CN111881961A - Power distribution network fault risk grade prediction method based on data mining - Google Patents

Power distribution network fault risk grade prediction method based on data mining Download PDF

Info

Publication number
CN111881961A
CN111881961A CN202010690948.XA CN202010690948A CN111881961A CN 111881961 A CN111881961 A CN 111881961A CN 202010690948 A CN202010690948 A CN 202010690948A CN 111881961 A CN111881961 A CN 111881961A
Authority
CN
China
Prior art keywords
data
fault
prediction
risk level
power distribution
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
CN202010690948.XA
Other languages
Chinese (zh)
Inventor
周佳威
冒烨颖
杨启明
王小蕾
马骏昶
董晓峰
顾佳
方琪
杭泱
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Suzhou Power Supply Co of State Grid Jiangsu Electric Power Co Ltd
Original Assignee
Suzhou Power Supply Co of State Grid Jiangsu Electric Power Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Suzhou Power Supply Co of State Grid Jiangsu Electric Power Co Ltd filed Critical Suzhou Power Supply Co of State Grid Jiangsu Electric Power Co Ltd
Priority to CN202010690948.XA priority Critical patent/CN111881961A/en
Publication of CN111881961A publication Critical patent/CN111881961A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
    • Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Abstract

The invention relates to a power distribution network fault risk grade prediction method based on data mining, which comprises the following steps of: step 1: acquiring various types of original data related to the power distribution network, including historical data and prediction data; step 2: processing the acquired data to obtain a data table containing historical data and prediction data of the labeled fault risk level; and step 3: analyzing the characteristic attribute of the historical data in the data table; and 4, step 4: training a fault risk grade prediction model by using the extracted historical data of the characteristic attributes and based on prediction data to obtain a trained fault risk grade prediction model; and 5: and predicting the risk level of the power distribution network fault by using the trained fault risk level prediction model. The method and the device can reliably predict the fault risk level of the power distribution network in a period of time in the future, are beneficial to eliminating hidden dangers in advance, reduce loss, reduce fault rate and improve the reliability of the power grid.

