CN117725981A - Power distribution network fault prediction method based on optimal time window mechanism - Google Patents
Power distribution network fault prediction method based on optimal time window mechanism Download PDFInfo
- Publication number
- CN117725981A CN117725981A CN202410176910.9A CN202410176910A CN117725981A CN 117725981 A CN117725981 A CN 117725981A CN 202410176910 A CN202410176910 A CN 202410176910A CN 117725981 A CN117725981 A CN 117725981A
- Authority
- CN
- China
- Prior art keywords
- representing
- hidden layer
- power distribution
- input
- time window
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 230000007246 mechanism Effects 0.000 title claims abstract description 28
- 238000003062 neural network model Methods 0.000 claims abstract description 34
- 239000013598 vector Substances 0.000 claims description 54
- 230000006870 function Effects 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 12
- 238000006243 chemical reaction Methods 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000001186 cumulative effect Effects 0.000 claims description 5
- 238000002372 labelling Methods 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims 2
- 229910000831 Steel Inorganic materials 0.000 description 4
- 239000010959 steel Substances 0.000 description 4
- 230000001052 transient effect Effects 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 2
- 239000004020 conductor Substances 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- 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
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of power distribution network fault prediction, and discloses a power distribution network fault prediction method based on an optimal time window mechanism. The power distribution network fault prediction method based on the optimal time window mechanism is characterized in that an optimal weight is found for the established neural network model, an optimal time window mechanism is established by matching historical data, and an optimal time window width is trainedInputting real-time data as network data, and using output value to generate sawThe tooth state is used for fault prediction, and when the tooth state is iterated for a plurality of times, the intervalWhen the pre-judging fault is about to occur, an alarm is sent out, and when the jagged state iterates for a time intervalWhen the method is used, the pre-judging fault does not occur, and an alarm is not sent out.
Description
Technical Field
The invention relates to the technical field of power distribution network fault prediction, in particular to a power distribution network fault prediction method based on an optimal time window mechanism.
Background
The aerial bare conductors in the Yunnan area have a proportion of more than 70%, the coverage rate of the Yunnan forest is high, natural protection areas are more, when extreme weather such as windy and rainy weather occurs, the bare conductors are extremely prone to ground faults, accidents caused by the faults of the power distribution network occur in the Yunnan mountain area each year, and how to safely and rapidly treat electric shock, mountain fire and personal casualties caused by the ground faults of the power distribution network in the Yunnan mountain area is a long-standing pain point and a long-standing problem.
Disclosure of Invention
(one) solving the technical problems
Aiming at the defects of the prior art, the invention provides a power distribution network fault prediction method based on an optimal time window mechanism, which has the advantages of predicting faults in time, giving an alarm and the like, and solves the technical problems.
(II) technical scheme
In order to achieve the above purpose, the present invention provides the following technical solutions: a power distribution network fault prediction method based on an optimal time window mechanism comprises the following steps:
s1, performing fault marking on historical data of a power distribution network, and establishing a neural network model;
s2, establishing an optimal time window mechanism for judging the fault state;
s3, training the model, and predicting the faults by matching with an optimal time window mechanism.
As a preferred technical solution of the present invention, the neural network model in step S1 includes six portions, and the six portions are sequentially connected, where the first portion is an input layer, the second portion is a first hidden layer, the third portion is a second hidden layer, the fourth portion is a third hidden layer, the fifth portion is an output layer, and the sixth portion is a recursive link.
As a preferred technical scheme of the present invention, the expression of the output result of the neural network model is as follows:
wherein,representing the number of input elements of the input layer,representing the number of iterations of the time,represents the firstThe value of the input of the individual network,the number of the nodes of the input layer is represented,the number of nodes of the first hidden layer is represented,the number of nodes of the second hidden layer is represented,representing input values to a networkAnd subscripts to the activation function of (1)Representing the first hidden layer in the second hidden layerPersonal value, subscriptRepresenting the first hidden layerPersonal value, subscriptRepresenting the first in the input layerThe value of the one of the values,representing the internal slaveTo the point ofIs summed up with the data of (a),representing the internal slaveTo the point ofIs subjected to a cumulative multiplication of the data of (a),representing the internal slaveTo the point ofIs summed.
As a preferable technical scheme of the invention, the expression of the integral input value of the neural network model is as follows:
wherein,representing the overall input values of the neural network model,input element representing input layer, subscriptRepresenting element number, item numberPersonal network input valueThe expression of (2) is as follows:
wherein, in the first iteration, i.e.In the time-course of which the first and second contact surfaces,，representing the time iteration number, superscriptRepresenting the transpose of the matrix,representing the last iteration, i.eThe output value of the time of day,representing the number of input layer nodes.
