CN117254587A

CN117254587A - Monitoring method and device for power distribution network, electronic equipment and storage medium

Info

Publication number: CN117254587A
Application number: CN202311191355.9A
Authority: CN
Inventors: 徐蕙; 马龙飞; 陆斯悦; 曾佳妮; 张禄; 李香龙; 王立永
Original assignee: State Grid Corp of China SGCC; State Grid Beijing Electric Power Co Ltd
Current assignee: State Grid Corp of China SGCC; State Grid Beijing Electric Power Co Ltd
Priority date: 2023-09-14
Filing date: 2023-09-14
Publication date: 2023-12-19

Abstract

The invention discloses a monitoring method and a device thereof for a power distribution network, electronic equipment and a storage medium, wherein the monitoring method comprises the following steps: the distribution transformer set contained in the target distribution network is determined, distribution transformer data of each distribution transformer are collected, distribution transformer characteristics corresponding to the distribution transformers are determined based on the distribution transformer data, the distribution transformer characteristics are input into a preset analysis model, distribution transformer states of the distribution transformers are output, and the normal operation of the target distribution network is determined under the condition that the distribution transformer states of all the distribution transformers are normal. The invention solves the technical problems that the monitoring efficiency of the power distribution network is low and the abnormal distribution transformer cannot be early warned in time in the related technology.

Description

Monitoring method and device for power distribution network, electronic equipment and storage medium

Technical Field

The invention relates to the field of artificial intelligence, in particular to a monitoring method and device of a power distribution network, electronic equipment and a storage medium.

Background

A distribution network refers to a network system for delivering electrical energy from a power plant to end users. The distribution network includes: and the power transmission line, the transformer substation, the distribution transformer, the distribution line and the like. The distribution network is capable of converting high-voltage electric energy into low-voltage electric energy suitable for household, industrial and commercial electricity, and ensuring safe, stable and reliable supply of electric energy. Therefore, monitoring and optimizing the operation quality of the power distribution network are one of the working focuses of the power company.

In the related art, for the detection and analysis of the abnormal points of the power distribution network, monitoring and maintenance are often carried out through power distribution network hardware facilities, so that a large amount of manpower resources are consumed, the monitoring efficiency is low, and early warning can not be carried out on the abnormal points of the power distribution network in time.

In view of the above problems, no effective solution has been proposed at present.

Disclosure of Invention

The embodiment of the invention provides a monitoring method and device of a power distribution network, electronic equipment and a storage medium, and aims to at least solve the technical problems that the monitoring efficiency of the power distribution network is low and abnormal power distribution transformers cannot be early warned in time in the related technology.

According to an aspect of an embodiment of the present invention, there is provided a monitoring method of a power distribution network, including: determining a distribution transformer set contained in a target power distribution network, wherein the distribution transformer set comprises: at least one distribution transformer; collecting distribution data of each distribution transformer, and determining distribution characteristics corresponding to the distribution transformers based on the distribution data; inputting the distribution characteristics into a preset analysis model, and outputting a distribution state of the distribution transformer, wherein the preset analysis model is an analysis model trained based on historical distribution data, and the distribution state comprises: normal state, abnormal state; and under the condition that the distribution transformer states of all the distribution transformers are the normal states, determining that the target distribution network normally operates.

Optionally, the monitoring method further comprises: collecting historical distribution data of each preset distribution transformer, wherein the preset distribution transformers are distribution transformers with overvoltage abnormal movements; and constructing an analysis model feature library based on all the historical distribution change data.

Optionally, the step of constructing an analysis model feature library based on all the historical transformation data includes: filtering noise of the historical configuration data to obtain target historical configuration data; dividing the target historical allocation data into a plurality of historical allocation sub-data associated with a preset dimension based on a preset dimension set; determining a numerical characteristic and a frequency characteristic based on the historical substation data, wherein the numerical characteristic is an index value characteristic obtained by calculation according to the historical substation data, and the frequency characteristic is a quantitative value characteristic obtained by statistics according to the historical substation data; and constructing the analysis model feature library based on all the numerical features and all the frequency features.

Optionally, after constructing the analysis model feature library based on all the historical transformation data, the method further comprises: constructing a random forest classification model, wherein the random forest classification model comprises: n decision trees, wherein N is a positive integer; inputting all data features in the analysis model feature library into the random forest classification model, and outputting a contribution value of each data feature to each decision tree, wherein the data features are numerical features or frequency features; determining an average contribution value of the data feature based on the contribution value of the data feature to each of the decision trees; and sorting all the average contribution values to obtain a sorting result, and selecting target data features indicated by the average contribution values within a preset range based on the sorting result to obtain a target data feature set.

Optionally, the step of constructing a random forest classification model includes: sampling is carried out for N times from the analysis model feature library based on a preset sampling strategy, and N training sample sets are obtained; training the decision tree by adopting the training sample set until all node attributes of each node in the decision tree belong to the same type; and constructing the random forest classification model based on the trained decision tree.

Optionally, after selecting the target data feature indicated by the average contribution value within a preset range based on the sorting result, obtaining a target data feature set, the method further includes: determining a set of classification algorithms, wherein the set of classification algorithms comprises: a plurality of classification algorithms, each of which corresponds to a classification frame; dividing the target data characteristic set into a training set and a testing set based on a preset proportion; and training the classification frame by adopting the training set until a loss value determined by a loss function corresponding to the classification frame is smaller than a preset loss threshold value, so as to obtain an initial classification model, wherein the loss function is constructed based on a predicted value and a labeling value, the predicted value is a value output by the initial classification model, and the labeling value is a value labeled for each target data characteristic in advance.

Optionally, after training the classification frame by using the training set until the loss value determined by the loss function corresponding to the classification frame is smaller than a preset loss threshold value, obtaining an initial classification model, the method further includes: and testing each initial classification model by adopting the test set to obtain a test result, wherein the test result comprises: presetting an index value; under the condition that the preset index value is smaller than a preset index threshold value, adjusting parameters corresponding to the initial classification model to obtain an adjusted initial classification model; continuing training the adjusted initial classification model by adopting the training set until the preset index value is greater than or equal to a preset index threshold value to obtain a target classification model; and sequencing the preset index values corresponding to all the target classification models, and characterizing the target classification model indicated by the maximum preset index value as the preset analysis model.

According to another aspect of the embodiment of the present invention, there is also provided a monitoring device for a power distribution network, including: a first determining unit, configured to determine a distribution transformer set included in a target power distribution network, where the distribution transformer set includes: at least one distribution transformer; the acquisition unit is used for acquiring distribution data of each distribution transformer and determining distribution characteristics corresponding to the distribution transformers based on the distribution data; the input unit is used for inputting the distribution characteristics to a preset analysis model and outputting the distribution state of the distribution transformer, wherein the preset analysis model is an analysis model trained based on historical distribution data, and the distribution state comprises: normal state, abnormal state; and the second determining unit is used for determining that the target power distribution network operates normally under the condition that the distribution transformer states of all the distribution transformers are the normal states.

