CN116581883A - Power distribution network line loss assessment method and device based on neural network - Google Patents
Power distribution network line loss assessment method and device based on neural network Download PDFInfo
- Publication number
- CN116581883A CN116581883A CN202310527559.9A CN202310527559A CN116581883A CN 116581883 A CN116581883 A CN 116581883A CN 202310527559 A CN202310527559 A CN 202310527559A CN 116581883 A CN116581883 A CN 116581883A
- Authority
- CN
- China
- Prior art keywords
- power data
- neural network
- training
- network model
- sample
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 191
- 238000003062 neural network model Methods 0.000 claims abstract description 131
- 230000008901 benefit Effects 0.000 claims abstract description 27
- 238000007781 pre-processing Methods 0.000 claims abstract description 17
- 239000011159 matrix material Substances 0.000 claims description 61
- 210000002569 neuron Anatomy 0.000 claims description 48
- 230000008569 process Effects 0.000 claims description 23
- 239000013598 vector Substances 0.000 claims description 21
- 238000010606 normalization Methods 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 18
- 238000010276 construction Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 10
- 238000011156 evaluation Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 description 10
- 238000004891 communication Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 5
- 230000005611 electricity Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Classifications
-
- 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
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
- G06N3/065—Analogue means
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- 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
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
- H02J13/00002—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by monitoring
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
- H02J13/00006—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by information or instructions transport means between the monitoring, controlling or managing units and monitored, controlled or operated power network element or electrical equipment
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The application discloses a power distribution network line loss assessment method and device based on a neural network, wherein the method comprises the following steps: the method comprises the steps of obtaining a plurality of pieces of electric power data of a power distribution network, inputting one or more pieces of electric power data serving as target characteristics into a pre-established neural network model to obtain benefit scores of the target characteristics, calculating influence scores of each piece of electric power data on the neural network model, determining the influence scores of the electric power data on damage of the power distribution network, preprocessing the data of each piece of electric power data by the neural network model to obtain a plurality of training sample sets, and training by taking the plurality of training sample sets as training samples. Therefore, the neural network model receives the input of each new energy power data, the strong expression capacity of the neural network is utilized, the complex nonlinear relation among the new energy power data is fully approximated, the fitting precision of the neural network model is higher, and the influence degree of each new energy power data on the line loss factor of the power distribution network is more accurate to evaluate.
Description
Technical Field
The application relates to the technical field of neural networks, in particular to a power distribution network line loss evaluation method and device based on a neural network.
Background
With the development of the power industry, the power industry steps into the stages of large power grids, large units, large capacity, extra-high voltage and smart power grids, so that the complexity of a power system is obviously increased. Meanwhile, in recent years, new energy units are also continuously applied to power systems, such as photovoltaic generator sets, wind turbine sets and the like, which are connected in series, so that larger impact is caused to safe and stable operation of a power grid. The installed capacity of new energy and the rapid increase of the power generation level enable the trend distribution of the power grid to change to a greater extent, and the line loss level of the power distribution network line also changes to a greater extent.
At present, alternating current power flow calculation is mainly adopted for a power distribution network, simulation analysis is carried out on the line loss of the power distribution network, and factors affecting the line loss of the power distribution network are evaluated based on a Mean-shift clustering method, but the influence of new energy on the line loss of the power distribution network is nonlinear and complex, fitting accuracy of the power flow calculation and Mean-shift clustering method is insufficient, and accuracy of line loss level evaluation of a power distribution network line is low.
Disclosure of Invention
In view of the above problems, the present application is provided to provide a method and an apparatus for estimating line loss of a power distribution network based on a neural network, so as to improve accuracy of line loss level estimation of a power distribution network line.
In order to achieve the above object, the following specific solutions are proposed:
a power distribution network line loss assessment method based on a neural network comprises the following steps:
acquiring a plurality of electric power data of a power distribution network;
taking one or more of various electric power data as a target feature, inputting the target feature into a pre-established neural network model, and outputting to obtain the benefit score of the target feature;
calculating the influence score of each piece of electric power data on the neural network model according to the income score of the target feature, and taking the influence score of each piece of electric power data on the neural network model as the influence score of the electric power data on the damage of the power distribution network;
the establishment process of the neural network model comprises the following steps:
carrying out data preprocessing on each item of electric power data to obtain a plurality of training sample sets;
and taking a plurality of batches of training sample sets as training samples, taking the reference line loss rate of the power distribution network as a training label, and training to obtain a neural network model, wherein the reference line loss rate of the power distribution network is calculated based on intelligent ammeter data of the power distribution network.
Optionally, the performing data preprocessing on each item of electric power data to obtain a plurality of training sample sets includes:
constructing a power data set according to each item of power data;
carrying out data standardization on each power data sample feature matrix in the power data set to obtain a power data set with the sample dimension eliminated;
and splitting the electric power data set according to a preset proportion to obtain a plurality of training sample sets.