As the inventionPreferably, the neural network model includes a weight vector between the second hidden layer and the third hidden layerThe expression is as follows:
wherein,representing a weight vector between the second hidden layer and the third hidden layer,representing a second hidden layerThe weight vectors of the respective corresponding nodes,representing the first hidden layer in the second hidden layerThe individual nodes are converted into weight vectors of the third hidden layer,representing the first hidden layer in the second hidden layerThe individual nodes are converted into weight vectors of the third hidden layer,representing the first hidden layer in the second hidden layerThe individual nodes are converted into weight vectors of a third hidden layer;
the neural network model also comprises a weight vector between the first hidden layer and the input layerThe expression of (2) is as follows:
wherein,representing the number of nodes of the first hidden layer,representing the input layerThe corresponding nodes are respectively converted into a first hidden layerThe weight vectors of the respective corresponding nodes,representing the first of the input layersConversion of the corresponding nodes into the first hidden layerThe weight vectors of the respective corresponding nodes,representing the first of the input layersConversion of the corresponding nodes into the first hidden layerThe weight vectors of the respective corresponding nodes,representing the first of the input layersConversion of the corresponding nodes into the first hidden layerThe weight vectors of the respective corresponding nodes,representing the first of the input layersConversion of the corresponding nodes into the first hidden layerWeight vectors of the corresponding nodes;
the neural network model also includes an output of the first hidden layerThe expression is as follows:
wherein,a variable representing the input level of the device,representing the internal slaveTo the point ofIs summed up with the data of (a),the weight vector between the first hidden layer and the input layer,represent the firstConversion of the corresponding nodes of the input layers into the first hidden layerThe weight vectors of the respective corresponding nodes,represent the firstAnd (3) fixing the connection weight between the first hidden layer and the second hidden layer of the neural network model to be 1 according to the network input value.
As a preferred technical scheme of the invention, the connection mode between the first hidden layer and the second hidden layer comprises full connection and partial connection, and when the connection mode between the first hidden layer and the second hidden layer is full connection, the connection combination number of the neurons isWhen the connection mode between the first hidden layer and the second hidden layer is partial connection, the number of connection combinations of the neurons is smaller thanThe network connection combination expression between the first hidden layer and the second hidden layer is as follows:
wherein,connecting the first hidden layer to the second hidden layerA collection of individual neurons that are selected from the group,representing the internal slaveTo the point ofIs summed up with the data of (a),connecting the first hidden layer to the whole second hidden layerA collection of individual neurons that are selected from the group,representing the internal slaveTo the point ofIs summed up with the data of (a),is thatThe number of the elements in the steel sheet is equal to the number of the elements in the steel sheet,is thatThe number of elements in the matrix;
the output of the second hidden layerThe expression is as follows:
wherein,representing the first hidden layer in the second hidden layerThe output value of the individual node(s),represents the first hidden layerThe number of output values is chosen to be the number of output values,variables representing the input layer correspond to the transformed expression of the first hidden layer output,representing a set of pairsThe data in (a) are multiplied,representing the internal slaveTo the point ofIs summed up with the data of (a),representing a cumulative multiplication after summation.
As a preferred embodiment of the present invention, the output of the third hidden layerThe expression is as follows:
wherein,represent the firstThe output value of the third hidden layer is iterated a number of times,the activation function is represented as a function of the activation,representing a collectionMiddle (f)The weight of the weight is calculated,representing a weight vector between the second hidden layer and the third hidden layer,representing the total output of the second hidden layer,representing the second hidden layerThe number of output values is chosen to be the number of output values,representing the internal slaveTo the point ofIs summed.
As a preferable technical scheme of the invention, the neural network model training set and the error functionThe expression is as follows:
wherein,a sample set representing a training is presented,represent the firstThe number of sets of inputs,represent the firstThe output set of the individual(s),the total number of elements representing the sample set,denoted as the firstThe number of elements to be added to the composition,represent the firstSample set of individual trainingThe output value of the third hidden layer is output for a number of iterations,representing a weight vector between the second hidden layer and the third hidden layer,represent the firstAll outputs of the second hidden layer in the training sample set,representing the number of pairs of internal partsAccording to the square of the square,representing an internal slaveTo the point ofIs summed up with the data of (a),representing error functionsConstant value of (a);
when training is carried out each time, iteration is carried out on the values, and the iteration expression is as follows:
wherein,represent the firstError function at multiple iterationsFor a pair ofIs used for the weight change amount of (a),representing a weight vector between the second hidden layer and the third hidden layer,represent the firstError function at multiple iterationsFor a pair ofIs used for the weight change amount of (a),representing a weight vector between the first hidden layer and the input layer,representing error functionsFor a pair ofThe deviation is calculated and the deviation is calculated,representing error functionsFor a pair ofThe deviation is calculated and the deviation is calculated,representing the learning rate of the training process.