Description

Power distribution network fault risk grade prediction method based on data mining
Technical Field
The invention relates to the technical field of power distribution network fault prediction, in particular to a power distribution network fault risk level prediction method based on data mining.
Background
The power distribution network is used as the last ring of power transmission to users by the power grid, the connection with the users is the closest, the influence on normal power utilization of the users is more direct, and the faults of the power distribution network directly cause the power loss of the users so as to influence normal production and life. In order to avoid risks brought by power loss, the causes of power distribution network faults need to be well understood and known, the inherent rules of the power distribution network faults are mastered, meanwhile, the power distribution network faults can be predicted to a certain extent, hidden dangers are eliminated in advance in a targeted mode, enough first-aid repair materials are configured, and first-aid repair personnel are arranged to be on site to guard, so that losses brought by the power distribution network faults are reduced to the greatest extent, the fault rate of the power distribution network is reduced, and the reliability of the power distribution network is improved.
Disclosure of Invention
The invention aims to provide a method for predicting the fault risk level of a power distribution network more reliably.
In order to achieve the purpose, the invention adopts the technical scheme that:
a power distribution network fault risk grade prediction method based on data mining comprises the following steps:
step 1: data acquisition: acquiring various types of original data related to the power distribution network, wherein the original data comprises historical data and prediction data;
step 2: data processing: processing the acquired data to obtain a data table containing the historical data and the predicted data with the labeled fault risk level;
and step 3: analyzing the characteristic attribute of the data: analyzing the characteristic attributes of the historical data in the data table, and extracting the characteristic attributes of the historical data;
and 4, step 4: training a fault risk grade prediction model: establishing a fault risk grade prediction model, and training the fault risk grade prediction model by using the historical data with the extracted characteristic attributes and based on the prediction data to obtain a trained fault risk grade prediction model;
and 5: and (3) predicting the failure risk level: and predicting the risk level of the power distribution network fault by using the trained fault risk level prediction model.
The power distribution network fault risk level prediction method based on data mining takes a feeder line in the power distribution network as an analysis object.
In the step 1, the type of the original data includes three types of feeder line self attribute, feeder line operation attribute and feeder line surrounding environment attribute.
The step 2 comprises the following substeps:
substep 2-1: merging the data tables: integrating the original data by taking a feeder line as a related vehicle to obtain an original data table;
substep 2-2: data cleaning: cleaning the abnormal items in the original data table to obtain a processed data table;
substeps 2-3: and (3) fault risk grade marking: and respectively judging and marking the risk level of each fault according to the historical data in the processed data table.
In the substep 2-3, the number of times of feeder failures and the time of failure power failure in the unit time are used as the judgment criteria of the failure risk level to judge and mark the risk level of each failure.
In the step 3, a RelieF algorithm is adopted to analyze the characteristic attributes of the historical data in the data table, and the characteristic attributes of the historical data are extracted.
The step 3 comprises the following substeps:
substep 3-1: randomly extracting a fault record d from the data table, and searching a most adjacent fault record H with the same fault risk level as the fault record d and a most adjacent fault record M with the different fault risk level from the fault record d;
substep 3-2: for each characteristic attribute, determining whether the characteristic attribute is beneficial to the fault risk level of the fault record d according to the magnitude relation of the characteristic difference between the fault record d and the fault record H and the fault record M respectively, so as to adjust the weight of the characteristic attribute to the fault risk level of the fault record d; if the characteristic difference between the fault record d and the fault record H is smaller than the characteristic difference between the fault record d and the fault record M, the characteristic attribute is beneficial to the fault risk level of the fault record d, the weight of the characteristic attribute to the fault risk level of the fault record d is increased, and otherwise, the weight of the characteristic attribute to the fault risk level of the fault record d is reduced;
repeating the substep 3-1 and the substep 3-2 for T times, and iterating to obtain the weight of each characteristic attribute corresponding to each fault risk level.
In the step 4, an AdaBoost algorithm is adopted to train the fault risk level prediction model.
The step 4 comprises the following substeps:
