CN110070111A - A kind of distribution network line classification method and system - Google Patents

Abstract

The invention discloses a method for classifying distribution network lines. The method includes collecting data of several distribution network lines and dividing them into a test sample set, a training sample set and a verification sample set; based on the training sample set, a K-center clustering algorithm is sequentially adopted and decision classification algorithm to generate K member classifiers; obtain several nearest neighbors of each test sample in the verification sample set; traverse the K member classifiers, if the member classifier can correctly classify the nearest neighbors of all test samples, the member The classifier is added to the optimal classifier set; the optimal classifier set is used to classify the distribution network lines to be classified. At the same time, the corresponding system is disclosed. The invention solves the problems of low classification accuracy, low efficiency and poor robustness of the existing distribution network lines.

Description

A kind of distribution network line classification method and system

technical field

The invention relates to a distribution network line classification method and system, belonging to the technical field of power grid operation automation.

Background technique

As users have higher and higher requirements for the distribution network, the State Grid Corporation of China has proposed to speed up the construction and transformation of the urban distribution network in recent years, strive to improve the safe operation level of the urban power grid, improve the power quality, and reduce the line loss. At the same time, it proposes to optimize the grid Structure and improve the automation and management level of urban distribution network.

An important part of distribution network construction is the operation and maintenance of distribution network lines. The principle of operation and maintenance is "important equipment, key operation and maintenance". Therefore, it is necessary to conduct grade evaluation on distribution network lines, and the premise of grade evaluation is to evaluate the distribution network. The existing power grid operation and maintenance systems mainly use the following methods: 1. Single classifier classification model; 2. Static classifier model; however, these commonly used methods have low accuracy, low efficiency, and robustness. Sex is not good.

SUMMARY OF THE INVENTION

The invention provides a distribution network line classification method and system, and solves the problems of low classification accuracy, low efficiency and poor robustness of the existing distribution network lines.

In order to solve the above-mentioned technical problems, the technical scheme adopted in the present invention is:

A distribution network line classification method, comprising,

Collect several distribution network line data and divide it into test sample set, training sample set and verification sample set;

Based on the training sample set, the K-center clustering algorithm and the decision-making classification algorithm are used in turn to generate K member classifiers;

Obtain several nearest neighbors of each test sample in the validation sample set;

Traverse K member classifiers, if the member classifier can correctly classify the nearest neighbors of all test samples, add the member classifier to the optimal classifier set;

Use the optimal classifier set to classify the distribution network lines to be classified.

Each sample in the test sample set, training sample set and verification sample set is the data of a distribution network line, and the distribution network line data includes the operation period, the monthly average load rate, the monthly average line loss rate, the annual number of faults, and the associated distribution. The number of annual failures of the transformer, the average load rate of the associated distribution transformer, and whether important users are included.

The process of generating K member classifiers is,

Use PAM to divide the training sample set to generate K training sample subsets;

Using the C4.5 classification algorithm, K training subsets are trained to generate K member classifiers.

Calculate the Euclidean distance from the test sample to each verification sample, and select several verification samples whose Euclidean distance is less than the threshold as the nearest neighbor.

A distribution network line classification system, comprising,

Collection module: collects data of several distribution network lines and divides them into test sample set, training sample set and verification sample set;

Member classifier generation module: Based on the training sample set, K-center clustering algorithm and decision classification algorithm are used in turn to generate K member classifiers;

The nearest neighbor acquisition module: to obtain several nearest neighbors of each test sample in the verification sample set;

Optimal classifier set generation module: traverse K member classifiers, if the member classifier can correctly classify the nearest neighbors of all test samples, add the member classifier to the optimal classifier set;

Classification module: Use the optimal classifier set to classify the distribution network lines to be classified.

Each sample in the test sample set, training sample set and verification sample set is the data of a distribution network line, and the distribution network line data includes the operation period, the monthly average load rate, the monthly average line loss rate, the annual number of faults, and the associated distribution. The number of annual failures of the transformer, the average load rate of the associated distribution transformer, and whether important users are included.