As a preferable technical scheme of the invention, the specific process of the step S2 is as follows:
s2.1, reading historical fault information;
s2.2, searching an optimal time window, searching a serrated time interval upwards according to the historical fault information, and continuously comparing with the historical data until determining an optimal time window width；
S2.3, the actual saw tooth shapeTime width of (2) and optimal time window widthA comparison is made.
As a preferable technical scheme of the invention, the process of the S3 is as follows:
s3.1, training a neural network model;
s3.2, taking the test set data as network input, and comparing the output value with the fault labeling value;
s3.3, inputting real-time data as network data, and predicting faults by using a state that the output value is saw-toothed:
when the jagged state iterates the time intervalWhen the pre-judging fault is about to happen, an alarm is sent out;
when the jagged state iterates the time intervalWhen the method is used, the pre-judging fault does not occur, and an alarm is not sent out;
wherein,indicating the width of the optimal time window,representing the jagged state iteration number interval.
Compared with the prior art, the invention provides a power distribution network fault prediction method based on an optimal time window mechanism, which has the following beneficial effects:
the invention searches the optimal weight of the established neural network model, establishes an optimal time window mechanism by matching with historical data, and trains out an optimal time window widthThen the real-time data is used as network data to input, and the output value is used to generate jagged state to make faultPrediction, when the jagged state iterates the time intervalWhen the pre-judging fault is about to occur, an alarm is sent out, and when the jagged state iterates for a time intervalWhen the method is used, the pre-judging fault does not occur, and an alarm is not sent out. Therefore, the effects of predicting the transient faults and eliminating hidden danger are achieved.
Drawings
FIG. 1 is a schematic diagram of a neural network according to the present invention;
FIG. 2 is a diagram illustrating a mechanism of a jagged optimal time window according to the present invention;
FIG. 3 is a flow chart of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1-3, the present invention provides the following technical solutions: a power distribution network fault prediction method based on an optimal time window mechanism comprises the following steps:
s1, performing fault marking on historical data of a power distribution network, and establishing a neural network model;
in step S1, the neural network model includes six parts, and the six parts are sequentially connected, where the first part is an input layer, the second part is a first hidden layer, the third part is a second hidden layer, the fourth part is a third hidden layer, the fifth part is an output layer, and the sixth part is a recursive link (which is a dotted line part in fig. 1).
The expression of the output result of the neural network model is as follows:
wherein,representing the number of input elements of the input layer,representing the number of iterations of the time,represents the firstThe value of the input of the individual network,the number of the nodes of the input layer is represented,the number of nodes of the first hidden layer is represented,the number of nodes of the second hidden layer is represented,representing input values to a networkAnd subscripts to the activation function of (1)Representing the first hidden layer in the second hidden layerPersonal value, subscriptRepresenting the first hidden layerPersonal value, subscriptRepresenting the first in the input layerThe value of the one of the values,representing the internal slaveTo the point ofIs summed up with the data of (a),representing the internal slaveTo the point ofIs subjected to a cumulative multiplication of the data of (a),representing the internal slaveTo the point ofIs summed.
The overall input value expression of the neural network model is as follows:
wherein,representing the overall input values of the neural network model,input element representing input layer, subscriptRepresenting element number, item numberPersonal network input valueThe expression of (2) is as follows:
wherein, in the first iteration, i.e.In the time-course of which the first and second contact surfaces,，representing the time iteration number, superscriptRepresenting the transpose of the matrix,representing the last iteration, i.eThe output value of the time of day,representing the number of input layer nodes.
The neural network model comprises a weight vector between the second hidden layer and the third hidden layerThe expression is as follows:
wherein,representing a weight vector between the second hidden layer and the third hidden layer,representing a second hidden layerThe weight vectors of the respective corresponding nodes,representing the first hidden layer in the second hidden layerThe individual nodes are converted into weight vectors of the third hidden layer,representing the first hidden layer in the second hidden layerThe individual nodes are converted into weight vectors of the third hidden layer,representing the first hidden layer in the second hidden layerThe individual nodes are converted into weight vectors of a third hidden layer;
the neural network model also comprises a weight vector between the first hidden layer and the input layerThe expression of (2) is as follows:
wherein,representing the number of nodes of the first hidden layer,representing the input layerThe corresponding nodes are respectively converted into a first hidden layerThe weight vectors of the respective corresponding nodes,representing the first of the input layersConversion of the corresponding nodes into the first hidden layerThe weight vectors of the respective corresponding nodes,representing the first of the input layersConversion of the corresponding nodes into the first hidden layerThe weight vectors of the respective corresponding nodes,representing the first of the input layersConversion of the corresponding nodes into the first hidden layerThe weight vectors of the respective corresponding nodes,representing the first of the input layersConversion of the corresponding nodes into the first hidden layerWeight vectors of corresponding nodes
Variables of input layerOutput corresponding to the first hidden layerThe expression is as follows:
wherein,a variable representing the input level of the device,representing the internal slaveTo the point ofIs summed up with the data of (a),the weight vector between the first hidden layer and the input layer,represent the firstConversion of the corresponding nodes of the input layers into the first hidden layerThe weight vectors of the respective corresponding nodes,represent the firstAnd (3) fixing the connection weight between the first hidden layer and the second hidden layer of the neural network model to be 1 according to the network input value.
