CN111191832A - Typhoon disaster power distribution network tower fault prediction method and system - Google Patents

Typhoon disaster power distribution network tower fault prediction method and system Download PDF

Info

Publication number
CN111191832A
CN111191832A CN201911352748.7A CN201911352748A CN111191832A CN 111191832 A CN111191832 A CN 111191832A CN 201911352748 A CN201911352748 A CN 201911352748A CN 111191832 A CN111191832 A CN 111191832A
Authority
CN
China
Prior art keywords
fault
tower
sample
samples
original
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.)
Pending
Application number
CN201911352748.7A
Other languages
Chinese (zh)
Inventor
郁琛
杨皖浙
倪明
谢云云
常康
吴涵
刘冰倩
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.)
State Grid Corp of China SGCC
Electric Power Research Institute of State Grid Fujian Electric Power Co Ltd
NARI Group Corp
Nari Technology Co Ltd
Original Assignee
State Grid Corp of China SGCC
Electric Power Research Institute of State Grid Fujian Electric Power Co Ltd
NARI Group Corp
Nari Technology 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 State Grid Corp of China SGCC, Electric Power Research Institute of State Grid Fujian Electric Power Co Ltd, NARI Group Corp, Nari Technology Co Ltd filed Critical State Grid Corp of China SGCC
Priority to CN201911352748.7A priority Critical patent/CN111191832A/en
Publication of CN111191832A publication Critical patent/CN111191832A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/067—Enterprise or organisation modelling
    • 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy 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
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Development Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Game Theory and Decision Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Primary Health Care (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Educational Administration (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a method and a system for predicting faults of a power distribution network tower in typhoon disasters, wherein real-time collected meteorological station data are input into a trained power distribution network tower fault prediction model based on a BP (back propagation) neural network, so that an output result of the neural network is obtained, and whether the tower has a trip fault or not is predicted; the training sample of the power distribution network tower fault prediction model based on the BP neural network comprises the following steps: selecting partial fault and non-fault samples from the new fault sample set and the non-fault sample set; the new fault sample set is formed by oversampling an original fault sample by adopting an SMOTE method; the new non-fault sample set is formed by undersampling the original non-fault sample by adopting an OSS method; the method can accurately predict the fault tripping of the power distribution network tower in the typhoon disaster occurrence process, can provide reference for typhoon disaster defense such as first-aid repair material allocation, rapid first-aid repair after failure and the like, and has certain theoretical value and engineering value.

