CN112819067A - Method, device, equipment and storage medium for processing bad data of power distribution network - Google Patents
Method, device, equipment and storage medium for processing bad data of power distribution network Download PDFInfo
- Publication number
- CN112819067A CN112819067A CN202110120172.2A CN202110120172A CN112819067A CN 112819067 A CN112819067 A CN 112819067A CN 202110120172 A CN202110120172 A CN 202110120172A CN 112819067 A CN112819067 A CN 112819067A
- Authority
- CN
- China
- Prior art keywords
- data
- bad
- preset
- time
- distribution network
- 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 53
- 238000012545 processing Methods 0.000 title claims abstract description 42
- 238000005259 measurement Methods 0.000 claims abstract description 55
- 238000012937 correction Methods 0.000 claims abstract description 19
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims description 33
- 230000008569 process Effects 0.000 claims description 18
- 238000007781 pre-processing Methods 0.000 claims description 14
- 230000006403 short-term memory Effects 0.000 claims description 14
- 238000003062 neural network model Methods 0.000 claims description 13
- 238000003672 processing method Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 10
- 230000015654 memory Effects 0.000 claims description 9
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 238000012986 modification Methods 0.000 claims description 3
- 230000004048 modification Effects 0.000 claims description 3
- 238000011176 pooling Methods 0.000 description 9
- 230000009471 action Effects 0.000 description 7
- 230000007787 long-term memory Effects 0.000 description 7
- 239000013598 vector Substances 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- G06N3/048—Activation functions
-
- 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
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Bioinformatics & Computational Biology (AREA)
- Primary Health Care (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The application discloses a method, a device, equipment and a storage medium for processing bad data of a power distribution network, which are used for acquiring measurement data in a preset time period in the power distribution network; inputting the measured data into a preset identification model for data classification, and outputting the measured data as normal data or bad data; determining the occurrence time t of bad data in the measured data based on the identification result, and inputting a preset number of normal data before the occurrence time t of the bad data into a preset data correction model to obtain predicted data at the time t; and replacing the bad data at the time t with the predicted data at the time t to obtain corrected measurement data. The method solves the technical problems that after the bad data are identified by adopting a convolutional neural network model, the rejected bad data are not supplemented or corrected, so that the data are incomplete, the observability of a power distribution network system is influenced, and the reliability of a power distribution network state estimation result is reduced.
Description
Technical Field
The present application relates to the field of data processing technologies, and in particular, to a method, an apparatus, a device, and a storage medium for processing bad data of a power distribution network.
Background
In recent years, a power distribution network system realizes a leap development, and meanwhile, the requirements of residents on the power supply quality, reliability and the like of a power grid are increasingly strict. However, because the real-time measurement system is not perfect, the measured data such as node voltage, line circuit, branch power, etc. obtained from the data acquisition and monitoring System (SCADA) may have errors, repetitions, deletions, etc., that is, bad data may appear in the obtained data. The existence of bad data can generate adverse effect on the running state of the power distribution network system, and further the performance of the power distribution network system is affected.
The traditional bad data identification method is mainly based on a state estimation method, and comprises a residual error search method, a non-quadratic criterion method, a zero residual error method, an estimation identification method and the like. These methods usually take weighted residuals or standard residuals as feature values, assume that they obey a certain probability distribution, and then set a threshold value according to a certain confidence level to perform hypothesis testing. After the position of the suspicious data is determined, the suspicious data are removed or weakened from the measured data set, and finally a new state estimation is generated. However, the methods have the common defects that residual pollution and residual inundation are easy to occur, so that missed detection or false detection is caused, and the identification effect is greatly reduced.
In order to solve the problems of the state estimation method, the prior art adopts a convolutional neural network to identify bad data. After a neural network model is trained, a result of identifying most types of measurement errors can be rapidly generated, and a safe and reliable data warehouse is formed in a control center. However, in the prior art, after the bad data is identified, the rejected bad data is not supplemented or corrected, so that the data is incomplete, the observability of a power distribution network system is influenced, and the reliability of a power distribution network state estimation result is reduced.
Disclosure of Invention
The application provides a processing method, a processing device, processing equipment and a storage medium for bad data of a power distribution network, and the processing device, the processing equipment and the storage medium are used for solving the technical problems that after the bad data are identified by a convolutional neural network model, the bad data after being eliminated are not supplemented or corrected, so that the data are incomplete, the observability of the power distribution network system is influenced, and the reliability of a power distribution network state estimation result is reduced.
In view of this, the first aspect of the present application provides a method for processing bad data of a power distribution network, including:
acquiring measurement data in a preset time period in a power distribution network, wherein the measurement data comprises node voltage, line current and branch power;
inputting the measured data into a preset identification model for data classification, and outputting an identification result of the measured data, wherein the identification result of the measured data comprises normal data and bad data;
determining the occurrence time t of bad data in the measured data based on the identification result, and inputting a preset number of normal data before the occurrence time t of the bad data into a preset data correction model to obtain predicted data at the time t;
and replacing the bad data at the moment t with the predicted data at the moment t to obtain the corrected measurement data.