Optionally, the monitoring device further comprises: the first acquisition module is used for acquiring historical distribution data of each preset distribution transformer, wherein the preset distribution transformers are distribution transformers with overvoltage abnormal movements; the first construction module is used for constructing an analysis model feature library based on all the historical distribution change data.

Optionally, the first building module includes: the first filtering submodule is used for filtering noise of the historical configuration data to obtain target historical configuration data; the first dividing sub-module is used for dividing the target historical allocation data into a plurality of historical allocation sub-data associated with the preset dimension based on a preset dimension set; the first determining submodule is used for determining numerical characteristics and frequency characteristics based on the historical substation data, wherein the numerical characteristics are index value characteristics obtained by calculation according to the historical substation data, and the frequency characteristics are quantitative value characteristics obtained by statistics according to the historical substation data; and the first construction submodule is used for constructing the analysis model feature library based on all the numerical features and all the frequency features.

Optionally, the monitoring device further comprises: the second construction module is used for constructing a random forest classification model after constructing an analysis model feature library based on all the historical transformation data, wherein the random forest classification model comprises the following components: n decision trees, wherein N is a positive integer; the first input module is used for inputting all data features in the analysis model feature library into the random forest classification model and outputting the contribution value of each data feature to each decision tree, wherein the data features are numerical features or frequency features; a first determining module for determining an average contribution value of the data feature based on the contribution value of the data feature to each of the decision trees; the first sorting module is used for sorting all the average contribution values to obtain a sorting result, and selecting target data features indicated by the average contribution values within a preset range based on the sorting result to obtain a target data feature set.

Optionally, the second building module includes: the first sampling submodule is used for sampling N times from the analysis model feature library based on a preset sampling strategy to obtain N training sample sets; a first training sub-module, configured to train the decision tree using the training sample set until all node attributes of each node in the decision tree belong to a same type; and the second construction submodule is used for constructing the random forest classification model based on the trained decision tree.

Optionally, the monitoring device further comprises: the second determining module is configured to determine a classification algorithm set after selecting, based on the sorting result, a target data feature indicated by the average contribution value within a preset range to obtain a target data feature set, where the classification algorithm set includes: a plurality of classification algorithms, each of which corresponds to a classification frame; the first dividing module is used for dividing the target data characteristic set into a training set and a testing set based on a preset proportion; the first training module is used for training the classification frame by adopting the training set until a loss value determined by a loss function corresponding to the classification frame is smaller than a preset loss threshold value, so as to obtain an initial classification model, wherein the loss function is constructed based on a predicted value and a labeling value, the predicted value is a value output by the initial classification model, and the labeling value is a value labeled for each target data characteristic in advance.

Optionally, the monitoring device further comprises: the first test module is configured to, after training the classification frame by using the training set until a loss value determined by a loss function corresponding to the classification frame is smaller than a preset loss threshold value, obtain initial classification models, test each of the initial classification models by using the test set, and obtain test results, where the test results include: presetting an index value; the first adjusting module is used for adjusting parameters corresponding to the initial classification model under the condition that the preset index value is smaller than a preset index threshold value to obtain the adjusted initial classification model; the second training module is used for continuing training the adjusted initial classification model by adopting the training set until the preset index value is greater than or equal to a preset index threshold value to obtain a target classification model; and the second sorting module is used for sorting the preset index values corresponding to all the target classification models and characterizing the target classification model indicated by the maximum preset index value as the preset analysis model.

According to another aspect of the embodiment of the present invention, there is further provided a computer readable storage medium, where the computer readable storage medium includes a stored computer program, and when the computer program runs, the device where the computer readable storage medium is controlled to execute the method for monitoring any one of the power distribution networks.

According to another aspect of the embodiment of the present invention, there is also provided an electronic device, including one or more processors and a memory, where the memory is configured to store one or more programs, and when the one or more programs are executed by the one or more processors, cause the one or more processors to implement the method for monitoring a power distribution network according to any one of the foregoing aspects.

In the method, a distribution transformer set contained in a target distribution network is determined, distribution data of each distribution transformer are collected, distribution characteristics corresponding to the distribution transformers are determined based on the distribution data, the distribution characteristics are input into a preset analysis model, distribution states of the distribution transformers are output, and the normal operation of the target distribution network is determined under the condition that the distribution states of all the distribution transformers are normal. In the method, the distribution transformers related to the target distribution network can be determined firstly, distribution transformer data of each distribution transformer are collected, corresponding distribution transformer characteristics are determined according to the distribution transformer data, the distribution transformer characteristics of each distribution transformer are analyzed by adopting a preset analysis model to obtain distribution transformer distribution states of the distribution transformers, if the distribution transformer distribution states of all the distribution transformers are normal states, the normal operation of the target distribution network is determined, otherwise, abnormal distribution transformers are early-warned, monitoring efficiency is improved, early-warned can be timely carried out on the abnormal distribution transformers, normal operation of the distribution network is effectively guaranteed, and further the technical problem that the monitoring efficiency of the distribution network is low and early-warned on the abnormal distribution transformers in time in related technologies is solved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiments of the invention and together with the description serve to explain the invention and do not constitute a limitation on the invention. In the drawings:

FIG. 1 is a flow chart of an alternative method of monitoring a power distribution network according to an embodiment of the present invention;

FIG. 2 is a schematic illustration of an alternative distribution network vulnerability analysis model feature library in accordance with an embodiment of the present invention;

FIG. 3 is a schematic diagram of an S-plot of an alternative Logistic function according to an embodiment of the invention;

FIG. 4 is a schematic illustration of an alternative distribution network vulnerability analysis model building process according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of an alternative classification algorithm-based distribution network vulnerability analysis method according to an embodiment of the present invention;

FIG. 6 is a schematic diagram of an alternative monitoring device for a power distribution network according to an embodiment of the present invention;

fig. 7 is a block diagram of a hardware structure of an electronic device (or a mobile device) for a monitoring method of a power distribution network according to an embodiment of the present invention.

Detailed Description

In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

It should be noted that, related information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present disclosure are information and data authorized by a user or sufficiently authorized by each party, and the collection, use and processing of related data need to comply with related laws and regulations and standards of related areas, and are provided with corresponding operation entries for the user to select authorization or rejection.

