model and the anomaly detection model to obtain the equipment status assessment results and anomaly detection results. Adjustment module: Based on the equipment status assessment results and the anomaly detection results, the module determines the performance of the data processing model, obtains the determination result, and adjusts the parameters of the data processing model based on the determination result.

Images