Optionally, data normalization is performed on each power data sample feature matrix in the power data set to obtain a power data set with the removed sample dimension, including:
performing data standardization processing on each element in each power data sample feature matrix in the power data set by the following formula to obtain standardized elements:
wherein ,xij X is the element of the ith row and the jth column in the characteristic matrix of the power data sample j For each element in the j-th column of the power data sample feature matrix, x' ij Is a standardized element obtained after data standardization treatment;
determining a power data sample feature matrix composed of the standardized elements as a standardized sample feature matrix;
and summarizing the standardized sample feature matrixes to obtain a power data set with the sample dimension eliminated.
Optionally, the building process of the neural network model further includes:
obtaining a test sample set, wherein the test sample set is obtained by splitting the electric power data set according to a preset proportion;
after the training sample sets are used as training samples and the power distribution network reference line loss rate is used as a training label, training to obtain the neural network model, the method further comprises the steps of:
and adjusting parameters of the neural network model through the test sample set to obtain the neural network model with the adjusted parameters.
Optionally, the constructing a power data set according to each item of power data includes:
dividing each piece of power data according to a preset time period to obtain a plurality of power data samples of the piece of power data in each time period;
constructing a power data sample vector of each power data in each period according to a plurality of power data samples of the power data in each period;
splicing the power data sample vectors of each item of power data in each period to obtain a power data sample feature matrix in the period;
and summarizing the characteristic matrix of the power data sample under each period to obtain a power data set.
Optionally, the training sample set is used as training samples, the reference line loss rate of the power distribution network is used as a training label, and the training is performed to obtain the neural network model, including:
Assigning a preset input weight to each neuron of an input layer of the neural network model, wherein the input weights obtained by assigning different neurons in the input layer are different;
inputting each row vector of each power data sample feature matrix in a current training sample set to each neuron of an input layer of the neural network model, and determining an output result of each neuron of an output layer of the neural network model under a first condition and a second condition, wherein the first condition is that the working probability of each neuron of an implicit layer of the neural network model is a preset probability, and the second condition is that a reference line loss rate of a power distribution network is used as a training label;
calculating a training loss value of the output result, and updating the input weight of each neuron on the input side based on the training loss value to obtain each updated input weight;
and taking the next training sample set of the current training sample set as a new current training sample set, distributing the corresponding input weight to each neuron of the input layer of the neural network model, and returning to execute the step of inputting each row of vector of each power data sample feature matrix in each training sample set to each neuron of the input layer of the neural network model until the next training sample set does not exist in the current training sample set, thereby obtaining the neural network model with the training completed.
Optionally, calculating a training loss value of the output result includes:
calculating a training loss value of the output result using the following formula:
wherein ,yk In order for the training tag to be a "training" tag,for the output result, n b The number of samples for the current set of training samples.
Optionally, calculating an impact score of each piece of power data on the neural network model according to the benefit score of the target feature includes:
calculating an impact score for each item of power data against the neural network model using:
wherein ,for the impact score of the power data p on the neural network model, S is the target feature, s|is the number of items of the power data in the target feature, m is the total number of the power data, v (S) is the benefit score of the target feature, and v (S) -v (S\p) is the contribution degree of the power data p to the benefit score of the target feature.
A neural network-based power distribution network line loss assessment device, comprising:
the power data acquisition unit is used for acquiring a plurality of power data of the power distribution network;
the profit score determining unit is used for taking one or more items of electric power data in various items of electric power data as target characteristics, inputting the target characteristics into a pre-established neural network model, and outputting the profit score of the target characteristics;
The damage influence score determining unit is used for calculating an influence score of each piece of electric power data for the neural network model according to the benefit score of the target feature, and taking the influence score of each piece of electric power data for the neural network model as the influence score of the electric power data for damage of the power distribution network;
the model building first unit is used for carrying out data preprocessing on each item of electric power data to obtain a plurality of training sample sets;
the model building second unit is used for taking a plurality of batches of training sample sets as training samples, taking the reference line loss rate of the power distribution network as a training label, and training to obtain a neural network model, wherein the reference line loss rate of the power distribution network is obtained by calculating based on intelligent ammeter data of the power distribution network.
Optionally, the modeling includes:
the power data set construction unit is used for constructing a power data set according to each item of power data;
the data normalization unit is used for performing data normalization on each power data sample feature matrix in the power data set to obtain a power data set with the sample dimension eliminated;
and the electric power data set splitting unit is used for splitting the electric power data set according to a preset proportion to obtain a plurality of training sample sets.
Optionally, the data normalization unit includes:
the first data normalization subunit is configured to perform data normalization processing on each element in each power data sample feature matrix in the power data set by using the following formula to obtain a normalized element:
wherein ,xij Is the number of electric powerX according to the elements of the ith row and the jth column in the sample feature matrix j For each element in the j-th column of the power data sample feature matrix, x' ij Is a standardized element obtained after data standardization treatment;
a second data normalization subunit, configured to determine a power data sample feature matrix composed of each normalized element, as a normalized sample feature matrix;
and the third data standardization subunit is used for summarizing the standardized sample feature matrixes to obtain a power data set with the removed sample size.
Optionally, the apparatus further comprises:
the test sample set acquisition unit is used for acquiring a test sample set, wherein the test sample set is obtained by splitting the electric power data set according to a preset proportion;
and the parameter adjustment unit is used for adjusting parameters of the neural network model through the test sample set after training to obtain the neural network model by taking a plurality of batches of training sample sets as training samples and taking the reference line loss rate of the power distribution network as training labels.