Description

Typhoon disaster power distribution network tower fault prediction method and system
Technical Field
The invention belongs to the technical field of disaster prevention and reduction of an electric power system, and particularly relates to a typhoon disaster power distribution network tower fault prediction method and system based on a BP neural network.
Background
The continuous deterioration of global greenhouse effect and ecological environment causes extreme natural disasters such as wind-waterlogging disasters to present increasingly high situation. The tower tripping fault caused by frequent natural disasters is an important reason for influencing the safe and stable operation of the power grid in China. Wherein, typhoon is the most serious natural disaster affecting coastal areas in China. Typhoon disasters in China mostly occur in summer and autumn, and the typhoon disaster early-warning system has the characteristics of strong burstiness and large destructive power, and causes serious social and economic hazards to provinces and cities in coastal areas of China. According to the current post-incident statistics of wind-waterlogging disasters of coastal areas such as Fujian provinces in China, the power distribution network tower is a high disaster causing object of typhoon disasters. According to the tower trip data of provinces and cities in coastal areas of China, tower faults caused by typhoon disasters account for more than 30% of total trip, and the safe and stable operation of a power grid, especially a power distribution network, is seriously influenced.
The power grid disaster historical data comprise power grid internal data, external meteorological data and the like, and meanwhile, the power grid disaster historical data have the characteristics of time-space attributes, multiple dimensions, multiple scales, uncertainty, strong periodicity, high attribute relevance and the like, and the meteorological environment data are analyzed and processed only by a traditional method to meet the difficulties. The neural network has extremely high superiority in pattern recognition, optimization calculation and nonlinear mapping. The typhoon path has certain regularity and has the characteristic of big data, and meanwhile, the high-voltage tripping early warning of the power distribution network tower is also a pattern recognition classification problem, so that the power grid fault early warning under the typhoon environment is very suitable for a neural network model. The BP neural network is a multilayer feedforward neural network trained according to an error back propagation algorithm and is the most widely applied neural network at present. However, the existing prediction method has large error and complex data processing.
Disclosure of Invention
The invention aims to provide a typhoon disaster power distribution network tower fault prediction method and system based on a BP neural network, and solves the problems of large error and complex data processing of the existing prediction method.
The technical solution for realizing the purpose of the invention is as follows: a method for predicting the fault of a power distribution network tower in a typhoon disaster comprises the following steps:
inputting real-time collected meteorological station data into a trained power distribution network tower fault prediction model based on a BP (back propagation) neural network to obtain an output result of the neural network and predict whether a tower has a trip fault;
the training sample of the power distribution network tower fault prediction model based on the BP neural network comprises the following steps: selecting partial fault and non-fault samples from the new fault sample set and the non-fault sample set;
the new fault sample set is formed by oversampling an original fault sample by adopting an SMOTE method;
the new non-fault sample set is formed by undersampling the original non-fault samples by adopting an OSS method.
Further, typhoon characteristic information of towers with faults and towers without faults is extracted from typhoon-related historical data to form an original fault sample set and an original non-fault sample set;
the typhoon characteristic information of each tower comprises: the maximum wind speed X1 born by each tower, the maximum value X2 of the included angle between the wind direction and the line trend where each tower is located when the maximum wind speed of each tower occurs, the average wind speed X3 of each tower in two minutes, the average included angle X4 of the wind direction and the line of each tower in two minutes, and the rainfall X5 of each tower.
Further, the historical data related to the typhoon comprises power distribution network towers, longitude and latitude information of the meteorological station, meteorological parameters at the towers, meteorological information of the meteorological station in a time period affected by the typhoon and fault information; the meteorological information comprises wind speed, wind direction and rainfall; the fault information comprises tower fault information under the influence of typhoon;
the method for calculating the data weight of different meteorological stations comprises the following steps of calculating meteorological parameters at a tower position by adopting the data weight of three meteorological stations closest to the tower:
Figure BDA0002335054350000021
in the formula (d)mIndicating the distance between the meteorological station m and the tower, m is 1,2,3, kmA data weight representing the mth weather station;
wind speed v at tower0The calculation method comprises the following steps:
Figure BDA0002335054350000022
in the formula, vmRepresenting wind speed of meteorological station m
Wind direction a at tower0The calculation method comprises the following steps:
Figure BDA0002335054350000023
in the formula, amRepresenting a weather station mThe wind direction of (a); a islineAnd the included angle between the line direction of the tower and the north direction of the earth is shown.
Further, a new fault sample set is formed by oversampling the original fault sample by using a SMOTE method, and the process is as follows:
(1) randomly extracting a part of samples in an original fault sample to form a fault sample set for oversampling, and searching a nearest neighbor fault samples of each sample x in the fault sample set;
(2) for each sample x, randomly selecting N fault samples from a nearest neighbor samples according to oversampling multiplying power N, and recording as y1,y2...,yNGenerating N incremental fault samples according to the following equation;
xnew＝x+rand*(yt-x)
wherein, t is 1,2 …, N; t is the number of oversampling times, ytThe t-th neighbor sample of sample x is represented, and rand represents a random number between 0 and 1; x is the number ofnewIndicating a newly added fault sample;
(3) and adding the added fault samples into the original fault samples to form new fault samples.
Further, the new non-fault sample set is formed by undersampling the original non-fault sample by using an OSS method, and the process is as follows:
(1) recording an original sample set as S, randomly selecting a non-fault sample in the original non-fault sample set and establishing a new data set C with all original fault samples;
(2) carrying out nearest neighbor classification on all non-fault samples in the S by using the samples of the data set C, comparing the classification results of the reclassified non-fault samples with the original classification results of the non-fault samples, and adding the non-fault samples with wrong comparison results into the data set C;
(3) for dataset C, the nearest neighbor sample of each sample in dataset C is traversed if there are two samples x belonging to different classesrAnd xw，xrAnd xwThe Euclidean distance therebetween is d (x)r,xw) Indicates that there is no other sample xlLet d (x)r,xl)<d(xr,xw) Or d (x)w,xl)<d(xr,xw) Then (x)r,xw) And forming a contact pair, and deleting the non-fault samples in the contact pair to form a new non-fault sample set.
Further, training is completed on the power distribution network tower fault prediction model based on the BP neural network, and the process comprises the following steps:
let the training sample be x1,x2…,xi,…,xnN samples, i 1,2, …, n, the ith sample x of the prediction modeliIs characterized by [ x ] as an inputi1,xi2…xip]TThere is p dimensions, corresponding to: maximum wind speed X borne by tower1Maximum value X of included angle between wind direction and line trend when maximum wind speed occurs2Two-minute average wind speed X of tower3Two-minute average included angle X between wind direction and line4Rainfall X5(ii) a The ith sample output is set to yiWhether typhoon tripping faults occur on the tower in the next period is shown; mixing X1、X2、X3、X4、X5Normalized to [0,1 respectively]Within the interval:
Figure BDA0002335054350000031
in the formula, u' represents X1、X2、X3、X4Or X5A normalized value; u represents X1、X2、X3、X4Or X5The original value of (a); u. ofminRepresents X1、X2、X3、X4Or X5A minimum value; u. ofmaxRepresents X1、X2、X3、X4Or X5Maximum value of (d);
the input layer has p +1 neurons, and the input samples are: [ x ] ofi1’,xi2'…xip’,-1]T；
