CN114997346A - False data identification method and device - Google Patents

False data identification method and device Download PDF

Info

Publication number
CN114997346A
CN114997346A CN202210942357.6A CN202210942357A CN114997346A CN 114997346 A CN114997346 A CN 114997346A CN 202210942357 A CN202210942357 A CN 202210942357A CN 114997346 A CN114997346 A CN 114997346A
Authority
CN
China
Prior art keywords
measurement value
actual measurement
data
false data
false
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210942357.6A
Other languages
Chinese (zh)
Inventor
范心明
史训涛
董镝
宋安琪
李新
张殷
李国伟
王俊波
唐琪
柯清派
蒋维
罗容波
黄静
陈邦发
刘石生
程志秋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
CSG Electric Power Research Institute
Foshan Power Supply Bureau of Guangdong Power Grid Corp
Original Assignee
CSG Electric Power Research Institute
Foshan Power Supply Bureau of Guangdong Power Grid Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by CSG Electric Power Research Institute, Foshan Power Supply Bureau of Guangdong Power Grid Corp filed Critical CSG Electric Power Research Institute
Priority to CN202210942357.6A priority Critical patent/CN114997346A/en
Publication of CN114997346A publication Critical patent/CN114997346A/en
Pending legal-status Critical Current

Links

Images

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
    • G06N3/08—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
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E10/00—Energy generation through renewable energy sources
    • Y02E10/70—Wind energy
    • Y02E10/76—Power conversion electric or electronic aspects

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Biophysics (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Primary Health Care (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The invention discloses a method and a device for identifying false data, wherein the method comprises the following steps: the method comprises the steps of obtaining an actual measurement value and data to be measured in a preset state estimation model, calculating to obtain false data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value, establishing a false data identification model based on the false data corresponding to the actual measurement value and the actual measurement value, inputting the data to be measured into the false data identification model, and identifying to obtain corresponding false data in the data to be measured. The method is beneficial to solving the technical problem that the existing identification method needs to spend a large amount of cost and energy when identifying the false data, and improves the identification efficiency of the false data.

Description

False data identification method and device
Technical Field
The invention relates to the technical field of power networks, in particular to a false data identification method and device.
Background
In recent years, the installed capacity of photovoltaic power generation and wind power generation is rapidly increased, and the uncertainty and randomness of power generation greatly influence the safe and stable operation of a power system. The power network and the natural Gas network are connected into an Integrated Electric-Gas Systems (IEGS) which can effectively solve the problem of wind and light power consumption. When wind and solar power generation is insufficient, the gas turbine can quickly respond to changes in demand and supply; when the wind and light power generation is abundant, the electric gas conversion equipment can convert the new energy power generation into natural gas and store the natural gas in a natural gas pipe network. Therefore, the coupling of power systems to natural gas systems is becoming increasingly intimate.
Accurate and reliable IEGS state estimation can provide operation data for the IEGS energy management system, and is the basis of IEGS operation and control. However, it has been found that hackers can avoid a Bad Data Detection (BDD) module by tampering with the sensor monitoring Data of the power system and injecting well-designed dummy Data, thereby bringing a deviation to the state estimation result and triggering an erroneous control decision, which is called a dummy Data Injection Attack (FDIA).
In the prior art, a large number of samples with labels are needed based on supervised learning, but in a real power system, a large amount of cost and energy are needed to obtain the data with the labels.
Therefore, in order to improve the efficiency of identifying the dummy data, it is necessary to construct a method for identifying the dummy data to solve the technical problem that the existing identification method needs to spend a lot of cost and effort when identifying the dummy data.
Disclosure of Invention
The invention provides a false data identification method and a false data identification device, which solve the technical problem that the existing identification method needs to spend a large amount of cost and energy when identifying false data.
In a first aspect, the present invention provides a method for identifying false data, including:
acquiring an actual measurement value and data to be measured in a preset state estimation model;
calculating to obtain false data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
establishing a false data identification model based on the actual measurement value and false data corresponding to the actual measurement value;
and inputting the data to be detected into the false data identification model, and identifying to obtain corresponding false data in the data to be detected.
