Station area phase line loss detection method, device and equipment based on clustering analysis

Abstract

The invention relates to the field of intelligent scheduling, and discloses a transformer area phase line loss detection method, device and equipment based on clustering analysis, and the method comprises the steps: obtaining transformer area training data, carrying out the preprocessing of the transformer area training data, carrying out the dimension reduction of the preprocessed transformer area training data, and carrying out the detection of the transformer area phase line loss. And carrying out clustering on the transformer area training data after dimension reduction based on a frost ice optimized density clustering RIME-DBSCAN algorithm so as to identify different types of transformer areas, carrying out line loss rate evaluation on the transformer area in each cluster based on a Stacking ensemble learning model, calculating the line loss rate of each phase, and comparing the line loss rate of each phase with a set threshold value so as to identify a phase and a section with abnormal line loss rates. Through the optimized data dimension reduction method and the improved RIME-DBSCAN clustering algorithm, accurate detection and management of the line loss rate of the transformer area are achieved, the data processing efficiency and accuracy can be improved, and the method can better adapt to the operation characteristics of different transformer areas.

Description

Technical Field

[0001] The present invention relates to the field of intelligent dispatching, and in particular to a method, device and equipment for detecting phase line loss in a substation based on cluster analysis. Background Art

[0002] In the management of power companies, line loss rate is an important indicator to measure the management level. It is also a comprehensive economic indicator that reflects the planning and design, operation and management level, and production technology level of the power system. At present, as my country's economy enters a new normal, the growth rate of electricity sales has slowed down, and market competition has intensified. The model of relying solely on the rapid growth of electricity sales to support the development of power grids and companies is unsustainable. With the deepening of power reform, the transmission and distribution prices of various provinces have been successively approved, and power companies have put forward higher requirements for further strengthening line loss management, tapping the potential for loss reduction, and improving operating efficiency.

[0003] Reducing line loss is an important measure for power grid companies to reduce carbon emissions and improve power grid efficiency. Accurate analysis of line loss anomalies has positive significance for lean line loss management. In the existing technology, the three-phase load and single-phase load mixed power network in the low-voltage distribution system is generally unbalanced due to the characteristics of power consumption and large load changes, which threatens the safe operation of equipment and increases line loss. Traditional line loss phase management mainly relies on month-end statistical analysis to guide loss reduction work. This method has rough data and cannot accurately locate the reasons for high line losses. The traditional theoretical line loss calculation method has limitations, it is difficult to accurately determine the cause of line loss, and the analysis of line loss causes is not intelligent enough. At present, each substation mainly obtains the value through the line loss rate calculation formula, and then analyzes it according to the reasonable line loss rate set manually. This method lacks scientific basis and it is difficult to achieve lean management goals. Summary of the invention

[0004] In view of this, the present application provides a substation phase line loss detection method based on cluster analysis to solve the technical problem that in the prior art, each substation mainly obtains a numerical value through a line loss rate calculation formula, and then analyzes it according to a reasonable line loss rate set manually, which lacks scientific basis, has high errors, and is difficult to achieve refined management.

[0005] According to a first aspect of the present application, a method for detecting phase line loss in a substation based on cluster analysis is provided, comprising:

[0006] Acquire the area training data, wherein the area training data includes the total meter power data, the household meter power data, the known line loss rate of each phase and the line loss abnormality mark;

[0007] Preprocessing the station area training data;

[0008] Performing dimensionality reduction processing on the pre-processed station area training data;

[0009] Clustering the reduced dimension area training data based on the frost-ice optimized density clustering RIME-DBSCAN algorithm to identify different categories of areas;

[0010] Based on the Stacking ensemble learning model, the line loss rate of each substation in each cluster is evaluated and the line loss rate of each phase is calculated;

[0011] Compare the line loss rate of each phase with the set threshold to identify the phases and sections with abnormal line loss rate.

[0012] According to a second aspect of the present application, a phase line loss detection device for a substation based on cluster analysis is provided, comprising:

[0013] A data acquisition module, which acquires the area training data, wherein the area training data includes the total meter power data, the household meter power data, the known line loss rate of each phase and the line loss abnormality mark;

[0014] A data processing module, used for preprocessing the station area training data;

[0015] A data dimension reduction module, used for performing dimension reduction processing on the pre-processed station area training data;

[0016] A data clustering module, used for clustering the dimension-reduced substation training data based on the frost-ice optimized density clustering RIME-DBSCAN algorithm to identify substations of different categories;

[0017] The line loss analysis module evaluates the line loss rate of each substation in each cluster and calculates the line loss rate of each phase based on the Stacking ensemble learning model.