substep 4-1: setting a series of same or different basic prediction algorithms, setting the initial weights of the basic prediction algorithms to be the same, and setting the initial weights of the historical data extracted with the characteristic attributes to be the same;
substep 4-2: selecting a first basic prediction algorithm as the currently selected basic prediction algorithm, and taking the historical data with the same initial weight of the extracted characteristic attributes as the current historical data;
substep 4-3: inputting current historical data into the currently selected basic prediction algorithm to obtain prediction data based on the currently selected basic prediction algorithm, and updating the weight of the currently calculated basic prediction algorithm according to the prediction data based on the currently selected basic prediction algorithm;
step 4-4: updating the weight of each piece of historical data according to whether the currently selected prediction data of the basic prediction algorithm is accurate or not;
and 4-5: judging whether a series of basic prediction algorithms are all updated in weight, if so, executing the step 4-6, otherwise, selecting the next basic prediction algorithm as the currently selected basic prediction algorithm, taking each piece of history data with the updated weight as the current history data, and returning to the step 4-3;
and 4-6: and constructing a trained fault risk level prediction model based on each basic prediction algorithm and the updated weight thereof, wherein the trained fault risk level prediction model is a weighted voting result of each basic prediction algorithm.
In the sub-step 4-3, the method of updating the weight of the currently calculated basic prediction algorithm based on the prediction data based on the currently selected basic prediction algorithm is: calculating the prediction error of the currently selected basic prediction algorithm
Figure BDA0002589337610000031
Wherein, thetat(i) A variable representing the correctness of the prediction of the ith data in the current historical data, theta, for the currently selected basic prediction algorithmt(i) 1 denotes prediction error, θt(i) When the classification is correct, m is the number of data in the current historical data, Dt(i) Weights before updating for the currently selected base prediction algorithm; recalculating updated weights for the currently selected base prediction algorithm
Figure BDA0002589337610000032
In sub-step 4-4, the method for updating the weight of each piece of history data includes: calculate each of saidUpdated weights of historical data
Figure BDA0002589337610000033
Wherein L ist(i) For prediction data based on the currently selected basic prediction algorithm, yiFor corresponding predicted data in said historical data, Zt+1Is a normalization factor.
Due to the application of the technical scheme, compared with the prior art, the invention has the following advantages: the method can reliably predict the fault risk level of the power distribution network in a period of time in the future, thereby helping a power grid operation maintenance unit to pertinently eliminate hidden dangers in advance, allocating enough first-aid repair materials, arranging the first-aid repair personnel to stay on spot to guard, reducing the loss caused by the power grid fault to the maximum extent, reducing the fault rate of the power distribution network and improving the reliability of the power distribution network.
Drawings
Fig. 1 is a schematic flow chart of a power distribution network fault risk level prediction method based on data mining.
Fig. 2 is a schematic flow diagram of a RelieF algorithm in the data mining-based power distribution network fault risk level prediction method of the present invention.
Fig. 3 is a schematic flow diagram of an AdaBoost algorithm in the data mining-based power distribution network fault risk level prediction method of the present invention.
Detailed Description
The invention will be further described with reference to examples of embodiments shown in the drawings to which the invention is attached.
The first embodiment is as follows: as shown in fig. 1, a power distribution network fault risk level prediction method based on data mining includes the following steps:
step 1: and (6) acquiring data.
The method comprises the steps of obtaining various types of related original data of the power distribution network, wherein data sources comprise a power distribution production management system, a power distribution automation system, a power distribution network geographic information system, an intelligent public distribution transformer monitoring system, a power distribution network intelligent auxiliary management and control system, a marketing business management system, a power utilization information acquisition system, an enterprise resource management system, an external meteorological system and the like in a power distribution network company.
The power distribution network fault risk grade prediction method based on data mining takes a feeder line in a power distribution network as an analysis object, namely a basic research unit. The types of the acquired original data include feeder line self attribute, feeder line operation attribute and feeder line surrounding environment attribute, which are specifically shown in the following table:
Figure BDA0002589337610000041
the intrinsic attribute of the feeder line is defined as the attribute of the feeder line, the attribute of the feeder line changing along with the operation of time is defined as the operation attribute of the feeder line, and the attribute of the physical environment where the feeder line is located is defined as the attribute of the surrounding environment, for example, the weather attribute belongs to the attribute of the surrounding environment.