The total number of hidden layer neurons is q +1, the kth hidden layer of the ith sampleNet input net of neuronsikComprises the following steps:
Figure BDA0002335054350000041
in the formula, wjkIs the weight, x, between the jth input layer neuron and the kth hidden layer neuronijInputting a value for a jth neuron of an ith sample of an input layer;
output m of kth hidden layer neuron of ith sampleikComprises the following steps:
Figure BDA0002335054350000042
the output layer is provided with a neuron, and the neuron and the hidden layer neuron use a weight value wkConnection, i-th sample network output netiComprises the following steps:
Figure BDA0002335054350000043
Figure BDA0002335054350000044
training of the BP neural network adopts a standard error back-propagation algorithm to adjust weight vectors between the hidden layer and the input layer and between the output layer and the hidden layer; when inputting a sample xiThen, the hidden layer output m is obtained by the forward propagation layer by layer of the networki1,mi2…miqOutput layer output is yiIf y isiAnd the actual output y of the sampleiIf the input error is inconsistent with the input error, the input error is retransmitted to the input layer by layer through the hidden layer, the output error is transmitted to each layer of neuron, the connection weight of each unit is adjusted based on the error, and the kth hidden layer neuron outputs the layer weight adjustment quantity delta wkComprises the following steps:
Δwk＝η1(yi'-yi)f'(yi)mik
＝η1(yi'-yi)yi(1-yi)mik,k＝1,2...q+1
weight adjustment Δ w between jth input layer neuron and kth hidden layer neuronjkComprises the following steps:
Δwjk＝η2(yi'-yi)yi(1-yi)wkmik(1-mik)xij,j＝1,2...p+1,k＝1,2...q+1
in the formula, η1And η2To the learning rate, f' (y)i) Is the partial differential of the output layer activation function; describing the fitting performance of the BP neural network by adopting an average mean square error e, and when e is smaller than a set threshold e0Or the training round r reaches the set iteration number rmaxAnd (5) ending the training:
Figure BDA0002335054350000051
and finally obtaining a power distribution network tower fault prediction model based on the BP neural network.
The utility model provides a typhoon calamity distribution network shaft tower fault prediction system which characterized in that includes:
the meteorological station data acquisition module is used for acquiring meteorological station data in real time;
the prediction module is used for inputting real-time collected meteorological station data into a trained power distribution network tower fault prediction model based on a BP (back propagation) neural network, obtaining an output result of the neural network and predicting whether the tower has a trip fault; the training sample of the power distribution network tower fault prediction model based on the BP neural network comprises the following steps: selecting partial fault and non-fault samples from the new fault sample set and the non-fault sample set; the new fault sample set is formed by oversampling an original fault sample by adopting an SMOTE method; the new non-fault sample set is formed by undersampling the original non-fault samples by adopting an OSS method.
Further, typhoon characteristic information of towers with faults and towers without faults is extracted from typhoon-related historical data to form an original fault sample set and an original non-fault sample set;
the typhoon characteristic information of each tower comprises: the maximum wind speed X1 born by each tower, the maximum value X2 of the included angle between the wind direction and the line trend where each tower is located when the maximum wind speed of each tower occurs, the average wind speed X3 of each tower in two minutes, the average included angle X4 of the wind direction and the line of each tower in two minutes, and the rainfall X5 of each tower.
Further, a new fault sample set is formed by oversampling the original fault sample by using a SMOTE method, and the process is as follows:
(1) randomly extracting a part of samples in an original fault sample to form a fault sample set for oversampling, and searching a nearest neighbor fault samples of each sample x in the fault sample set;
(2) for each sample x, randomly selecting N fault samples from a nearest neighbor samples according to oversampling multiplying power N, and recording as y1,y2...,yNGenerating N incremental fault samples according to the following equation;
xnew＝x+rand*(yt-x)
wherein, t is 1,2 …, N; t is the number of oversampling times, ytThe t-th neighbor sample of sample x is represented, and rand represents a random number between 0 and 1; x is the number ofnewIndicating a newly added fault sample;
(3) and adding the added fault samples into the original fault samples to form new fault samples.
The new non-fault sample set is formed by undersampling the original non-fault sample by adopting an OSS method, and the process is as follows:
(1) recording an original sample set as S, randomly selecting a non-fault sample in the original non-fault sample set and establishing a new data set C with all original fault samples;
(2) carrying out nearest neighbor classification on all non-fault samples in the S by using the samples of the data set C, comparing the classification results of the reclassified non-fault samples with the original classification results of the non-fault samples, and adding the non-fault samples with wrong comparison results into the data set C;
(3) for data set C, the nearest neighbor sample of each sample in data set C is traversed, if there are two samples belonging to different categoriesSample x of classrAnd xw，xrAnd xwThe Euclidean distance therebetween is d (x)r,xw) Indicates that there is no other sample xlLet d (x)r,xl)<d(xr,xw) Or d (x)w,xl)<d(xr,xw) Then (x)r,xw) And forming a contact pair, and deleting the non-fault samples in the contact pair to form a new non-fault sample set.
A computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computing device, cause the computing device to perform any of the methods of typhoon disaster power distribution network tower failure prediction methods described previously.
The invention achieves the following beneficial effects:
1) according to the power distribution network tower fault prediction model under the typhoon disaster based on the BP neural network and the meteorological station data collected in real time, the effect of predicting the power distribution network line fault with higher precision is achieved;
2) according to the prediction of the fault tripping of the power distribution network tower in the typhoon disaster occurrence process, the method can provide reference for typhoon disaster defense such as emergency repair material allocation, rapid emergency repair after a fault and the like, and has certain theoretical value and engineering value.
Drawings
Fig. 1 is a flow chart of a method for predicting a fault of a power distribution network tower in a typhoon disaster in an embodiment of the invention;
FIG. 2 is a flow chart of another method for predicting the fault of the power distribution network tower in the typhoon disaster in the embodiment of the invention;
FIG. 3 is a diagram of a BP neural network structure according to an embodiment of the present invention;
FIG. 4 is a unipolar sigmoid function image according to an embodiment of the present invention;
FIG. 5 is a flowchart illustrating a BP neural network training process according to an embodiment of the present invention;
FIG. 6 is a diagram illustrating an error distribution of a BP neural network training sample according to an embodiment of the present invention;
fig. 7 is an error distribution diagram of a BP neural network test sample according to an embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
Example 1:
as shown in fig. 1, a method for predicting a fault of a power distribution network tower in a typhoon disaster includes the following steps:
inputting real-time collected meteorological station data into a trained power distribution network tower fault prediction model based on a BP (back propagation) neural network to obtain an output result of the neural network and predict whether a tower has a trip fault;
the training sample of the power distribution network tower fault prediction model based on the BP neural network comprises the following steps: selecting partial fault and non-fault samples from the new fault sample set and the non-fault sample set;
the new fault sample set is formed by oversampling an original fault sample by adopting an SMOTE method;
the new non-fault sample set is formed by undersampling the original non-fault samples by adopting an OSS method.
Further, typhoon characteristic information of towers with faults and towers without faults is extracted from typhoon-related historical data to form an original fault sample set and an original non-fault sample set;
the typhoon characteristic information of each tower comprises: the maximum wind speed X1 born by each tower, the maximum value X2 of the included angle between the wind direction and the line trend where each tower is located when the maximum wind speed of each tower occurs, the average wind speed X3 of each tower in two minutes, the average included angle X4 of the wind direction and the line of each tower in two minutes, and the rainfall X5 of each tower.