[0018] The abnormal warning module compares the line loss rate of each phase with the set threshold to identify the phases and sections with abnormal line loss rate.

[0019] According to a third aspect of the present application, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the above-mentioned method for detecting phase line loss in a substation based on cluster analysis are implemented.

[0020] According to a fourth aspect of the present application, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned substation phase line loss detection method based on cluster analysis are implemented.

[0021] By means of the above technical scheme, the present application provides a method, device, equipment and medium for detecting phase line loss in a substation based on cluster analysis, which obtains substation training data, pre-processes the substation training data, performs dimensionality reduction processing on the pre-processed substation training data, clusters the reduced dimensionality substation training data based on the frost-optimized density clustering RIME-DBSCAN algorithm to identify substations of different categories, and evaluates the line loss rate of each substation in each cluster based on the Stacking ensemble learning model, calculates the line loss rate of each phase, and compares the line loss rate of each phase with the set threshold to identify the phases and sections with abnormal line loss rate. Through the optimized data dimensionality reduction method and the improved RIME-DBSCAN clustering algorithm, accurate detection and management of substation line loss rate can be achieved. The present invention can not only improve data processing efficiency and accuracy, but also better adapt to the operating characteristics of different substations, and provide scientific and reliable technical support for line loss management of power grid enterprises.

[0022] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application and to implement it in accordance with the contents of the specification, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0024] Figure 1 A schematic diagram of an application scenario of a phase line loss detection method in a substation based on cluster analysis provided in an embodiment of the present application is shown;

[0025] Figure 2 A schematic diagram of a flow chart of a phase line loss detection method for a transformer substation based on cluster analysis provided in an embodiment of the present application is shown;

[0026] Figure 3 A schematic diagram of a method for data dimensionality reduction processing provided in an embodiment of the present application is shown;

[0027] Figure 4 A schematic diagram of a method flow for clustering station area data provided in an embodiment of the present application is shown;

[0028] Figure 5 A schematic diagram of the process of training the stacking ensemble learning model for the sub-station area provided in the embodiment of the present application is shown;

[0029] Figure 6A schematic structural diagram of a phase line loss detection device in a substation based on cluster analysis provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0030] The specific implementation of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0031] The present invention provides a method for detecting phase line loss in a substation based on cluster analysis, which can be applied to Figure 1 In the scenario shown, the power consumption information collection system is mainly composed of the system master station (the master station is equipped with a data storage server and a data analysis server), a communication channel, a collection terminal and an electric energy meter. The collection objects are mainly divided into two categories: one is the dedicated transformer users with high voltage power supply, and the other is the general industrial and commercial users and residential users with low voltage power supply. The second category of users are usually concentrated under the public distribution transformer, and the public transformer collection terminal or low-voltage concentrator is used to realize centralized meter reading. The data types collected by the power consumption information collection system are diverse, including data on power, power quality, AC sampling, working conditions, line loss, event records and archive information. The power meter collection is based on high-speed power line carrier (HPLC). The power consumption information collection system is used to obtain the daily frozen data of the power meter in the substation area. Combined with the total power information of the substation area, the three-phase comprehensive line loss of the substation area can be calculated, and the daily line loss analysis and management can be carried out. For the low-voltage distribution network substation area, some substation areas have internal exchange and collection measurement points, and the total power information of the substation area can be directly obtained; while some substation areas do not have internal exchange and collection measurement points, and the total power information must be obtained through the corresponding public transformer terminal. Considering that the household meters in the substation are divided into single-phase meters and three-phase meters, for single-phase meters, the actual phase information of the electric energy meter can be obtained by reading the phase information of the carrier module on the electric energy meter, and the data items of the three-phase meter should include the daily frozen data of the phase. A substation phase line loss detection method based on clustering analysis provided by an embodiment of the present invention is executed in the data analysis server, by obtaining substation training data, preprocessing the substation training data, performing dimensionality reduction processing on the preprocessed substation training data, clustering the reduced dimensionality substation training data based on the frost-ice optimized density clustering RIME-DBSCAN algorithm to identify substations of different categories, and based on the Stacking ensemble learning model, evaluating the line loss rate of each substation in each cluster, calculating the line loss rate of each phase, and comparing the line loss rate of each phase with the set threshold to identify the phases and sections with abnormal line loss rate. Through the optimized data dimension reduction method and the improved RIME-DBSCAN clustering algorithm, accurate detection and management of the line loss rate of the substation area can be achieved. The present invention can not only improve the data processing efficiency and accuracy, but also better adapt to the operating characteristics of different substations, and provide scientific and reliable technical support for the line loss management of power grid enterprises.