Member classifier generation module includes PAM module and C4.5 module;

PAM module: Use PAM to divide the training sample set to generate K training sample subsets;

C4.5 module: The C4.5 classification algorithm is used to train K training subsets to generate K member classifiers.

The process of obtaining the nearest neighbor by the nearest neighbor acquisition module is as follows:

Calculate the Euclidean distance from the test sample to each verification sample, and select several verification samples whose Euclidean distance is less than the threshold as the nearest neighbor.

A computer-readable storage medium storing one or more programs comprising instructions that, when executed by a computing device, cause the computing device to perform a distribution line classification method.

A computing device comprising one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the The one or more programs include instructions for performing the distribution line classification method.

The beneficial effects achieved by the invention are as follows: the invention solves the problems of low classification accuracy, low efficiency and poor robustness of the existing distribution network lines.

Description of drawings

Fig. 1 is the flow chart of the present invention;

Figure 2 is a comparison diagram of the three classification methods.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings. The following examples are only used to illustrate the technical solutions of the present invention more clearly, and cannot be used to limit the protection scope of the present invention.

As shown in Figure 1, a method for classifying distribution network lines includes the following steps:

Step 1: Collect several distribution network line data and divide them into test sample set, training sample set and verification sample set.

The ratio of training sample set, test sample set and verification sample set is 8:1:1. Each sample of test sample set, training sample set and verification sample set is the data of a distribution network line. The distribution network line data includes Years of operation, monthly average load rate, monthly average line loss rate, annual failure times, annual failure times of associated distribution transformers, average load rate of associated distribution transformers, and whether important users are included.

Operation period: obtain the distribution network line operation date from the production management system, operation period = current operation time - operation date;

Monthly average load rate: obtain the real-time current from the dispatching automation system, and obtain the line rated current from the production management system, load rate = real-time current/rated current;

Monthly average line loss rate: Obtain the current month's power loss and current month's power of distribution network lines from the electricity consumption information collection system, and the monthly average line loss rate = current month's power loss/current month's power;

Annual number of failures: Obtain the annual number of failures from the scheduling automation system;

The annual number of faults associated with distribution transformers: obtain the information of all distribution transformers associated with the distribution network lines from the production management system, and obtain the annual number of faults from the dispatch automation system;

The average load rate of the associated distribution transformers: obtain all distribution transformer information associated with the distribution network lines from the production management system, and obtain the load and capacity from the electricity consumption information collection system, load rate = load/capacity;

Whether to include important users: Obtain all distribution and transformer information associated with distribution network lines from the production management system, and check whether important users are included from the marketing system.

Step 2: Based on the training sample set, the K-center clustering algorithm and the decision-making classification algorithm are sequentially used to generate K member classifiers.

The specific process is: using PAM to divide the training sample set to generate K training sample subsets; using the C4.5 classification algorithm to train the K training subsets to generate K member classifiers.

PAM divides the training samples into K center point clustering. According to the Calinski-Harabasz criterion, the K value is calculated from the inter-class variance, intra-class variance and complexity, and multiple sub-training sample sets are generated.

The PAM division is as follows:

(1) arbitrarily select K training sample objects from the training sample set as the initial representative objects;

(2) assigning each remaining object to the set represented by the nearest representative object;

(3) Randomly select a non-representative object and calculate the total cost of exchanging the representative object with the non-representative object Among them, p is a point in the space, which is a non-representative object in the representative set C _J , O _j is the representative object in the representative set C _J , and k is the number of representative objects in the representative set C _J ;

(4) If the total cost is less than 0, replace the representative object with a non-representative object, and generate K representative object sets;

(5) Repeat steps (2), (3), (4) until each set does not change.

The C4.5 classification algorithm is a series of algorithms used in machine learning and data mining classification problems with the goal of supervised learning: given a data set, each tuple in it can be described by a set of attribute values, Each tuple belongs to one of a mutually exclusive category; the goal is to learn, through learning, to find a mapping from attribute values to categories that can be used to classify new entities with unknown categories.

Step 3: Obtain several nearest neighbors of each test sample in the verification sample set.