The acquired original data comprises historical data and prediction data according to time, the historical data can be used for training a prediction model subsequently, and the prediction data is used for predicting the power distribution network fault.
The self attribute of the feeder line can be regarded as inherent unchanged within a period of time, and no prediction data exists; the prediction data of the feeder line operation attribute is operation prediction data obtained by using a conventional power load prediction method, such as a regression analysis method or an elastic coefficient method; the ambient environment attribute uses weather forecast information as prediction data.
Step 2: and (6) data processing.
And processing the acquired data to obtain a data table containing historical data and prediction data marking the fault risk level.
Step 2 comprises the following substeps:
substep 2-1: and merging the data tables.
And integrating the original data of different data sources by taking the feeder line as a related vehicle to obtain an original data table.
Substep 2-2: and (6) data cleaning.
And cleaning the abnormal entries in the original data table to obtain a processed data table. Specifically, the method comprises the steps of removing repeated values, manually supplementing empty values and manually correcting abnormal values.
Substeps 2-3: and (4) marking the fault risk level.
And respectively judging and marking the risk level of each fault according to the historical data in the processed data table, so as to obtain the historical data containing the marked fault risk level, and further obtain the data table containing the historical data and the prediction data marked with the fault risk level. In the substep, the number of feeder failures and the failure blackout time in unit time are used as the judgment criteria of the failure risk level to judge and mark the risk level of each failure, as shown in the following table:
risk rating Number of monthly failures Accumulated power failure time of month
1 ＝1 [0,60]
2 ≥2 [60,240]
3 ≥4 ≥240
And step 3: and analyzing the characteristic attribute of the data.
And analyzing the characteristic attributes of the historical data in the data table by adopting a RelieF algorithm, and extracting the characteristic attributes of the historical data.
As shown in fig. 2, step 3 comprises the following substeps:
substep 3-1: randomly extracting a fault record D from a training set D formed by a data table, searching a most adjacent fault record H with the same (same) fault risk level as the fault record D and a most adjacent fault record M with the different (different) fault risk level from the fault record D from the data table, recording the fault record H as nearhit, and recording the fault record M as nearmiss.
Substep 3-2: and for each characteristic attribute, determining whether the characteristic attribute is beneficial to the fault risk level of the fault record d according to the size relationship of the characteristic difference between the fault record d and the fault records H and M, so as to adjust the weight of the characteristic attribute to the fault risk level of the fault record d.
The specific rule is as follows: if the characteristic difference between the fault record d and the fault record H is smaller than the characteristic difference between the fault record d and the fault record M on a certain characteristic attribute, the characteristic attribute is beneficial to the fault risk level of the fault record d, and the weight of the characteristic attribute to the fault risk level of the fault record d should be increased; conversely, it is not beneficial to reduce the weight of the feature attribute to the failure risk level of the failure record d.
The weights of the final characteristic attributes can be obtained by iterating the process for T times, and the larger the weight is, the stronger the correlation between the characteristic attributes and the fault risk level is. And K samples are extracted from the same-class records and the different-class records, so that the RelieF algorithm for processing the multi-class problems is realized. Based on the above, repeating the substep 3-1 and the substep 3-2 for T times, the weights of the characteristic attributes corresponding to the fault risk levels can be obtained through iteration, and further the weight average value of the characteristic attributes (characteristic vectors) can be calculated to output the optimal fault feature set.
In the above scheme, the characteristic difference can be calculated by the following formula:
Figure BDA0002589337610000061
wherein diff (a, X, Y) is a difference value (i.e. a feature difference) between the vector X and the vector Y on the feature attribute a, the vector X and the vector Y respectively correspond to two fault records requiring feature difference calculation, such as a fault record d and a fault record H or a fault record d and a fault record M, max (a) is a maximum value of the feature attribute a, and min (a) is a minimum value of the feature attribute a.
The corresponding weight W can be updated and calculated for the characteristic attribute a in the training set Da', the calculation method is as follows:
Figure BDA0002589337610000062