Optionally, the power data set construction unit includes:
the first power data set construction subunit is used for dividing each power data item according to a preset time period by the power data set construction unit to obtain a plurality of power data samples of the power data item in each time period;
a second power data set construction subunit for constructing a power data sample vector for each power data under each period from a plurality of power data samples for the power data under the period;
the third electric power data set construction subunit is used for splicing electric power data sample vectors of all electric power data in each time period to obtain an electric power data sample feature matrix in the time period;
and the fourth power data set construction subunit is used for summarizing the power data sample feature matrixes under each period to obtain a power data set.
Optionally, the modeling includes:
an initial weight distribution unit, configured to distribute a preset input weight to each neuron in an input layer of the neural network model, where input weights obtained by distribution of different neurons in the input layer are different;
the model training unit is used for inputting each row vector of each power data sample feature matrix in the current batch of training sample sets to each neuron of an input layer of the neural network model, and determining an output result of the neuron of an output layer of the neural network model under a first condition and a second condition, wherein the first condition is that the working probability of each neuron of an implicit layer of the neural network model is a preset probability, and the second condition is that the reference line loss rate of a power distribution network is used as a training label;
The training loss calculation unit is used for calculating a training loss value of the output result, updating the input weight of each neuron on the input side based on the training loss value, and obtaining each updated input weight;
and the weight reassigning unit is used for taking the next training sample set of the current training sample set as a new current training sample set, assigning an input weight corresponding to each neuron of the input layer of the neural network model, and returning to the step executed by the model training unit until the next training sample set does not exist in the current training sample set, so as to obtain the neural network model for completing training.
Optionally, the training loss calculation unit includes:
a training loss calculation subunit, configured to calculate a training loss value of the output result using the following formula:
wherein ,yk In order for the training tag to be a "training" tag,for the output result, n b The number of samples for the current set of training samples.
Optionally, the damage impact score determining unit includes:
a damage impact score determination subunit for calculating an impact score for each item of power data for the neural network model using the following formula:
wherein ,for the impact score of the power data p on the neural network model, S is the target feature, s|is the number of items of the power data in the target feature, m is the total number of the power data, v (S) is the benefit score of the target feature, and v (S) -v (S\p) is the contribution degree of the power data p to the benefit score of the target feature.
By means of the technical scheme, one or more pieces of electric power data in the electric power distribution network are used as target characteristics, the target characteristics are input into a pre-established neural network model, the profit score of the target characteristics is output, the influence score of each piece of electric power data on the neural network model is calculated according to the profit score of the target characteristics, the influence score of each piece of electric power data on the neural network model is used as the influence score of each piece of electric power data on the damage of the electric power distribution network, the data preprocessing is carried out on each piece of electric power data in the establishment process of the neural network model, a plurality of batches of training sample sets are obtained, a plurality of batches of training sample sets are used as training samples, the reference line loss rate of the electric power distribution network is used as a training label, the reference line loss rate of the electric power distribution network is obtained through training, and the reference line loss rate of the electric power distribution network is calculated based on intelligent ammeter data of the electric power distribution network. Therefore, the neural network model receives the input of each new energy power data, the complex nonlinear relation among the new energy power data is fully approximated by utilizing the strong expression capability of the neural network, and the fitting precision of the neural network model is higher, so that the influence degree of each new energy power data on the line loss factor of the power distribution network is more accurate to evaluate.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the application. Also, like reference numerals are used to designate like parts throughout the figures. In the drawings:
fig. 1 is a schematic flow chart of evaluating line loss of a power distribution network according to an embodiment of the present application;
fig. 2 is a schematic flow chart of evaluating line loss of a power distribution network according to an embodiment of the present application;
FIG. 3 is a schematic illustration of a neural network for training a neural network model for evaluating line loss of a power distribution network according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of an apparatus for evaluating line loss of a power distribution network according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application 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 application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
The scheme of the application can be realized based on the terminal with the data processing capability, and the terminal can be a computer, a server, a cloud end and the like.
Next, as described in connection with fig. 1, a method for evaluating a line loss of a power distribution network based on a neural network according to the present application may include the following steps:
step S110, acquiring a plurality of electric power data of the power distribution network.
Specifically, each power data of the power distribution network can be power data affecting the line loss of the power distribution network, and the power data can include conventional unit output data, wind power output data, photovoltaic output data, off-grid power data, load level data and the like.
And step S120, taking one or more pieces of electric power data in various pieces of electric power data as target characteristics, inputting the target characteristics into a pre-established neural network model, and outputting a benefit score for obtaining the target characteristics.
Specifically, in a pre-established neural network model, various electric power data have been trained. If the target feature comprises not less than one item of power data, the items of power data in the target feature can represent a combination participating in cooperation for a power distribution network operation event, and the profit score of the target feature can represent the maximum profit obtained by the coordinated cooperation of the items of power data in the target feature.
The process of establishing the neural network model may include:
s1, carrying out data preprocessing on each item of electric power data to obtain a plurality of training sample sets.