[0032] The present invention is described in detail below through specific embodiments.

[0033] Embodiment 1:

[0034] like Figure 2 As shown, a phase line loss detection method based on cluster analysis in a substation is provided in an embodiment of the present invention, comprising:

[0035] Step 201, obtaining station area training data;

[0036] The substation training data includes the total meter power data, household meter power data, known line loss rates of each phase and abnormal line loss marks;

[0037] Step 202, pre-processing the station area training data;

[0038] Among them, in the first embodiment of the present invention, the mutual information method is used to identify and process outliers and missing values ​​in the data to improve the integrity and consistency of the data. The mutual information in the first embodiment of the present invention is a statistic that measures the mutual dependence between two variables. By calculating the mutual information between the feature and the target variable, outliers and missing values ​​in the data can be identified and processed, thereby improving the integrity and consistency of the data. Step 202 may specifically include:

[0039] Step 202-1, preliminarily fill missing values ​​and remove obvious outliers in the station area training data;

[0040] Among them, missing values ​​can be filled by using mean filling and other similar methods.

[0041] Step 202-2, discretizing the continuous variables in the station area training data (for example, using equal frequency binning or equal width binning);

[0042] Among them, the equal frequency binning method can be used to discretize continuous variables.

[0043] Step 202-3, calculate the mutual information I(X;Y) between each feature and the target variable;

[0044] Among them, the mutual information of each data point is calculated, and the points with mutual information significantly lower than the average are identified as outliers. Mutual information is a statistic that measures the mutual dependence between two variables. By calculating the mutual information between the feature and the target variable, outliers and missing values ​​in the data can be identified and processed, thereby improving the integrity and consistency of the data. In the first embodiment of the present invention, the mutual information is equal to

[0045]

[0046] Among them, p(x,y) is the joint probability distribution, and p(x) and p(y) are marginal probability distributions.

[0047] Step 202-4: Based on the mutual information calculation result, identify points whose mutual information is significantly lower than the average value as outliers, and identify features whose mutual information is significantly lower than the average value that may have missing values;

[0048] Step 202-5: For the identified outliers, delete, modify or replace them with adjacent values ​​to fix the outlier problem;

[0049] Among them, according to the mutual information calculation results, points whose mutual information is significantly lower than the average are identified as outliers. In order to improve efficiency, the quartile method (IQR) or Z score method can be used to further screen outliers.

[0050] Step 202-6: For features with missing values, data filling is performed;

[0051] Among them, the data graph filling processing includes mean filling, median filling or K neighbor filling. Mean filling uses the mean of the feature to fill the missing value, median filling uses the median of the feature to fill the missing value, and KNN filling uses the nearest neighbor method to fill the missing value.

[0052] Step 203, performing dimensionality reduction processing on the pre-processed station area training data;

[0053] Step 204: clustering the dimension-reduced substation training data based on the frost-ice optimized density clustering RIME-DBSCAN algorithm to identify substations of different categories;

[0054] Step 205: Based on the Stacking ensemble learning model, the line loss rate of each substation in each cluster is evaluated, and the line loss rate of each phase is calculated;

[0055] Step 206: Compare the line loss rate of each phase with a set threshold value to identify the phase and section with abnormal line loss rate.

[0056] The present invention provides a substation phase line loss detection method based on cluster analysis, which obtains substation training data, pre-processes the substation training data, performs dimensionality reduction processing on the pre-processed substation training data, clusters the substation training data after dimensionality reduction based on the frost optimized density clustering RIME-DBSCAN algorithm to identify substations of different categories, and evaluates the line loss rate of each substation in each cluster based on the Stacking ensemble learning model, calculates the line loss rate of each phase, and compares the line loss rate of each phase with the set threshold to identify the phases and sections with abnormal line loss rate. Through the optimized data dimensionality reduction method and the improved RIME-DBSCAN clustering algorithm, accurate detection and management of substation line loss rate can be achieved. The present invention can not only improve data processing efficiency and accuracy, but also better adapt to the operating characteristics of different substations, and provide scientific and reliable technical support for line loss management of power grid enterprises.