The invention provides a distribution point network weak point analysis method based on a classification algorithm, which aims at the problem that the distribution transformer (namely a distribution transformer) is abnormal in voltage caused by heavy overload, unreasonable gear and unbalanced three-phase load, obtains data characteristics through calculation of data such as historical three-phase voltage, current and load, and then inputs the data characteristics into a distribution network weak point analysis model for learning so as to early warn the distribution weak point and provide data support for power distribution network optimization work.

According to the method, analysis can be carried out based on distribution and transformation voltage fluctuation record data, characteristics of data of a history fluctuation distribution transformer in a fluctuation prophase are explored aiming at voltage abnormality caused by heavy overload, unreasonable gear and three-phase load unbalance, a distribution network weak point analysis model characteristic library is built based on data of three-phase voltage, current, load and the like, a random forest classification model is utilized to obtain characteristic importance of the distribution network weak point analysis model, and characteristics in the characteristic library are screened. Afterwards, a distribution network weak point analysis model is built based on classification algorithms such as Logistic regression (Logistic regression, which is a classification model in machine learning), random forest, GBDT (Gradient Boosting Decision Tree, namely gradient lifting decision), XGBoost (Exterme Gradient Boosting, namely limit gradient lifting), lightGBM (Light Gradient Boosting Machine, namely a distributed gradient lifting framework based on a decision tree algorithm), fusion model and the like, distribution transformation characteristics are input into the model, whether distribution transformation is a weak point or not, and index values such as accuracy, recall rate, precision rate, F1 value (namely the harmonic average of the accuracy and recall rate) and the like of each model can be calculated to evaluate the model, and performance and classification capability of the model are measured.

The present invention will be described in detail with reference to the following examples.

Example 1

According to an embodiment of the present invention, there is provided an embodiment of a method of monitoring a power distribution network, it being noted that the steps shown in the flowchart of the figures may be performed in a computer system, such as a set of computer executable instructions, and that although a logical sequence is shown in the flowchart, in some cases the steps shown or described may be performed in a different order than what is shown or described herein.

Fig. 1 is a flowchart of an alternative method for monitoring a power distribution network according to an embodiment of the present invention, as shown in fig. 1, the method including the steps of:

step S101, determining a distribution transformer set included in the target power distribution network, where the distribution transformer set includes: at least one distribution transformer.

Step S102, collecting distribution transformer data of each distribution transformer, and determining distribution transformer characteristics corresponding to the distribution transformers based on the distribution transformer data.

Step S103, inputting the distribution characteristics into a preset analysis model, and outputting the distribution state of the distribution transformer, wherein the preset analysis model is an analysis model trained based on historical distribution data, and the distribution state comprises: normal state, abnormal state.

Step S104, under the condition that the distribution transformer states of all distribution transformers are normal, determining that the target distribution network operates normally.

Through the steps, the distribution transformer set contained in the target distribution network can be determined, distribution transformer data of each distribution transformer are collected, distribution transformer characteristics corresponding to the distribution transformers are determined based on the distribution transformer data, the distribution transformer characteristics are input into a preset analysis model, distribution transformer states of the distribution transformers are output, and the normal operation of the target distribution network is determined under the condition that the distribution transformer states of all the distribution transformers are normal. In the embodiment of the invention, the distribution transformer related to the target distribution network can be determined first, then the distribution transformer data of each distribution transformer are collected, the corresponding distribution transformer characteristics are determined according to the distribution transformer data, then the distribution transformer characteristics of each distribution transformer are analyzed by adopting a preset analysis model to obtain the distribution transformer state of the distribution transformer, if the distribution transformer states of all the distribution transformers are normal states, the normal operation of the target distribution network is determined, otherwise, the abnormal distribution transformers are early-warned, so that the monitoring efficiency is improved, the abnormal distribution transformers can be early-warned in time, the normal operation of the distribution network is effectively ensured, and the technical problem that the monitoring efficiency of the distribution network is low and the abnormal distribution transformers cannot be early-warned in time in the related art is solved.

Embodiments of the present invention will be described in detail with reference to the following steps.

In the embodiment of the invention, analysis can be performed based on the voltage fluctuation record list of the distribution transformer, and main abnormality reasons comprise the following categories: the fault of the equipment body and the accessory parts, the fault of the secondary line, the fault of the line, heavy overload, unreasonable gear, unbalanced three-phase load and the like, wherein the fault of the equipment body and the accessory parts has a plurality of influencing factors, and the fault cause is difficult to be positioned by exploring the voltage data fluctuation before and after the fault of the equipment; the abnormal acquired data caused by the fault of the secondary wiring can be solved by fastening wiring afterwards; the voltage can be automatically recovered after the fault of the line is eliminated due to the abnormality caused by the fault of the line, so that the three types of distribution and transformation voltage abnormal changes are not suitable for early warning by analyzing voltage and current load data. The voltage abnormality caused by heavy overload, unreasonable gear and unbalanced three-phase load can be analyzed based on data such as historical three-phase voltage, current and load, so that the embodiment performs data analysis on the voltage abnormality caused by heavy overload, unreasonable gear and unbalanced three-phase load, and inputs the acquired data into a distribution network weak point analysis model for learning so as to early warn the distribution weak point.

Optionally, the monitoring method further comprises: collecting historical distribution data of each preset distribution transformer, wherein the preset distribution transformers are distribution transformers with overvoltage abnormal movements; and constructing an analysis model feature library based on all the historical configuration data.

In the embodiment of the invention, the front-stage data characteristics of the distribution transformer with voltage fluctuation (namely, the preset distribution transformer is a distribution transformer with overvoltage fluctuation) (namely, the historical distribution data of each preset distribution transformer is collected, for example, the data can be collected based on an electricity consumption information collection system, a PMS system (power production management system, namely, an electric power production management system), a marketing system, an external weather system and the like), then the distribution of the voltage abnormality conditions of the distribution transformer with different weather and different months is analyzed according to the historical distribution data, and the number of 7-day abnormality times, the number of 15-day abnormality times, the number of 7-day three-phase unbalance times, the number of 7-day heavy overload times and the like before the distribution transformer with voltage fluctuation are statistically analyzed (namely, an analysis model characteristic library is constructed based on all the historical distribution data).

In the embodiment of the invention, the preset distribution transformer is a distribution transformer with overvoltage fluctuation, wherein the voltage fluctuation is caused by heavy overload, unreasonable gear or unbalanced three-phase load.

Optionally, the step of constructing an analysis model feature library based on all the historical configuration data includes: filtering noise of the historical configuration data to obtain target historical configuration data; dividing the target historical transformer substation data into a plurality of historical transformer substation data associated with the preset dimension based on the preset dimension set; determining a numerical characteristic and a frequency characteristic based on the historical substation data, wherein the numerical characteristic is an index value characteristic obtained by calculation according to the historical substation data, and the frequency characteristic is a numerical value characteristic obtained by statistics according to the historical substation data; and constructing an analysis model feature library based on all the numerical features and all the frequency features.