The connection mode between the first hidden layer and the second hidden layer comprises full connection and partial connection, and when the connection mode between the first hidden layer and the second hidden layer is full connection, the connection combination number of the neurons isWhen the connection mode between the first hidden layer and the second hidden layer is partial connection, the number of connection combinations of the neurons is smaller thanThe network connection combination expression between the first hidden layer and the second hidden layer is as follows:
wherein,connecting the first hidden layer to the second hidden layerA collection of individual neurons that are selected from the group,representing the internal slaveTo the point ofIs summed up with the data of (a),connecting the first hidden layer to the whole second hidden layerA collection of individual neurons that are selected from the group,representing the internal slaveTo the point ofIs summed up with the data of (a),is thatThe number of the elements in the steel sheet is equal to the number of the elements in the steel sheet,is thatThe number of elements in the matrix;
output of the second hidden layerThe expression is as follows:
wherein,representing the first hidden layer in the second hidden layerThe output value of the individual node(s),represents the first hidden layerThe number of output values is chosen to be the number of output values,variables representing the input layer correspond to the transformed expression of the first hidden layer output,representing a set of pairsThe data in (a) are multiplied,representing the internal slaveTo the point ofIs summed up with the data of (a),representing a cumulative multiplication after summation.
Output of the third hidden layerThe expression is as follows:
wherein,represent the firstThe output value of the third hidden layer is iterated a number of times,the activation function is represented as a function of the activation,representing a collectionMiddle (f)The weight of the weight is calculated,representing a weight vector between the second hidden layer and the third hidden layer,representing the total output of the second hidden layer,representing the second hidden layerThe number of output values is chosen to be the number of output values,representing the internal slaveTo the point ofIs summed.
Neural network model training set and error functionThe expression is as follows:
wherein,a sample set representing a training is presented,represent the firstThe number of sets of inputs,represent the firstThe output set of the individual(s),the total number of elements representing the sample set,denoted as the firstThe number of elements to be added to the composition,represent the firstSample set of individual trainingThe output value of the third hidden layer is output for a number of iterations,representing a weight vector between the second hidden layer and the third hidden layer,represent the firstAll outputs of the second hidden layer in the training sample set,representing the squaring of the internal data,representing an internal slaveTo the point ofIs summed up with the data of (a),representing error functionsConstant value of (a);
when training is carried out each time, iteration is carried out on the value, and a specific iteration expression is as follows:
wherein,represent the firstError function at multiple iterationsFor a pair ofIs used for the weight change amount of (a),representing a weight vector between the second hidden layer and the third hidden layer,represent the firstError function at multiple iterationsFor a pair ofIs used for the weight change amount of (a),representing a weight vector between the first hidden layer and the input layer,representing error functionsFor a pair ofThe deviation is calculated and the deviation is calculated,representing error functionsFor a pair ofThe deviation is calculated and the deviation is calculated,the learning rate representing the training process is usually a fixed value;
s2, an optimal time window mechanism is established, and faults occurring in the power distribution network are mainly classified into permanent faults and transient faults. For permanent faults, historical fault data is used as network input, and an optimal time window is searched: finding out the sawtooth-shaped output state value, namely, the output value at the previous moment is in a fault state, the current moment is in a normal state, the next moment is in a fault state, and the like, training out an optimal time window widthDegree ofAs long as the zigzag time width (i.e., the zigzag iteration number interval) is greater than or equal to the optimal time window width, the method can be used as a basis for judging the transient fault, so as to predict the transient fault and eliminate hidden danger, as shown in fig. 2;
s3, training the model, and predicting faults by matching with an optimal time window mechanism;
model training, namely dividing historical fault information data into a training set and a testing set by using the thought of supervised learning, and training the neural network by using the training set. The historical data is subjected to fault labeling ('0' is represented as a fault state, '1' is represented as a normal state), then the historical data is used as different network inputs, the neural network model is trained, and the optimal weight is found out on the basis of meeting certain errors.
And secondly, verifying the model, namely taking the test set data as network input, and comparing the output value with the fault labeling value, so as to verify the accuracy and timeliness of the model.
Finally, the fault prediction is carried out by taking real-time data as network data to be input and using the state that the output value is in a saw tooth shape to carry out the fault prediction:
when the jagged state iterates the time intervalWhen the pre-judging fault is about to happen, an alarm is sent out;
when the jagged state iterates the time intervalWhen the method is used, the pre-judging fault does not occur, and an alarm is not sent out.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (10)
1. A power distribution network fault prediction method based on an optimal time window mechanism is characterized by comprising the following steps of: the method comprises the following steps:
s1, performing fault marking on historical data of a power distribution network, and establishing a neural network model;
s2, establishing an optimal time window mechanism for judging the fault state;
s3, training the model, and predicting the faults by matching with an optimal time window mechanism.