[0057] Embodiment 2:

[0058] In order to reduce the data dimension, retain the relative distance between data, improve the generalization ability of the later model training, and accelerate the convergence of the model, a series of optimizations are performed on step 203 to improve the dimensionality reduction effect, such as Figure 3 As shown, including:

[0059] Dimensionality reduction method 1: Use entropy weight method combined with grey correlation analysis to extract indicator features that are highly correlated with line loss rate. The specific steps include:

[0060] Step 203-11, normalize the area training data according to the indicator characteristic attributes; wherein the indicator characteristic attributes include positive attributes and negative attributes, the positive attribute indicates a positive correlation with the line loss rate, the negative attribute indicates a negative correlation with the line loss rate, and the indicator characteristic a of the positive attribute is ij The normalization process is Negative attribute indicator characteristics a ij The normalization process is a ij is the jth sample value of the ith feature, b ij is the standardized value;

[0061] Step 203-12: Calculate the entropy value of the jth sample of the i-th feature index And calculate the weight of the jth sample according to the entropy value Among them, n is the number of samples of the i-th feature, m is the number of features, and p ij equal k is equal to ln(m) is the natural logarithm of m;

[0062] Step 203-13: Select each phase line loss rate as the reference sequence X0, and the data of the i-th characteristic index as the comparison sequence X0. i ;

[0063] Step 203-14: Calculate the grey correlation coefficient between the jth sample of the ith characteristic index and the line loss rate Among them, Δ ij Equal to |x0-x ij |, is the maximum value of the absolute difference, is the minimum value of the absolute difference, and is the resolution coefficient;

[0064] Step 203-15: Calculate the grey correlation between each characteristic index and the line loss rate using the grey correlation and weight. Among them, ω j is the weight of the jth sample, ξ ij is the grey correlation coefficient between the jth sample of the ith characteristic index and the line loss rate, and n is the number of samples of the ith characteristic.

[0065] To better understand the specific implementation of the first dimension reduction processing method, for example, assume that there are three electrical characteristic indicators x1, x2, x3, x4 and a line loss rate x0, and the standardized data of each indicator is as follows:

[0066] x0=[0.2,0.3,0.4,0.5]

[0067] x1=[0.1,0.2,0.3,0.4]

[0068] x2=[0.3,0.3,0.3,0.3]

[0069] x3=[0.4,0.3,0.2,0.1]

[0070] 1. Calculate the absolute difference:

[0071] Δ 01 =∣0.2-0.1,0.3-0.2,0.4-0.3,0.5-0.4∣=[0.1,0.1,0.1,0.1]=[Δ 11 , Δ 12 , Δ 13 , Δ 14 ]

[0072] Δ 02 =∣0.2-0.3,0.3-0.3,0.4-0.3,0.5-0.3∣=[0.1,0,0.1,0.2]=[Δ 21 , Δ 22 , Δ 23 , Δ 24 ]

[0073] Δ 03 =∣0.2-0.4,0.3-0.3,0.4-0.2,0.5-0.1∣=[0.2,0,0.2,0.4]=[Δ 31 , Δ 32 , Δ 33 , Δ 34 ]

[0074] 2. Calculate the minimum and maximum absolute differences:

[0075] Δ min =0,Δ max =0.4

[0076] 3. Calculate the correlation coefficient (take ρ = 0.5):

[0077] ξ 11 ,ξ 12 ,ξ 13 ,ξ 14 =0+0.5×0.4 / 0.1+0.5×0.4=0.2 / 0.3=0.67

[0078] ξ 21 =0.1+0.5×0.4 / 0.1+0.5×0.4=0.3 / 0.2≈0.67,ξ 22 =1,ξ 23 =0.67,ξ 24 =0.2 / 0.4=0.5

[0079] ξ 31 =0+0.5×0.4 / 0.2+0.5×0.4=0.2 / 0.4=0.5,ξ 32 =1,ξ 33 =0.5,ξ 34 =0.2 / 0.4=0.5

[0080] 4. Calculate the correlation:

[0081] Through the above steps, the correlation between each electrical characteristic index and the line loss rate can be calculated, and it can be determined which characteristic indexes have a significant impact on the line loss rate, thereby providing a basis for subsequent line loss rate anomaly detection and optimization.

[0082] The dimensionality reduction processing method provided by the present invention is a combination of the entropy weight method and the grey correlation analysis through an improved grey correlation analysis. The weight of each feature index is calculated through the entropy weight method, which improves the accurate evaluation of the importance of each index. The entropy value and the grey correlation analysis are combined to make the weight distribution more objective and avoid the interference of subjective factors. The improved grey correlation degree calculation method combines the entropy weight method and the grey correlation analysis to make the calculation result more accurate and can more truly reflect the association between the feature and the target variable. Through the improved grey correlation analysis technology, key features can be more effectively identified, providing a more reliable basis for subsequent analysis and decision-making, especially in complex systems and high-dimensional data environments.