In the embodiment of the invention, a distribution network weak point analysis model feature library (namely an analysis model feature library) is constructed according to collected historical distribution transformer data, and specifically comprises the following steps: firstly, various voltage abnormal characteristics are extracted and noise is filtered (namely noise of historical distribution data is filtered to obtain target historical distribution data), and conversion from qualitative description to quantitative calculation of data characteristics is realized. And then calculating numerical characteristics and frequency characteristics from data dimensions (namely preset dimensions) of current, voltage, load, distribution network architecture, temperature, humidity, weather and the like according to the target historical distribution data (namely dividing the target historical distribution data into a plurality of historical distribution sub-data associated with the preset dimensions based on a preset dimension set, and determining the numerical characteristics and the frequency characteristics based on the historical distribution sub-data). The numerical characteristics are index value characteristics obtained by calculation according to historical configuration transformer sub-data, and comprise voltage standard deviation, current average value, three-phase unbalance rate, load rate, temperature average value and the like; the frequency characteristics are obtained according to statistics of historical distribution substation data, and the frequency characteristics comprise distribution voltage abnormal times, current 0 times, overload times, abnormal times of transformers under the lines to which the distribution substation belongs, rain times and the like. And then, constructing an analysis model feature library according to all the numerical features and all the frequency features.

FIG. 2 is a schematic diagram of an optional distribution network weak point analysis model feature library according to an embodiment of the present invention, as shown in FIG. 2, a data source for collecting data may be a PMS, marketing, mining, weather, etc., after collecting data from these data sources, the data may be divided according to dimensions (for example, dimensions of current, voltage, load, distribution network architecture, temperature, humidity, weather, etc.), where based on the data associated with the voltage dimensions, a mean, a maximum, a minimum, a median, an extremely poor, a variance, a standard deviation, a mean of two-phase low voltage and an extremely poor, a maximum, a minimum, a mean of absolute value average of voltage differences, an extremely poor, numerical features of average of voltage front-back differential fluctuation, a maximum, a minimum, etc., and a number of 0/empty times of voltage, a number of voltage less than 198, a frequency feature of abnormal times of distribution voltage of about 7/15 days, etc. may be determined; the numerical characteristics such as the mean value, the maximum value, the minimum value, the variance, the standard deviation, the three-phase unbalance rate and the like and the frequency characteristics such as the current of 0/empty frequency and the three-phase unbalance frequency of approximately 7/15 days can be determined based on the data related to the current dimension; the numerical characteristics such as the load rate and the like can be determined based on the data related to the load dimension, and the frequency characteristics such as the idle load/light load/overload frequency and the like in the period of approximately 7/15 days; based on the data related to the distribution structure dimension, the frequency characteristics such as the abnormal frequency of the transformation under the line to which the distribution transformer belongs in the period of approximately 7/15 days can be determined; based on the data related to the humidity dimension, the numerical characteristics such as the mean value, the maximum value, the minimum value and the like can be determined; based on the data related to the temperature dimension, the numerical characteristics such as the mean value, the maximum value, the minimum value and the like, and the frequency characteristics such as the frequency that the temperature is less than 0 degree in the near 7/15 days and the frequency that the temperature is greater than 30 degrees in the near 7/15 days can be determined; frequency characteristics such as the wind/rain/snow times of approximately 7/15 days can be determined based on the data associated with the meteorological dimensions.

Optionally, after constructing the analysis model feature library based on all the historical configuration data, the method further comprises: constructing a random forest classification model, wherein the random forest classification model comprises: n decision trees, wherein N is a positive integer; inputting all data features in the analysis model feature library into a random forest classification model, and outputting the contribution value of each data feature to each decision tree, wherein the data features are numerical features or frequency features; determining an average contribution value of the data feature based on the contribution value of the data feature to each decision tree; and sorting all the average contribution values to obtain a sorting result, and selecting target data features indicated by the average contribution values in a preset range based on the sorting result to obtain a target data feature set.

In the embodiment of the invention, the characteristics of the distribution network weak point analysis model are also required to be screened, the characteristics of the characteristic importance degree can be output by utilizing the random forest classification model (namely, according to how much the characteristics make contributions on each tree in the random forest, the average value is taken, and then the contribution sizes among the characteristics are compared), the characteristic importance degree of the distribution network weak point analysis model is obtained, and the characteristics with higher characteristic importance degree are extracted, specifically: a random forest classification model may be constructed (the random forest classification model includes N (positive integer) decision trees), then all data features (which may be numerical features or frequency features) in the analysis model feature library are input to the random forest classification model for feature importance analysis to obtain a contribution value of each data feature to each decision tree, then a corresponding average contribution value is calculated according to the contribution value of each data feature to each decision tree, all average contribution values are ranked to obtain a ranking result, and a target data feature indicated by the average contribution value within a preset range (for example, the first 100 bits) may be selected according to the ranking result, so as to obtain a target data feature set (that is, a feature with a larger average contribution value may be selected as a training data feature of the distribution network weak point analysis model (i.e., the classification model)).

In the embodiment of the invention, the random forest is a classifier comprising a plurality of decision trees, and the output class is the mode of the class output by the decision trees. Each decision tree is a classifier, for an input sample, N trees have N classification results, and random forests integrate all classification voting results, designating the class with the highest voting frequency as the final output.

In the embodiment of the invention, the information, entropy and information gain are the basis for determining the feature selection sequence when the decision tree classifies by using the features. If a collection of classified things can be divided into multiple categories, then the information for a certain category can be defined as follows:

I(X＝xi)＝-log ₂ p(x _i )；

wherein I (x) is information representing a random variable, p (x) _i ) Meaning when x _i Probability of occurrence, x _i Representing a certain random variable.

Entropy is used to measure uncertainty, when the greater the entropy, x=x _i The greater the uncertainty of (c), the smaller the opposite. For classification problems in machine learning, the greater the entropy, i.e., the greater the uncertainty of this class, and vice versa.

The information gain is an index used to select a feature in the decision tree algorithm, the greater the information gain, the better the selectivity of this feature.

Optionally, the step of constructing a random forest classification model includes: sampling is carried out for N times from an analysis model feature library based on a preset sampling strategy, and N training sample sets are obtained; training the decision tree by adopting a training sample set until all node attributes of each node in the decision tree belong to the same type; and constructing a random forest classification model based on the trained decision tree.