It can be appreciated that the influence of each item of power data of the power distribution network on the line loss of the power distribution network is nonlinear and complex, and the sample sizes of different pieces of power data are different, so that data preprocessing is required for each item of power data, the data preprocessing can be data standardization processing, and then a plurality of training sample sets of each item of power data after the data standardization processing are constructed so as to train the neural network model batch by batch.
S2, taking a plurality of batches of training sample sets as training samples, taking the reference line loss rate of the power distribution network as a training label, and training to obtain a neural network model.
Specifically, the reference line loss rate of the power distribution network may be calculated based on smart meter data of the power distribution network.
The calculation mode of the reference line loss rate of the power distribution network can be (electricity purchase quantity-electricity sales quantity)/electricity purchase quantity multiplied by 100%, and the electricity purchase quantity and the electricity sales quantity can be measured by an intelligent ammeter of the power distribution network.
And step S130, calculating the influence score of each piece of electric power data on the neural network model according to the benefit score of the target feature, and taking the influence score of each piece of electric power data on the neural network model as the influence score of the electric power data on the damage of the power distribution network.
According to the power distribution network line loss assessment method based on the neural network, through the acquisition of a plurality of pieces of power data of a power distribution network, one or more pieces of power data in the pieces of power data are used as target characteristics, the target characteristics are input into a pre-established neural network model, the gain score of the target characteristics is output and obtained, the influence score of each piece of power data on the neural network model is calculated according to the gain score of the target characteristics, the influence score of each piece of power data on the neural network model is used as the influence score of each piece of power data on damage of the power distribution network, wherein the establishment process of the neural network model is to conduct data preprocessing on each piece of power data to obtain a plurality of training sample sets, a plurality of batches of training sample sets are used as training samples, the power distribution network reference line loss rate is used as a training label, the power distribution network model is obtained through training, and the power distribution network reference line loss rate is obtained through calculation based on intelligent ammeter data of the power distribution network. Therefore, the neural network model receives the input of each new energy power data, the complex nonlinear relation among the new energy power data is fully approximated by utilizing the strong expression capability of the neural network, and the fitting precision of the neural network model is higher, so that the influence degree of each new energy power data on the line loss factor of the power distribution network is more accurate to evaluate.
In some embodiments of the present application, a process for preprocessing data of each electric power data to obtain a plurality of training sample sets mentioned in the foregoing embodiments is described, where the process may include:
s1, constructing a power data set according to each item of power data.
It will be appreciated that the power data set may comprise a plurality of data points, each of which may comprise a respective item of power data for the same study period.
Specifically, the process of constructing a power data set according to each item of power data may include:
s11, dividing each piece of power data according to a preset time period to obtain a plurality of power data samples of the piece of power data in each time period.
Specifically, the preset period may represent a unit time corresponding to the number of samples of the power data divided in the unit time.
For example, for a sample in a day, if the preset period is 1 hour, 24 power data samples may be divided, each power data sample having a duration of 1 hour.
S12, constructing a power data sample vector of each power data in each period according to a plurality of power data samples of the power data in each period.
And S13, splicing the power data sample vectors of each item of power data in each period to obtain a power data sample feature matrix in the period.
The number of rows of the power data sample feature matrix may be equal to the total number of terms of the power data.
For example, each power data is 5 power data of conventional machine set output data, wind power output data, photovoltaic output data, off-grid power data and load level data, and then the power data sample feature matrix is X, and the vector of each row of Xn is the number of unit samples of each power data in each time period and is also equal to the number of rows of the power data sample characteristic matrix X i A power data sample vector representing the ith power data. Vector of each column of the power data sample feature matrix X>m is power dataThe total number of terms is also equal to the number of columns of the power data sample feature matrix X, i.e., 5.
And S14, summarizing the characteristic matrix of the power data sample under each period to obtain a power data set.
And S2, carrying out data standardization on each power data sample feature matrix in the power data set to obtain a power data set with the sample dimension eliminated.
Specifically, the process of performing data standardization on each power data sample feature matrix in the power data set to obtain the power data set with the removed sample dimension may include:
S21, carrying out data standardization processing on each element in each electric power data sample feature matrix in the electric power data set by the following formula to obtain standardized elements:
wherein ,xij X is the element of the ith row and the jth column in the characteristic matrix of the power data sample j For each element in the j-th column of the power data sample feature matrix, x' ij Is a standardized element obtained after data standardization processing.
S22, determining a power data sample feature matrix composed of the standardized elements as a standardized sample feature matrix.
It will be appreciated that the object of the data normalization process is a single element in the power data sample feature matrix in the power data set, so that the individual elements after the data normalization process may be recombined to obtain the normalized sample feature matrix.
And S23, summarizing the characteristic matrixes of the standardized samples to obtain a power data set with the removed sample dimension.
S3, splitting the electric power data set according to a preset proportion to obtain a plurality of training sample sets.
Specifically, the electric power data set is split according to a preset proportion, and a test sample set can be obtained in addition to a plurality of training sample sets.
The preset proportion can be customized, and represents the proportion of the power data set divided into sample sets for training and used for testing, for example, a plurality of training sample sets account for 80% of the power data set, and a plurality of test sample sets account for 20% of the power data set.