2. The power distribution network fault prediction method based on the optimal time window mechanism according to claim 1, wherein the power distribution network fault prediction method is characterized by comprising the following steps of: the neural network model in step S1 includes six parts, and the six parts are sequentially connected, where the first part is an input layer, the second part is a first hidden layer, the third part is a second hidden layer, the fourth part is a third hidden layer, the fifth part is an output layer, and the sixth part is a recursive link.
3. The power distribution network fault prediction method based on the optimal time window mechanism according to claim 2, wherein the power distribution network fault prediction method is characterized by: the expression of the output result of the neural network model is as follows:
；
wherein,representing the number of input elements of the input layer, +.>Representing the number of time iterations, +.>Represents->Personal network input value,/->Representing the number of input layer nodes, < >>Representing the number of nodes of the first hidden layer, +.>Representing the number of nodes of the second hidden layer, +.>Representing input values to the network->Is described in (2), and subscript +.>Representing the +.f in the second hidden layer>Personal value, subscript->Representing the +.f in the first hidden layer>Personal value, subscript->Representing the +.>Personal value (s)/(s)>Representing the interior slave->To->Summation of the data of (2) is performed, < >>Representing the interior slave->To->Is cumulatively multiplied by the data of (2)>Representing the interior slave->To->Is summed.
4. A power distribution network fault prediction method based on an optimal time window mechanism according to claim 3, wherein: the overall input value expression of the neural network model is as follows:
；
wherein,representing the overall input value of the neural network model, +.>Input element representing input layer, subscript +.>Representing element number, < ->Personal network input value +.>The expression of (2) is as follows:
；
wherein, in the first iteration, i.e.When (I)>，/>Represents the time iteration number, superscript +.>Representing the transpose of the matrix>Representing the last iteration, i.e +.>Output value of time>Representing the number of input layer nodes.
5. The power distribution network fault prediction method based on the optimal time window mechanism according to claim 2, wherein the power distribution network fault prediction method is characterized by: the neural network model includes a weight vector between the second hidden layer and the third hidden layerThe expression is as follows:
；
wherein,representing a weight vector between the second hidden layer and the third hidden layer +.>Representing +.>Weight vectors of the respective nodes, +.>Representing the +.>The individual nodes are converted into weight vectors of the third hidden layer,>representing the +.>The individual nodes are converted into weight vectors of the third hidden layer,>representing the +.>The individual nodes are converted into weight vectors of a third hidden layer;
the neural network model also comprises a weight vector between the first hidden layer and the input layerThe expression of (2) is as follows:
；
wherein,representing the number of nodes of the first hidden layer, +.>Representing +.>The corresponding nodes are respectively converted into a first hidden layer +.>Weight vectors of the respective nodes, +.>Representing the%>The corresponding nodes are converted into the first hidden layer +.>Weight vectors of the respective nodes, +.>Representing the%>The corresponding nodes are converted into the first hidden layer +.>Weight vectors of the respective nodes, +.>Representing the%>The corresponding nodes are converted into the first hidden layer +.>Weight vectors of the respective nodes, +.>Representing the%>The corresponding nodes are converted into the first hidden layer +.>Weight vectors of the corresponding nodes;
the neural network model also includes an output of the first hidden layerThe expression is as follows:
；
wherein,variable representing input layer, +_>Representing the interior slave->To->Number of (2)According to the summation, the->Weight vector between first hidden layer and input layer,>indicate->The corresponding node of the input layer is converted into the first hidden layer->Weight vectors of the respective nodes, +.>Indicate->And (3) fixing the connection weight between the first hidden layer and the second hidden layer of the neural network model to be 1 according to the network input value.
6. The power distribution network fault prediction method based on the optimal time window mechanism according to claim 5, wherein the power distribution network fault prediction method is characterized by: the connection mode between the first hidden layer and the second hidden layer comprises full connection and partial connection, and when the connection mode between the first hidden layer and the second hidden layer is full connection, the connection combination number of the neurons isWhen the connection mode between the first hidden layer and the second hidden layer is partial connection, the connection combination number of the neurons is less than +.>The network connection combination expression between the first hidden layer and the second hidden layer is as follows:
；
wherein,connecting the first hidden layer to the second hidden layer>A collection of individual neurons, < >>Representing the interior slave->To->Summation of the data of (2) is performed, < >>Connecting the first hidden layer for the whole second hidden layer>A collection of individual neurons, < >>Representing the interior slave->To->Summation of the data of (2) is performed, < >>Is->The number of elements in->Is->The number of elements in the matrix;
the output of the second hidden layerThe expression is as follows:
；
wherein,representing the +.>Output value of individual node,/>Represents the +.o. of the first hidden layer>The number of output values is chosen to be the number of output values,the variable representing the input layer corresponds to the conversion expression of the first hidden layer output,/for>Representing the set +.>Data in (a) are multiplied by +.>Representing the interior slave->To->Is summed up with the data of (a),representing a cumulative multiplication after summation.