Optionally, calculating to obtain the false data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value, where the calculating includes:
calculating to obtain a state variable corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
calculating to obtain an attack vector corresponding to the actual measurement value based on the state variable and the actual measurement value;
and calculating the sum of the attack vector and the actual measurement value to obtain the false data corresponding to the actual measurement value.
Optionally, calculating a state variable corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value, where the calculating includes:
generating a simulated attack vector corresponding to the actual measurement value and different attack strengths;
calculating to-be-measured data according to the to-be-attacked vector and the actual measurement value;
and inputting the data to be measured into the preset state estimation model to obtain a state variable corresponding to the actual measurement value.
Optionally, establishing a false data identification model based on the actual measurement value and false data corresponding to the actual measurement value, including:
establishing a preliminary neural network model based on the actual measurement value and the false data corresponding to the actual measurement value;
dividing the actual measurement value and the false data corresponding to the actual measurement value into training data and verification data;
and training and verifying the preliminary neural network model based on the training data and the verification data to obtain the false data identification model.
Optionally, training and verifying the preliminary neural network model based on the training data and the verification data to obtain the false data recognition model, including:
initializing parameters of the preliminary neural network model;
training the preliminary neural network model based on the training data and the parameters to obtain a trained neural network model;
and verifying the trained false data identification model based on the training data to obtain the false data identification model.
In a second aspect, the present invention provides an apparatus for identifying false data, comprising:
the acquisition module is used for acquiring an actual measurement value and data to be measured in a preset state estimation model;
the calculation module is used for calculating to obtain false data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
the establishing module is used for establishing a false data identification model based on the actual measurement value and the false data corresponding to the actual measurement value;
and the identification module is used for inputting the data to be detected into the false data identification model and identifying to obtain corresponding false data in the data to be detected.
Optionally, the calculation module comprises:
the variable submodule is used for calculating to obtain a state variable corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
the attack submodule is used for calculating to obtain an attack vector corresponding to the actual measurement value based on the state variable and the actual measurement value;
and the false submodule is used for calculating the sum of the attack vector and the actual measurement value to obtain false data corresponding to the actual measurement value.
Optionally, the variable submodule includes:
the generating unit is used for generating a pseudo-attack vector corresponding to the actual measurement value and different attack strengths;
the calculation unit is used for calculating to-be-measured data according to the to-be-attacked vector and the actual measurement value;
and the variable unit is used for inputting the data to be measured into the preset state estimation model to obtain the state variable corresponding to the actual measurement value.
Optionally, the establishing module includes:
the establishing submodule is used for establishing a preliminary neural network model based on the actual measuring value and the false data corresponding to the actual measuring value;
the dividing submodule is used for dividing the actual measuring value and the false data corresponding to the actual measuring value into training data and verification data;
and the training submodule is used for training and verifying the preliminary neural network model based on the training data and the verification data to obtain the false data identification model.
Optionally, the training submodule includes:
an initialization unit, configured to initialize parameters of the preliminary neural network model;
a training unit, configured to train the preliminary neural network model based on the training data and the parameters, to obtain a trained neural network model;
and the verification unit is used for verifying the trained false data identification model based on the training data to obtain the false data identification model.
According to the technical scheme, the invention has the following advantages: the invention provides a false data identification method, which comprises the steps of obtaining an actual measurement value and to-be-detected data in a preset state estimation model, calculating to obtain false data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value, establishing a false data identification model based on the false data corresponding to the actual measurement value and the actual measurement value, inputting the to-be-detected data into the false data identification model, and identifying to obtain the corresponding false data in the to-be-detected data.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without inventive exercise.
FIG. 1 is a flowchart illustrating a first embodiment of a method for identifying false data according to the present invention;
FIG. 2 is a flowchart illustrating a second embodiment of a method for identifying false data according to the present invention;
FIG. 3 is a block diagram of an embodiment of a device for identifying false data according to the present invention.
Detailed Description
The embodiment of the invention provides a false data identification method and device, which are used for solving the technical problem that the existing identification method needs to spend a large amount of cost and energy when identifying false data.
In order to make the objects, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the embodiments described below are only a part of the embodiments of the present invention, 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 invention.