According to the power distribution network line loss evaluation method based on the neural network, through data preprocessing of all electric power data in a data standardization mode, influence caused by difference of dimensions among samples can be eliminated through all electric power data, training data of more fitting model training is provided for subsequent neural network model training, and accuracy of subsequent neural network model training is improved.
In some embodiments of the present application, the process of training to obtain the neural network model is described by using a plurality of training sample sets as training samples and using the reference line loss rate of the power distribution network as training labels, where the process may include:
s1, distributing preset input weights to each neuron of an input layer of the neural network model.
Wherein, the input weights obtained by the allocation of different neurons in the input layer are different.
Specifically, a gaussian distribution with an average value of 1 and a standard deviation of 0.1 may be selected, and a plurality of preset input weights may be determined according to the number of columns of the feature matrix of each power data sample.
It will be appreciated that each column in the power data sample feature matrix is input to a corresponding neuron in the input layer, and thus each input weight may be assigned to a neuron corresponding to each column in the power data sample feature matrix.
S2, inputting each row of vector of each power data sample feature matrix in the current training sample set to each neuron of the input layer of the neural network model, and determining the output result of the neurons of the output layer of the neural network model under a first condition and a second condition.
Specifically, the first condition is that the working probability of each neuron in the hidden layer of the neural network model is a preset probability, so that each neuron in the hidden layer has the possibility of working with the preset probability, and the preset probability can be customized, for example, 0.5. The connection between layers is reduced through a Dropout technology, a preset probability is given to neurons of the hidden layers, and the situation that part of neurons of the hidden layers work and part of neurons of the hidden layers do not work occurs, so that generalization of a neural network model is enhanced. While neurons of both the input and output layers function properly.
The neural network model can be trained by using Mini-batch and input into the neural network model batch by batch to obtain the output result of each batch.
For example, as shown in fig. 2, the power data sample feature matrix is used as an input feature matrix, and comprises m rows and n columns, each row of vectors is input to each neuron corresponding to an input layer of the neural network model, the input layer is transferred to one neuron of an output layer through an hidden layer, and part of neurons in the hidden layer do not work due to the setting of preset probability, so that the neural network model has stronger generalization
And the second condition is that the reference line loss rate of the power distribution network is used as a training label.
S3, calculating a training loss value of the output result, and updating the input weight of each neuron on the input side based on the training loss value to obtain each updated input weight.
Specifically, the process of calculating the training loss value of the output result may include:
calculating a training loss value of the output result using the following formula:
wherein ,yk In order for the training tag to be a "training" tag,for the output result, n b For the reason ofNumber of samples of the previous training sample set.
S4, judging whether the current training sample set has the next training sample set, if so, executing S5, and if not, determining that the neural network model has completed training.
It will be appreciated that if there is no next training sample set, it may be indicative that all training sample sets have been taken to train the neural network model, and then it may be determined that the neural network model has completed training.
Further, after the neural network model is trained, parameters of the trained neural network model can be adjusted through the test sample set, and the neural network model with the adjusted parameters is obtained.
S5, taking the next training sample set of the current training sample set as a new current training sample set, distributing the corresponding input weight to each neuron of the input layer of the neural network model, and returning to the execution step S2.
It will be appreciated that if there is a next training sample set, it may be indicated that the neural network model has not yet completed training, and the next training sample set needs to be input into each neuron of the input layer, so as to continue training the neural network model.
In some embodiments of the present application, the process of calculating the impact score of each piece of power data on the neural network model according to the benefit score of the target feature mentioned in the above embodiments is described, and the process may include:
Calculating an impact score for each item of power data against the neural network model using:
wherein ,for the impact score of power data p on the neural network model, S is the target feature, |S| is the number of items of power data in the target feature, m is the total number of pieces of power data, v (S) is the benefit score of the target feature, and v (S) -v (S\p) is the contribution degree of the power data p to the benefit score of the target feature.
The device for realizing the power distribution network line loss assessment based on the neural network provided by the embodiment of the application is described below, and the device for realizing the power distribution network line loss assessment based on the neural network described below and the method for realizing the power distribution network line loss assessment based on the neural network described above can be referred to correspondingly.
Referring to fig. 3, fig. 3 is a schematic structural diagram of a power distribution network line loss evaluation device based on a neural network according to an embodiment of the present application.
As shown in fig. 3, the apparatus may include:
the power data acquisition unit 11 is used for acquiring a plurality of power data of the power distribution network;
a benefit score determining unit 12, configured to take one or more of the power data as a target feature, input the target feature into a pre-established neural network model, and output a benefit score for obtaining the target feature;
A damage impact score determining unit 13, configured to calculate an impact score of each piece of power data for the neural network model according to the benefit score of the target feature, and take the impact score of each piece of power data for the neural network model as an impact score of the power data for damage of the power distribution network;
a first model building unit 121, configured to perform data preprocessing on each item of electric power data, so as to obtain a plurality of training sample sets;
the model building second unit 122 is configured to train to obtain a neural network model by taking a plurality of batches of the training sample sets as training samples and a reference line loss rate of the power distribution network as training labels, where the reference line loss rate of the power distribution network is calculated based on smart meter data of the power distribution network.