Optionally, the inputting the measured data into a preset identification model for data classification, and outputting the identification result of the measured data further includes:
preprocessing the measurement data through a preset formula, wherein the preset formula is as follows:
wherein,for the j-th dimension data in the i-th measurement data,the j-th dimension data in the ith type of measured data after preprocessing is obtained; i is 1,2 and 3, which respectively represent node voltage, line current and branch power; j is 1,2, …, Ni,NiThe total dimensionality of the data in the ith type of measured data.
Optionally, the configuration process of the preset identification model is as follows:
acquiring a measured data training set, wherein the label of each training sample in the measured data training set is the normal data or the bad data;
inputting the measured data training set into a preset convolutional neural network for training, and outputting a prediction result corresponding to each training sample;
calculating a loss value through a loss function based on the prediction result corresponding to each training sample and the label;
and updating the parameters of the preset convolutional neural network by a gradient descent method based on the loss value until the loss value is lower than a preset threshold value, so as to obtain the preset identification model.
Optionally, the preset data correction model is a Q learning-long and short term memory neural network model;
correspondingly, the inputting a preset number of normal data before the time t when the bad data appears into a preset data correction model to obtain the predicted data at the time t includes:
inputting n normal data before the time t when the bad data appears into the Q learning-long short-term memory neural network model, enabling the Q learning-long short-term memory neural network model to process the n normal data based on the hidden layer state and the unit state of the previous n times t-1, t-2, t.
Optionally, the determining, based on the identification result, a time t when bad data occurs in the measured data, inputting a preset number of normal data before the time t when the bad data occurs to a preset data correction model to obtain predicted data at the time t, and then further including:
inputting the prediction data at the moment t into the preset identification model for data classification, and outputting the identification result of the prediction data at the moment t;
correspondingly, the replacing the bad data at the time t with the predicted data at the time t to obtain the corrected measurement data includes:
and when the identification result of the predicted data at the moment t is the normal data, replacing the bad data at the moment t with the predicted data at the moment t to obtain the corrected measured data.
This application second aspect provides a processing apparatus of distribution network bad data, includes:
the system comprises an acquisition unit, a processing unit and a control unit, wherein the acquisition unit is used for acquiring measurement data in a preset time period in the power distribution network, and the measurement data comprises node voltage, line current and branch power;
the first classification unit is used for inputting the measured data into a preset identification model for data classification and outputting an identification result of the measured data, wherein the identification result of the measured data comprises normal data and bad data;
the input unit is used for determining the occurrence time t of bad data in the measured data based on the identification result, and inputting a preset number of normal data before the occurrence time t of the bad data into a preset data correction model to obtain predicted data at the time t;
and the replacing unit is used for replacing the bad data at the moment t with the predicted data at the moment t to obtain the corrected measurement data.
Optionally, the method further includes:
the preprocessing unit is used for preprocessing the measured data through a preset formula, wherein the preset formula is as follows:
wherein,for the j-th dimension data in the i-th measurement data,the j-th dimension data in the ith type of measured data after preprocessing is obtained; i is 1,2 and 3, which respectively represent node voltage, line current and branch power; j is 1,2, …, Ni,NiThe total dimensionality of the data in the ith type of measured data.
Optionally, the method further includes:
the second classification unit is used for inputting the prediction data at the moment t into the preset identification model for data classification and outputting the identification result of the prediction data at the moment t;
correspondingly, the replacement unit is specifically configured to:
and when the identification result of the predicted data at the moment t is the normal data, replacing the bad data at the moment t with the predicted data at the moment t to obtain the corrected measured data.
The third aspect of the application provides a processing device for bad data of a power distribution network, which comprises a processor and a memory;
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is used for executing the processing method of the bad data of the power distribution network according to any one of the first aspect according to the instructions in the program code.
A fourth aspect of the present application provides a computer-readable storage medium, which is used for storing program codes, wherein the program codes are used for executing the processing method of bad data of the power distribution network according to any one of the first aspects.
According to the technical scheme, the method has the following advantages:
the application provides a processing method of bad data of a power distribution network, which comprises the following steps: acquiring measurement data in a preset time period in a power distribution network, wherein the measurement data comprises node voltage, line current and branch power; inputting the measured data into a preset identification model for data classification, and outputting an identification result of the measured data, wherein the identification result of the measured data comprises normal data and bad data; determining the occurrence time t of bad data in the measured data based on the identification result, and inputting a preset number of normal data before the occurrence time t of the bad data into a preset data correction model to obtain predicted data at the time t; and replacing the bad data at the time t with the predicted data at the time t to obtain corrected measurement data.