[0083] Dimensionality reduction processing method 2: Dimensionality reduction processing is performed on the station area training data through T-distributed random neighbor embedding t-SNE, and the specific steps include:

[0084] Step 203-21, calculate the similarity between the training data points in the station area and the similarity between the high-dimensional space data points Among them, x i 、x j For any two data in the i-th feature and the j-th feature in the high-dimensional space, the parameter σ i is the variance of the Gaussian distribution, the denominator is the summation term used to normalize the similarity, the purpose is to ensure that the sum of the similarities is 1, ||x i -x j || 2 Represents the square of the Euclidean distance between data;

[0085] Step 203-22: randomly select feature indicators corresponding to the data point positions in the initial low-dimensional space, and calculate the similarity between the data points in the low-dimensional space Among them, x i 、x j are any two data of the i-th feature and the j-th feature in the low-dimensional space;

[0086] Step 203-23, iteratively update the high-dimensional space to a low-dimensional space, and calculate the similarity between the data points in the high-dimensional space and the similarity between the data points in the low-dimensional space to minimize the KL divergence To update the target, until the predetermined number of iterations or convergence conditions are reached, where P is the joint probability distribution of each data point in the high-dimensional space, and Q is the joint probability distribution in the low-dimensional space.

[0087] The second dimensionality reduction processing method provided by the present invention is to perform dimensionality reduction processing on high-dimensional feature data through t-SNE (t-Distributed Stochastic Neighbor Embedding) to extract key indicator features, so that the subsequent clustering algorithm can more effectively identify different types of substations.

[0088] Embodiment three:

[0089] In order to achieve the frost optimized density clustering RIME-DBSCAN algorithm to improve the clustering effect, such as Figure 4 As shown, including:

[0090] Step 401: Perform multi-dimensional clustering on the data using the DBSCAN algorithm;

[0091] The DBSCAN clustering algorithm process in this step is a conventional process in the prior art, specifically including judging the number of points in the neighborhood of each data point ∈ (distance metric, used to locate points near any point) as the core point. If the number is not less than MinPts (the minimum number of points that are clustered together for an area that is considered dense), then p is the core point. Starting from the core point, all points in its ∈ neighborhood are marked as the same cluster. For each point that is directly accessible by density, repeat the process until no new points are added to the cluster.

[0092] Step 402, optimizing clustering results based on soft frost search, hard frost penetration and active greedy selection mechanism of frost ice optimization algorithm RIME;

[0093] Among them, in this step, the soft frost search strategy is used to conduct global exploration to simulate the process of frost ice growing freely and slowly on the surface, and the hard frost piercing mechanism is used to conduct in-depth search in the local range to simulate the process of frost ice growing consistently in a fixed direction. The active greedy selection mechanism is used to compare the fitness values ​​before and after the update, retain the better solution, realize population update, and retain the better solution. The soft frost search strategy uses the strong randomness and coverage of frost particles to enable the algorithm to quickly cover the entire search space and avoid falling into the local optimal solution. The hard frost piercing mechanism improves the convergence of the algorithm and the ability to jump out of the local optimum. The soft frost search update is performed according to the following formula:

[0094]

[0095] is the updated position of the jth dimension of the ith particle, R best,j is the optimal value of the jth dimension, r1 and r2 are random numbers, U bij and L bijare the upper and lower bounds of the jth dimension of the ith particle, respectively. θ represents the direction angle, which is usually used to determine the update direction and simulate the growth of frost in random directions. θ is usually a random number uniformly distributed in the interval [0,2π], which is used to simulate various possible directions in the growth process. β represents the step coefficient, which controls the size of the update step. β is usually a random number uniformly distributed in the interval [0,1], which is used to simulate the random change of the step in the growth process. h represents the distance proportional factor between the current particle and the optimal particle, which determines the range of the update. h is obtained by calculating the position difference between the current particle and the global optimal particle, which is used to ensure that the particle moves in the direction of the optimal solution. E is a control parameter, which is usually a preset threshold, which is used to decide whether to update the particle position.

[0096] The frost penetration mechanism works according to the following formula:

[0097]

[0098] in, is the updated position of the jth dimension of the ith particle, R best,j is the optimal value of the jth dimension, r3 is a random number, F norm (S i ) is the standardized value of the current individual fitness.