In a first embodiment, referring to fig. 1, fig. 1 is a flowchart illustrating a first method for identifying false data according to a first embodiment of the present invention, including:
step S101, acquiring an actual measurement value and data to be measured in a preset state estimation model;
step S102, calculating to obtain false data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
in the embodiment of the invention, a pseudo-attack vector corresponding to the actual measurement value and different attack strengths is generated, pseudo-measurement data is obtained by calculation according to the pseudo-attack vector and the actual measurement value, the pseudo-measurement data is input into the preset state estimation model to obtain a state variable corresponding to the actual measurement value, an attack vector corresponding to the actual measurement value is obtained by calculation based on the state variable and the actual measurement value, and the sum of the attack vector and the actual measurement value is calculated to obtain the false data corresponding to the actual measurement value.
Step S103, establishing a false data identification model based on the actual measurement value and the false data corresponding to the actual measurement value;
in the embodiment of the invention, a preliminary neural network model is established based on the actual measurement value and the false data corresponding to the actual measurement value, the actual measurement value and the false data corresponding to the actual measurement value are divided into training data and verification data, parameters of the preliminary neural network model are initialized, the preliminary neural network model is trained based on the training data and the parameters to obtain a trained neural network model, and the trained false data recognition model is verified based on the training data to obtain the false data recognition model.
And step S104, inputting the data to be detected into the false data identification model, and identifying to obtain corresponding false data in the data to be detected.
According to the false data identification method provided by the embodiment of the invention, the actual measurement value and the data to be detected in the preset state estimation model are obtained, the false data corresponding to the actual measurement value is obtained through calculation according to the preset state estimation model and the actual measurement value, the false data identification model is established based on the false data corresponding to the actual measurement value and the actual measurement value, the data to be detected is input into the false data identification model, and the corresponding false data in the data to be detected is obtained through identification.
In a second embodiment, referring to fig. 2, fig. 2 is a flowchart illustrating a method for identifying false data according to the present invention, including:
step S201, acquiring an actual measurement value and data to be measured in a preset state estimation model;
in the embodiment of the present invention, an actual measurement value in a preset state estimation model and at least one measured measurement value are obtained as data to be measured.
In specific implementation, the state estimation model is determined according to the state variables and the measurement values of the power network model, the state variables and the measurement values of the natural gas network model and the coupling element model, and a preset state estimation model is obtained.
The measurement values of the power network model comprise node injection active power, node injection reactive power, initial end active power of a voltage amplitude measurement circuit, initial end reactive power of the voltage amplitude measurement circuit, tail end active power of the voltage amplitude measurement circuit and tail end reactive power of the voltage amplitude measurement circuit; the state variables of the power network model include all node voltage magnitudes and phase angles of the balancing nodes.
The power network model comprises a node power balance equation which is specifically as follows:
Figure 666661DEST_PATH_IMAGE001
；
wherein, P i 、Q i Respectively injecting active power and reactive power into the node i; v i Is the voltage amplitude of node i; theta ij Is the phase angle difference of the nodes i and j; g ij 、B ij Respectively are the mutual conductance and mutual susceptance of the nodes i and j.
The state variables of the natural gas network model comprise the gas pressures of all nodes and the natural gas flow of all compressor branches; the measurements of the natural gas network model include node gas pressure, node injected natural gas flow, and branch natural gas flow.
The natural gas flow equation of the pipeline is as follows:
Figure 397856DEST_PATH_IMAGE002
；
Figure 217651DEST_PATH_IMAGE003
；
wherein phi is a Panhandle A equation; s ij Is the flow direction of the natural gas flow; II type i Is the gas pressure p of a natural gas network node i i Square of (ii), i.e. pi =
Figure 831035DEST_PATH_IMAGE004
；ΔΠ ij Is the difference of the pressure squares between the natural gas points i, j, i.e. Δ Π ij =
Figure 895943DEST_PATH_IMAGE005
；K g Is a pipeline constant that is related to the natural gas pipeline length, efficiency coefficient, natural gas pipeline diameter.
The non-pipeline branch equation is specifically as follows:
Figure 983110DEST_PATH_IMAGE006
；
wherein, w 1 、w 2 、w 3 D is a parameter representing the compressor operation mode; II type c,in Is the square of the compressor inlet air pressure; II type c,ou t is the square of the compressor outlet pressure. When w is 1 、w 2 、w 3 D take different values, the compressor works in different modes.