In the embodiment of the invention, the construction process of the random forest classification model is as follows:

(1) Taking m samples from the original training set by using random back sampling, and performing n_tree sampling altogether to generate n_tree training sets (namely, performing N times of sampling from an analysis model feature library based on a preset sampling strategy to obtain N training sample sets, wherein the preset sampling strategy refers to taking a preset number (such as m) samples from each random back sampling).

(2) For the n_tree training sets, n_tree decision tree models are trained (i.e., the decision tree is trained using a training sample set) respectively.

(3) For a single decision tree model, assuming that the number of training sample features is n, the best feature can be selected for splitting according to indexes such as information gain and the like during each splitting.

(4) Each decision tree splits in this way until all training examples for that node belong to the same class (i.e., until all node attributes for each node in the decision tree are of the same type). Pruning is not required in the splitting process of the decision tree.

(5) And forming a random forest by the generated multiple decision trees, and voting to determine a final classification result according to the multiple tree classifiers (namely, constructing a random forest classification model based on the trained decision trees).

In the embodiment of the invention, all the features of the distribution network weak point analysis model feature library are input into a random forest model for training, model parameters are continuously adjusted to continuously improve the accuracy of the model so as to acquire the feature importance of the random forest and extract the features with higher feature importance, and table 1 is an optional feature importance display result, as shown in table 1:

TABLE 1

Optionally, after selecting the target data feature indicated by the average contribution value within the preset range based on the sorting result, obtaining the target data feature set, the method further includes: determining a set of classification algorithms, wherein the set of classification algorithms comprises: a plurality of classification algorithms, each classification algorithm corresponding to a classification frame; dividing the target data characteristic set into a training set and a testing set based on a preset proportion; training a classification frame by using a training set until a loss value determined by a loss function corresponding to the classification frame is smaller than a preset loss threshold value, and obtaining an initial classification model, wherein the loss function is constructed based on a predicted value and a labeling value, the predicted value is a value output by the initial classification model, and the labeling value is a value labeled for each target data characteristic in advance.

In the embodiment of the invention, a distribution network weak point analysis model (namely a preset analysis model) needs to be constructed according to the screened target data characteristic set, and the method specifically comprises the following steps: a set of classification algorithms may be first, the set of classification algorithms comprising: a plurality of classification algorithms (e.g., logistic regression, random forest, GBDT, XGBoost, lightGBM, fusion model, etc.) each correspond to a classification framework to build a distribution network vulnerability analysis model based on these classification algorithms. Then, based on a preset proportion (for example, 8:2), dividing the target data feature set into a training set and a testing set (namely, after feature screening is carried out on data such as voltage data, current data and the like before the distribution transformer which is subject to the abnormality, the screened feature data is divided into the training set and the testing set), then training a classification frame by adopting the training set until a loss value determined by a loss function corresponding to the classification frame is smaller than a preset loss threshold value (which can be set according to actual conditions), so as to obtain an initial classification model, wherein the loss function is constructed based on a predicted value and a labeling value, the predicted value is a value output by the initial classification model, and the labeling value is a value labeled for each target data feature in advance (for example, a corresponding value can be labeled for each feature manually in advance).

Taking a Logistic regression model as an example, how to determine parameters during training is analyzed.

For the Logistic regression model, the statistical method used in analyzing the classification variables is a Log-linear model (Log-linear-mode 1). The analysis model of the weak point of the distribution network researches whether the distribution transformer is a weak point or not, and can be divided into two types of yes and no, and the two types are called binary variables. When one binary variable in the logarithmic model is taken as a dependent variable, the logarithmic linear model becomes a Logistic regression model.

Assuming that Y is a random variable and obeys two-point distribution, when the formulation becomes a weak point, the value of Y is 1, and when the formulation is not a weak point, the value of Y is 0.

Assuming independent variables affecting experimental resultsIs X= (X) ₁ ，X ₂ ，…，X _p ) Given X, the probability P of y=1 is:

wherein beta is ₀ ，β ₁ ，β ₂ ，…，β _p As a parameter, the above formula is a Logistic function, which has an S distribution.

FIG. 3 is a schematic diagram of an S-plot of an alternative Logistic function according to an embodiment of the present invention, as shown in FIG. 3, with w on the abscissa (from- +to+), and P on the ordinate (from 0 to 1), showing the values of P when w takes different values, and when w=w ₀ ，p＝0.5。

Order theIt can be seen from fig. 3 that when w goes to minus infinity,/is >When w goes to positive infinity, ++>The range of values for the Logistic function varies from 0 to 1, regardless of any value taken by w. This property of the Logistic function ensures that the probability estimated by the Logistic model is not greater than 1 or less than 0. At the same time, the shape of the function is suitable for researching probability, when w moves from minus infinity to right, the function value is slowly increased firstly and is close to w ₀ The time begins to increase rapidly, the increasing speed then begins to slow down gradually, and finally the function value tends to be 1 when w tends to be positive to infinity. The S-curve of the Logistic function shows that the contribution of w varies for a case with a very small value of w, and then the corresponding probability increases very rapidly in the intermediate phase, but after the value of w increases to a certain extent, the probability remains at a nearly constant level, which means that w is at P (y=1|X) is closer to 0 or 1 than when P (y= 1|X) is in the intermediate stage.

From the definition of Logistic regression, the probability pi of a formulation becoming a weak point is:

therefore, the probability that the strain is not a weak point is:

it can be derived that

/>

Ratio ofFor the advantage ratio of the event, the log transformation of the above formula is as follows:

The logarithm of the dominance ratio is called the Logit function (i.e., g (x ₁ ，x ₂ ，…，x _n ) The Logit function transformation produces a parameter β ₀ ，β ₁ ，β ₂ ，…，β _p The problem of fitting the parameters of the Logistic regression model is converted to fitting the parameters of the linear model, and the embodiment can estimate the parameters using Maximum Likelihood (MLE).

Taking the XGBoost model as an example, how to train the model is analyzed.

XGBoost is an extensible system based on TreeBoosting (i.e., an ensemble learning algorithm) method, the base decision maker of XGBoost is serial, and a loss function Taylor (i.e., an infinite series developed at a certain point to approach a function) needs to be developed to a second order, and an objective function needs to be constructed to be optimized, and for a quadratic function with respect to a leaf node predicted value, its minimum value can be taken at the symmetry axis.

Based on the characteristic of XGBoost, the prediction loss of the t-th tree in the XGBoost can be calculated, and for a certain node of the t-th tree, the reduction of the loss function after splitting can be calculated, if and only if the reduction of the loss function after splitting is smaller than a certain threshold value, the node is split, and after all nodes of all trees in the XGBoost are split, the training of the XGBoost model is completed.