According to the method, after the preset identification model is adopted to identify the bad data in the measured data, the preset number of normal data before the time t when the bad data appears are input into the preset data correction model to predict the normal measured data at the time t, the predicted data at the time t are obtained, then the predicted data at the time t is used for replacing the bad data at the time t, the corrected measured data are obtained, the bad data are eliminated, and the integrity of the data is also ensured.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive exercise.
Fig. 1 is a schematic flow chart of a method for processing bad data of a power distribution network according to an embodiment of the present application;
fig. 2 is another schematic flow chart of a method for processing bad data of a power distribution network according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of a processing apparatus for bad data of a power distribution network according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of a preset identification model according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of a Q learning-long short term memory neural network model according to an embodiment of the present application.
Detailed Description
The application provides a processing method, a processing device, processing equipment and a storage medium for bad data of a power distribution network, and the processing device, the processing equipment and the storage medium are used for solving the technical problems that after the bad data are identified by a convolutional neural network model, the bad data after being eliminated are not supplemented or corrected, so that the data are incomplete, the observability of the power distribution network system is influenced, and the reliability of a power distribution network state estimation result is reduced.
In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
For easy understanding, please refer to fig. 1, an embodiment of a method for processing bad data of a power distribution network provided in the present application includes:
Acquiring historical and real-time data of node voltage in the power distribution network for one year from a data acquisition and monitoring system, and constructing a node voltage data matrix:
in the formula, V (i, t) is a measured voltage value of the ith node in the distribution network at time t, i is 1, 2. T is 1,2, T is the total number of sampling times per day; the number of data samples N is 365 because 365 days a year.
Acquiring historical and real-time data of line current in the power distribution network for one year, and constructing a line current data matrix:
in the formula, I (L, t) is a measured current value of the first line in the power distribution network at time t, wherein L is 1, 2. T is 1,2, T is the total number of sampling times per day; the number of data samples N is 365.
Acquiring historical and real-time data of branch power in the power distribution network for one year, and constructing a branch power data matrix:
in the formula, P (K, t) is a measured power value of the kth branch in the power distribution network at time t, K is 1,2, and K is the total number of branches in the power distribution network; t is 1,2, T is the total number of sampling times per day; the number of data samples N is 365.
And 102, inputting the measured data into a preset identification model for data classification, and outputting an identification result of the measured data, wherein the identification result of the measured data comprises normal data and bad data.
The preset identification module is used for identifying whether the input measurement data is normal data or bad data, and includes a plurality of convolution layers, a plurality of pooling layers, a full-link layer and a Softmax classifier, please refer to fig. 4. The preset identification module processes the input measurement data as follows:
1) and (3) rolling layers: the convolution layer carries out convolution operation to each dimension data of input through a plurality of convolution kernel effect successive layers to obtain every layer of convolution layer's different characteristic data through nonlinear activation function, export through the last layer of convolution kernel of combination, and then can extract the contact successive layer of hiding in data inside, the concrete calculation process of convolution layer is:
in the formula,the output of the mth layer convolution layer of the jth dimension data in the ith type measurement data; i is 1,2 and 3, which respectively represent node voltage, line current and branch power; j is 1,2, …, Ni,p=1,2,…,Ni,NiThe total dimension of the data in the ith type measured data;inputting the mth layer convolution layer of jth dimension data in the ith type measurement data;the m-th layer convolution kernel is the p-th dimension data and the j-th dimension data in the ith type of measurement data;the offset of the mth layer convolution layer of the jth dimension data in the ith class data.
2) A pooling layer: the pooling layer is a dimension reduction operation, namely, the preset identification model can pay more attention to the features rather than specific positions of the features, so that the learning process comprises more degrees of freedom and can tolerate the micro displacement of some features; and dimension reduction can be carried out on the features, so that the model can extract wider features, the parameter quantity in the learning process of the preset identification model is reduced, and the calculation process of the pooling layer is as follows:
in the formula,the output of the mth layer pooling layer of the jth dimension data in the ith type of measurement data; i is 1,2 and 3, which respectively represent node voltage, line current and branch power; j is 1,2, …, Ni,NiMeasuring the total dimension of the data in the data set of the ith type;the weight of the mth layer pooling layer of the jth dimension data in the ith type of measurement data; down (#) is a down-sampling function;inputting the mth layer pooling layer of the jth dimension data in the ith type of measurement data;bias of mth layer pooling layer for jth dimension data of ith type of metrology data.
3) Full connection layer: the full-connection layer performs full-connection on the characteristics output by the last convolution layer, and the specific processing process is as follows:
in the formula,the j-dimension characteristic vector in the i-type measurement data is obtained; i is 1,2 and 3, which respectively represent node voltage, line current and branch power; j is 1,2, …, Ni,NiThe total dimension of the data in the ith type measured data; b0Is an offset; omega0Is a weight; f (-) is the activation function.