The specific process is: calculating the Euclidean distance from the test sample to each verification sample, and selecting several verification samples whose Euclidean distance is less than the threshold as the nearest neighbors.

Step 4, traverse the K member classifiers, if the member classifier can correctly classify the nearest neighbors of all test samples, the member classifier is added to the optimal classifier set.

Step 5: Use the optimal classifier set to classify the distribution network lines to be classified.

According to the characteristics of the distribution network line to be classified, the classifier combination is adaptively selected from the optimal classifier set or the classifier weight is assigned to perform the final dynamic combination for classification.

In order to verify the above method, do the following experiments:

The 2018 operation data of 1,600 distribution network lines in each of the three provinces were selected as the training, verification and test data of the model. For each distribution network line, the operating years, monthly average load rate, monthly average line loss rate, and annual failure rate were selected. There are seven characteristic variables, including the number of times, the annual failure times of the associated distribution transformers, the average load rate of the associated distribution transformers, and whether important users are included.

The sample data is divided into three groups: training, validation, and testing. The training data set accounts for 80% of the total data samples, with a total of 1280 records. The test data set and the validation data set each account for 10% of the total, with 160 records each. The dataset was discretized. The experimental comparison methods include the above methods and the current conventional model methods, namely the dynamic combination classification DDC-CD (ie the method of the present invention), the single classifier method (C4.5), the static combination method AdaBoost (with C4.5 as the member classification) machine learning algorithms). The three methods are compared in terms of data classification accuracy, classification efficiency, and robustness, as shown in Tables 1, 2 and Figure 2.

Table 1 Comparison of classification accuracy of three classification methods (%)

Table 2 Classification robustness comparison of three classification methods (times)

The results in Table 1 show that the classification accuracy of AdaBoost is better than C4.5, and the classification accuracy of DCC-CD is better than that of AdaBoost and C4.5. This shows that DCC-CD can make full use of the prediction information provided by the base classifier, effectively improve the classification accuracy, and the combined superiority is much greater than that of AdaBoost.

The results in Figure 2 show that the classification efficiency of AdaBoost on the three sets of data sources is better than C4.5, and the classification efficiency of dynamic classification DCC-CD is better than that of AdaBoost and the base classifier C4.5.

In addition, the 10 classification accuracy intervals of the cross-validation method are counted, and the accuracy intervals are selected as (0, R-3) and (R+3, 100), where R is the average accuracy. From Table 2, we can see that C4.5 has 5 out of ten validations, while AdaBoost has 3 and DCC-CD has only 1. The results show that DCC-CD outperforms C4.5 and AdaBoost in terms of classification robustness.

To sum up, the above method adopts combined classification to solve the problems of low classification accuracy, low efficiency and poor robustness of existing distribution network lines.

A distribution network line classification system, comprising:

Collection module: collects data of several distribution network lines and divides them into test sample set, training sample set and verification sample set. Each sample in the test sample set, training sample set and verification sample set is the data of a distribution network line, and the distribution network line data includes the operation period, the monthly average load rate, the monthly average line loss rate, the annual number of faults, and the associated distribution. The number of annual failures of the transformer, the average load rate of the associated distribution transformer, and whether important users are included.

Member classifier generation module: Based on the training sample set, K-center clustering algorithm and decision classification algorithm are used in turn to generate K member classifiers.

Member classifier generation module includes PAM module and C4.5 module; PAM module: uses PAM to divide the training sample set to generate K training sample subsets; C4.5 module: adopts C4.5 classification algorithm to train K training samples The subset is trained to generate K member classifiers.

The nearest neighbor acquisition module: acquires several nearest neighbors of each test sample in the verification sample set.

The process of obtaining the nearest neighbors by the nearest neighbor acquiring module is: calculating the Euclidean distance from the test sample to each verification sample, and selecting several verification samples whose Euclidean distance is less than the threshold as the nearest neighbors.

Optimal classifier set generation module: traverse K member classifiers, if the member classifier can correctly classify the nearest neighbors of all test samples, the member classifier is added to the optimal classifier set.

Classification module: Use the optimal classifier set to classify the distribution network lines to be classified.