7. The power distribution network fault prediction method based on the optimal time window mechanism according to claim 6, wherein the power distribution network fault prediction method is characterized by: the output of the third hidden layerThe expression is as follows:
；
wherein,indicate->Iterating the output value of the third hidden layer, < >>Representing an activation function->Representation set->Middle->Personal weight(s)>Representing a weight vector between the second hidden layer and the third hidden layer +.>Representing the total output of the second hidden layer, +.>Represents the +.>Output value->Representing the interior slave->To->Is summed.
8. The power distribution network fault prediction method based on the optimal time window mechanism according to claim 7, wherein: the neural network model training set and the error functionThe expression is as follows:
；
；
wherein,sample set representing training, ++>Indicate->Input set->Indicate->Output set of individuals, < >>Total number of elements representing sample set, +.>Denoted as +.>Element(s)>Indicate->Sample set of individual training->Outputting the output value of the third hidden layer by iteration, < ->Representing a weight vector between the second hidden layer and the third hidden layer +.>Indicate->All outputs of the second hidden layer in the sample set of the training,/o>Representing squaring of internal data,/->Representing the internal slave->To->Summation of the data of (2) is performed, < >>Representing error function->Constant value of (a);
when training is carried out each time, iteration is carried out on the values, and the iteration expression is as follows:
；
；
wherein,indicate->Error function at iteration times>For->Weight change of (2)Quantity (S)>Representing a weight vector between the second hidden layer and the third hidden layer +.>Indicate->Error function at iteration times>For->Weight change amount of ∈10->Representing a weight vector between the first hidden layer and the input layer,/>Representing error function->For->Performing partial guide calculation and->Representing error function->For->Performing partial guide calculation and->Representing the learning rate of the training process.
9. The power distribution network fault prediction method based on the optimal time window mechanism according to claim 1, wherein the power distribution network fault prediction method is characterized by comprising the following steps of: the specific process of the step S2 is as follows:
s2.1, reading historical fault information;
s2.2, searching an optimal time window, searching a serrated time interval upwards according to the historical fault information, and continuously comparing with the historical data until determining an optimal time window width；
S2.3, the actual saw-tooth time width and the optimal time window widthA comparison is made.
10. The power distribution network fault prediction method based on the optimal time window mechanism according to claim 9, wherein: the process of S3 is as follows:
s3.1, training a neural network model;
s3.2, taking the test set data as network input, and comparing the output value with the fault labeling value;
s3.3, inputting real-time data as network data, and predicting faults by using a state that the output value is saw-toothed:
when the jagged state iterates the time intervalWhen the pre-judging fault is about to happen, an alarm is sent out;
when the jagged state iterates the time intervalWhen the method is used, the pre-judging fault does not occur, and an alarm is not sent out;
wherein,represents the optimal time window width,/->Representing the jagged state iteration number interval.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410176910.9A CN117725981B (en) | 2024-02-08 | 2024-02-08 | Power distribution network fault prediction method based on optimal time window mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410176910.9A CN117725981B (en) | 2024-02-08 | 2024-02-08 | Power distribution network fault prediction method based on optimal time window mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117725981A true CN117725981A (en) | 2024-03-19 |
| CN117725981B CN117725981B (en) | 2024-04-30 |
Family
ID=90202007
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410176910.9A Active CN117725981B (en) | 2024-02-08 | 2024-02-08 | Power distribution network fault prediction method based on optimal time window mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117725981B (en) |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080250265A1 (en) * | 2007-04-05 | 2008-10-09 | Shu-Ping Chang | Systems and methods for predictive failure management |
| CN102062831A (en) * | 2010-10-29 | 2011-05-18 | 昆明理工大学 | Single-phase permanent fault recognition method for extra-high voltage AC transmission line |