4) Softmax classifier: the fully-connected output is processed to identify the type of the various measurement data.
Furthermore, considering that certain correlation and difference exist between input measurement data, and the measurement data are not acquired from the same equipment, and thus have different physical meanings and dimensions, if proper processing is not performed, the identification effect of the preset identification model is affected. Therefore, in order to eliminate dimensions, weaken the physical significance of data, weaken the individual difference of the data, and balance the data dimension, so that the data are in a similar scale, the embodiment of the present application preprocesses the measured data, and before inputting the measured data into a preset identification model for data classification and outputting the identification result of the measured data, the measured data can be preprocessed by a preset formula, wherein the preset formula is as follows:
wherein,for the j-th dimension data in the i-th measurement data,the j-th dimension data in the ith type of measured data after preprocessing is obtained; i is 1,2 and 3, which respectively represent node voltage, line current and branch power; j is 1,2, …, Ni,NiThe total dimensionality of the data in the ith type of measured data.
Further, the configuration process of the preset identification model comprises the following steps:
and A1, acquiring a measured data training set, wherein the label of each training sample in the measured data training set is normal data or poor data.
The data format of the training sample is the same as the format of the measured data to be tested, and is different in that the training sample is labeled, i.e., normal data or bad data is known for the training sample. Assume that a training set of metrology data is obtained asContains N types of training samples, wherein dj∈RL×1The training sample is labeledYj∈[1,N]。
And A2, inputting the measured data training set into a preset convolutional neural network for training, and outputting a prediction result corresponding to each training sample.
Before the measured data training set is input to the preset convolutional neural network for training, the measured data training set may be preprocessed, and the specific process of preprocessing may refer to the preprocessing process in the foregoing steps, which is not described herein again.
The training sample is subjected to feature extraction and processing through a convolutional layer, a pooling layer and a full-connection layer in a preset convolutional neural network, and a Softmax classifier predicts the probability P (Y) of the training sample in a certain classj=N|dj) The specific calculation process is as follows:
wherein λ ═ λ1,λ2,…,λN]TAnd characterizing the parameter vector of the Softmax classifier, wherein U is a final prediction result, and the sum of all prediction results is 1.
A3, calculating a loss value by a loss function based on the prediction result and the label corresponding to each training sample.
The loss function is:
wherein j is 1,2, …, Ni,NiMeasuring the total dimensionality of the data in the data training set for the ith type; 1 {. is an indicative function, i.e. 1 when the class attribute is correct, and 0 otherwise; delta is an attenuation coefficient, and certain parameters of the classifier are set to be 0, so that the generalization capability is improved; m is 1,2, …, M is preset convolution spiritAnd outputting the total layer number through the network.
And A4, updating the parameters of the preset convolutional neural network by a gradient descent method based on the loss value until the loss value is lower than a preset threshold value, and obtaining a preset identification model.
And updating the parameters of the preset convolutional neural network by a gradient descent method based on the loss value until the loss value J (lambda) is lower than a preset threshold value, and taking the latest parameter-updated preset convolutional neural network as a preset identification model.
And 103, determining the occurrence time t of the bad data in the measured data based on the identification result, and inputting a preset number of normal data before the occurrence time t of the bad data into a preset data correction model to obtain predicted data at the time t.
The preset data modification model is a Q learning-long short term memory neural network model, please refer to fig. 5. Inputting n normal data before the time t when the bad data appears into a Q learning-long and short term memory neural network model, enabling the Q learning-long and short term memory neural network model to process the n normal data based on the hidden layer state and the unit state of the previous n times t-1, t-2, t. The specific processing process of the Q learning-long and short term memory neural network model on the input data is as follows:
1) determining the occurrence time t of bad data in the measured data based on the identification result, and acquiring n normal data before the occurrence time t of the bad dataAnd obtaining an input vector which corresponds to one input unit of the long and short term memory network in the Q learning-long and short term memory neural network model.
2)[ht-1,Ct-1],[ht-2,Ct-2],…,[ht-n,Ct-n]Respectively the current time input vector in the long-short term memory neural networkLast moment hidden layer state ht-1,ht-2,…,ht-nAnd cell state C at the previous timet-1,Ct-2,…,Ct-n. The hidden layer state and the unit state correspond to the other two input units of the long-term and short-term memory neural network, the interior of the hidden layer state and the unit state is composed of three gates, namely a forgetting gate FtInput door ItAnd an output gate OtTo control the discarding and inheriting of information, the calculation formulas of the three gates are respectively as follows:
wherein, sigma is a Sigmoid activation function; wF,WI,WORespectively are weight matrixes of a forgetting gate, an input gate and an output gate;indicating that layer state h is to be hiddent-1And an input vectorConnecting to obtain a new vector; b isF、BI、BORespectively are offset vectors of the forgetting gate, the input gate and the output gate.