Further, the historical data related to the typhoon comprises power distribution network towers, longitude and latitude information of the meteorological station, meteorological parameters at the towers, meteorological information of the meteorological station in a time period affected by the typhoon and fault information; the meteorological information comprises wind speed, wind direction and rainfall; the fault information comprises tower fault information under the influence of typhoon;
the method for calculating the data weight of different meteorological stations comprises the following steps of calculating meteorological parameters at a tower position by adopting the data weight of three meteorological stations closest to the tower:
Figure BDA0002335054350000071
in the formula (d)mIndicating the distance between the meteorological station m and the tower, m is 1,2,3, kmA data weight representing the mth weather station;
wind speed v at tower0The calculation method comprises the following steps:
Figure BDA0002335054350000072
in the formula, vmRepresenting the wind speed of the meteorological station m;
wind direction a at tower0The calculation method comprises the following steps:
Figure BDA0002335054350000081
in the formula, amRepresenting the wind direction of the weather station m; a islineAnd the included angle between the line direction of the tower and the north direction of the earth is shown.
Further, a new fault sample set is formed by oversampling the original fault sample by using a SMOTE method, and the process is as follows:
(1) randomly extracting a part of samples in an original fault sample to form a fault sample set for oversampling, and searching a nearest neighbor fault samples of each sample x in the fault sample set;
(2) for each sample x, randomly selecting N fault samples from a nearest neighbor samples according to oversampling multiplying power N, and recording as y1,y2...,yNGenerating N incremental fault samples according to the following equation;
xnew＝x+rand*(yt-x)
wherein, t is 1,2 …, N;t is the number of oversampling times, ytThe t-th neighbor sample of sample x is represented, and rand represents a random number between 0 and 1; x is the number ofnewIndicating a newly added fault sample;
(3) and adding the added fault samples into the original fault samples to form new fault samples.
Further, the new non-fault sample set is formed by undersampling the original non-fault sample by using an OSS method, and the process is as follows:
(1) recording an original sample set as S, randomly selecting a non-fault sample in the original non-fault sample set and establishing a new data set C with all original fault samples;
(2) carrying out nearest neighbor classification on all non-fault samples in the S by using the samples of the data set C, comparing the classification results of the reclassified non-fault samples with the original classification results of the non-fault samples, and adding the non-fault samples with wrong comparison results into the data set C;
(3) for dataset C, the nearest neighbor sample of each sample in dataset C is traversed if there are two samples x belonging to different classesrAnd xw，xrAnd xwThe Euclidean distance therebetween is d (x)r,xw) Indicates that there is no other sample xlLet d (x)r,xl)<d(xr,xw) Or d (x)w,xl)<d(xr,xw) Then (x)r,xw) And forming a contact pair, and deleting the non-fault samples in the contact pair to form a new non-fault sample set.
Further, training is completed on the power distribution network tower fault prediction model based on the BP neural network, and the process comprises the following steps:
let the training sample be x1,x2…,xi,…,xnN samples, i 1,2, …, n, the ith sample x of the prediction modeliIs characterized by [ x ] as an inputi1,xi2…xip]TThere is p dimensions, corresponding to: maximum wind speed X borne by tower1Maximum value X of included angle between wind direction and line trend when maximum wind speed occurs2Tower two minute averagingWind speed X3Two-minute average included angle X between wind direction and line4Rainfall X5(ii) a The ith sample output is set to yiWhether typhoon tripping faults occur on the tower in the next period is shown; mixing X1、X2、X3、X4、X5Normalized to [0,1 respectively]Within the interval:
Figure BDA0002335054350000091
in the formula, u' represents X1、X2、X3、X4Or X5A normalized value; u represents X1、X2、X3、X4Or X5The original value of (a); u. ofminRepresents X1、X2、X3、X4Or X5A minimum value; u. ofmaxRepresents X1、X2、X3、X4Or X5Maximum value of (d);
the input layer has p +1 neurons, and the input samples are: [ x ] ofi1’,xi2'…xip’,-1]T；
The total number of hidden neurons is q +1, the net input net of the kth hidden neuron in the ith sampleikComprises the following steps:
Figure BDA0002335054350000092
in the formula, wjkIs the weight, x, between the jth input layer neuron and the kth hidden layer neuronijInputting a value for a jth neuron of an ith sample of an input layer;
output m of kth hidden layer neuron of ith sampleikComprises the following steps:
Figure BDA0002335054350000093
the output layer is provided with a neuron, and the neuron and the hidden layer neuron use a weight value wkConnection, i-th sample network output netiComprises the following steps:
Figure BDA0002335054350000094
Figure BDA0002335054350000095
training of the BP neural network adopts a standard error back-propagation algorithm to adjust weight vectors between the hidden layer and the input layer and between the output layer and the hidden layer; when inputting a sample xiThen, the hidden layer output m is obtained by the forward propagation layer by layer of the networki1,mi2…miqOutput layer output is yiIf y isiAnd the actual output y of the sampleiIf the input error is inconsistent with the input error, the input error is retransmitted to the input layer by layer through the hidden layer, the output error is transmitted to each layer of neuron, the connection weight of each unit is adjusted based on the error, and the kth hidden layer neuron outputs the layer weight adjustment quantity delta wkComprises the following steps:
Δwk＝η1(yi'-yi)f'(yi)mik
＝η1(yi'-yi)yi(1-yi)mik,k＝1,2...q+1
weight adjustment Δ w between jth input layer neuron and kth hidden layer neuronjkComprises the following steps:
Δwjk＝η2(yi'-yi)yi(1-yi)wkmik(1-mik)xij,j＝1,2...p+1,k＝1,2...q+1
in the formula, η1And η2To the learning rate, f' (y)i) Is the partial differential of the output layer activation function; describing the fitting performance of the BP neural network by adopting an average mean square error e, and when e is smaller than a set threshold e0Or the training round r reaches the set iteration number rmaxAnd (5) ending the training:
Figure BDA0002335054350000101
and finally obtaining a power distribution network tower fault prediction model based on the BP neural network.
Example 2:
as shown in fig. 2, a method for predicting a fault of a power distribution network tower in a typhoon disaster includes the following steps:
step 1, collecting historical data related to typhoon and calculating meteorological parameters at a tower;
the historical data related to the typhoon comprises longitude and latitude information of a power distribution network tower and a meteorological station, and meteorological information and fault information of the meteorological station in a period influenced by the typhoon; the meteorological information comprises wind speed, wind direction and rainfall; the fault information includes tower fault information (i.e., trip information) that occurs under the influence of typhoons.
The method for calculating the data weight of different meteorological stations comprises the following steps of calculating meteorological parameters at a tower position by adopting the data weight of three meteorological stations closest to the tower:
Figure BDA0002335054350000102
in the formula (d)mIndicating the distance between the meteorological station m and the tower, m is 1,2,3, kmRepresenting the data weight of the mth weather station.
Wind speed v at tower0The calculation method comprises the following steps:
Figure BDA0002335054350000103
in the formula, vmRepresenting the wind speed of the meteorological station m;
wind direction a at tower0The calculation method comprises the following steps:
Figure BDA0002335054350000111
in the formula, amRepresenting the wind direction of the weather station m; a islineThe included angle between the line direction of the tower and the north direction of the earth can be expressedAnd the latitude and longitude information of the tower and adjacent towers on the same line are calculated.
Step 2, extracting typhoon characteristic information of towers with faults and towers without faults from typhoon-related historical data to form an original fault sample set and an original non-fault sample set to form an original sample set;