| US20190146021A1 (en) * | 2016-02-19 | 2019-05-16 | General Electric Technology Gmbh | Apparatus for determination of a ground fault and associated method |
| CN112051481A (en) * | 2020-08-12 | 2020-12-08 | 华中科技大学 | Alternating current-direct current hybrid power grid fault area diagnosis method and system based on LSTM |
| CN112230100A (en) * | 2020-09-29 | 2021-01-15 | 山东大学 | Slow-development permanent fault early warning method and system |
| CN112542823A (en) * | 2020-11-05 | 2021-03-23 | 上海合凯电气科技有限公司 | Reclosing control method and system and reclosing control equipment |
| CN113589098A (en) * | 2021-07-12 | 2021-11-02 | 国网河南省电力公司灵宝市供电公司 | Power grid fault prediction and diagnosis method based on big data drive |
| CN114156831A (en) * | 2021-11-22 | 2022-03-08 | 昆明理工大学 | Photoelectric combined instantaneous fault discrimination method |
| CN114429248A (en) * | 2022-03-31 | 2022-05-03 | 山东德佑电气股份有限公司 | Transformer apparent power prediction method |
| US20220221852A1 (en) * | 2021-01-14 | 2022-07-14 | University Of Louisiana At Lafayette | Method and architecture for embryonic hardware fault prediction and self-healing |
| CN115221769A (en) * | 2021-04-15 | 2022-10-21 | 广州中国科学院先进技术研究所 | Fault prediction method, system, electronic equipment and storage medium |
| CN116361624A (en) * | 2023-03-31 | 2023-06-30 | 成都理工大学 | Error feedback-based large-range ground subsidence prediction method and system |
| CN116400172A (en) * | 2023-05-22 | 2023-07-07 | 合肥工业大学 | Cloud-edge cooperative power distribution network fault detection method and system based on random matrix |
| CN116592993A (en) * | 2023-04-11 | 2023-08-15 | 辽宁科技大学 | Mechanical vibration fault diagnosis method based on deep learning |
| CN117118856A (en) * | 2023-08-23 | 2023-11-24 | 中国电信股份有限公司技术创新中心 | Knowledge graph completion-based network fault reasoning method and related equipment |
| CN117434384A (en) * | 2023-11-07 | 2024-01-23 | 南方电网科学研究院有限责任公司 | Power distribution network insulation fault identification method and related device |
| CN117517876A (en) * | 2024-01-04 | 2024-02-06 | 昆明理工大学 | Fault positioning method, fault positioning equipment and storage medium for direct current transmission line |
-
2024
- 2024-02-08 CN CN202410176910.9A patent/CN117725981B/en active Active
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080250265A1 (en) * | 2007-04-05 | 2008-10-09 | Shu-Ping Chang | Systems and methods for predictive failure management |
| CN102062831A (en) * | 2010-10-29 | 2011-05-18 | 昆明理工大学 | Single-phase permanent fault recognition method for extra-high voltage AC transmission line |
| US20190146021A1 (en) * | 2016-02-19 | 2019-05-16 | General Electric Technology Gmbh | Apparatus for determination of a ground fault and associated method |
| CN112051481A (en) * | 2020-08-12 | 2020-12-08 | 华中科技大学 | Alternating current-direct current hybrid power grid fault area diagnosis method and system based on LSTM |
| WO2022068074A1 (en) * | 2020-09-29 | 2022-04-07 | 山东大学 | Early warning method and system for slowly developing permanent fault |
| CN112230100A (en) * | 2020-09-29 | 2021-01-15 | 山东大学 | Slow-development permanent fault early warning method and system |
| CN112542823A (en) * | 2020-11-05 | 2021-03-23 | 上海合凯电气科技有限公司 | Reclosing control method and system and reclosing control equipment |
| US20220221852A1 (en) * | 2021-01-14 | 2022-07-14 | University Of Louisiana At Lafayette | Method and architecture for embryonic hardware fault prediction and self-healing |
| CN115221769A (en) * | 2021-04-15 | 2022-10-21 | 广州中国科学院先进技术研究所 | Fault prediction method, system, electronic equipment and storage medium |
| CN113589098A (en) * | 2021-07-12 | 2021-11-02 | 国网河南省电力公司灵宝市供电公司 | Power grid fault prediction and diagnosis method based on big data drive |
| CN114156831A (en) * | 2021-11-22 | 2022-03-08 | 昆明理工大学 | Photoelectric combined instantaneous fault discrimination method |
| CN114429248A (en) * | 2022-03-31 | 2022-05-03 | 山东德佑电气股份有限公司 | Transformer apparent power prediction method |
| CN116361624A (en) * | 2023-03-31 | 2023-06-30 | 成都理工大学 | Error feedback-based large-range ground subsidence prediction method and system |
| CN116592993A (en) * | 2023-04-11 | 2023-08-15 | 辽宁科技大学 | Mechanical vibration fault diagnosis method based on deep learning |
| CN116400172A (en) * | 2023-05-22 | 2023-07-07 | 合肥工业大学 | Cloud-edge cooperative power distribution network fault detection method and system based on random matrix |
| CN117118856A (en) * | 2023-08-23 | 2023-11-24 | 中国电信股份有限公司技术创新中心 | Knowledge graph completion-based network fault reasoning method and related equipment |
| CN117434384A (en) * | 2023-11-07 | 2024-01-23 | 南方电网科学研究院有限责任公司 | Power distribution network insulation fault identification method and related device |