wherein, WaIs the weight before updating the characteristic attribute a, K is the number of records nearest to the fault record d, T is the number of iterations, HkRepresenting the closest homogeneous record set, M, from the fault record dkRepresenting the heterogeneous record set nearest to the fault record d, wherein N (M) is the ratio of the number of M records in the total number of records, class (d) represents the record class to which the record d belongs, and N (class (d)) is the ratio of the number of samples of the record class to which the record d belongs in the total number of samples.
And 4, step 4: and (5) training a fault risk grade prediction model.
And establishing a fault risk level prediction model, and training the fault risk level prediction model by using the extracted historical data of the characteristic attributes and based on prediction data by adopting an AdaBoost algorithm to obtain the trained fault risk level prediction model.
As shown in fig. 3, step 4 comprises the following substeps:
substep 4-1: setting a series of same or different basic prediction algorithms, setting the initial weights of the basic prediction algorithms to be the same, and setting the initial weights of the historical data extracted with the characteristic attributes to be the same. The established fault risk level prediction model is obtained based on a series of basic prediction algorithms and their corresponding weights. In this example, the Cart decision Tree prediction algorithm was chosen.
Substep 4-2: and selecting the first basic prediction algorithm as the currently selected basic prediction algorithm, and taking the historical data with the same initial weight after the characteristic attributes are extracted as the current historical data.
Substep 4-3: inputting the current historical data into the currently selected basic prediction algorithm in the classifier to obtain prediction data based on the currently selected basic prediction algorithm, and updating the weight of the currently calculated basic prediction algorithm according to the prediction data based on the currently selected basic prediction algorithm.
In this substep, the method of updating the weight of the currently calculated basic prediction algorithm based on the prediction data based on the currently selected basic prediction algorithm is:
firstly, the prediction error theta of the currently selected basic prediction algorithm t is calculatedtThe calculation method comprises the following steps:
Figure BDA0002589337610000071
wherein, thetat(i) A variable representing the correctness of the prediction of the ith data in the current historical data, θ, for the currently selected basic prediction algorithmt(i) 1 denotes prediction error, θt(i) When the classification is correct, m is the number of data in the current historical data, Dt(i) The pre-update weights for the currently selected base prediction algorithm.
Then calculating the updated weight alpha of the currently selected basic prediction algorithm ttThe calculation method comprises the following steps:
Figure BDA0002589337610000072
step 4-4: and updating the weight of each piece of historical data according to whether the prediction data of the currently selected basic prediction algorithm is accurate or not.
In sub-step 4-4, the method for updating the weight of each piece of historical data comprises: calculating updated weight D of each piece of historical datat+1The calculation method comprises the following steps:
Figure BDA0002589337610000073
wherein L ist(i) For prediction data based on the currently selected underlying prediction algorithm, yiFor corresponding predictive data in the historical data, Zt+1To normalize the factor so that
Figure BDA0002589337610000074
After each basic prediction algorithm, the mispredicted recording weight will be increased and the correctly predicted recording weight will be decreased.
And 4-5: judging whether a series of basic prediction algorithms have updated weights, if so, executing the step 4-6, otherwise, selecting the next basic prediction algorithm as the currently selected basic prediction algorithm, taking each piece of historical data of which the weights are updated at the time as the current historical data, returning to the step 4-3, namely, taking the historical data of which the weights are updated at the time as the input of the next basic prediction algorithm, and circularly executing the steps.
And 4-6: and constructing a trained fault risk level prediction model based on each basic prediction algorithm and the updated weight thereof, wherein the trained fault risk level prediction model is a weighted voting result of each basic prediction algorithm.
And 5: and predicting the failure risk level.
And predicting the risk level of the power distribution network fault by using the trained fault risk level prediction model to obtain the fault risk level of the power distribution network in a period of time in the future. The prediction result of the fault risk level can help a power grid operation maintenance unit to pertinently eliminate hidden dangers in advance, configure enough first-aid repair materials, arrange the first-aid repair personnel to perform stationing and protection, reduce loss caused by power grid faults to the maximum extent, reduce the fault rate of a power distribution network and improve the reliability of the power grid.
The above embodiments are merely illustrative of the technical ideas and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the contents of the present invention and implement the present invention, and not to limit the protection scope of the present invention. All equivalent changes and modifications made according to the spirit of the present invention should be covered within the protection scope of the present invention.