[0099] The greedy selection mechanism is performed according to the following formula:

[0100]

[0101]

[0102] Step 403: Delete particles by fitness difference, generate new particles by MCMC method, and repeat the steps of particle updating and fitness calculation until a predetermined number of iterations or convergence condition is reached;

[0103] Among them, the MCMC method is the Markov Chain Monte Carlo method (MCMC). In this step, the individual fitness differences are monitored, particles are deleted, and the quality of the particle group is maintained. At the same time, the MCMC method is used to generate new particles to further improve the global search ability of the algorithm. According to the fitness differences of the particles, particles are deleted to maintain the quality of the particle group. The fitness variance formula is:

[0104]

[0105] Among them, F i is the fitness value of the th particle, is the fitness mean. The Markov Chain Monte Carlo (MCMC) method is used to generate new particles to improve the global search ability of the algorithm. The MCMC acceptance probability formula is:

[0106]

[0107] Among them, π(x) is the target distribution, q(x|x′) is the proposed distribution (conditional probability distribution) of generating state x′ from state x, and the fitness of each particle is calculated for the DBSCAN algorithm. The fitness is usually related to the compactness and separation of the clusters. The fitness calculation function is

[0108]

[0109]

[0110] Among them, ∈ i is the compactness of the ith cluster, δ is the separation of the ith cluster, and in the compactness calculation, |C i | is the number of samples in cluster Ci, and is the centroid of cluster Ci.

[0111] Step 404: When the number of iterations reaches a predetermined value or the fitness change is less than a certain threshold, the iteration is stopped and the optimal parameters are output.

[0112] The iteration termination conditions include reaching the maximum number of iterations or the fitness change being less than a predetermined threshold, and outputting the parameter combination corresponding to the particle with the highest fitness, i.e., the optimal clustering parameters.

[0113] The frost-optimized density clustering RIME-DBSCAN algorithm provided by the present invention is combined with the soft frost search strategy of RIME to enhance the global search capability. The hard frost puncture mechanism effectively improves the local optimization capability. The DBSCAN clustering algorithm is combined with the fitness difference deletion strategy to improve the clustering effect.

[0114] Embodiment 4:

[0115] In order to calculate the line loss rate of each phase and identify the phase and section with abnormal line loss according to the final line loss rate prediction value, a substation area line loss anomaly detection model based on the substation area Stacking ensemble learning model is proposed in the fourth embodiment of the present invention, such as Figure 5 As shown, the training steps of the model include:

[0116] Step 501: For each base learner, use the training data to train and obtain the base learner model f1, f2, ..., f n ;

[0117] Among them, the first layer base learners can include models such as linear regression, random forest, and gradient boosting tree.

[0118] Step 502: Use the base learner model to predict the training data and obtain the first-layer prediction output Z = {f1(X), f2(X), ..., f n (X)};

[0119] Where X is the training data;

[0120] Step 503: Use the first layer prediction output Z as a feature and the line loss rate Y as a target to train the meta-learner model g;

[0121] Step 504: Use the meta-learner model to predict the test data to obtain the final line loss rate prediction value

[0122] Among them, the meta-learner, as the second-layer learner, generally uses linear regression or other simple models to learn and predict the output of the first-layer base learner.

[0123] The substation area line loss anomaly detection model based on the Stacking ensemble learning model provided by the present invention realizes line loss rate evaluation for the substations in each cluster (that is, one model corresponds to one type of substation), which can effectively solve the problem of unsatisfactory analysis and detection caused by large differences in line loss rates among substations, and improve the accuracy and reliability of line loss rate evaluation.

[0124] Further, as Figures 2 to 5 In the specific implementation of the method, an embodiment of the present invention provides a phase line loss detection device based on cluster analysis, such as Figure 6 As shown, the device comprises:

[0125] The data acquisition module 610 acquires the area training data, wherein the area training data includes the total meter power data, the household meter power data, the known line loss rate of each phase and the line loss abnormality mark;

[0126] The data processing module 620 is used to pre-process the station area training data;

[0127] A data dimension reduction module 630 is used to perform dimension reduction processing on the pre-processed station area training data;

[0128] A data clustering module 640 is used to cluster the reduced dimension area training data based on the frost ice optimized density clustering RIME-DBSCAN algorithm to identify different categories of areas;

[0129] The line loss analysis module 650 evaluates the line loss rate of each substation in each cluster based on the Stacking ensemble learning model and calculates the line loss rate of each phase;

[0130] The abnormal warning module 660 compares the line loss rate of each phase with a set threshold to identify the phase and section with abnormal line loss rate.