Optionally, the modeling includes:
the power data set construction unit is used for constructing a power data set according to each item of power data;
the data normalization unit is used for performing data normalization on each power data sample feature matrix in the power data set to obtain a power data set with the sample dimension eliminated;
and the electric power data set splitting unit is used for splitting the electric power data set according to a preset proportion to obtain a plurality of training sample sets.
Optionally, the data normalization unit includes:
the first data normalization subunit is configured to perform data normalization processing on each element in each power data sample feature matrix in the power data set by using the following formula to obtain a normalized element:
wherein ,xij X is the element of the ith row and the jth column in the characteristic matrix of the power data sample j For each element in the j-th column of the power data sample feature matrix, x' ij Is a standardized element obtained after data standardization treatment;
a second data normalization subunit, configured to determine a power data sample feature matrix composed of each normalized element, as a normalized sample feature matrix;
and the third data standardization subunit is used for summarizing the standardized sample feature matrixes to obtain a power data set with the removed sample size.
Optionally, the apparatus further comprises:
the test sample set acquisition unit is used for acquiring a test sample set, wherein the test sample set is obtained by splitting the electric power data set according to a preset proportion;
and the parameter adjustment unit is used for adjusting parameters of the neural network model through the test sample set after training to obtain the neural network model by taking a plurality of batches of training sample sets as training samples and taking the reference line loss rate of the power distribution network as training labels.
Optionally, the power data set construction unit includes:
the first power data set construction subunit is used for dividing each power data item according to a preset time period by the power data set construction unit to obtain a plurality of power data samples of the power data item in each time period;
a second power data set construction subunit for constructing a power data sample vector for each power data under each period from a plurality of power data samples for the power data under the period;
the third electric power data set construction subunit is used for splicing electric power data sample vectors of all electric power data in each time period to obtain an electric power data sample feature matrix in the time period;
and the fourth power data set construction subunit is used for summarizing the power data sample feature matrixes under each period to obtain a power data set.
Optionally, the modeling includes:
an initial weight distribution unit, configured to distribute a preset input weight to each neuron in an input layer of the neural network model, where input weights obtained by distribution of different neurons in the input layer are different;
the model training unit is used for inputting each row vector of each power data sample feature matrix in the current batch of training sample sets to each neuron of an input layer of the neural network model, and determining an output result of the neuron of an output layer of the neural network model under a first condition and a second condition, wherein the first condition is that the working probability of each neuron of an implicit layer of the neural network model is a preset probability, and the second condition is that the reference line loss rate of a power distribution network is used as a training label;
The training loss calculation unit is used for calculating a training loss value of the output result, updating the input weight of each neuron on the input side based on the training loss value, and obtaining each updated input weight;
and the weight reassigning unit is used for taking the next training sample set of the current training sample set as a new current training sample set, assigning an input weight corresponding to each neuron of the input layer of the neural network model, and returning to the step executed by the model training unit until the next training sample set does not exist in the current training sample set, so as to obtain the neural network model for completing training.
Optionally, the training loss calculation unit includes:
a training loss calculation subunit, configured to calculate a training loss value of the output result using the following formula:
wherein ,yk In order for the training tag to be a "training" tag,for the output result, n b The number of samples for the current set of training samples.
Optionally, the damage impact score determining unit includes:
a damage impact score determination subunit for calculating an impact score for each item of power data for the neural network model using the following formula:
wherein ,for the impact score of the power data p on the neural network model, S is the target feature, S is the number of terms of the power data in the target feature, m is the total number of the power data, v (S) is the benefit score of the target feature, and v (S) -v (S\p) is the contribution of the power data p to the benefit score of the target featureDegree.
The power distribution network line loss evaluation device based on the neural network provided by the embodiment of the application can be applied to power distribution network line loss evaluation equipment based on the neural network, such as a terminal: cell phones, computers, etc. Optionally, fig. 4 shows a hardware structure block diagram of a power distribution network line loss evaluation device based on a neural network, and referring to fig. 4, the hardware structure of the power distribution network line loss evaluation device based on the neural network may include: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4;
in the embodiment of the application, the number of the processor 1, the communication interface 2, the memory 3 and the communication bus 4 is at least one, and the processor 1, the communication interface 2 and the memory 3 complete the communication with each other through the communication bus 4;
processor 1 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present application, etc.;
The memory 3 may comprise a high-speed RAM memory, and may further comprise a non-volatile memory (non-volatile memory) or the like, such as at least one magnetic disk memory;
wherein the memory stores a program, the processor is operable to invoke the program stored in the memory, the program operable to:
acquiring a plurality of electric power data of a power distribution network;
taking one or more of various electric power data as a target feature, inputting the target feature into a pre-established neural network model, and outputting to obtain the benefit score of the target feature;
calculating the influence score of each piece of electric power data on the neural network model according to the income score of the target feature, and taking the influence score of each piece of electric power data on the neural network model as the influence score of the electric power data on the damage of the power distribution network;
the establishment process of the neural network model comprises the following steps:
carrying out data preprocessing on each item of electric power data to obtain a plurality of training sample sets;
and taking a plurality of batches of training sample sets as training samples, taking the reference line loss rate of the power distribution network as a training label, and training to obtain a neural network model, wherein the reference line loss rate of the power distribution network is calculated based on intelligent ammeter data of the power distribution network.