Optionally, after training the classification frame by using the training set until the loss value determined by the loss function corresponding to the classification frame is smaller than the preset loss threshold value, obtaining the initial classification model, the method further includes: testing each initial classification model by adopting a test set to obtain a test result, wherein the test result comprises: presetting an index value; under the condition that the preset index value is smaller than the preset index threshold value, adjusting parameters corresponding to the initial classification model to obtain an adjusted initial classification model; continuing training the adjusted initial classification model by adopting the training set until the preset index value is greater than or equal to the preset index threshold value to obtain a target classification model; and sequencing the preset index values corresponding to all the target classification models, and characterizing the target classification model indicated by the maximum preset index value as a preset analysis model.

In the embodiment of the invention, a test set can be used for model prediction, the accuracy, recall, precision and F1 value of each model are output to evaluate the model, the performance and classification capacity of the model are measured, and then the model with the best performance and classification capacity is selected for subsequent analysis, specifically: each initial classification model may be tested using a test set to obtain test results (the test results include preset index values, e.g., accuracy, recall, precision, F1 values, etc.). If an initial classification model with the preset index value smaller than the preset index threshold value (which can be set according to the actual situation) exists, parameters corresponding to the initial classification model can be adjusted to obtain an adjusted initial classification model. And then, continuing training the adjusted initial classification model by adopting the training set until the preset index value is greater than or equal to the preset index threshold value so as to obtain the target classification model. And then, sorting the preset index values corresponding to all the target classification models, and characterizing the target classification model indicated by the maximum preset index value as a preset analysis model.

Table 2 is an alternative model evaluation result, as shown in table 2:

TABLE 2

FIG. 4 is a schematic diagram of an alternative distribution network vulnerability analysis model construction process according to an embodiment of the present invention, as shown in FIG. 4, the screened feature data may be divided into a training set (i.e., training set) (80%) and a validation set (20%), then training models (e.g., logistic regression, random forest, GBDT, XGBoost, lightGBM, fusion models of the foregoing model components, etc.) using the training set, and then validating the application model (i.e., trained model) using the validation set to achieve model assessment, where the assessment results include: accuracy, precision, recall, F1 values, confusion matrix, etc.

In the embodiment of the invention, a distribution transformer set used by a target power distribution network can be determined first, and the distribution transformer set comprises: at least one distribution transformer.

In the embodiment of the invention, the distribution transformer data of each distribution transformer can be collected from the power consumption information collection system, the PMS system, the marketing system, the external weather system and other systems, and then the distribution transformer characteristics are screened from the distribution transformer data according to the characteristics screened before the training model (namely, the distribution transformer characteristics corresponding to the distribution transformer are determined based on the distribution transformer data).

In the embodiment of the invention, the distribution characteristics are input into a preset analysis model (namely, an analysis model trained based on historical distribution data) for analysis, so as to obtain the distribution state (normal state or abnormal state) of the distribution transformer.

In the embodiment of the invention, if the distribution transformer states of all the distribution transformers are normal states, the normal operation of the target distribution network can be determined, and if the distribution transformers in abnormal states exist, the alarm can be given out in time to remind related personnel to overhaul so as to avoid the problem of the operation of the distribution network.

The following detailed description is directed to alternative embodiments.

Fig. 5 is a schematic diagram of an alternative classification algorithm-based distribution network weak point analysis method according to an embodiment of the present invention, as shown in fig. 5, the classification algorithm-based distribution network weak point analysis method includes the following steps:

(1) And searching the characteristics of the early data of the distribution transformer with abnormal movement to analyze the distribution of the voltage abnormal conditions of the distribution transformer with different weather and different months, and counting the abnormal times of the distribution transformer with abnormal movement of the voltage, the three-phase unbalance times of the distribution transformer with abnormal movement of the voltage, the heavy overload times of the distribution transformer with abnormal movement of the distribution transformer with different months, the heavy overload times of the distribution transformer with abnormal movement of the voltage, the distribution transformer with abnormal movement of the distribution transformer with different months, the distribution transformer with abnormal voltage.

(2) And constructing a distribution network weak point analysis model feature library, and extracting various voltage abnormal features and filtering noise to realize the conversion of data features from qualitative description to quantitative calculation. Based on the data acquired from the electricity consumption information acquisition system, the PMS system, the marketing system, the external weather system and other systems, numerical characteristics and frequency characteristics can be calculated from data dimensions such as current, voltage, load, distribution network architecture, temperature, humidity, weather and the like, wherein the numerical characteristics comprise voltage standard deviation, current average value, three-phase unbalance rate, load rate, temperature average value and the like; the frequency characteristics comprise the abnormal times of the distribution and transformation voltage, the times of current being 0, the overload times, the abnormal times of transformers under the lines to which the distribution and transformation belongs, the raining times and the like.

(3) The feature screening of the distribution network weak point analysis model can utilize the feature importance of the random forest classification model to output (i.e. based on how much each feature contributes to each tree in the random forest, average value is taken, and then the contribution between the features is compared), all the features (numerical features and frequency features) of the distribution network weak point analysis model feature library are input into the random forest model for training, model parameters are continuously adjusted to continuously improve the accuracy of the model, the feature importance of the distribution network weak point analysis model is obtained, and the feature with larger feature importance is extracted.

(4) The method comprises the steps of constructing and evaluating a distribution network weak point analysis model, dividing the data into a training set and a test set based on data such as voltage, current and the like before distribution transformation with abnormal movement, then respectively constructing the distribution network weak point analysis model based on classification algorithms such as Logistic regression, random forest, GBDT, XGBoost, lightGBM, fusion model and the like, inputting the training set data into the model for training, carrying out model prediction by using the test set, and outputting accuracy, recall rate, precision rate and F1 value of each model to measure the performance and classification capability of the model.

According to the embodiment of the invention, analysis is carried out based on distribution and transformation voltage abnormal record data, the abnormal voltage early-stage data characteristics of historical abnormal distribution transformation are analyzed aiming at voltage abnormality caused by heavy overload, unreasonable gear and three-phase load unbalance, a distribution network weak point analysis model characteristic library is constructed based on three-phase voltage, current, load and other data, then a random forest classification model is utilized to obtain the characteristic importance of the distribution network weak point analysis model, characteristic library characteristics are screened, distribution network weak point analysis models are built based on classification algorithms such as Logistic regression, random forest, GBDT, XGBoost, lightGBM and fusion model respectively, the screened characteristics are input into the model, whether the output distribution transformation is weak point or not, the accuracy rate, recall rate, precision rate and F1 value of each model can be calculated, the performance and classification capacity of the model are measured, and data support is provided for distribution network optimization.

The following describes in detail another embodiment.