The node flow balance equation is specifically as follows:
Figure 147375DEST_PATH_IMAGE007
；
wherein, A g Is a node-branch incidence matrix of the natural gas network, if i is the initial node of the pipeline k, A g (i, k) =1, if i is the end node of the pipeline k, Ag (i, k) = -1, and the other elements are 0; f. of p Is the flow vector of all natural gas pipelines; b is g Is the correlation matrix between the natural gas node and the compressor, if i is the start node of compressor k, then B g (i, k) =1, if i is the last node of compressor k, then B g (i, k) = -1, and the other elements are 0; f. of p Is the natural gas flow vector of all natural gas pipelines; f. of c Is the natural gas flow vector of all compressor branches; finj is the injected natural gas flow vector of all nodes of the natural gas network; f. of G Is the natural gas flow vector provided by all gas sources; f. of L Is the natural gas flow vector taken by all natural gas loads; f. of GPG Is the natural gas flow vector consumed by all gas turbines.
The loop pressure equation is specifically:
Figure 451318DEST_PATH_IMAGE008
；
wherein, C g Is a branch-loop incidence matrix, if branch i is in loop k, then C g (i, k) =1, and the other elements are 0.
The electro-pneumatic coupling model includes a GPG model, a P2G model, and a compressor model. Wherein, the GPG model generates electricity by consuming natural gas, and satisfies the following formula:
Figure 54338DEST_PATH_IMAGE009
；
wherein, f GPG Is the natural gas flow consumed by the gas turbine; p is GPG Is the active power produced by the gas turbine; eta GPG Is the power generation efficiency of the gas turbine; h gas Is the heating value of natural gas.
The P2G equipment converts electric energy into natural gas, and the model thereof is specifically as follows:
Figure 432229DEST_PATH_IMAGE010
；
wherein, P P2G Is the active power consumed by the P2G device; f. of P2G Is the natural gas flow rate of the P2G plant into the natural gas network; eta P2G Is the efficiency of the P2G device.
The compressor model specifically comprises:
Figure 95292DEST_PATH_IMAGE011
；
wherein, T s Is the compressor temperature; f. of c Is the natural gas flow through the compressor; eta c Is the efficiency of the compressor; p c,in 、P c,out Inlet and outlet air pressures of the compressor respectively; alpha is the polytropic exponent of the compressor.
The compressors can be classified into the following two types according to the driving type of the compressor: electrically driven compressors and natural gas driven compressors.
1) When the power of the compressor is provided by the generator, the consumed active power is:
Figure 119486DEST_PATH_IMAGE012
；
2) when the compressor power is provided by the gas turbine, the active power consumed is:
Figure 526197DEST_PATH_IMAGE013
；
the compressor can be a coupling element in IEGS, but it consumes little real power and natural gas whether the compressor is driven electrically or by natural gas, and the coupling constraints of the compressor are ignored in this specification.
The measurement equation of the power network is as follows:
Figure 86491DEST_PATH_IMAGE014
；
wherein, P ij 、Q ij Respectively the active and reactive power of the starting end of the branch i-j.
For a natural gas network, the state variables are the gas pressures of all nodes and the natural gas flow of all compressor branches, i.e. x g =[p i T f c T ] T The quantity measured includes node pressure, node injected natural gas flow, branch natural gas flow, i.e. z g =[p i T f i T f ij T ] T The measurement equation for the natural gas network is:
Figure 592559DEST_PATH_IMAGE015
；
the presence of the compressor and coupling elements in the IEGS requires the following equality constraints to be satisfied:
Figure 871093DEST_PATH_IMAGE016
；
for IEGS, the metrology model is specifically:
Figure 317380DEST_PATH_IMAGE017
；
wherein z is e Measuring the power grid quantity; h is e A power grid measurement equation is obtained; z is a radical of g Measuring the air network quantity; h is g Is a gas network measurement equation; g represents component constraints, mainly including compressor constraints, GPG and P2G plant coupling constraints; s represents a zero injection constraint; e.g. of a cylinder e 、e g The error vectors are measured by the power grid and the gas grid respectively.