dividing the typhoon occurrence time into discrete time periods at intervals of 30min, and respectively counting the typhoon information of each tower in each time period, wherein the typhoon information of each tower comprises: the maximum wind speed X1 (the maximum value of the instantaneous wind speed in a given time period) borne by each tower, the maximum value X2 of the included angle between the wind direction and the line where each tower is located when the maximum wind speed occurs, the average wind speed of each tower in two minutes (the average value of the average wind speed of two minutes in a 1-hour time period, namely the monitoring information given by the weather station is the average wind speed of two minutes, the average value of all the average wind speeds of two minutes given in the hour) X3, the average included angle (the average value of the average included angle of two minutes in the 1-hour time period) X4 between the wind direction of each tower and the line, and the rainfall of each tower X5. The typhoon information of the tower with the fault forms an original fault sample set, and the typhoon information of the tower without the fault forms an original non-fault sample set.
Step 3, resampling the original fault sample set and the original non-fault sample set to form a new fault sample set and a non-fault sample set;
step 3-1, oversampling the original fault sample by adopting a SMOTE (synthetic minority oversampling technology) method to form a new fault sample set:
(1) randomly extracting 50% (or other proportion, for example, 50%) of samples in the original fault sample to form a fault sample set for oversampling, and for each sample x in the fault sample set, searching a (usually, 5) nearest neighbor fault samples;
(2) for each sample x, randomly selecting N fault samples from a nearest neighbor samples according to oversampling multiplying power N, and recording as y1,y2...,yNGenerating N incremental fault samples according to the following equation;
xnew＝x+rand*(yt-x)
wherein, t is 1,2 …, N; t is the number of oversampling times, ytThe t-th adjacent sample of the sample x is represented, N is oversampling multiplying power, and rand represents a random number between 0 and 1; x is the number ofnewIndicating a newly added fault sample; x represents an original fault sample to be sampled currently;
(3) and adding the added fault samples into the original fault samples to form new fault samples.
Step 3-2, the original non-fault sample is subjected to undersampling by adopting OSS (single-side selection algorithm) to form a new non-fault sample set, and the method comprises the following steps:
(1) recording the original sample set obtained in the step (2) as S, randomly selecting one non-fault sample from the original non-fault sample set and establishing a new data set C with all original fault samples;
(2) carrying out nearest neighbor classification on all non-fault samples in the S by using samples of the data set C, comparing the classification results of the reclassified non-fault samples with the original classification results of the non-fault samples, and adding the non-fault samples with wrong comparison results into the data set C so as to delete 'safe samples' far away from the fault samples in the non-fault sample set;
the nearest neighbor classification is to find a sample with the nearest European distance in C for each non-fault sample in S, and take the class of the sample as a new class of the non-fault sample;
(3) for dataset C, the nearest neighbor sample of each sample in dataset C is traversed if there are two samples x belonging to different classesrAnd xw，xrAnd xwThe Euclidean distance therebetween is d (x)r,xw) Indicates that there is no other sample xlLet d (x)r,xl)<d(xr,xw) Or d (x)w,xl)<d(xr,xw) Then (x)r,xw) Forming a link pair, deleting the non-fault samples in the link pair, thereby deleting the noise samples in the non-fault sample set within the distribution range of the fault samples and the boundary samples at the classification boundary to form a new link pairA non-failure sample set.
Step 4, selecting partial fault and non-fault samples from the new fault sample set and the non-fault sample set as training samples to train a power distribution network tower fault prediction model based on the BP neural network;
step 4-1, let the training sample be x1,x2…,xi,…,xnN samples, i is 1,2, …, n, i indicates the sample number; prediction model ith sample xiIs characterized by [ x ] as an inputi1,xi2…xip]TIn the present model, p is taken as 5, which corresponds to: maximum wind speed X borne by tower1Maximum value X of included angle between wind direction and line trend when maximum wind speed occurs2Two-minute average wind speed X of tower3Two-minute average included angle X between wind direction and line4Rainfall X5. The ith sample output is set to yiAnd the method indicates whether the tower has typhoon trip fault (if the tower has fault, y)iIs 1, not failed, yiIs 0). Mixing X1、X2、X3、X4、X5Normalized to [0,1 respectively]Within the interval:
Figure BDA0002335054350000121
in the formula, u' represents X1、X2、X3、X4Or X5A normalized value; u represents X1、X2、X3、X4Or X5The original value of (a); u. ofminRepresents X1、X2、X3、X4Or X5A minimum value; u. ofmaxRepresents X1、X2、X3、X4Or X5Is measured.
As shown in fig. 3, the input layer has p +1 neurons, where p neurons correspond to the p-dimensional input features after sample normalization, respectively, and another neuron is designed for the threshold of the hidden layer activation function, and its value is-1, then the input sample is:
[xi1’,xi2'…xip’,-1]T。
the number of hidden layer neurons is q +1, wherein q neurons are connected with the input layer by a weight value, and the weight value between the jth input layer neuron and the kth hidden layer neuron is wjkIf another neuron is also set for the output layer activation function, and its value is always-1, then the net input net of the kth hidden layer neuron of the ith sampleikComprises the following steps:
Figure BDA0002335054350000131
in the formula, xijInputting a value for a jth neuron of an ith sample of an input layer;
as shown in fig. 4, the hidden layer activation function adopts a unipolar sigmoid function.
The output m of the kth hidden layer neuron of the ith sampleikComprises the following steps:
Figure BDA0002335054350000132
the output layer is provided with a neuron, and the neuron and the hidden layer neuron use a weight value wkThe connection and activation functions also adopt sigmoid functions, and the network output net of the ith sampleiComprises the following steps:
Figure BDA0002335054350000133
Figure BDA0002335054350000134
yioutput for the ith sample;
and 4-2, training the BP neural network, and adjusting weight vectors between the hidden layer and the input layer and between the output layer and the hidden layer by adopting a standard error back-propagation algorithm. When inputting a sample xiThen, the hidden layer output m is obtained by the forward propagation layer by layer of the networki1,mi2…miqThe output layer outputsyiIf y isiAnd the actual output y of the sampleiIf the output error is inconsistent with the preset error, the error back transmission stage is started, the output error is back transmitted to the input layer by layer through the hidden layer, the error is transmitted to each layer of neuron, the connection weight of each unit is adjusted based on the error, and the weight adjustment quantity delta w of the output layer of the kth hidden layer neuron is adjusted according to the rulekComprises the following steps:
Δwk＝η1(yi'-yi)f'(yi)mik
＝η1(yi'-yi)yi(1-yi)mik,k＝1,2...q+1
weight adjustment Δ w between jth input layer neuron and kth hidden layer neuronjkComprises the following steps:
Δwjk＝η2(yi'-yi)yi(1-yi)wkmik(1-mik)xij,j＝1,2...p+1,k＝1,2...q+1
in the formula, η1And η2To the learning rate, f' (y)i) The partial differential of the activation function of the output layer can be adaptively adjusted in the training process according to the error convergence condition so as to accelerate the convergence speed. After the learning rate is adjusted once, if the total error is increased, the adjustment is invalid, and the learning rate is multiplied by a coefficient smaller than 1; if the total error drops, the adjustment is valid, multiplying the learning rate by a factor greater than 1.
Inputting all samples into the network to perform one-time weight adjustment is regarded as one-time training, and normal samples and fault samples are input in a random sequence in one-time adjustment so as to prevent the weight from being adjusted in the same direction all the time. Describing the fitting performance of the BP neural network by adopting an average mean square error e, and when e is smaller than a set threshold e0Or the training round r reaches the set iteration number rmaxAnd (5) ending the training:
Figure BDA0002335054350000141
and finally obtaining a power distribution network tower fault prediction model based on the BP neural network.
And 5, inputting real-time collected meteorological station data into a trained power distribution network tower fault prediction model based on the BP neural network to obtain an output result of the neural network, and predicting whether the tower has a trip fault.