Example two

The monitoring device for a power distribution network provided in this embodiment includes a plurality of implementation units, each implementation unit corresponding to each implementation step in the first embodiment.

Fig. 6 is a schematic diagram of an alternative monitoring device for an electrical distribution network according to an embodiment of the present invention, as shown in fig. 6, the monitoring device may include: a first determination unit 60, an acquisition unit 61, an input unit 62, a second determination unit 63, wherein,

a first determining unit 60, configured to determine a distribution transformer set included in the target power distribution network, where the distribution transformer set includes: at least one distribution transformer;

the acquisition unit 61 is configured to acquire distribution data of each distribution transformer, and determine distribution characteristics corresponding to the distribution transformers based on the distribution data;

the input unit 62 is configured to input the distribution characteristics to a preset analysis model, and output a distribution state of the distribution transformer, where the preset analysis model is an analysis model trained based on historical distribution data, and the distribution state includes: normal state, abnormal state;

the second determining unit 63 is configured to determine that the target power distribution network is operating normally when all distribution transformer distribution states are normal states.

The above-mentioned monitoring device may determine, through the first determining unit 60, a distribution transformer set included in the target distribution network, collect distribution transformer data of each distribution transformer through the collecting unit 61, determine distribution transformer characteristics corresponding to the distribution transformers based on the distribution transformer data, input the distribution transformer characteristics to a preset analysis model through the input unit 62, output distribution transformer distribution states, and determine, through the second determining unit 63, that the target distribution network operates normally under the condition that all distribution transformer distribution states are normal states. In the embodiment of the invention, the distribution transformer related to the target distribution network can be determined first, then the distribution transformer data of each distribution transformer are collected, the corresponding distribution transformer characteristics are determined according to the distribution transformer data, then the distribution transformer characteristics of each distribution transformer are analyzed by adopting a preset analysis model to obtain the distribution transformer state of the distribution transformer, if the distribution transformer states of all the distribution transformers are normal states, the normal operation of the target distribution network is determined, otherwise, the abnormal distribution transformers are early-warned, so that the monitoring efficiency is improved, the abnormal distribution transformers can be early-warned in time, the normal operation of the distribution network is effectively ensured, and the technical problem that the monitoring efficiency of the distribution network is low and the abnormal distribution transformers cannot be early-warned in time in the related art is solved.

Optionally, the monitoring device further comprises: the first acquisition module is used for acquiring historical distribution data of each preset distribution transformer, wherein the preset distribution transformers are distribution transformers with overvoltage abnormal movements; the first construction module is used for constructing an analysis model feature library based on all the historical configuration data.

Optionally, the first building module includes: the first filtering sub-module is used for filtering noise of the historical configuration data to obtain target historical configuration data; the first dividing sub-module is used for dividing the target historical allocation data into a plurality of historical allocation sub-data associated with the preset dimension based on the preset dimension set; the first determining sub-module is used for determining numerical characteristics and frequency characteristics based on the historical substation data, wherein the numerical characteristics are index value characteristics obtained by calculation according to the historical substation data, and the frequency characteristics are numerical value characteristics obtained by statistics according to the historical substation data; and the first construction submodule is used for constructing an analysis model feature library based on all the numerical features and all the frequency features.

Optionally, the monitoring device further comprises: the second construction module is used for constructing a random forest classification model after constructing an analysis model feature library based on all the historical configuration data, wherein the random forest classification model comprises the following components: n decision trees, wherein N is a positive integer; the first input module is used for inputting all data features in the analysis model feature library into the random forest classification model and outputting the contribution value of each data feature to each decision tree, wherein the data features are numerical features or frequency features; the first determining module is used for determining an average contribution value of the data feature based on the contribution value of the data feature to each decision tree; the first sorting module is used for sorting all the average contribution values to obtain a sorting result, and selecting target data features indicated by the average contribution values in a preset range based on the sorting result to obtain a target data feature set.

Optionally, the second building block comprises: the first sampling submodule is used for sampling N times from the analysis model feature library based on a preset sampling strategy to obtain N training sample sets; the first training submodule is used for training the decision tree by adopting the training sample set until all node attributes of each node in the decision tree belong to the same type; and the second construction submodule is used for constructing a random forest classification model based on the trained decision tree.

Optionally, the monitoring device further comprises: the second determining module is configured to determine a classification algorithm set after selecting, based on the sorting result, a target data feature indicated by an average contribution value within a preset range to obtain a target data feature set, where the classification algorithm set includes: a plurality of classification algorithms, each classification algorithm corresponding to a classification frame; the first dividing module is used for dividing the target data characteristic set into a training set and a testing set based on a preset proportion; the first training module is used for training the classification frame by adopting the training set until the loss value determined by the loss function corresponding to the classification frame is smaller than a preset loss threshold value, so as to obtain an initial classification model, wherein the loss function is constructed based on a predicted value and a labeling value, the predicted value is a value output by the initial classification model, and the labeling value is a value labeled for each target data characteristic in advance.

Optionally, the monitoring device further comprises: the first test module is used for testing each initial classification model by adopting a test set after training the classification frame by adopting the training set until the loss value determined by the loss function corresponding to the classification frame is smaller than a preset loss threshold value to obtain the initial classification model, and obtaining a test result, wherein the test result comprises: presetting an index value; the first adjusting module is used for adjusting parameters corresponding to the initial classification model under the condition that the preset index value is smaller than the preset index threshold value to obtain an adjusted initial classification model; the second training module is used for continuing training the adjusted initial classification model by adopting the training set until the preset index value is greater than or equal to the preset index threshold value to obtain a target classification model; the second sorting module is used for sorting preset index values corresponding to all the target classification models and characterizing the target classification model indicated by the maximum preset index value as a preset analysis model.

The monitoring device may further include a processor and a memory, where the first determining unit 60, the collecting unit 61, the input unit 62, the second determining unit 63, etc. are stored as program units, and the processor executes the program units stored in the memory to implement corresponding functions.

The processor includes a kernel, and the kernel fetches a corresponding program unit from the memory. The kernel can be set to one or more, and the normal operation of the target distribution network is determined by adjusting the kernel parameters under the condition that the distribution transformer states of all distribution transformers are normal.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), which includes at least one memory chip.

The present application also provides a computer program product adapted to perform, when executed on a data processing device, a program initialized with the method steps of: the distribution transformer set contained in the target distribution network is determined, distribution transformer data of each distribution transformer are collected, distribution transformer characteristics corresponding to the distribution transformers are determined based on the distribution transformer data, the distribution transformer characteristics are input into a preset analysis model, distribution transformer states of the distribution transformers are output, and the normal operation of the target distribution network is determined under the condition that the distribution transformer states of all the distribution transformers are normal.