[0131] In an embodiment of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of implementing a method for detecting phase line loss in a substation based on cluster analysis include:

[0132] Acquire the area training data, wherein the area training data includes the total meter power data, the household meter power data, the known line loss rate of each phase and the line loss abnormality mark;

[0133] Preprocessing the station area training data;

[0134] Performing dimensionality reduction processing on the pre-processed station area training data;

[0135] Clustering the reduced dimension area training data based on the frost-ice optimized density clustering RIME-DBSCAN algorithm to identify different categories of areas;

[0136] Based on the Stacking ensemble learning model, the line loss rate of each substation in each cluster is evaluated and the line loss rate of each phase is calculated;

[0137] Compare the line loss rate of each phase with the set threshold to identify the phases and sections with abnormal line loss rate.

[0138] In an embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0139] Acquire the area training data, wherein the area training data includes the total meter power data, the household meter power data, the known line loss rate of each phase and the line loss abnormality mark;

[0140] Preprocessing the station area training data;

[0141] Performing dimensionality reduction processing on the pre-processed station area training data;

[0142] Clustering the reduced dimension area training data based on the frost-ice optimized density clustering RIME-DBSCAN algorithm to identify different categories of areas;

[0143] Based on the Stacking ensemble learning model, the line loss rate of each substation in each cluster is evaluated and the line loss rate of each phase is calculated;

[0144] Compare the line loss rate of each phase with the set threshold to identify the phases and sections with abnormal line loss rate.

[0145] It should be noted that, in the above embodiments, only the phase line loss of the substation is used as an example to illustrate the principles and implementation steps of the embodiments of the present invention, and no specific limitation is made to the actual application scenarios, such as low-voltage substations, high-voltage substations, active substations, and passive substations. The technical solution of the present invention can also be applied to extended scenarios of line loss analysis and evaluation of various substations. For the functions or steps that can be implemented by the computer-readable storage medium or computer device, please refer to the aforementioned method embodiments. To avoid repetition, they will not be described one by one here.

[0146] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0147] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0148] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A phase line loss detection method based on cluster analysis, characterized in that: include: Acquire the area training data, wherein the area training data includes the total meter power data, the household meter power data, the known line loss rate of each phase and the line loss abnormality mark; Preprocessing the station area training data; Performing dimensionality reduction processing on the pre-processed station area training data; Clustering the reduced dimension area training data based on the frost-ice optimized density clustering RIME-DBSCAN algorithm to identify different categories of areas; Based on the Stacking ensemble learning model, the line loss rate of each substation in each cluster is evaluated and the line loss rate of each phase is calculated; Compare the line loss rate of each phase with the set threshold to identify the phases and sections with abnormal line loss rate.

2. The method for detecting phase line loss in a substation based on cluster analysis according to claim 1, characterized in that: The step of preprocessing the station area training data includes: Preliminarily filling missing values ​​and removing obvious outliers in the station area training data; Discretizing continuous variables in the station area training data; Calculate the mutual information I(X;Y) between each feature and the target variable; According to the mutual information calculation results, points with mutual information significantly lower than the average are identified as outliers, and features with mutual information significantly lower than the average may have missing values; For the identified outliers, fix the outlier problem by choosing to delete, modify, or replace them with neighboring values; For features that are identified to have missing values, data filling processing is performed, wherein the data graph filling processing includes mean filling, median filling or K-nearest neighbor filling.

3. The method for detecting phase line loss in a substation based on cluster analysis according to claim 1, characterized in that: The step of performing dimensionality reduction processing on the pre-processed station area training data comprises: The entropy weight method is combined with grey correlation analysis to extract the index features that are highly correlated with the line loss rate, or the dimensionality reduction processing of the training data of the substation area is performed through T-distributed random neighbor embedding t-SNE.