The preset state estimation model is constructed based on a weighted least square method, and the preset state estimation model is as follows:
Figure 669864DEST_PATH_IMAGE018
；
wherein minJ (x) is a state estimation model constructed based on a weighted least square method, R e Rg is respectively e e 、e g The covariance matrix of (a); r is e Measuring the residual error, r, for the grid quantity e =z e －h e (x e )；r g Measuring the residual error, r, for the air grid quantity g =z g －h g (x g )。
For ease of analysis, the above formula is written in a compact form as follows:
Figure 409150DEST_PATH_IMAGE019
；
wherein, x = [ x ] e T x g T ] T ；r=[r e T r g T ] T ；c(·)=[g T (·) s T (·)] T ；
Figure 847085DEST_PATH_IMAGE020
。
Step S202, calculating to obtain a state variable corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
in an optional embodiment, calculating a state variable corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value includes:
generating a simulated attack vector corresponding to the actual measurement value and different attack strengths;
calculating to-be-measured data according to the to-be-attacked vector and the actual measurement value;
and inputting the data to be measured into the preset state estimation model to obtain a state variable corresponding to the actual measurement value.
In the embodiment of the invention, simulated attack vectors corresponding to different attack strengths with the actual measurement value are generated, simulated measurement data are obtained by calculation according to the simulated attack vectors and the actual measurement value, and the simulated measurement is carried outInputting measured data into the preset state estimation model to obtain a state variable x corresponding to the actual measurement value a Said state variable x a Satisfies a preset threshold.
In a specific implementation, the state variable x a The maximum normalized residual error of (a) satisfies a preset threshold, specifically including:
Figure 595598DEST_PATH_IMAGE021
；
wherein the content of the first and second substances,
Figure 864905DEST_PATH_IMAGE022
is a preset threshold.
In order to solve a preset state estimation model, firstly, a Lagrangian function is constructed:
Figure 978355DEST_PATH_IMAGE023
；
wherein the content of the first and second substances,
Figure 464438DEST_PATH_IMAGE024
the corresponding lagrange multiplier vector is constrained for the equation.
Order to
Figure 688745DEST_PATH_IMAGE025
For x and
Figure 78139DEST_PATH_IMAGE024
has a first partial derivative of zero:
Figure 424806DEST_PATH_IMAGE026
；
where H (), C () are the jacobian matrices of the quantity measurements and the equality constraints, respectively, on all state variables.
The above system of nonlinear equations can be solved by the gauss-newton method, each iteration solving the following system of linear equations:
Figure 571754DEST_PATH_IMAGE027
；
wherein, the first and the second end of the pipe are connected with each other,
Figure 662070DEST_PATH_IMAGE028
，x k is the state variable of the kth iteration;
Figure 843652DEST_PATH_IMAGE029
。
the inverse of the coefficient matrix of equation (22) at the kth iteration is:
Figure 862686DEST_PATH_IMAGE030
；
in the last iteration before convergence, the state variable x at this time is assumed k Close to the true value x. Taylor expansion is carried out on the residual error r at the state truth value x and high-order terms are ignored, so that a linear relation between the residual error r and a measurement error vector e can be obtained:
Figure 824826DEST_PATH_IMAGE031
；
where S is the residual sensitivity matrix.
Considering that the measurement error vector follows a normal distribution, i.e. e i ~N(0,R ii ) So the residuals also follow a normal distribution
Figure 390936DEST_PATH_IMAGE032
With a covariance matrix of
Figure 489342DEST_PATH_IMAGE033
Therefore, the normalized residual error of the measurement zi is specifically:
Figure 177813DEST_PATH_IMAGE034
；
wherein the content of the first and second substances,
Figure 299352DEST_PATH_IMAGE035
is composed of
Figure 230006DEST_PATH_IMAGE036
The ith diagonal element of (1).
At this time, the normalized residual r N Subject to a standard normal distribution, i.e. r i N N (0, 1). Thus, a hypothesis test can be performed based on this if the residual r is normalized i N If the following formula is satisfied, the measurement is judged to be normal data, otherwise, the measurement is judged to be bad data:
Figure 448498DEST_PATH_IMAGE037
；
Figure 245552DEST_PATH_IMAGE022
is a threshold value at a certain confidence level.