According to another aspect of the embodiment of the present invention, there is also provided a computer readable storage medium, where the computer readable storage medium includes a stored computer program, and when the computer program runs, the device on which the computer readable storage medium is located is controlled to execute the method for monitoring the power distribution network.

According to another aspect of the embodiment of the present invention, there is also provided an electronic device, including one or more processors and a memory, where the memory is configured to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method for monitoring a power distribution network described above.

Fig. 7 is a block diagram of a hardware structure of an electronic device (or a mobile device) for a monitoring method of a power distribution network according to an embodiment of the present invention. As shown in fig. 7, the electronic device may include one or more processors 702 (shown in fig. 7 as 702a, 702b, 702 n), a memory 704 for storing data (the processor 702 may include, but is not limited to, a microprocessor MCU, or a programmable logic device FPGA, etc.). In addition, the method may further include: a display, an input/output interface (I/0 interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of the I/0 interface), a network interface, a keyboard, a power supply, and/or a camera. It will be appreciated by those of ordinary skill in the art that the configuration shown in fig. 7 is merely illustrative and is not intended to limit the configuration of the electronic device described above. For example, the electronic device may also include more or fewer components than shown in FIG. 7, or have a different configuration than shown in FIG. 7.

The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.

In the foregoing embodiments of the present invention, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.

In the several embodiments provided in the present application, it should be understood that the disclosed technology content may be implemented in other manners. The above-described embodiments of the apparatus are merely exemplary, and the division of the units, for example, may be a logic function division, and may be implemented in another manner, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some interfaces, units or modules, or may be in electrical or other forms.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-0nly Memory (ROM), a random access Memory (RAM, random Access Memory), a removable hard disk, a magnetic disk, or an optical disk, or the like, which can store program codes.

The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.

Claims

1. A method for monitoring a power distribution network, comprising:

determining a distribution transformer set contained in a target power distribution network, wherein the distribution transformer set comprises: at least one distribution transformer;

collecting distribution data of each distribution transformer, and determining distribution characteristics corresponding to the distribution transformers based on the distribution data;

inputting the distribution characteristics into a preset analysis model, and outputting a distribution state of the distribution transformer, wherein the preset analysis model is an analysis model trained based on historical distribution data, and the distribution state comprises: normal state, abnormal state;

and under the condition that the distribution transformer states of all the distribution transformers are the normal states, determining that the target distribution network normally operates.

2. The method of monitoring according to claim 1, further comprising:

Collecting historical distribution data of each preset distribution transformer, wherein the preset distribution transformers are distribution transformers with overvoltage abnormal movements;

and constructing an analysis model feature library based on all the historical distribution change data.

3. The method of monitoring of claim 2, wherein the step of constructing an analytical model feature library based on all of the historical transformation data comprises:

filtering noise of the historical configuration data to obtain target historical configuration data;

dividing the target historical allocation data into a plurality of historical allocation sub-data associated with a preset dimension based on a preset dimension set;

determining a numerical characteristic and a frequency characteristic based on the historical substation data, wherein the numerical characteristic is an index value characteristic obtained by calculation according to the historical substation data, and the frequency characteristic is a quantitative value characteristic obtained by statistics according to the historical substation data;

and constructing the analysis model feature library based on all the numerical features and all the frequency features.

4. The method of monitoring of claim 2, further comprising, after constructing an analytical model feature library based on all of the historical transformation data:

Constructing a random forest classification model, wherein the random forest classification model comprises: n decision trees, wherein N is a positive integer;

inputting all data features in the analysis model feature library into the random forest classification model, and outputting a contribution value of each data feature to each decision tree, wherein the data features are numerical features or frequency features;

determining an average contribution value of the data feature based on the contribution value of the data feature to each of the decision trees;

and sorting all the average contribution values to obtain a sorting result, and selecting target data features indicated by the average contribution values within a preset range based on the sorting result to obtain a target data feature set.

5. The method of monitoring of claim 4, wherein the step of constructing a random forest classification model comprises:

sampling is carried out for N times from the analysis model feature library based on a preset sampling strategy, and N training sample sets are obtained;

training the decision tree by adopting the training sample set until all node attributes of each node in the decision tree belong to the same type;

And constructing the random forest classification model based on the trained decision tree.

6. The method according to claim 4, further comprising, after selecting the target data feature indicated by the average contribution value within a preset range based on the sorting result, obtaining a target data feature set:

determining a set of classification algorithms, wherein the set of classification algorithms comprises: a plurality of classification algorithms, each of which corresponds to a classification frame;

dividing the target data characteristic set into a training set and a testing set based on a preset proportion;

and training the classification frame by adopting the training set until a loss value determined by a loss function corresponding to the classification frame is smaller than a preset loss threshold value, so as to obtain an initial classification model, wherein the loss function is constructed based on a predicted value and a labeling value, the predicted value is a value output by the initial classification model, and the labeling value is a value labeled for each target data characteristic in advance.

7. The method according to claim 6, further comprising, after training the classification frame using the training set until a loss value determined by a loss function corresponding to the classification frame is less than a preset loss threshold, obtaining an initial classification model:

And testing each initial classification model by adopting the test set to obtain a test result, wherein the test result comprises: presetting an index value;

under the condition that the preset index value is smaller than a preset index threshold value, adjusting parameters corresponding to the initial classification model to obtain an adjusted initial classification model;

continuing training the adjusted initial classification model by adopting the training set until the preset index value is greater than or equal to a preset index threshold value to obtain a target classification model;

and sequencing the preset index values corresponding to all the target classification models, and characterizing the target classification model indicated by the maximum preset index value as the preset analysis model.

8. A monitoring device for an electrical distribution network, comprising:

a first determining unit, configured to determine a distribution transformer set included in a target power distribution network, where the distribution transformer set includes: at least one distribution transformer;

the acquisition unit is used for acquiring distribution data of each distribution transformer and determining distribution characteristics corresponding to the distribution transformers based on the distribution data;

the input unit is used for inputting the distribution characteristics to a preset analysis model and outputting the distribution state of the distribution transformer, wherein the preset analysis model is an analysis model trained based on historical distribution data, and the distribution state comprises: normal state, abnormal state;

And the second determining unit is used for determining that the target power distribution network operates normally under the condition that the distribution transformer states of all the distribution transformers are the normal states.

9. A computer readable storage medium, characterized in that the computer readable storage medium comprises a stored computer program, wherein the computer program, when run, controls a device in which the computer readable storage medium is located to perform the method of monitoring a power distribution network according to any one of claims 1 to 7.

10. An electronic device comprising one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of monitoring the power distribution network of any of claims 1-7.