Inputting real-time collected meteorological station data into trained neural network to generate tower sample x to be predicted0＝(x01,x02…x0p) The sample output y is obtained by forward propagation through the network0After, y0The value interval of (2) is (0, 1). Taking 0.5 as a judgment threshold value, if y is output0If the tower is more than or equal to 0.5, the tower is more inclined to the fault class, so that the tower is predicted to be classified as the fault class, and the fault early warning of the typhoon is sent out; when y is0<And when 0.5 hour, no fault early warning signal is sent out.
Example 3:
a typhoon disaster power distribution network tower fault prediction system comprises:
the meteorological station data acquisition module is used for acquiring meteorological station data in real time;
the prediction module is used for inputting real-time collected meteorological station data into a trained power distribution network tower fault prediction model based on a BP (back propagation) neural network, obtaining an output result of the neural network and predicting whether the tower has a trip fault; the training sample of the power distribution network tower fault prediction model based on the BP neural network comprises the following steps: selecting partial fault and non-fault samples from the new fault sample set and the non-fault sample set; the new fault sample set is formed by oversampling an original fault sample by adopting an SMOTE method; the new non-fault sample set is formed by undersampling the original non-fault samples by adopting an OSS method.
Extracting typhoon characteristic information of towers with faults and towers without faults from typhoon-related historical data to form an original fault sample set and an original non-fault sample set;
the typhoon characteristic information of each tower comprises: the maximum wind speed X1 born by each tower, the maximum value X2 of the included angle between the wind direction and the line trend where each tower is located when the maximum wind speed of each tower occurs, the average wind speed X3 of each tower in two minutes, the average included angle X4 of the wind direction and the line of each tower in two minutes, and the rainfall X5 of each tower.
The new fault sample set is formed by oversampling the original fault sample by adopting a SMOTE method, and the process is as follows:
(1) randomly extracting a part of samples in an original fault sample to form a fault sample set for oversampling, and searching a nearest neighbor fault samples of each sample x in the fault sample set;
(2) for each sample x, randomly selecting N fault samples from a nearest neighbor samples according to oversampling multiplying power N, and recording as y1,y2...,yNGenerating N incremental fault samples according to the following equation;
xnew＝x+rand*(yt-x)
wherein, t is 1,2 …, N; t is the number of oversampling times, ytThe t-th neighbor sample of sample x is represented, and rand represents a random number between 0 and 1; x is the number ofnewIndicating a newly added fault sample;
(3) and adding the added fault samples into the original fault samples to form new fault samples.
The new non-fault sample set is formed by undersampling the original non-fault sample by adopting an OSS method, and the process is as follows:
(1) recording an original sample set as S, randomly selecting a non-fault sample in the original non-fault sample set and establishing a new data set C with all original fault samples;
(2) carrying out nearest neighbor classification on all non-fault samples in the S by using the samples of the data set C, comparing the classification results of the reclassified non-fault samples with the original classification results of the non-fault samples, and adding the non-fault samples with wrong comparison results into the data set C;
(3) for dataset C, the nearest neighbor sample of each sample in dataset C is traversed if there are two samples x belonging to different classesrAnd xw，xrAnd xwThe Euclidean distance therebetween is d (x)r,xw) Indicates that there is no other sample xlLet d (x)r,xl)<d(xr,xw) Or d (x)w,xl)<d(xr,xw) Then (x)r,xw) And forming a contact pair, and deleting the non-fault samples in the contact pair to form a new non-fault sample set.
Example 4:
a computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computing device, cause the computing device to perform any of the above-described methods of typhoon disaster power distribution network tower failure prediction.
A computing device, comprising, in combination,
one or more processors, memory, and one or more programs stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the above-described typhoon disaster power distribution network tower failure prediction methods.
Example 5:
by taking pole tower fault conditions in the process of typhoon disasters in 2015-2016 year in certain city of Fujian province in China as an example, the effectiveness of the power distribution network pole tower fault prediction model based on the BP neural network provided by the invention is verified, typhoon information around the pole tower in a time period before an accident occurs is extracted, the typhoon information comprises a maximum wind speed X1, a maximum value X2 of an included angle between a wind direction and a line trend, an average wind speed X3 of two minutes, an average included angle X4 of the wind direction and the line of two minutes, and rainfall X5, and specific data are shown in Table 1. In addition, 14 thousands of tower cases which do not have faults in the typhoon influence time period are collected, and the two types of samples form original samples.
TABLE 1 statistics of tower typhoon characteristics of tripping operation
Figure BDA0002335054350000161
Figure BDA0002335054350000171
First, 14 ten thousand normal samples are randomly undersampled, and the number of the normal samples is reduced to 300. SMOTE oversampling is carried out on 38 fault samples, and the number of the few types of samples is expanded to 216; and performing noise elimination processing on the normal samples, and further reducing the number of the normal samples to 216. And processing the obtained training sample set based on a resampling algorithm to obtain 432 samples in total. And (3) verifying the effectiveness of the method by using part of historical samples as test samples. 292 samples (normal samples: fault samples ═ 1:1) are input into the neural network to train the network, and the remaining 140 samples are used as subsequent prediction samples. A power distribution network tower fault prediction model based on a BP neural network is established according to the method, and an attached figure 5 is a neural network training flow chart. For 292 simulation training samples, the first 146 are fault samples, the last 146 are normal samples, and the prediction output y of the samples isiAnd the actual output yi' Difference yi-yi' | is as shown in FIG. 6.
The results of the training sample fault early warning obtained after the prediction output classification is performed by using 0.5 as a threshold are shown in table 2.
TABLE 2 BP neural network training sample failure early warning result
Figure BDA0002335054350000172
And (3) evaluating the accuracy of the model by adopting the forecasting accuracy FDR, the false alarm rate FAR and the total forecasting accuracy PA:
Figure BDA0002335054350000173
Figure BDA0002335054350000174
Figure BDA0002335054350000175
in the formula, FP and FN are the numbers of the fault samples classified as normal and fault, respectively, and TP and TF are the numbers of the normal samples classified as normal and fault, respectively. The FDR index represents the proportion of early warning success in samples with typhoon faults actually, the FAR index represents the proportion of false failure early warning in samples without typhoon faults, and the PA index represents the proportion of early warning success in all samples. The early warning accuracy of the training samples is shown in table 3.
TABLE 3 training sample Fault Warning results
Figure BDA0002335054350000181
For 140 prediction samples (of which the first 70 are fault samples and the last 70 are normal samples), the output error of the distribution network tower fault prediction model based on the BP neural network is shown in fig. 7, and the trip fault early warning result is shown in table 3.
TABLE 4 early warning results of test sample failures
Figure BDA0002335054350000182
As can be seen from fig. 6 and table 4, the prediction accuracy of the pole tower fault prediction model based on the BP neural network reaches 90%, the false alarm rate is 17.14%, the total prediction accuracy is 86.43%, most of unpredictable faults are in the case of low wind speed, and the pole towers in the tripping false alarm samples are all located in the zones with high wind speed and are in the dangerous state of tripping.
In conclusion, by adopting the power distribution network tower fault prediction model based on the BP neural network, tower faults in the process of typhoon disasters in a certain city of Fujian province are predicted, so that higher prediction accuracy and lower false alarm rate are obtained, and references can be provided for typhoon disaster prevention such as first-aid repair material allocation and rapid first-aid repair after faults.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.