3) In the Q learning mechanism, Qk[(ht-1,Ct-1)k,ak],Qk[(ht-2,Ct-2)k,ak],…,Qk[(ht-n,Ct-n)k,ak]Are respectively a passing state [ ht-1,Ct-1]k,[ht-2,Ct-2]k,…,[ht-n,Ct-n]kAnd select action akThereafter, a cumulative prize expectation is obtained. Wherein, akRefers to the step length (control action) corresponding to the hidden layer state and the unit state during the state update in the kth iteration. The specific expressions of the cumulative prize expectation are as follows:
...
wherein, alpha is a learning factor; r ((h)t-i,Ct-i)k,(ht-i,Ct-i)k+1,ak) I is 1,2, …, n +1 is the immediate reward function value; a is an action set; gamma is a discount factor.
4)Q*[(ht-1,Ct-1)r,ar],Q*[(ht-2,Ct-2)r,ar],…,Q*[(ht-n,Ct-n)r,ar]All adopt an optimal strategy pi*The corresponding maximum accumulated reward expectation value is expressed as follows:
Q*[(ht-1,Ct-1)r,ar]=maxQπ[(ht-1,Ct-1)k,ak];
Q*[(ht-2,Ct-2)r,ar]=maxQπ[(ht-2,Ct-2)k,ak];
...
Q*[(ht-n,Ct-n)r,ar=maxQπ[(ht-n,Ct-n)k,ak];
in the formula, pi is all strategy sets, and r belongs to kmax,kmaxAnd representing the iteration times corresponding to the optimal strategy as the maximum iteration times. Policy refers to performing action arAnd then, a series of actions which maximize the accumulated reward expectation value are taken, and if the accumulated reward expectation value obtained by the actions is maximum, the actions are called an optimal strategy.
5)[ht-1,Ct-1]best,[ht-2,Ct-2]best,…,[ht-n,Ct-n]bestThe optimal hidden layer state and the optimal unit state before the time t in the long-short term memory neural network are respectively processed through the full connection layer, so that the optimal prediction result of the node voltage/line current/branch power at the time t can be calculated, and the prediction data at the time t can be obtained.
And 104, replacing the bad data at the time t with the predicted data at the time t to obtain corrected measurement data.
And replacing the bad data at the moment t with the predicted data at the moment t to obtain the complete corrected measurement data.
In the embodiment of the application, after the preset identification model is adopted to identify the bad data in the measured data, the preset number of normal data before the time t when the bad data appears is input into the preset data correction model to predict the normal measured data at the time t to obtain the predicted data at the time t, then the predicted data at the time t is used for replacing the bad data at the time t to obtain the corrected measured data, namely the bad data is eliminated, and the integrity of the data is also ensured.
The foregoing is an embodiment of a method for processing bad data of a power distribution network provided by the present application, and the following is another embodiment of a method for processing bad data of a power distribution network provided by the present application.
Referring to fig. 2, a method for processing bad data of a power distribution network according to an embodiment of the present application includes:
And 203, determining the occurrence time t of the bad data in the measured data based on the identification result, and inputting a preset number of normal data before the occurrence time t of the bad data into a preset data correction model to obtain the predicted data at the time t.
The specific contents of steps 201 to 203 are the same as those of steps 101 to 103, and the details of steps 201 to 203 are not repeated herein.
And 204, inputting the prediction data at the moment t into a preset identification model for data classification, and outputting an identification result of the prediction data at the moment t.
And step 205, when the identification result of the predicted data at the time t is normal data, replacing the bad data at the time t with the predicted data at the time t to obtain corrected measured data.
And inputting the predicted data at the time t into a preset identification model again for identification so as to determine whether the predicted data predicted by the preset data correction model is normal or not, so as to achieve the purpose of checking the predicted data, if the predicted data is normal data, replacing the bad data at the time t with the predicted data at the time t to obtain corrected measured data, if the predicted data is bad data, indicating that the predicted data still has problems, discarding the data correspondingly (namely n input data are changed into n +1 data), and then predicting again until the data are normal and then taking the data as the corrected data.
In the embodiment of the application, after identifying bad data in measured data by adopting a preset identification model, inputting a preset number of normal data before the time t when the bad data appears into a preset data correction model to predict normal measured data at the time t to obtain predicted data at the time t, and then replacing the bad data at the time t with the predicted data at the time t to obtain corrected measured data, namely the bad data are eliminated and the integrity of the data is ensured, so that the technical problems that after identifying the bad data by adopting a convolutional neural network model in the prior art, the bad data after elimination is not supplemented or corrected, the data is incomplete, the observability of a power distribution network system is influenced, and the reliability of a power distribution network state estimation result is reduced are solved;
furthermore, after the prediction data at the time t is obtained, the prediction data is input into a preset identification model to check the prediction data, whether the prediction data is normal or not is determined, if the prediction data is normal, the prediction data replaces the bad data at the time t, and the reliability of the corrected measurement data is further ensured.
The foregoing is another embodiment of the method for processing bad data of the power distribution network provided by the present application, and the following is an embodiment of a device for processing bad data of the power distribution network provided by the present application.
Referring to fig. 3, an apparatus for processing bad data of a power distribution network according to an embodiment of the present application includes:
an obtaining unit 301, configured to obtain measurement data in a preset time period in a power distribution network, where the measurement data includes node voltage, line current, and branch power;
a first classification unit 302, configured to input the measurement data into a preset identification model for data classification, and output an identification result of the measurement data, where the identification result of the measurement data includes normal data and bad data;
an input unit 303, configured to determine a time t when bad data occurs in the measured data based on the identification result, and input a preset number of normal data before the time t when the bad data occurs to the preset data correction model to obtain predicted data at the time t;
a replacing unit 304, configured to replace the bad data at the time t with the predicted data at the time t, and obtain corrected measurement data.
As a further improvement, the method further comprises the following steps:
the preprocessing unit 305 is configured to preprocess the measurement data according to a preset formula, where the preset formula is:
wherein,for the j-th dimension data in the i-th measurement data,the j-th dimension data in the ith type of measured data after preprocessing is obtained; i is 1,2 and 3, which respectively represent node voltage, line current and branch power; j is 1,2, …, Ni,NiThe total dimensionality of the data in the ith type of measured data.
As a further improvement, the method further comprises the following steps:
the second classification unit 306 is configured to input the prediction data at the time t to a preset identification model for data classification, and output an identification result of the prediction data at the time t;
accordingly, the substitution unit 304 is specifically configured to:
and when the identification result of the predicted data at the time t is normal data, replacing the bad data at the time t with the predicted data at the time t to obtain corrected measured data.
In the embodiment of the application, after the preset identification model is adopted to identify the bad data in the measured data, the preset number of normal data before the time t when the bad data appears is input into the preset data correction model to predict the normal measured data at the time t to obtain the predicted data at the time t, then the predicted data at the time t is used for replacing the bad data at the time t to obtain the corrected measured data, namely the bad data is eliminated, and the integrity of the data is also ensured.
The embodiment of the application also provides processing equipment for the bad data of the power distribution network, and the equipment comprises a processor and a memory;
the memory is used for storing the program codes and transmitting the program codes to the processor;
the processor is used for executing the processing method of the bad data of the power distribution network in the processing method embodiment of the bad data of the power distribution network according to the instructions in the program codes.
The embodiment of the application also provides a computer-readable storage medium, which is used for storing a program code, and the program code is used for executing the processing method of the bad data of the power distribution network in the processing method embodiment of the bad data of the power distribution network.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for executing all or part of the steps of the method described in the embodiments of the present application through a computer device (which may be a personal computer, a server, or a network device). And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (10)
1. A processing method for bad data of a power distribution network is characterized by comprising the following steps:
acquiring measurement data in a preset time period in a power distribution network, wherein the measurement data comprises node voltage, line current and branch power;
inputting the measured data into a preset identification model for data classification, and outputting an identification result of the measured data, wherein the identification result of the measured data comprises normal data and bad data;
determining the occurrence time t of bad data in the measured data based on the identification result, and inputting a preset number of normal data before the occurrence time t of the bad data into a preset data correction model to obtain predicted data at the time t;
and replacing the bad data at the moment t with the predicted data at the moment t to obtain the corrected measurement data.
2. The method for processing bad data of the power distribution network according to claim 1, wherein the step of inputting the measured data into a preset identification model for data classification and outputting the identification result of the measured data further comprises:
preprocessing the measurement data through a preset formula, wherein the preset formula is as follows:
wherein,for the j-th dimension data in the i-th measurement data,is the i-th class quantity after the pretreatmentMeasuring j-th dimension data in the data; i is 1,2 and 3, which respectively represent node voltage, line current and branch power; j is 1,2, …, Ni,NiThe total dimensionality of the data in the ith type of measured data.
3. The method for processing the bad data of the power distribution network according to claim 1, wherein the configuration process of the preset identification model is as follows:
acquiring a measured data training set, wherein the label of each training sample in the measured data training set is the normal data or the bad data;
inputting the measured data training set into a preset convolutional neural network for training, and outputting a prediction result corresponding to each training sample;
calculating a loss value through a loss function based on the prediction result corresponding to each training sample and the label;
and updating the parameters of the preset convolutional neural network by a gradient descent method based on the loss value until the loss value is lower than a preset threshold value, so as to obtain the preset identification model.
4. The method for processing the bad data of the power distribution network according to claim 1, wherein the preset data modification model is a Q learning-long short term memory neural network model;
correspondingly, the inputting a preset number of normal data before the time t when the bad data appears into a preset data correction model to obtain the predicted data at the time t includes:
inputting n normal data before the time t when the bad data appears into the Q learning-long short-term memory neural network model, enabling the Q learning-long short-term memory neural network model to process the n normal data based on the hidden layer state and the unit state of the previous n times t-1, t-2, t.
5. The method for processing the bad data of the power distribution network according to claim 1, wherein the determining a time t of occurrence of the bad data in the measured data based on the identification result inputs a preset number of normal data before the time t of occurrence of the bad data into a preset data correction model to obtain predicted data of the time t, and then further comprises:
inputting the prediction data at the moment t into the preset identification model for data classification, and outputting the identification result of the prediction data at the moment t;
correspondingly, the replacing the bad data at the time t with the predicted data at the time t to obtain the corrected measurement data includes:
and when the identification result of the predicted data at the moment t is the normal data, replacing the bad data at the moment t with the predicted data at the moment t to obtain the corrected measured data.
6. A processing apparatus of distribution network bad data, characterized by comprising:
the system comprises an acquisition unit, a processing unit and a control unit, wherein the acquisition unit is used for acquiring measurement data in a preset time period in the power distribution network, and the measurement data comprises node voltage, line current and branch power;
the first classification unit is used for inputting the measured data into a preset identification model for data classification and outputting an identification result of the measured data, wherein the identification result of the measured data comprises normal data and bad data;
the input unit is used for determining the occurrence time t of bad data in the measured data based on the identification result, and inputting a preset number of normal data before the occurrence time t of the bad data into a preset data correction model to obtain predicted data at the time t;
and the replacing unit is used for replacing the bad data at the moment t with the predicted data at the moment t to obtain the corrected measurement data.
7. The device for processing the bad data of the power distribution network according to claim 6, further comprising:
the preprocessing unit is used for preprocessing the measured data through a preset formula, wherein the preset formula is as follows:
wherein,for the j-th dimension data in the i-th measurement data,the j-th dimension data in the ith type of measured data after preprocessing is obtained; i is 1,2 and 3, which respectively represent node voltage, line current and branch power; j is 1,2, …, Ni,NiThe total dimensionality of the data in the ith type of measured data.
8. The device for processing the bad data of the power distribution network according to claim 6, further comprising:
the second classification unit is used for inputting the prediction data at the moment t into the preset identification model for data classification and outputting the identification result of the prediction data at the moment t;
correspondingly, the replacement unit is specifically configured to:
and when the identification result of the predicted data at the moment t is the normal data, replacing the bad data at the moment t with the predicted data at the moment t to obtain the corrected measured data.
9. The device for processing the bad data of the power distribution network is characterized by comprising a processor and a memory;
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is used for executing the processing method of the bad data of the power distribution network according to any one of claims 1 to 5 according to the instructions in the program code.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium is used for storing program codes, and the program codes are used for executing the processing method of bad data of the power distribution network according to any one of claims 1-5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110120172.2A CN112819067A (en) | 2021-01-28 | 2021-01-28 | Method, device, equipment and storage medium for processing bad data of power distribution network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110120172.2A CN112819067A (en) | 2021-01-28 | 2021-01-28 | Method, device, equipment and storage medium for processing bad data of power distribution network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112819067A true CN112819067A (en) | 2021-05-18 |
Family
ID=75859870
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110120172.2A Pending CN112819067A (en) | 2021-01-28 | 2021-01-28 | Method, device, equipment and storage medium for processing bad data of power distribution network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112819067A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113515512A (en) * | 2021-06-22 | 2021-10-19 | 国网辽宁省电力有限公司鞍山供电公司 | Quality control and improvement method for industrial internet platform data |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106684865A (en) * | 2017-01-04 | 2017-05-17 | 北京清能互联科技有限公司 | Reactive daily curve bad data identification and correction method |
CN107591811A (en) * | 2017-10-17 | 2018-01-16 | 中国农业大学 | Var Optimization Method in Network Distribution and device |
US20180157838A1 (en) * | 2016-12-07 | 2018-06-07 | General Electric Company | Feature and boundary tuning for threat detection in industrial asset control system |
CN108199891A (en) * | 2018-01-12 | 2018-06-22 | 东北大学 | A kind of cps network attack discrimination methods based on artificial neural network multi-angle comprehensive decision |
CN109659933A (en) * | 2018-12-20 | 2019-04-19 | 浙江工业大学 | A kind of prediction technique of power quality containing distributed power distribution network based on deep learning model |
CN109993361A (en) * | 2019-03-29 | 2019-07-09 | 国网上海市电力公司 | A kind of power distribution network operation trend prediction technique based on PMU |
CN111080484A (en) * | 2019-12-21 | 2020-04-28 | 国网山东省电力公司泰安供电公司 | Method and device for monitoring abnormal data of power distribution network |
CN111709790A (en) * | 2020-06-19 | 2020-09-25 | 广东电网有限责任公司 | Method, device, equipment and storage medium for identifying abnormal electricity price of day-ahead market |
CN111881124A (en) * | 2020-07-24 | 2020-11-03 | 贵州电网有限责任公司 | Data processing method and system based on state estimation of improved algorithm |
-
2021
- 2021-01-28 CN CN202110120172.2A patent/CN112819067A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180157838A1 (en) * | 2016-12-07 | 2018-06-07 | General Electric Company | Feature and boundary tuning for threat detection in industrial asset control system |
CN106684865A (en) * | 2017-01-04 | 2017-05-17 | 北京清能互联科技有限公司 | Reactive daily curve bad data identification and correction method |
CN107591811A (en) * | 2017-10-17 | 2018-01-16 | 中国农业大学 | Var Optimization Method in Network Distribution and device |
CN108199891A (en) * | 2018-01-12 | 2018-06-22 | 东北大学 | A kind of cps network attack discrimination methods based on artificial neural network multi-angle comprehensive decision |
CN109659933A (en) * | 2018-12-20 | 2019-04-19 | 浙江工业大学 | A kind of prediction technique of power quality containing distributed power distribution network based on deep learning model |
CN109993361A (en) * | 2019-03-29 | 2019-07-09 | 国网上海市电力公司 | A kind of power distribution network operation trend prediction technique based on PMU |
CN111080484A (en) * | 2019-12-21 | 2020-04-28 | 国网山东省电力公司泰安供电公司 | Method and device for monitoring abnormal data of power distribution network |
CN111709790A (en) * | 2020-06-19 | 2020-09-25 | 广东电网有限责任公司 | Method, device, equipment and storage medium for identifying abnormal electricity price of day-ahead market |
CN111881124A (en) * | 2020-07-24 | 2020-11-03 | 贵州电网有限责任公司 | Data processing method and system based on state estimation of improved algorithm |
Non-Patent Citations (3)
Title |
---|
叶学勇: "基于神经网路的电力系统不良数据修正", 《电网技术》 * |
吴京秋 等: "间隙统计在电力系统不良数据辨识中的应用", 《南京工程学院学报(自然科学版)》 * |
周福举等: "基于BP神经网络的线损不良数据辨识及其预测", 《节能》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113515512A (en) * | 2021-06-22 | 2021-10-19 | 国网辽宁省电力有限公司鞍山供电公司 | Quality control and improvement method for industrial internet platform data |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN116757534B (en) | Intelligent refrigerator reliability analysis method based on neural training network | |
CN109490814B (en) | Metering automation terminal fault diagnosis method based on deep learning and support vector data description | |
CN113702895B (en) | Online quantitative evaluation method for error state of voltage transformer | |
CN111813084A (en) | Mechanical equipment fault diagnosis method based on deep learning | |
CN112379269A (en) | Battery abnormity detection model training and detection method and device thereof | |
CN113486578A (en) | Method for predicting residual life of equipment in industrial process | |
CN111833583B (en) | Training method, device, equipment and medium for power data anomaly detection model | |
CN108919059A (en) | A kind of electric network failure diagnosis method, apparatus, equipment and readable storage medium storing program for executing | |
CN116910493A (en) | Construction method and device of equipment fault diagnosis model based on multi-source feature extraction | |
CN111046961A (en) | Fault classification method based on bidirectional long-and-short-term memory unit and capsule network | |
CN115169430A (en) | Cloud network end resource multidimensional time sequence anomaly detection method based on multi-scale decoding | |
CN117407770A (en) | High-voltage switch cabinet fault mode classification and prediction method based on neural network | |
CN115184054A (en) | Mechanical equipment semi-supervised fault detection and analysis method, device, terminal and medium | |
CN113988210A (en) | Method and device for restoring distorted data of structure monitoring sensor network and storage medium | |
CN116956197B (en) | Deep learning-based energy facility fault prediction method and device and electronic equipment | |
CN112819067A (en) | Method, device, equipment and storage medium for processing bad data of power distribution network | |
CN115017970A (en) | Migration learning-based gas consumption behavior anomaly detection method and system | |
CN113935413A (en) | Distribution network wave recording file waveform identification method based on convolutional neural network | |
CN118035866A (en) | Fault diagnosis method for aluminum extrusion equipment based on expert network | |
CN117368789A (en) | Power failure co-location method and system based on multi-source heterogeneous data | |
CN116384223A (en) | Nuclear equipment reliability assessment method and system based on intelligent degradation state identification | |
CN112115984A (en) | Tea garden abnormal data correction method and system based on deep learning and storage medium | |
CN115794805A (en) | Medium-low voltage distribution network measurement data supplementing method | |
CN112766503B (en) | Convolutional neural network model fault tolerance method based on key layer rollback mechanism | |
CN112598186B (en) | Improved LSTM-MLP-based small generator fault prediction 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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210518 |