Claims (10)

1. A power distribution network fault risk grade prediction method based on data mining is characterized by comprising the following steps: the power distribution network fault risk level prediction method based on data mining comprises the following steps:
step 1: data acquisition: acquiring various types of original data related to the power distribution network, wherein the original data comprises historical data and prediction data;
step 2: data processing: processing the acquired data to obtain a data table containing the historical data and the predicted data with the labeled fault risk level;
and step 3: analyzing the characteristic attribute of the data: analyzing the characteristic attributes of the historical data in the data table, and extracting the characteristic attributes of the historical data;
and 4, step 4: training a fault risk grade prediction model: establishing a fault risk grade prediction model, and training the fault risk grade prediction model by using the historical data with the extracted characteristic attributes and based on the prediction data to obtain a trained fault risk grade prediction model;
and 5: and (3) predicting the failure risk level: and predicting the risk level of the power distribution network fault by using the trained fault risk level prediction model.
2. The data mining-based power distribution network fault risk level prediction method of claim 1, characterized in that: the power distribution network fault risk level prediction method based on data mining takes a feeder line in the power distribution network as an analysis object.
3. The data mining-based power distribution network fault risk level prediction method of claim 2, characterized in that: in the step 1, the type of the original data includes three types of feeder line self attribute, feeder line operation attribute and feeder line surrounding environment attribute.
4. The data mining-based power distribution network fault risk level prediction method of claim 2, characterized in that: the step 2 comprises the following substeps:
substep 2-1: merging the data tables: integrating the original data by taking a feeder line as a related vehicle to obtain an original data table;
substep 2-2: data cleaning: cleaning the abnormal items in the original data table to obtain a processed data table;
substeps 2-3: and (3) fault risk grade marking: and respectively judging and marking the risk level of each fault according to the historical data in the processed data table.
5. The data mining-based power distribution network fault risk level prediction method of claim 4, wherein: in the substep 2-3, the number of times of feeder failures and the time of failure power failure in the unit time are used as the judgment criteria of the failure risk level to judge and mark the risk level of each failure.
6. The data mining-based power distribution network fault risk level prediction method of claim 2, characterized in that: in the step 3, a RelieF algorithm is adopted to analyze the characteristic attributes of the historical data in the data table, and the characteristic attributes of the historical data are extracted.
7. The data mining-based power distribution network fault risk level prediction method of claim 6, wherein: the step 3 comprises the following substeps:
substep 3-1: randomly extracting a fault record d from the data table, and searching a most adjacent fault record H with the same fault risk level as the fault record d and a most adjacent fault record M with the different fault risk level from the fault record d;
substep 3-2: for each characteristic attribute, determining whether the characteristic attribute is beneficial to the fault risk level of the fault record d according to the magnitude relation of the characteristic difference between the fault record d and the fault record H and the fault record M respectively, so as to adjust the weight of the characteristic attribute to the fault risk level of the fault record d; if the characteristic difference between the fault record d and the fault record H is smaller than the characteristic difference between the fault record d and the fault record M, the characteristic attribute is beneficial to the fault risk level of the fault record d, the weight of the characteristic attribute to the fault risk level of the fault record d is increased, and otherwise, the weight of the characteristic attribute to the fault risk level of the fault record d is reduced;
repeating the substep 3-1 and the substep 3-2 for T times, and iterating to obtain the weight of each characteristic attribute corresponding to each fault risk level.
8. The data mining-based power distribution network fault risk level prediction method of claim 2, characterized in that: in the step 4, an AdaBoost algorithm is adopted to train the fault risk level prediction model.
9. The data mining-based power distribution network fault risk level prediction method of claim 8, wherein: the step 4 comprises the following substeps:
substep 4-1: setting a series of same or different basic prediction algorithms, setting the initial weights of the basic prediction algorithms to be the same, and setting the initial weights of the historical data extracted with the characteristic attributes to be the same;
substep 4-2: selecting a first basic prediction algorithm as the currently selected basic prediction algorithm, and taking the historical data with the same initial weight of the extracted characteristic attributes as the current historical data;
substep 4-3: inputting current historical data into the currently selected basic prediction algorithm to obtain prediction data based on the currently selected basic prediction algorithm, and updating the weight of the currently calculated basic prediction algorithm according to the prediction data based on the currently selected basic prediction algorithm;
step 4-4: updating the weight of each piece of historical data according to whether the currently selected prediction data of the basic prediction algorithm is accurate or not;
and 4-5: judging whether a series of basic prediction algorithms are all updated in weight, if so, executing the step 4-6, otherwise, selecting the next basic prediction algorithm as the currently selected basic prediction algorithm, taking each piece of history data with the updated weight as the current history data, and returning to the step 4-3;
and 4-6: and constructing a trained fault risk level prediction model based on each basic prediction algorithm and the updated weight thereof, wherein the trained fault risk level prediction model is a weighted voting result of each basic prediction algorithm.
10. The data mining-based power distribution network fault risk level prediction method of claim 9, wherein: in the sub-step 4-3, the method of updating the weight of the currently calculated basic prediction algorithm based on the prediction data based on the currently selected basic prediction algorithm is: calculating the prediction error of the currently selected basic prediction algorithm
Figure FDA0002589337600000031
Wherein, thetat(i) A variable representing the correctness of the prediction of the ith data in the current historical data, theta, for the currently selected basic prediction algorithmt(i) 1 denotes prediction error, θt(i) When the classification is correct, m is the number of data in the current historical data, Dt(i) Weights before updating for the currently selected base prediction algorithm; recalculating updated weights for the currently selected base prediction algorithm
Figure FDA0002589337600000032
In sub-step 4-4, the method for updating the weight of each piece of history data includes: calculating the updated weight of each piece of historical data
Figure FDA0002589337600000033
Wherein L ist(i) For prediction data based on the currently selected basic prediction algorithm, yiFor corresponding predicted data in said historical data, Zt+1Is a normalization factor.
CN202010690948.XA 2020-07-17 2020-07-17 Power distribution network fault risk grade prediction method based on data mining Withdrawn CN111881961A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010690948.XA CN111881961A (en) 2020-07-17 2020-07-17 Power distribution network fault risk grade prediction method based on data mining