| CN117517876A (en) * | 2024-01-04 | 2024-02-06 | 昆明理工大学 | Fault positioning method, fault positioning equipment and storage medium for direct current transmission line |
Non-Patent Citations (8)
| Title |
|---|
| DHIYA AL-JUMEILY等: "Predicting Physical Time Series Using Dynamic Ridge Polynomial Neural Networks", PLOS ONE, vol. 9, no. 8, 26 August 2014 (2014-08-26), pages 1 - 15 * |
| FEI DENG等: "The novel characteristics for training Ridge Polynomial neural network based on Lagrange multiplier", ALEXANDRIA ENGINEERING JOURNAL, vol. 67, 15 March 2023 (2023-03-15), pages 93 - 103 * |
| IMAN NIKOOFEKR等: "Nature of fault determination on transmission lines for single phase autoreclosing applications", THE INSTITUTION OF ENGINEERING AND TECHNOLOGY, vol. 12, no. 4, 15 January 2018 (2018-01-15), pages 903 - 911, XP006104153, DOI: 10.1049/iet-gtd.2017.1058 * |
| VIT-AP UNIVERSITY, AMRAVATI等: "Higher Order Neural Network and Its Applications: A Comprehensive Survey", ANALYTICS AND NETWORKING, 31 January 2018 (2018-01-31), pages 695 - 709 * |
| 严秋问;江修波;蔡金锭;: "基于瞬时性故障时频分析的配网绝缘状态监测", 电气开关, no. 02, 15 April 2016 (2016-04-15), pages 39 - 43 * |
| 俞嘉 等: "基于瞬时故障分析的配电网在线绝缘监测", 电气自动化, vol. 44, no. 3, 30 May 2022 (2022-05-30), pages 103 - 106 * |
| 梁林;江亚群;黄纯;: "带并联电抗器的超高压输电线路单相故障识别", 电力系统及其自动化学报, no. 08, 15 August 2016 (2016-08-15), pages 32 - 37 * |
| 邓飞 等: "一种基于链表法和时间延时的配电网故障定位方法", 昆明理工大学学报(自然科学版), vol. 44, no. 2, 15 April 2019 (2019-04-15), pages 63 - 68 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117725981B (en) | 2024-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Abdel-Basset et al. | An efficient teaching-learning-based optimization algorithm for parameters identification of photovoltaic models: Analysis and validations | |
| Jiang | Computation of monthly mean daily global solar radiation in China using artificial neural networks and comparison with other empirical models | |
| Celik | Artificial neural network modelling and experimental verification of the operating current of mono-crystalline photovoltaic modules | |
| Achour et al. | Hybrid model for estimating monthly global solar radiation for the Southern of Algeria:(Case study: Tamanrasset, Algeria) | |
| CN113283155B (en) | Near-surface air temperature estimation method, system, storage medium and equipment | |
| CN110417011A (en) | A kind of online dynamic secure estimation method based on mutual information Yu iteration random forest | |
| CN113505923B (en) | Regional power grid short-term load prediction method and system | |
| CN113723669B (en) | Power transmission line icing prediction method based on Informmer model | |
| CN114676822A (en) | Multi-attribute fusion air quality forecasting method based on deep learning | |
| Matera et al. | Hourly forecasting of the photovoltaic electricity at any latitude using a network of artificial neural networks | |
| CN110807508B (en) | Bus peak load prediction method considering complex weather influence | |
| CN116796168A (en) | CNN-BiLSTM high-altitude multi-factor power transmission line audible noise prediction method based on multi-head attention mechanism | |
| CN116383658A (en) | BP neural network-based solar panel fault diagnosis method and device | |
| Nelega et al. | Prediction of power generation of a photovoltaic power plant based on neural networks | |
| CN110084403B (en) | Multi-period joint probability density prediction method and system for current-carrying capacity of overhead conductor | |
| Abd Rahman et al. | Artificial neural network forecasting performance with missing value imputations | |
| CN114021758A (en) | Operation and maintenance personnel intelligent recommendation method and device based on fusion of gradient lifting decision tree and logistic regression | |
| Chen et al. | Data-driven distribution network topology identification considering correlated generation power of distributed energy resource | |
| CN117725981A (en) | Power distribution network fault prediction method based on optimal time window mechanism | |
| Omer et al. | Adaptive boosting and bootstrapped aggregation based ensemble machine learning methods for photovoltaic systems output current prediction | |
| CN113177717B (en) | Quick evaluation method for toughness of power transmission system based on influence increment sensitivity | |
| CN115577854A (en) | Quantile regression wind speed interval prediction method based on EEMD-RBF combination | |
| Achour et al. | Smart model for accurate estimation of solar radiation | |
| CN113779892A (en) | Wind speed and wind direction prediction method | |
| CN106570561B (en) | A kind of insoluble sediment density forecasting system of insulator surface and method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |