A concentrator electric data anomaly diagnosis method based on DBSCAN+XGBoost model

Abstract

The application relates to the technical field of urban lighting, and discloses a concentrator electric data anomaly diagnosis method based on a DBSCAN+XGBoost model, which comprises the following operation steps: data acquisition, real-time consumption and storage, data segmentation, data standardization, principal component analysis and dimension reduction, implementation of a DBSCAN density clustering method, and implementation of an XGboost method; the method further comprises training and optimizing the XGBoost model after setting the XGBoost parameters, using a test set to evaluate the performance of the model, subsequently using the trained XGBoost model to predict whether each sample belongs to an abnormal situation, and analyzing and identifying the cause of the energy consumption anomaly. Compared with the prior art, the application has the advantages that the electric data anomaly diagnosis in the field of urban landscape lighting and urban road lighting is more accurate and efficient, and a series of problems, such as poor stability and reliability of the urban lighting system in the prior art, are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of urban lighting, in particular to a concentrator electric data anomaly diagnosis method based on DBSCAN+XGBoost model. BACKGROUND

[0002] DBSCAN is a density-based clustering algorithm, and density-based clustering is to find high-density areas separated by low-density areas, so first discuss the definition of density, the density of a particular point in the data set can be estimated by counting the points within the Eps radius of the particular point (including itself), based on this measure, in DBSCAN, points are divided into three categories: points inside the dense area (core points), points on the edge of the dense area (boundary points) and points in the sparse area (noise or background points).

[0003] XGBoost, which is an efficient gradient boosting decision tree algorithm, has improved the original GBDT, making the model effect greatly improved, as a forward additive model, its core is to use the integrated idea-Boosting idea, to integrate multiple weak learners into a strong learner through a certain method, that is, to use multiple trees to make decisions together, and the result of each tree is the difference between the target value and the prediction result of all previous trees, and all the results are added to get the final result, so as to improve the effect of the whole model.

[0004] In the prior art, the Chinese patent document with publication number CN111783818A proposes a precise marketing method based on Xgboost and DBSCAN, the precise marketing method based on Xgboost and DBSCAN classifies customers through Xgboost algorithm, classifies customers into target customers and non-target customers, and recommends products to customers through DBSCAN algorithm, recommends corresponding financial products to target customers, reduces labor cost, improves marketing efficiency and marketing success rate, therefore, using Xgboost and DBSCAN algorithm technology has gradually become an efficient method to solve life problems.

[0005] As an important part of urban infrastructure, urban lighting not only provides convenience for residents' night activities, but also enhances the aesthetics and safety of the city. The normal operation of the urban lighting system is crucial to maintaining the normal order of the city and the quality of life of residents. However, with the continuous development of the city and the increasing demand for electricity, the urban lighting system has become more and more complex and is vulnerable to various power problems, such as current overload, voltage fluctuation, etc. In order to ensure the stable operation of the urban lighting system, the monitoring and anomaly diagnosis of power data are particularly important. The traditional monitoring method mainly relies on manual inspection, which is time-consuming and inefficient and difficult to meet the real-time requirements. In recent years, with the rapid development of data science and artificial intelligence technology, using advanced technologies such as DBSCAN and XGBoost to process power data of urban lighting systems will gradually become an innovative and efficient means. Therefore, the application discloses a concentrator electric data anomaly diagnosis method based on a DBSCAN+XGBoost model. SUMMARY

[0006] (I) Technical problems solved

[0007] In view of the deficiencies in the prior art, the application provides a concentrator electric data anomaly diagnosis method based on a DBSCAN+XGBoost model, which has the advantages of making the electric data anomaly diagnosis in the field of urban landscape lighting and urban road lighting more accurate and efficient, and solving the problems of poor stability and reliability of the urban lighting system in the prior art.

[0008] (II) Technical solutions

[0009] To achieve the above-mentioned purpose, the application provides the following technical solutions: a concentrator electric data anomaly diagnosis method based on a DBSCAN+XGBoost model, comprising the following operation steps:

[0010] S1, data acquisition: through the 4G module installed on the centralized controller device, the electric data acquisition work is realized, and the acquired electric data is transmitted to the MQTT message transmission protocol, and then stored in the message queue according to the subscription message topic mode;

[0011] S2, real-time consumption and storage: introducing Spark Structured Streaming technology, real-time consumption of data obtained from Kafka data stream and generation of historical data;

[0012] S3, data segmentation: dividing the data into multiple different subsets, and selecting appropriate models or model parameters for each subset, the subsets including A-phase data subset, B-phase data subset and C-phase data subset;

[0013] S4, data standardization: after feature extraction of the data, the data is standardized to ensure the scale consistency between different features, which can make the values of different features distributed in similar ranges;

[0014] S5, principal component analysis and dimension reduction: the most representative information is extracted from the multi-dimensional standardized data after S4 processing, the dimension is reduced and sufficient data variability is retained;

[0015] S6, implementation of DBSCAN density clustering method: after selecting appropriate DBSCAN parameters, based on the selected parameters, using HDBSCAN algorithm to calculate the density clustering of the data after principal component analysis and dimension reduction, obtaining the calculation result, according to the calculation result, the samples are divided into different clusters, each cluster represents a group of samples with similar characteristics, analyzing the characteristics, size and distribution of each cluster to obtain insights into energy consumption anomalies;

[0016] S7, implementation of XGboost method: after setting XGBoost parameters, training and tuning XGBoost model, and using test set to evaluate the performance of the model, then using the trained XGBoost model to predict whether each sample belongs to abnormal situation, analyzing and identifying the causes of energy consumption anomalies.

[0017] Preferably, the historical data includes time series data and timing strategy data, which covers key parameters such as A / B / C phase current, A / B / C phase voltage, A / B / C phase active power.

[0018] Preferably, in the S1 data collection stage, the power consumption data is converted into the following vector form X:

[0019] Formula: X =

[0020] [[T1, VA1, VB1, VC1, VAB1, VBC1, VCA1, IA1, IB1, IC1, PA1, PB1, PC1],

[0021] …, …,

[0022] [Tm, VAm, VBm, VCm, VAB1, VBC1, VCA1, IAm, IBm, ICm, PAm, PBm, PCm]].

[0023] Preferably, the first column parameter T1…Tm represents the time stamp reported by the central controller device at 1…m time points;

[0024] The parameters VA1…VAm, VB1…VBm, VC1…VCm in columns 2-4 represent the A-phase instantaneous voltage, B-phase instantaneous voltage, and C-phase instantaneous voltage reported by the centralized controller device every 5 min within the T1-Tm time range;

[0025] The parameters VAB1…VABm, VBC1…VBCm, VCA1…VCAm in columns 5-7 represent the AB-phase instantaneous voltage, BC-phase instantaneous voltage, and CA-phase instantaneous voltage reported by the centralized controller device every 5 min within the T1-Tm time range.

[0026] The parameters IA1…IAm, IB1…IBm, IC1…ICm in columns 8-10 represent the A-phase instantaneous current, B-phase instantaneous current, and C-phase instantaneous current reported by the centralized controller device every 5 min within the T1-Tm time range.

[0027] Preferably, the parameters PA1…PAm, PB1…PBm, PC1…PCm in columns 11-13 represent the A-phase instantaneous power, B-phase instantaneous power, and C-phase instantaneous power reported by the centralized controller device every 5 min within the T1-Tm time range.

[0028] Preferably, the A-phase data subset includes the timestamp (T1-Tm), A-phase instantaneous voltage (VA1-VAm), AB-phase instantaneous voltage (VAB1-VABm), A-phase instantaneous current (IA1-IAm), and A-phase instantaneous power (PA1-PAm).

[0029] Preferably, the B-phase data subset includes the timestamp (T1-Tm), B-phase instantaneous voltage (VB1-VBm), BC-phase instantaneous voltage (VBC1-VBCm), B-phase instantaneous current (IB1-IBm), and B-phase instantaneous power (PB1-PBm).

[0030] Preferably, the C-phase data subset includes the timestamp (T1-Tm), C-phase instantaneous voltage (VC1-VCm), CA-phase instantaneous voltage (VCA1-VCAm), C-phase instantaneous current (IC1-ICm), and C-phase instantaneous power (PC1-PCm).

[0031] Preferably, data standardization is performed in S4 using the StandardScaler method, specifically using the following formula:

[0032] z = x - μ / σ

[0033] where x represents the original feature value, μ represents the mean of the feature, σ represents the standard deviation of the feature, and z represents the standardized feature value.

[0034] Preferably, in the S6, the HDBSCAN algorithm automatically identifies core samples, edge samples and noise samples to determine the cluster assignment of each sample.

[0035] (III) Beneficial Effects

[0036] Compared with the prior art, the present application provides a concentrator electric data anomaly diagnosis method based on a DBSCAN+XGBoost model, which has the following beneficial effects: by combining the high-density clustering-based spatial noise (HDBSCAN) algorithm with machine learning technology, the normal and abnormal data in the historical power data can be more accurately distinguished and labeled, in addition, as a density clustering method, the HDBSCAN algorithm can effectively identify different clustering clusters in the data, thereby distinguishing abnormal patterns from normal patterns, and as a gradient boosting decision tree algorithm, XGBoost can learn complex patterns from known data and make fast and accurate predictions when new data arrives, thus, this innovative technology based on the HDBSCAN algorithm and the XGBoost model provides an advanced method for power data anomaly diagnosis of urban lighting systems, which can improve the stability, reliability and maintenance efficiency of urban lighting systems by predicting and accurately identifying abnormal conditions in power data in real time, thereby providing more comfortable and safe night environments for urban residents and optimizing urban power supply efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 The present application is a schematic diagram of the operation principle and operation steps. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0039] As introduced in the background, there are deficiencies in the prior art. In order to solve the above technical problems, the present application proposes a concentrator electric data anomaly diagnosis method based on a DBSCAN+XGBoost model.

[0040] In a typical embodiment of the present application, as shown in Figure 1 A concentrator electric data anomaly diagnosis method based on a DBSCAN+XGBoost model includes the following operation steps:

[0041] S1, data collection: through the 4G module installed on the centralized controller device, the power consumption data collection work is realized, these collected power consumption data are transmitted to the MQTT message transmission protocol, and then stored in the message queue according to the subscription message theme, this step ensures the efficient caching and transmission of power consumption data;

[0042] S2, real-time consumption and storage: introduce Spark Structured Streaming technology, real-time consumption of data obtained from Kafka data stream and generate historical data, these data are efficiently stored in Hadoop distributed file system (HDFS) to ensure the real-time and reliability of data;

[0043] S3, data segmentation: divide the data into multiple different subsets, and select appropriate models or model parameters for each subset, the subsets include A-phase data subset, B-phase data subset and C-phase data subset, the advantage of this is to solve the problem that the generalization ability of the model is limited due to the inconsistent devices under A, B and C three-phase lines, which helps to improve the prediction performance of the model under different lines;

[0044] S4, data standardization: after feature extraction, the data is standardized to ensure the scale consistency between different features, and the standardization can make the values of different features distributed in similar ranges;

[0045] S5, principal component analysis and dimension reduction: extract the most representative information from the standardized data after S4 processing, reduce the dimension and retain enough data variability;

[0046] S6, implementation of DBSCAN density clustering method: after selecting appropriate DBSCAN parameters, based on the selected parameters, using HDBSCAN algorithm to calculate the density clustering of the principal component analysis and dimension reduction data, get the calculation result, according to the calculation result, the samples are divided into different clusters, each cluster represents a group of samples with similar characteristics, analyze the characteristics, size and distribution of each cluster to obtain insights about energy consumption anomalies;

[0047] S7, Implementation of XGboost method: after setting the XGBoost parameters, the XGBoost model is trained and optimized, and the performance of the model is evaluated using the test set, and then the trained XGBoost model is used to predict whether each sample belongs to an abnormal situation, analyze and identify the reason for the energy consumption anomaly, the current, voltage and power parameters of the historical data are processed by using the DBSCAN (density-based spatial clustering of applications with noise) algorithm, since the core of the algorithm is density clustering, which can effectively separate normal and abnormal data, thereby providing valuable basis for subsequent analysis, thereby making the electric data anomaly diagnosis in the field of urban landscape lighting and urban road lighting more accurate and efficient.

[0048] As a preferred embodiment in the present embodiment, the historical data includes time series data and timing strategy data, the time series data covers key parameters such as A / B / C phase current, A / B / C phase voltage, A / B / C phase active power, in the S1 data collection stage, the power consumption data is converted into the following vector form X:

[0049] Formula: X =

[0050] [[T1, VA1, VB1, VC1, VAB1, VBC1, VCA1, IA1, IB1, IC1, PA1, PB1, PC1],

[0051] …, …,

[0052] [Tm, VAm, VBm, VCm, VAB1, VBC1, VCA1, IAm, IBm, ICm, PAm, PBm, PCm]].

[0053] As a preferred embodiment in the present embodiment, the first column parameters T1…Tm represent the time stamps reported by the centralized controller device at 1…m time points;

[0054] The second to fourth column parameters VA1…VAm, VB1…VBm, and VC1…VCm represent A-phase instantaneous voltage, B-phase instantaneous voltage, and C-phase instantaneous voltage reported by the centralized controller device every 5 min within the T1-Tm time range;

[0055] The fifth to seventh column parameters VAB1…VABm, VBC1…VBCm, and VCA1…VCAm represent AB-phase instantaneous voltage, BC-phase instantaneous voltage, and CA-phase instantaneous voltage reported by the centralized controller device every 5 min within the T1-Tm time range;

[0056] The parameters IA1...IAm, IB1...IBm, IC1...ICm in columns 8-10 represent the A-phase instantaneous current, B-phase instantaneous current, C-phase instantaneous current reported by the centralized controller device every 5 min in the T1-Tm time range;

[0057] The parameters PA1...PAm, PB1...PBm, PC1...PCm in columns 11-13 represent the A-phase instantaneous power, B-phase instantaneous power, C-phase instantaneous power reported by the centralized controller device every 5 min in the T1-Tm time range;

[0058] The A-phase data subset includes the time stamp (T1-Tm), A-phase instantaneous voltage (VA1-VAm), AB-phase instantaneous voltage (VAB1-VABm), A-phase instantaneous current (IA1-IAm), A-phase instantaneous power (PA1-PAm), the B-phase data subset includes the time stamp (T1-Tm), B-phase instantaneous voltage (VB1-VBm), BC-phase instantaneous voltage (VBC1-VBCm), B-phase instantaneous current (IB1-IBm), B-phase instantaneous power (PB1-PBm), and the C-phase data subset includes the time stamp (T1-Tm), C-phase instantaneous voltage (VC1-VCm), CA-phase instantaneous voltage (VCA1-VCAm), C-phase instantaneous current (IC1-ICm), C-phase instantaneous power (PC1-PCm), in addition, the data vector of each subset is as follows:

[0059] A-phase data subset:

[0060] [[T1, VA1, VAB1, IA1, PA1, 'A'],

[0061] [T2, VA2, VAB2, IA2, PA2, 'A'], ...

[0063] [Tm, VAm, VABm, IAm, PAm, 'A']]

[0064] B-phase data subset:

[0065] [[T1, VB1, VBC1, IB1, PB1, 'B'],

[0066] [T2, VB2, VBC2, IB2, PB2, 'B'], ...

[0068] [Tm, VBm, VBCm, IBm, PBm, 'B']]

[0069] C-phase data subset:

[0070] [[T1, VC1, VCA1, IC1, PC1, 'C'],

[0071] [T2,VC2,VCA2,IC2,PC2,'C'], ...

[0073] [Tm,VCm,VCAm,ICm,PCm,'C']]

[0074] In using these subset data, we can choose different models or model parameters according to the characteristics and needs of each subset to optimize the generalization ability and prediction performance of the model.

[0075] As a preferred embodiment in this embodiment, the StandardScaler method is used for data standardization in S4, which helps to improve the stability and performance of the model. The specific formula is as follows:

[0076] z = x - μ / σ

[0077] Where x represents the original feature value, μ represents the mean of the feature, σ represents the standard deviation of the feature, and z represents the standardized feature value.

[0078] The following is a more detailed description of the principal component analysis and dimensionality reduction process:

[0079] In the principal component analysis (PCA) stage, we use the standardized data, i.e. the feature vector X after data standardization, and the goal of principal component analysis is to project the original data onto a new orthogonal feature space, where each new feature (principal component) is a linear combination of the original features, and these principal components are sorted in descending order of the degree of explaining data variability, thus retaining the key information of the data. Assuming there are m samples and n features, the feature matrix is X (dimension m x n), where each row represents a sample and each column represents a feature. The steps of principal component analysis are as follows:

[0080] First, data standardization: calculate the mean μj and standard deviation σj of each feature, and then standardize the features using the following formula:

[0081]

[0082] Second, calculate the covariance matrix: the elements cij of the covariance matrix C are calculated as follows:

[0083]

[0084] Where ˉxi and xˉj are the means of the i-th column and the j-th column, respectively.

[0085] Third, eigenvalue decomposition: calculate the eigenvalues λ1, λ2,..., λn of the covariance matrix and the corresponding eigenvectors v1, v2,..., vn.

[0086] Fourth step, selecting principal components: arrange the eigenvectors in descending order according to the size of the eigenvalue, and select the first n eigenvectors whose cumulative contribution rate reaches a predetermined threshold, wherein the cumulative contribution rate is defined as:

[0087]

[0088] By selecting the principal components whose cumulative contribution rate exceeds 85%, we ensure that sufficient data variability is retained during dimension reduction, and help to alleviate the dimension disaster problem, improve the efficiency and accuracy of subsequent data analysis and model establishment;

[0089] Fifth step, data dimension reduction: multiply the standardized eigenvectors X by the selected principal component eigenvectors to obtain the dimension-reduced data;

[0090] In S6, the HDBSCAN algorithm automatically identifies core samples, edge samples and noise samples to determine the cluster assignment of each sample. The following describes the implementation steps of the hierarchical clustering-based density peak scanning algorithm (DBSCAN) and how to apply it to concentrator energy consumption anomaly reason diagnosis. The steps of the method are as follows:

[0091] DBSCAN parameter setting

[0092] Selecting appropriate DBSCAN parameters can ensure accurate identification of clusters and identification of anomalies in the data during density clustering. These parameters include the minimum number of samples and the distance threshold. The present application uses an exhaustive method to calculate the DBSCAN parameters, wherein is the neighborhood radius (0.1, 20), and the minimum sample number range defined by the density is (3, 16). The best parameter combination is obtained by loop calculation;

[0093] Density clustering calculation

[0094] Based on the selected parameters, we use the HDBSCAN algorithm to perform density clustering calculation on the principal component analysis dimension-reduced data. Since HDBSCAN involves complex algorithm formulas, we mainly describe the implementation process:

[0095] The flow of DBSCAN algorithm is as follows:

[0096] Input: sample set D = (x 1,x 2,...,x m), neighborhood parameters (∈,MinPts), sample distance measurement method

[0097] Output: cluster division C.

[0098] First step: initialization

[0099] Initialize the core point set Ω = ∈,

[0100] Initialize the cluster number k = 0,

[0101] Initialize the unvisited sample set Γ = D,

[0102] Cluster division

[0103] Second step: Calculate the core point set Ω = ∈

[0104] for j = 1, 2,... m, do;

[0105] Find the ∈ neighborhood sub-sample set N ∈ (xi) of sample xi by distance measurement

[0106] if N ∈ (xi) ≥ MinPts then

[0107] Add sample xi to the core object sample set: Ω = Ω ∪ x

[0108] end if

[0109] end for

[0110] Third step: Clustering

[0111] while Ω ≠ ∈ do:

[0112] Randomly select a core point o in the core point set Ω,

[0113] Initialize the current cluster core object queue Ωcur = o,

[0114] Initialize the class serial number k = k + 1,

[0115] Initialize the current cluster sample set Ck = o,

[0116] Update the unvisited sample set Γ = Γ - o

[0117] while Ωcur ≠ ∈ do:

[0118] Take the first sample q of Ωcur

[0119] Calculate the ∈-neighborhood N ∈ (q) of q by neighborhood distance threshold ∈

[0120] if |N ∈ (q)| ≥ MinPts then

[0121] Let

[0122] Add to the queue Ωcur

[0123] Add

[0124]

[0125] end if

[0126] endwhile

[0127] Step 4: Output C = C1, C2,..., Ck.

[0128] Implementation of XGBoost Method

[0129] In the following section, we will describe the implementation steps of the XGBoost method in detail, as well as how to apply it to the concentrator energy consumption anomaly reason diagnosis:

[0130] XGBoost Parameter Settings

[0131] Selecting appropriate XGBoost parameters is crucial to ensure the performance of the model, these parameters include learning rate, number of trees, and tree depth;

[0132] XGBoost Model Training and Tuning

[0133] After data preparation and feature engineering, we can start training the XGBoost model, this stage includes the following steps:

[0134] Dataset division: divide the dataset into training set and test set, so as to use the training set to train the model, and use the test set to evaluate the model performance.

[0135] Model training: use the training set to train the XGBoost model, in the training process, the model will adjust the weight according to the training data step by step, in order to minimize the loss function;

[0136] Parameter tuning: use techniques such as cross-validation to adjust the model parameters to obtain better performance, parameter tuning can help prevent overfitting and improve the generalization ability of the model;

[0137] Feature importance analysis: during the training process, XGBoost can calculate the importance score of the features, by analyzing the importance of the features, we can understand which features have more influence on the prediction performance of the model, so as to optimize the feature engineering strategy;

[0138] Model evaluation and application

[0139] After the model training is completed, we will use the test set to evaluate the performance of the model, common evaluation indicators include accuracy, precision, recall, F1 score, by evaluating the performance of the model, we can understand the prediction performance of the model on unseen data, and then in the concentrator energy consumption anomaly reason diagnosis, we can use the trained XGBoost model to predict whether each sample belongs to an abnormal situation, and then analyze and identify the reason for the energy consumption anomaly.

[0140] While embodiments of the present application have been shown and described with reference to particular embodiments thereof, it will be understood by those skilled in the art that various changes in form and details can be made therein without departing from the spirit and scope of the application. The scope of the application is defined by the appended claims and their equivalents.

Claims

1. A method for diagnosing concentrator electrical data anomalies based on the DBSCAN+XGBoost model, characterized in that, The following steps are included: S1. Data Acquisition: Electricity consumption data is collected through a 4G module installed on the centralized controller device. The collected electricity consumption data is transmitted to the MQTT message transmission protocol and then stored in a message queue according to the subscription message topic. S2, Real-time Consumption and Storage: Introducing Spark Structured Streaming technology to consume data obtained from Kafka data streams in real time and generate historical data; S3. Data Segmentation: Divide the data into multiple different subsets and select an appropriate model or model parameters for each subset. The subsets include the A-phase data subset, the B-phase data subset, and the C-phase data subset. S4. Data Standardization: After feature extraction, the data is standardized to ensure scale consistency between different features. The standardization enables the values ​​of different features to be distributed within a similar range. S5. Principal Component Analysis and Dimensionality Reduction: Extract the most representative information from the multidimensional standardized data after processing in S4, reduce dimensionality and retain sufficient data variability; S6. Implementation of DBSCAN density clustering method: After selecting appropriate DBSCAN parameters, the HDBSCAN algorithm is used to perform density clustering calculation on the data after principal component analysis dimensionality reduction based on the selected parameters. The calculation results are obtained, and the samples are divided into different clusters according to the calculation results. Each cluster represents a group of samples with similar characteristics. The characteristics, size and distribution of each cluster are analyzed to obtain insights about energy consumption anomalies. S7. Implementation of the XGBoost method: After setting the XGBoost parameters, train and fine-tune the XGBoost model, and use the test set to evaluate the model's performance. Then, use the trained XGBoost model to predict whether each sample belongs to an abnormal situation, and analyze and identify the causes of energy consumption anomalies.

2. The method for concentrator electrical data anomaly diagnosis based on the DBSCAN+XGBoost model according to claim 1, characterized in that: The historical data includes time-series data and timing strategy data. The time-series data covers key parameters such as A / B / C phase current, A / B / C phase voltage, and A / B / C phase active power.

3. The method for concentrator electrical data anomaly diagnosis based on the DBSCAN+XGBoost model according to claim 1, characterized in that: During the S1 data acquisition phase, the electricity consumption data is converted into the following vector form X: Formula: X = [[T1,VA1,VB1,VC1,VAB1,VBC1,VCA1,IA1,IB1,IC1,PA1,PB1,PC1], …，…， [Tm,VAm,VBm,VCm,VAB1,VBC1,VCA1,IAm,IBm,ICm,PAm,PBm,PCm]].

4. The method for concentrator electrical data anomaly diagnosis based on the DBSCAN+XGBoost model according to claim 3, characterized in that: The parameters in column 1, T1…Tm, represent the timestamps reported by the centralized controller device at times 1…m. The parameters VA1…VAm, VB1…VBm, and VC1…VCm in columns 2 to 4 represent the instantaneous voltage of phase A, phase B, and phase C reported by the centralized controller device every 5 minutes within the time range of T1 to Tm. The parameters VAB1…VABm, VBC1…VBCm, and VCA1…VCAm in columns 5 to 7 represent the instantaneous voltage between phases AB, BC, and CA reported by the centralized controller every 5 minutes within the time range of T1 to Tm. The parameters IA1…IAm, IB1…IBm, and IC1…ICm in columns 8 to 10 represent the instantaneous currents of phase A, phase B, and phase C reported by the centralized controller device every 5 minutes within the time range of T1 to Tm.

5. The method for concentrator electrical data anomaly diagnosis based on the DBSCAN+XGBoost model according to claim 3, characterized in that: The parameters in columns 11 to 13, PA1…PAm, PB1…PBm, and PC1…PCm, represent the instantaneous power of phase A, phase B, and phase C reported by the centralized controller device every 5 minutes within the time range of T1 to Tm.

6. The method for concentrator electrical data anomaly diagnosis based on the DBSCAN+XGBoost model according to claim 3, characterized in that: The A-phase data subset includes timestamps (T1~Tm), A-phase instantaneous voltage (VA1~VAm), AB-phase instantaneous voltage (VAB1~VABm), A-phase instantaneous current (IA1~IAm), and A-phase instantaneous power (PA1~PAm).

7. The method for concentrator electrical data anomaly diagnosis based on the DBSCAN+XGBoost model according to claim 3, characterized in that: The B-phase data subset includes timestamps (T1~Tm), B-phase instantaneous voltages (VB1~VBm), BC-phase instantaneous voltages (VBC1~VBCm), B-phase instantaneous currents (IB1~IBm), and B-phase instantaneous power (PB1~PBm).

8. The method for concentrator electrical data anomaly diagnosis based on the DBSCAN+XGBoost model according to claim 3, characterized in that: The C-phase data subset includes timestamps (T1~Tm), C-phase instantaneous voltages (VC1~VCm), CA phase-to-phase instantaneous voltages (VCA1~VCAm), C-phase instantaneous currents (IC1~ICm), and C-phase instantaneous power (PC1~PCm).

9. The method for concentrator electrical data anomaly diagnosis based on the DBSCAN+XGBoost model according to claim 1, characterized in that: In step S4, the StandardScaler method is used to standardize the data, specifically using the following formula: z = x - μ / σ Where x represents the original feature value, μ represents the mean of the feature, σ represents the standard deviation of the feature, and z represents the standardized feature value.

10. The method for concentrator electrical data anomaly diagnosis based on the DBSCAN+XGBoost model according to claim 9, characterized in that: In S6, the HDBSCAN algorithm automatically identifies core samples, edge samples, and noise samples to determine the cluster assignment for each sample.

Images