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010690948.XA CN111881961A (en) 2020-07-17 2020-07-17 Power distribution network fault risk grade prediction method based on data mining

Publications (1)

Publication Number Publication Date
CN111881961A true CN111881961A (en) 2020-11-03

Family

ID=73156045

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010690948.XA Withdrawn CN111881961A (en) 2020-07-17 2020-07-17 Power distribution network fault risk grade prediction method based on data mining

Country Status (1)

Country Link
CN (1) CN111881961A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112749904A (en) * 2021-01-14 2021-05-04 国网湖南省电力有限公司 Power distribution network fault risk early warning method and system based on deep learning
CN114372596A (en) * 2022-03-21 2022-04-19 广东电网有限责任公司佛山供电局 Power data analysis method and system based on data fusion
CN115334560A (en) * 2022-08-19 2022-11-11 中国联合网络通信集团有限公司 Method, device and equipment for monitoring base station abnormity and computer readable storage medium
CN115713239A (en) * 2022-12-07 2023-02-24 中国南方电网有限责任公司超高压输电公司广州局 Operation and maintenance risk prediction method and device, computer equipment and storage medium

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112749904A (en) * 2021-01-14 2021-05-04 国网湖南省电力有限公司 Power distribution network fault risk early warning method and system based on deep learning
CN112749904B (en) * 2021-01-14 2023-06-27 国网湖南省电力有限公司 Power distribution network fault risk early warning method and system based on deep learning
CN114372596A (en) * 2022-03-21 2022-04-19 广东电网有限责任公司佛山供电局 Power data analysis method and system based on data fusion
CN115334560A (en) * 2022-08-19 2022-11-11 中国联合网络通信集团有限公司 Method, device and equipment for monitoring base station abnormity and computer readable storage medium
CN115713239A (en) * 2022-12-07 2023-02-24 中国南方电网有限责任公司超高压输电公司广州局 Operation and maintenance risk prediction method and device, computer equipment and storage medium
CN115713239B (en) * 2022-12-07 2024-04-02 中国南方电网有限责任公司超高压输电公司广州局 Operation and maintenance risk prediction method, device, computer equipment and storage medium

Similar Documents

Publication Publication Date Title
CN111881961A (en) Power distribution network fault risk grade prediction method based on data mining
EP3043447B1 (en) Distribution transformer heavy loading and overloading mid-term and short-term pre-warning analytics model
CN103095494B (en) A kind of electric power telecommunication network risk assessment method
CN110807550B (en) Distribution transformer overload recognition and early warning method based on neural network and terminal equipment
CN111476435B (en) Charging pile load prediction method based on density peak value
CN106779129A (en) A kind of Short-Term Load Forecasting Method for considering meteorologic factor
CN111488896B (en) Distribution line time-varying fault probability calculation method based on multi-source data mining
CN206312210U (en) A kind of status assessing system of Distribution Network Equipment
CN105956788A (en) Dynamic management control method for cost of power transmission and transformation project
CN110969306B (en) Power distribution low-voltage transformer area load prediction method and device based on deep learning
Xue et al. Adaptive ultra-short-term wind power prediction based on risk assessment
CN111461921A (en) Load modeling typical user database updating method based on machine learning
CN114386753A (en) Equipment risk comprehensive analysis early warning method based on main transformer load condition
CN116502160A (en) Automatic electric quantity data acquisition system
CN110059938B (en) Power distribution network planning method based on association rule driving
CN116227637A (en) Active power distribution network oriented refined load prediction method and system
CN113205223A (en) Electric quantity prediction system and prediction method thereof
CN112508254B (en) Method for determining investment prediction data of transformer substation engineering project
CN112734274B (en) Low-carbon power grid operation leading influence factor mining and comprehensive evaluation method
CN115905319B (en) Automatic identification method and system for abnormal electricity fees of massive users
CN111967919A (en) System and method for analyzing electricity consumption behavior of residents based on autoregressive and adaptive boosting algorithm
CN117394529A (en) SCADA-based auxiliary decision method and system for main distribution network loop-closing reverse power supply control conditions
CN109829115B (en) Search engine keyword optimization method
CN104143117A (en) Method for extracting correlation coefficient between special loads and day loads of power grid
CN114676931B (en) Electric quantity prediction system based on data center technology

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WW01 Invention patent application withdrawn after publication
WW01 Invention patent application withdrawn after publication

Application publication date: 20201103