Performing a state estimation by using the state estimation model to obtain a state variable x a 。
Step S203, calculating to obtain an attack vector corresponding to the actual measurement value based on the state variable and the actual measurement value;
it should be noted that the attack strength may represent the percentage of the actual measurement value of the tampered data, for example: the attack strength may be 5%, 8%, 10% of the actual measurement value. The attack strength can also be set to other percentages, and can be determined according to actual conditions.
In the embodiment of the invention, the variable x is determined according to the state a And calculating an attack vector a according to the actual measurement value.
Step S204, calculating the sum of the attack vector and the actual measurement value to obtain false data corresponding to the actual measurement value;
in the embodiment of the invention, the attack vector a and the actual measurement value are superposed to obtain corresponding false data.
Step S205, establishing a preliminary neural network model based on the actual measurement value and the false data corresponding to the actual measurement value;
step S206, dividing the actual measurement value and the false data corresponding to the actual measurement value into training data and verification data;
step S207, training and verifying the preliminary neural network model based on the training data and the verification data to obtain a false data identification model;
in an optional embodiment, training and verifying the preliminary neural network model based on the training data and the verification data to obtain a false data recognition model, includes:
initializing parameters of the preliminary neural network model;
training the preliminary neural network model based on the training data and the parameters to obtain a trained neural network model;
and verifying the trained false data identification model based on the training data to obtain the false data identification model.
In the embodiment of the invention, the parameters of the preliminary neural network model are initialized, the preliminary neural network model is trained based on the training data and the parameters to obtain the trained neural network model, and the trained false data recognition model is verified based on the training data to obtain the false data recognition model.
In a specific implementation, an Adaptive Auto Encoder (AAE) is used to detect spurious data. The AAE includes three training phases, a reconstruction phase, a regularization phase, and a semi-supervised classification phase. The embodiments of the present specification train the network with an Adam optimizer at each stage. In addition, batch standardization is added in the neural network to solve the problem of internal covariate shift in the deep neural network, and the training times are reduced under the condition of achieving the same precision, so that the training time is reduced.
1) A reconstruction stage: at this stage, the AAE may be considered as an autoencoder. For a single layer encoder, input X i And output Y i The relationship of (c) can be written as:
Figure 916705DEST_PATH_IMAGE038
；
wherein W is the weight of the encoder; b is an offset; f () represents an activation function. For a single layer decoder, we can write:
Figure 886935DEST_PATH_IMAGE039
；
AAE updates the encoder and decoder such that the mean square error between the input X and the reconstructed X' is minimized, and the loss function is:
Figure 461398DEST_PATH_IMAGE040
；
2) a regularization stage: at this time, generators (encoders) and discriminators in two countermeasure networks are in a game stage, the generators aim to generate samples which can not be distinguished by the discriminators as far as possible, and the discriminators aim to distinguish real samples from generated samples. The objective function of the countermeasure network corresponding to the classification variable Y is:
Figure 491671DEST_PATH_IMAGE041
；
similarly, the objective function of the countermeasure network corresponding to the hidden layer variable Z is:
Figure 650120DEST_PATH_IMAGE042
；
3) a semi-supervised classification stage: this stage only trains labeled samples. The self-encoder updates the part of the encoder q (Y | X) such that the cross-entropy loss function is minimal:
Figure 860259DEST_PATH_IMAGE043
；
and S208, inputting the data to be detected into the false data identification model, and identifying to obtain corresponding false data in the data to be detected.
The false data identification method provided by the embodiment of the invention comprises the steps of obtaining an actual measurement value and to-be-detected data in a preset state estimation model, calculating to obtain false data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value, establishing a false data identification model based on the false data corresponding to the actual measurement value and the actual measurement value, inputting the to-be-detected data into the false data identification model, and identifying to obtain the corresponding false data in the to-be-detected data.