Alternatively, the refinement function and the extension function of the program may be described with reference to the above.
The embodiment of the present application also provides a storage medium storing a program adapted to be executed by a processor, the program being configured to:
acquiring a plurality of electric power data of a power distribution network;
taking one or more of various electric power data as a target feature, inputting the target feature into a pre-established neural network model, and outputting to obtain the benefit score of the target feature;
calculating the influence score of each piece of electric power data on the neural network model according to the income score of the target feature, and taking the influence score of each piece of electric power data on the neural network model as the influence score of the electric power data on the damage of the power distribution network;
the establishment process of the neural network model comprises the following steps:
carrying out data preprocessing on each item of electric power data to obtain a plurality of training sample sets;
and taking a plurality of batches of training sample sets as training samples, taking the reference line loss rate of the power distribution network as a training label, and training to obtain a neural network model, wherein the reference line loss rate of the power distribution network is calculated based on intelligent ammeter data of the power distribution network.
Alternatively, the refinement function and the extension function of the program may be described with reference to the above.
Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
In the present specification, each embodiment is described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and may be combined according to needs, and the same similar parts may be referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. The power distribution network line loss assessment method based on the neural network is characterized by comprising the following steps of:
acquiring a plurality of electric power data of a power distribution network;
taking one or more of various electric power data as a target feature, inputting the target feature into a pre-established neural network model, and outputting to obtain the benefit score of the target feature;
calculating the influence score of each piece of electric power data on the neural network model according to the income score of the target feature, and taking the influence score of each piece of electric power data on the neural network model as the influence score of the electric power data on the damage of the power distribution network;
The establishment process of the neural network model comprises the following steps:
carrying out data preprocessing on each item of electric power data to obtain a plurality of training sample sets;
and taking a plurality of batches of training sample sets as training samples, taking the reference line loss rate of the power distribution network as a training label, and training to obtain a neural network model, wherein the reference line loss rate of the power distribution network is calculated based on intelligent ammeter data of the power distribution network.
2. The method of claim 1, wherein the performing data preprocessing on the power data to obtain a plurality of training sample sets includes:
constructing a power data set according to each item of power data;
carrying out data standardization on each power data sample feature matrix in the power data set to obtain a power data set with the sample dimension eliminated;
and splitting the electric power data set according to a preset proportion to obtain a plurality of training sample sets.
3. The method of claim 2, wherein data normalization of each power data sample feature matrix in the power data set to obtain a sample size eliminated power data set comprises:
performing data standardization processing on each element in each power data sample feature matrix in the power data set by the following formula to obtain standardized elements:
wherein ,xij X is the element of the ith row and the jth column in the characteristic matrix of the power data sample j For each element in the j-th column of the power data sample feature matrix, x' ij Is a standardized element obtained after data standardization treatment;
determining a power data sample feature matrix composed of the standardized elements as a standardized sample feature matrix;
and summarizing the standardized sample feature matrixes to obtain a power data set with the sample dimension eliminated.
4. The method of claim 2, wherein the neural network model building process further comprises:
obtaining a test sample set, wherein the test sample set is obtained by splitting the electric power data set according to a preset proportion;
after the training sample sets are used as training samples and the power distribution network reference line loss rate is used as a training label, training to obtain the neural network model, the method further comprises the steps of:
and adjusting parameters of the neural network model through the test sample set to obtain the neural network model with the adjusted parameters.
5. The method of claim 2, wherein constructing a power data set from the plurality of power data comprises:
Dividing each piece of power data according to a preset time period to obtain a plurality of power data samples of the piece of power data in each time period;
constructing a power data sample vector of each power data in each period according to a plurality of power data samples of the power data in each period;
splicing the power data sample vectors of each item of power data in each period to obtain a power data sample feature matrix in the period;
and summarizing the characteristic matrix of the power data sample under each period to obtain a power data set.
6. The method according to claim 2, wherein the training with a plurality of training sample sets as training samples and a power distribution network reference line loss rate as training labels to obtain a neural network model includes:
assigning a preset input weight to each neuron of an input layer of the neural network model, wherein the input weights obtained by assigning different neurons in the input layer are different;
inputting each row vector of each power data sample feature matrix in a current training sample set to each neuron of an input layer of the neural network model, and determining an output result of each neuron of an output layer of the neural network model under a first condition and a second condition, wherein the first condition is that the working probability of each neuron of an implicit layer of the neural network model is a preset probability, and the second condition is that a reference line loss rate of a power distribution network is used as a training label;
Calculating a training loss value of the output result, and updating the input weight of each neuron on the input side based on the training loss value to obtain each updated input weight;
and taking the next training sample set of the current training sample set as a new current training sample set, distributing the corresponding input weight to each neuron of the input layer of the neural network model, and returning to execute the step of inputting each row of vector of each power data sample feature matrix in each training sample set to each neuron of the input layer of the neural network model until the next training sample set does not exist in the current training sample set, thereby obtaining the neural network model with the training completed.
7. The method of claim 6, wherein calculating the training loss value for the output result comprises:
calculating a training loss value of the output result using the following formula:
wherein ,yk In order for the training tag to be a "training" tag,for the output result, n b The number of samples for the current set of training samples.
8. The method of claim 1, wherein calculating an impact score for each piece of power data for the neural network model based on the benefit score for the target feature comprises:
Calculating an impact score for each item of power data against the neural network model using:
wherein ,for the impact score of the power data p on the neural network model, S is the target feature, s|is the number of items of the power data in the target feature, m is the total number of the power data, v (S) is the benefit score of the target feature, and v (S) -v (S\p) is the contribution degree of the power data p to the benefit score of the target feature.
9. The utility model provides a distribution network line loss evaluation device based on neural network which characterized in that includes:
the power data acquisition unit is used for acquiring a plurality of power data of the power distribution network;
the profit score determining unit is used for taking one or more items of electric power data in various items of electric power data as target characteristics, inputting the target characteristics into a pre-established neural network model, and outputting the profit score of the target characteristics;
the damage influence score determining unit is used for calculating an influence score of each piece of electric power data for the neural network model according to the benefit score of the target feature, and taking the influence score of each piece of electric power data for the neural network model as the influence score of the electric power data for the damage of the power distribution network;
The model building first unit is used for carrying out data preprocessing on each item of electric power data to obtain a plurality of training sample sets;
the model building second unit is used for taking a plurality of batches of training sample sets as training samples, taking the reference line loss rate of the power distribution network as a training label, and training to obtain a neural network model, wherein the reference line loss rate of the power distribution network is obtained by calculating based on intelligent ammeter data of the power distribution network.
10. The apparatus of claim 9, wherein the model builds a first unit comprising:
the power data set construction unit is used for constructing a power data set according to each item of power data;
the data normalization unit is used for performing data normalization on each power data sample feature matrix in the power data set to obtain a power data set with the sample dimension eliminated;
and the electric power data set splitting unit is used for splitting the electric power data set according to a preset proportion to obtain a plurality of training sample sets.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310527559.9A CN116581883A (en) | 2023-05-10 | 2023-05-10 | Power distribution network line loss assessment method and device based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310527559.9A CN116581883A (en) | 2023-05-10 | 2023-05-10 | Power distribution network line loss assessment method and device based on neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116581883A true CN116581883A (en) | 2023-08-11 |
Family
ID=87535234
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310527559.9A Pending CN116581883A (en) | 2023-05-10 | 2023-05-10 | Power distribution network line loss assessment method and device based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116581883A (en) |
-
2023
- 2023-05-10 CN CN202310527559.9A patent/CN116581883A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sun et al. | Using Bayesian deep learning to capture uncertainty for residential net load forecasting | |
| CN108280552B (en) | Power load prediction method and system based on deep learning and storage medium | |
| CN107506868B (en) | Method and device for predicting short-time power load | |
| Cheng et al. | Enhanced state estimation and bad data identification in active power distribution networks using photovoltaic power forecasting | |
| US20230118702A1 (en) | Method, device and computer readable storage medium for estimating SOC of lithium battery | |
| CN111898247B (en) | Landslide displacement prediction method, landslide displacement prediction equipment and storage medium | |
| CN112834927A (en) | Lithium battery residual life prediction method, system, device and medium | |
| CN114490065A (en) | Load prediction method, device and equipment | |
| CN110634082A (en) | Low-frequency load shedding system operation stage prediction method based on deep learning | |
| CN115358437A (en) | Power supply load prediction method based on convolutional neural network | |
| CN114202174A (en) | Electricity price risk grade early warning method and device and storage medium | |
| CN111612648B (en) | Training method and device for photovoltaic power generation prediction model and computer equipment | |
| Jamshidi et al. | Using artificial neural networks and system identification methods for electricity price modeling | |
| CN115271282A (en) | Customer value determination method and device based on fuzzy logic | |
| CN112561303A (en) | Power system dynamic analysis method based on integrated learning and power grid topological change | |
| Tee et al. | Short-term load forecasting using artificial neural networks | |
| CN112039111A (en) | Method and system for participating in peak regulation capacity of power grid by new energy microgrid | |
| CN116581883A (en) | Power distribution network line loss assessment method and device based on neural network | |
| CN112581311B (en) | Method and system for predicting long-term output fluctuation characteristics of aggregated multiple wind power plants | |
| CN112508734B (en) | Method and device for predicting power generation capacity of power enterprise based on convolutional neural network | |
| CN115935284A (en) | Power grid abnormal voltage detection method, device, equipment and storage medium | |
| CN115146715A (en) | Power utilization potential safety hazard diagnosis method, device, equipment and storage medium | |
| CN114091908A (en) | Power distribution network comprehensive evaluation method, device and equipment considering multi-mode energy storage station | |
| CN113469570A (en) | Information quality evaluation model construction method, device, equipment and storage medium | |
| CN113435915B (en) | Method, device, equipment and storage medium for detecting electricity stealing behavior of user |
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 |