Claims (10)

1. A fault prediction method for a power distribution network tower in a typhoon disaster is characterized by comprising the following steps:
inputting real-time collected meteorological station data into a trained power distribution network tower fault prediction model based on a BP (back propagation) neural network to obtain an output result of the neural network and predict whether a tower has a trip fault;
the training sample of the power distribution network tower fault prediction model based on the BP neural network comprises the following steps: selecting partial fault and non-fault samples from the new fault sample set and the non-fault sample set;
the new fault sample set is formed by oversampling an original fault sample by adopting an SMOTE method;
the new non-fault sample set is formed by undersampling the original non-fault samples by adopting an OSS method.
2. The typhoon disaster power distribution network tower fault prediction method according to claim 1, wherein typhoon feature information of towers with faults and towers without faults is extracted from typhoon related historical data to form an original fault sample set and an original non-fault sample set;
the typhoon characteristic information of each tower comprises: the maximum wind speed X1 born by each tower, the maximum value X2 of the included angle between the wind direction and the line trend where each tower is located when the maximum wind speed of each tower occurs, the average wind speed X3 of each tower in two minutes, the average included angle X4 of the wind direction and the line of each tower in two minutes, and the rainfall X5 of each tower.
3. The method for predicting the fault of the power distribution network tower caused by the typhoon disaster as claimed in claim 2, wherein the historical data related to the typhoon comprises longitude and latitude information of the power distribution network tower and a meteorological station, meteorological parameters at the tower, meteorological information of the meteorological station in a period influenced by the typhoon and fault information; the meteorological information comprises wind speed, wind direction and rainfall; the fault information comprises tower fault information under the influence of typhoon;
the method for calculating the data weight of different meteorological stations comprises the following steps of calculating meteorological parameters at a tower position by adopting the data weight of three meteorological stations closest to the tower:
Figure FDA0002335054340000011
in the formula (d)mIndicating the distance between the meteorological station m and the tower, m is 1,2,3, kmA data weight representing the mth weather station;
wind speed v at tower0The calculation method comprises the following steps:
Figure FDA0002335054340000012
in the formula, vmRepresenting the wind speed of the meteorological station m;
wind direction a at tower0The calculation method comprises the following steps:
Figure FDA0002335054340000021
in the formula, amRepresenting the wind direction of the weather station m; a islineAnd the included angle between the line direction of the tower and the north direction of the earth is shown.
4. The method for predicting the fault of the power distribution network tower caused by the typhoon disaster according to the claim 1, wherein the new fault sample set is formed by oversampling the original fault sample by adopting a SMOTE method, and the process is as follows:
(1) randomly extracting a part of samples in an original fault sample to form a fault sample set for oversampling, and searching a nearest neighbor fault samples of each sample x in the fault sample set;
(2) for each sample x, randomly selecting N fault samples from a nearest neighbor samples according to oversampling multiplying power N, and recording as y1,y2...,yNGenerating N incremental fault samples according to the following equation;
xnew＝x+rand*(yt-x)
wherein, t is 1,2 …, N; t is the number of oversampling times, ytThe t-th neighbor sample of sample x is represented, and rand represents a random number between 0 and 1; x is the number ofnewIndicating a newly added fault sample;
(3) and adding the added fault samples into the original fault samples to form new fault samples.
5. The method for predicting the fault of the power distribution network tower caused by the typhoon disaster according to the claim 1, wherein the new non-fault sample set is formed by undersampling the original non-fault sample by adopting an OSS method, and the process is as follows:
(1) recording an original sample set as S, randomly selecting a non-fault sample in the original non-fault sample set and establishing a new data set C with all original fault samples;
(2) carrying out nearest neighbor classification on all non-fault samples in the S by using the samples of the data set C, comparing the classification results of the reclassified non-fault samples with the original classification results of the non-fault samples, and adding the non-fault samples with wrong comparison results into the data set C;
(3) for dataset C, the nearest neighbor sample of each sample in dataset C is traversed if there are two samples x belonging to different classesrAnd xw，xrAnd xwThe Euclidean distance therebetween is d (x)r,xw) Indicates that there is no other sample xlLet d (x)r,xl)<d(xr,xw) Or d (x)w,xl)<d(xr,xw) Then (x)r,xw) And forming a contact pair, and deleting the non-fault samples in the contact pair to form a new non-fault sample set.
6. The method for predicting the fault of the power distribution network tower in the typhoon disaster according to claim 1, wherein a power distribution network tower fault prediction model based on a BP neural network is trained and completed, and the process comprises the following steps:
let the training sample be x1,x2…,xi,…,xnN samples, i 1,2, …, n, the ith sample x of the prediction modeliIs characterized by [ x ] as an inputi1,xi2…xip]TThere is p dimensions, corresponding to: maximum wind speed X borne by tower1Maximum value X of included angle between wind direction and line trend when maximum wind speed occurs2Two-minute average wind speed X of tower3Two-minute average included angle X between wind direction and line4Rainfall X5(ii) a The ith sample output is set to yiWhether typhoon tripping faults occur on the tower in the next period is shown; mixing X1、X2、X3、X4、X5Normalized to [0,1 respectively]Within the interval:
Figure FDA0002335054340000031
in the formula, u' represents X1、X2、X3、X4Or X5A normalized value; u represents X1、X2、X3、X4Or X5The original value of (a); u. ofminRepresents X1、X2、X3、X4Or X5A minimum value; u. ofmaxRepresents X1、X2、X3、X4Or X5Maximum value of (d);
the input layer has p +1 neurons, and the input samples are: [ x ] ofi1’,xi2'…xip’,-1]T；
The total number of hidden neurons is q +1, the net input net of the kth hidden neuron in the ith sampleikComprises the following steps:
Figure FDA0002335054340000032
in the formula, wjkIs the weight, x, between the jth input layer neuron and the kth hidden layer neuronijInputting a value for a jth neuron of an ith sample of an input layer;
output m of kth hidden layer neuron of ith sampleikComprises the following steps:
Figure FDA0002335054340000033
the output layer is provided with a neuron, and the neuron and the hidden layer neuron use a weight value wkConnection, i-th sample network output netiComprises the following steps:
Figure FDA0002335054340000034
Figure FDA0002335054340000035
training of the BP neural network adopts a standard error back-propagation algorithm to adjust weight vectors between the hidden layer and the input layer and between the output layer and the hidden layer; when inputting a sample xiThen, the hidden layer output m is obtained by the forward propagation layer by layer of the networki1,mi2…miqOutput layer output is yiIf y isiAnd the actual output y of the sampleiIf the input error is inconsistent with the input error, the input error is retransmitted to the input layer by layer through the hidden layer, the output error is transmitted to each layer of neuron, the connection weight of each unit is adjusted based on the error, and the kth hidden layer neuron outputs the layer weight adjustment quantity delta wkComprises the following steps:
Δwk＝η1(yi'-yi)f'(yi)mik
＝η1(yi'-yi)yi(1-yi)mik,k＝1,2...q+1
weight adjustment Δ w between jth input layer neuron and kth hidden layer neuronjkComprises the following steps:
Δwjk＝η2(yi'-yi)yi(1-yi)wkmik(1-mik)xij,
j＝1,2...p+1,k＝1,2...q+1
in the formula, η1And η2To the learning rate, f' (y)i) Is the partial differential of the output layer activation function; describing the fitting performance of the BP neural network by adopting an average mean square error e, and when e is smaller than a set threshold e0Or the training round r reaches the set iteration number rmaxAnd (5) ending the training:
Figure FDA0002335054340000041
and finally obtaining a power distribution network tower fault prediction model based on the BP neural network.
7. The utility model provides a typhoon calamity distribution network shaft tower fault prediction system which characterized in that includes:
the meteorological station data acquisition module is used for acquiring meteorological station data in real time;
the prediction module is used for inputting real-time collected meteorological station data into a trained power distribution network tower fault prediction model based on a BP (back propagation) neural network, obtaining an output result of the neural network and predicting whether the tower has a trip fault; the training sample of the power distribution network tower fault prediction model based on the BP neural network comprises the following steps: selecting partial fault and non-fault samples from the new fault sample set and the non-fault sample set; the new fault sample set is formed by oversampling an original fault sample by adopting an SMOTE method; the new non-fault sample set is formed by undersampling the original non-fault samples by adopting an OSS method.
8. The system for predicting the fault of the typhoon disaster power distribution network tower according to claim 7, wherein typhoon characteristic information of towers with fault and towers without fault is extracted from typhoon related historical data to form an original fault sample set and an original non-fault sample set;
the typhoon characteristic information of each tower comprises: the maximum wind speed X1 born by each tower, the maximum value X2 of the included angle between the wind direction and the line trend where each tower is located when the maximum wind speed of each tower occurs, the average wind speed X3 of each tower in two minutes, the average included angle X4 of the wind direction and the line of each tower in two minutes, and the rainfall X5 of each tower.
9. The system for predicting the fault of the power distribution network tower caused by the typhoon disaster according to the claim 1, wherein the new fault sample set is formed by oversampling the original fault sample by adopting a SMOTE method, and the process is as follows:
(1) randomly extracting a part of samples in an original fault sample to form a fault sample set for oversampling, and searching a nearest neighbor fault samples of each sample x in the fault sample set;
(2) for each sample x, randomly selecting N fault samples from a nearest neighbor samples according to oversampling multiplying power N, and recording as y1,y2...,yNGenerating N incremental fault samples according to the following equation;
xnew＝x+rand*(yt-x)
wherein, t is 1,2 …, N; t is the number of oversampling times, ytThe t-th neighbor sample of sample x is represented, and rand represents a random number between 0 and 1; x is the number ofnewIndicating a newly added fault sample;
(3) and adding the added fault samples into the original fault samples to form new fault samples.
The new non-fault sample set is formed by undersampling the original non-fault sample by adopting an OSS method, and the process is as follows:
(1) recording an original sample set as S, randomly selecting a non-fault sample in the original non-fault sample set and establishing a new data set C with all original fault samples;
(2) carrying out nearest neighbor classification on all non-fault samples in the S by using the samples of the data set C, comparing the classification results of the reclassified non-fault samples with the original classification results of the non-fault samples, and adding the non-fault samples with wrong comparison results into the data set C;
(3) for dataset C, the nearest neighbor sample of each sample in dataset C is traversed if there are two samples x belonging to different classesrAnd xw，xrAnd xwThe Euclidean distance therebetween is d (x)r,xw) To representAnd there is no other sample xlLet d (x)r,xl)<d(xr,xw) Or d (x)w,xl)<d(xr,xw) Then (x)r,xw) And forming a contact pair, and deleting the non-fault samples in the contact pair to form a new non-fault sample set.
10. A computer readable storage medium storing one or more programs, characterized in that: the one or more programs include instructions which, when executed by a computing device, cause the computing device to perform any of the typhoon disaster power distribution network tower failure prediction methods of claims 1-6.
CN201911352748.7A 2019-12-25 2019-12-25 Typhoon disaster power distribution network tower fault prediction method and system Pending CN111191832A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911352748.7A CN111191832A (en) 2019-12-25 2019-12-25 Typhoon disaster power distribution network tower fault prediction method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911352748.7A CN111191832A (en) 2019-12-25 2019-12-25 Typhoon disaster power distribution network tower fault prediction method and system