A computer-readable storage medium storing one or more programs comprising instructions that, when executed by a computing device, cause the computing device to perform a distribution line classification method.

A computing device comprising one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the The one or more programs include instructions for performing the distribution line classification method.

As will be appreciated by those skilled in the art, the embodiments of the present application may be provided as a method, a system, or a computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It will be understood that each process and/or block in the flowchart illustrations and/or block diagrams, and combinations of processes and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

The above are only examples of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are included in the application for pending approval of the present invention. within the scope of the claims.

Claims (10)

1. a kind of distribution line classification method, it is characterised in that: including,

Several distribution network line data are acquired, and are divided into test sample collection, training sample set and verifying sample set；

Based on training sample set, K central cluster algorithm and Decision Classfication algorithm are successively used, generates K member classifiers；

Obtain several nearest-neighbors of each test sample in verifying sample set；

Traverse K member classifiers, the nearest-neighbors of all test samples if member classifiers can correctly classify, the member Classifier is added in optimum classifier set；

Classified with optimum classifier collection to distribution line to be sorted.

2. a kind of distribution line classification method according to claim 1, it is characterised in that: test sample collection, training sample Each sample of collection and verifying sample set is a distribution network line data, and distribution network line data include putting into operation to put down the time limit, the moon Equal load factor, monthly average line loss per unit, the annual number of stoppages, the annual number of stoppages for being associated with distribution transforming, the average load for being associated with distribution transforming Rate and whether include responsible consumer.

3. a kind of distribution line classification method according to claim 1, it is characterised in that: generate K member classifiers' Process is,

Training sample set is divided using PAM, generates K training sample subset；

Using C4.5 sorting algorithm, K training subset is trained, generates K member classifiers.

4. a kind of distribution line classification method according to claim 1, it is characterised in that: calculate test sample to each verifying The Euclidean distance of sample selects Euclidean distance to be less than several verifying samples of threshold value as nearest-neighbors.

5. a kind of distribution line categorizing system, it is characterised in that: including,

Acquisition module: several distribution network line data of acquisition, and it is divided into test sample collection, training sample set and verifying sample This collection；

Member classifiers' generation module: being based on training sample set, successively uses K central cluster algorithm and Decision Classfication algorithm, raw At K member classifiers；

Nearest-neighbors obtain module: obtaining several nearest-neighbors of each test sample in verifying sample set；

Optimum classifier set generation module: K member classifiers of traversal, all test specimens if member classifiers can correctly classify This nearest-neighbors then the member classifiers are added in optimum classifier set；

Categorization module: classified with optimum classifier collection to distribution line to be sorted.

6. a kind of distribution line categorizing system according to claim 5, it is characterised in that: test sample collection, training sample Each sample of collection and verifying sample set is a distribution network line data, and distribution network line data include putting into operation to put down the time limit, the moon Equal load factor, monthly average line loss per unit, the annual number of stoppages, the annual number of stoppages for being associated with distribution transforming, the average load for being associated with distribution transforming Rate and whether include responsible consumer.

7. a kind of distribution line categorizing system according to claim 5, it is characterised in that: member classifiers' generation module packet Include PAM module and C4.5 module；

PAM module: dividing training sample set using PAM, generates K training sample subset；

C4.5 module: using C4.5 sorting algorithm, be trained to K training subset, generates K member classifiers.

8. a kind of distribution line categorizing system according to claim 5, it is characterised in that: nearest-neighbors obtain module and obtain The process of nearest-neighbors is,

Calculate test sample to it is each verifying sample Euclidean distance, select Euclidean distance less than threshold value several verifying samples as Nearest-neighbors.

9. a kind of computer readable storage medium for storing one or more programs, it is characterised in that: one or more of journeys Sequence include instruction, described instruction when executed by a computing apparatus so that the calculatings equipment execution according to claim 1 to 4 institutes Method either in the method stated.

10. a kind of calculating equipment, it is characterised in that: including,

One or more processors, memory and one or more programs, wherein one or more programs are stored in described deposit It in reservoir and is configured as being executed by one or more of processors, one or more of programs include for executing basis The instruction of method either in method described in Claims 1-4.