4. The method for detecting phase line loss in a substation based on cluster analysis according to claim 3, characterized in that: The step of extracting index features highly correlated with line loss rate by using entropy weight method combined with grey correlation analysis technology includes: According to the indicator characteristic attributes, the training data of the substation area is normalized, wherein the indicator characteristic attributes include positive attributes and negative attributes, the positive attribute indicates a positive correlation with the line loss rate, the negative attribute indicates a negative correlation with the line loss rate, and the indicator characteristic a of the positive attribute is ij The normalization process is Negative attribute indicator characteristics a ij The normalization process is a ij is the jth sample value of the ith feature, b ij is the standardized value; Calculate the entropy value of the jth sample of the i-th feature index And calculate the weight of the jth sample according to the entropy value Among them, n is the number of samples of the i-th feature, m is the number of features, and p ij equal k is equal to ln(m) is the natural logarithm of m; Select each phase line loss rate as the reference sequence X0, and the data of the i-th characteristic index as the comparison sequence X i ; Calculate the grey correlation coefficient between the jth sample of the ith characteristic index and the line loss rate Among them, Δ ij Equal to |x0-x ij |, is the maximum value of the absolute difference, is the minimum value of the absolute difference, and is the resolution coefficient; Using grey correlation and weight, calculate the grey correlation between each characteristic index and line loss rate Among them, ω j is the weight of the jth sample, ξ ij is the grey correlation coefficient between the jth sample of the ith characteristic index and the line loss rate, and n is the number of samples of the ith characteristic.

5. The method for detecting phase line loss in a substation based on cluster analysis according to claim 3, characterized in that: The step of performing dimensionality reduction processing on the station area training data by T-distributed random neighbor embedding t-SNE comprises: Calculate the similarity between the training data points in the station area and the similarity between the high-dimensional space data points Among them, x i 、x j For any two data in the i-th feature and the j-th feature in the high-dimensional space, the parameter σ i is the variance of the Gaussian distribution, the denominator is the summation term used to normalize the similarity, the purpose is to ensure that the sum of the similarities is 1, ||x i -x j || 2 Represents the square of the Euclidean distance between data; Randomly select feature indicators to form the data point positions in the initial low-dimensional space, and calculate the similarity between the data points in the low-dimensional space Among them, x i 、x j are any two data in the i-th feature and the j-th feature in the low-dimensional space; Iteratively update the high-dimensional space to a low-dimensional space, and calculate the similarity between the data points in the high-dimensional space and the similarity between the data points in the low-dimensional space to minimize the KL divergence To update the target, until the predetermined number of iterations or convergence condition is reached, where P is the joint probability distribution of each data point in the high-dimensional space, and Q is the joint probability distribution in the low-dimensional space.

6. The method for detecting phase line loss in a substation based on cluster analysis according to claim 1, characterized in that: The step of clustering the dimension-reduced station area training data based on the RIME-DBSCAN algorithm to identify stations of different categories includes: Use DBSCAN algorithm to perform multidimensional clustering on data; Soft frost search, hard frost penetration and active greedy selection mechanism based on frost ice optimization algorithm RIME optimize clustering results; Particles are deleted by fitness differences, and new particles are generated using the MCMC method. The steps of particle update and fitness calculation are repeated until the predetermined number of iterations or convergence conditions are reached; When the number of iterations reaches a predetermined value or the fitness change is less than a certain threshold, the iteration is stopped and the optimal parameters are output.

7. The method for detecting phase line loss in a substation based on cluster analysis according to claim 1, characterized in that: The steps of evaluating the line loss rate of each substation in each cluster and calculating the line loss rate of each phase based on the Stacking ensemble learning model include: For each base learner, use the training data to train and obtain the base learner model f1,f2,…,f n ; Use the base learner model to predict the training data and get the first layer prediction output Z = {f1(X), f2(X), …, f n (X)}, where X is the training data; Use the first layer prediction output Z as the feature and the line loss rate Y as the target to train the meta-learner model g; Use the meta-learner model to predict the test data and obtain the final line loss rate prediction value 8. A phase line loss detection device for a substation based on cluster analysis, characterized in that: include: A data acquisition module, which acquires the area training data, wherein the area training data includes the total meter power data, the household meter power data, the known line loss rate of each phase and the line loss abnormality mark; A data processing module, used for preprocessing the station area training data; A data dimension reduction module, used for performing dimension reduction processing on the pre-processed station area training data; A data clustering module, used for clustering the dimension-reduced substation training data based on the frost-ice optimized density clustering RIME-DBSCAN algorithm to identify substations of different categories; The line loss analysis module evaluates the line loss rate of each substation in each cluster and calculates the line loss rate of each phase based on the Stacking ensemble learning model. The abnormal warning module compares the line loss rate of each phase with the set threshold to identify the phases and sections with abnormal line loss rate.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the phase line loss detection method based on cluster analysis according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the phase line loss detection method based on cluster analysis in a substation are implemented as claimed in any one of claims 1 to 7.

Images