Publications (1)

Publication Number Publication Date
CN111191832A true CN111191832A (en) 2020-05-22

Family

ID=70707470

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911352748.7A Pending CN111191832A (en) 2019-12-25 2019-12-25 Typhoon disaster power distribution network tower fault prediction method and system

Country Status (1)

Country Link
CN (1) CN111191832A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113689053A (en) * 2021-09-09 2021-11-23 国网安徽省电力有限公司电力科学研究院 Strong convection weather overhead line power failure prediction method based on random forest
CN113837477A (en) * 2021-09-27 2021-12-24 西安交通大学 Data dual-drive power grid fault prediction method, device and equipment under typhoon disaster
CN116257761A (en) * 2023-05-15 2023-06-13 北谷电子股份有限公司 Training method and system for tower fault probability prediction model
CN116520817A (en) * 2023-07-05 2023-08-01 贵州宏信达高新科技有限责任公司 ETC system running state real-time monitoring system and method based on expressway

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001351057A (en) * 2000-06-07 2001-12-21 Kajima Corp Predictive system and method and recording medium
CN103557884A (en) * 2013-09-27 2014-02-05 杭州银江智慧城市技术集团有限公司 Multi-sensor data fusion early warning method for monitoring electric transmission line tower
CN109558893A (en) * 2018-10-31 2019-04-02 华南理工大学 Fast integration sewage treatment method for diagnosing faults based on resampling pond
CN110045197A (en) * 2019-02-27 2019-07-23 国网福建省电力有限公司 A kind of Distribution Network Failure method for early warning

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001351057A (en) * 2000-06-07 2001-12-21 Kajima Corp Predictive system and method and recording medium
CN103557884A (en) * 2013-09-27 2014-02-05 杭州银江智慧城市技术集团有限公司 Multi-sensor data fusion early warning method for monitoring electric transmission line tower
CN109558893A (en) * 2018-10-31 2019-04-02 华南理工大学 Fast integration sewage treatment method for diagnosing faults based on resampling pond
CN110045197A (en) * 2019-02-27 2019-07-23 国网福建省电力有限公司 A kind of Distribution Network Failure method for early warning

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
CHEN YU ET AL: "BP Neural Network Based Fault Prediction of Distribution Network during Typhoon Disaster", 《2019 IEEE INNOVATIVE SMART GRID TECHNOLOGIES ASIA》 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113689053A (en) * 2021-09-09 2021-11-23 国网安徽省电力有限公司电力科学研究院 Strong convection weather overhead line power failure prediction method based on random forest
CN113689053B (en) * 2021-09-09 2024-03-29 国网安徽省电力有限公司电力科学研究院 Strong convection weather overhead line power failure prediction method based on random forest
CN113837477A (en) * 2021-09-27 2021-12-24 西安交通大学 Data dual-drive power grid fault prediction method, device and equipment under typhoon disaster
WO2023045278A1 (en) * 2021-09-27 2023-03-30 西安交通大学 Data dual-drive method, apparatus, and device for predicting power grid failure during typhoon
CN113837477B (en) * 2021-09-27 2023-06-27 西安交通大学 Method, device and equipment for predicting power grid faults under typhoon disasters driven by data
CN116257761A (en) * 2023-05-15 2023-06-13 北谷电子股份有限公司 Training method and system for tower fault probability prediction model
CN116520817A (en) * 2023-07-05 2023-08-01 贵州宏信达高新科技有限责任公司 ETC system running state real-time monitoring system and method based on expressway
CN116520817B (en) * 2023-07-05 2023-08-29 贵州宏信达高新科技有限责任公司 ETC system running state real-time monitoring system and method based on expressway

Similar Documents

Publication Publication Date Title
CN111191832A (en) Typhoon disaster power distribution network tower fault prediction method and system
CN109270407B (en) Extra-high voltage direct current transmission line fault reason identification method based on multi-source information fusion
CN103117546B (en) A kind of Ultrashort-term slide prediction method for wind power
CN107124394A (en) A kind of powerline network security postures Forecasting Methodology and system
CN105354643A (en) Risk prediction evaluation method for wind power grid integration
CN109782124A (en) A kind of main adapted integration Fault Locating Method and system based on gradient descent algorithm
CN106611243A (en) Residual correction method for wind speed prediction based on GARCH (Generalized ARCH) model
CN116021981A (en) Method, device, equipment and storage medium for predicting ice coating faults of power distribution network line
Wang et al. Pm2. 5 prediction based on neural network
CN104574211A (en) Power grid dispatching operating risk early warning method and system based on risk source
CN111127242A (en) Power system reliability dynamic real-time assessment method based on small sample data
CN106056305A (en) Power generation system reliability rapid assessment method based on state clustering
Rohaimi et al. 3 Hours ahead of time flood water level prediction using NNARX structure: Case study pahang
CN104834816A (en) Short-term wind speed prediction method
CN112668184A (en) Method and device for calculating wind power outage risk value under typhoon disaster and storage medium
CN113689053B (en) Strong convection weather overhead line power failure prediction method based on random forest
CN103646095B (en) The reliability of a kind of common cause failure based on data-driven judges system and method
CN114266370A (en) Method and system for generating fault handling plan of power grid equipment in typhoon meteorological environment on line and storage medium
CN114358415A (en) Typhoon season overhead line trip prediction method based on interactive hidden Markov model
CN114492578A (en) Electrical equipment operation data classification method based on BP neural network
Sun et al. Reliability prediction of distribution network based on PCA-GA-BP neural network
CN114049014A (en) Method, device and system for evaluating operation state of offshore wind turbine generator
CN113988210A (en) Method and device for restoring distorted data of structure monitoring sensor network and storage medium
Li et al. Prediction Algorithm of Wind Waterlogging Disaster in Distribution Network Based on Multi‐Source Data Fusion
CN113054653A (en) Power system transient stability evaluation method based on VGGNet-SVM

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20200522

RJ01 Rejection of invention patent application after publication