Referring to fig. 3, fig. 3 is a block diagram illustrating an embodiment of a device for identifying false data according to the present invention, including:
an obtaining module 301, configured to obtain an actual measurement value and data to be measured in a preset state estimation model;
a calculating module 302, configured to calculate to obtain false data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
the establishing module 303 is configured to establish a false data identification model based on the actual measurement value and the false data corresponding to the actual measurement value;
and the identification module 304 is configured to input the data to be detected into the false data identification model, and identify to obtain corresponding false data in the data to be detected.
In an alternative embodiment, the calculation module 302 includes:
the variable submodule is used for calculating to obtain a state variable corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
the attack submodule is used for calculating to obtain an attack vector corresponding to the actual measurement value based on the state variable and the actual measurement value;
and the false submodule is used for calculating the sum of the attack vector and the actual measurement value to obtain false data corresponding to the actual measurement value.
In an alternative embodiment, the variable submodule includes:
the generating unit is used for generating a pseudo-attack vector corresponding to the actual measurement value and different attack strengths;
the calculation unit is used for calculating to-be-measured data according to the to-be-attacked vector and the actual measurement value;
and the variable unit is used for inputting the data to be measured into the preset state estimation model to obtain the state variable corresponding to the actual measurement value.
In an alternative embodiment, the establishing module 303 includes:
the establishing submodule is used for establishing a preliminary neural network model based on the actual measurement value and the false data corresponding to the actual measurement value;
the dividing submodule is used for dividing the actual measuring value and the false data corresponding to the actual measuring value into training data and verification data;
and the training submodule is used for training and verifying the preliminary neural network model based on the training data and the verification data to obtain the false data identification model.
In an alternative embodiment, the training submodule includes:
an initial unit, configured to initialize parameters of the preliminary neural network model;
a training unit, configured to train the preliminary neural network model based on the training data and the parameters, to obtain a trained neural network model;
and the verification unit is used for verifying the trained false data identification model based on the training data to obtain the false data identification model.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the embodiments provided in the present application, it should be understood that the method and apparatus disclosed in the present invention can 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 invention 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 as 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 invention may be embodied in the form of a software product, which is stored in a readable storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned readable storage medium includes: a U-disk, a removable hard disk, a Read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, an optical disk, or other various media capable of storing program codes.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will 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 of the embodiments of the present invention.

Claims (10)

1. A method for identifying spurious data, comprising:
acquiring an actual measurement value and data to be measured in a preset state estimation model;
calculating to obtain false data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
establishing a false data identification model based on the actual measurement value and the false data corresponding to the actual measurement value;
and inputting the data to be detected into the false data identification model, and identifying to obtain corresponding false data in the data to be detected.
2. The method for identifying the dummy data according to claim 1, wherein the calculating the dummy data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value comprises:
calculating to obtain a state variable corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
calculating to obtain an attack vector corresponding to the actual measurement value based on the state variable and the actual measurement value;
and calculating the sum of the attack vector and the actual measurement value to obtain the false data corresponding to the actual measurement value.
3. The method for identifying the false data according to claim 2, wherein calculating the state variable corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value includes:
generating a simulated attack vector corresponding to the actual measurement value and different attack strengths;
calculating to-be-measured data according to the to-be-attacked vector and the actual measurement value;
and inputting the data to be measured into the preset state estimation model to obtain a state variable corresponding to the actual measurement value.
4. The method for identifying the dummy data according to claim 1, wherein the establishing a dummy data identification model based on the actual measurement value and the dummy data corresponding to the actual measurement value comprises:
establishing a preliminary neural network model based on the actual measurement value and the false data corresponding to the actual measurement value;
dividing the actual measurement value and the false data corresponding to the actual measurement value into training data and verification data;
and training and verifying the preliminary neural network model based on the training data and the verification data to obtain the false data identification model.
5. The method for recognizing the spurious data according to claim 4, wherein training and verifying the preliminary neural network model based on the training data and the verification data to obtain the spurious data recognition model comprises:
initializing parameters of the preliminary neural network model;
training the preliminary neural network model based on the training data and the parameters to obtain a trained neural network model;
and verifying the trained false data identification model based on the training data to obtain the false data identification model.
6. An apparatus for identifying false data, comprising:
the acquisition module is used for acquiring an actual measurement value and data to be measured in a preset state estimation model;
the calculation module is used for calculating to obtain false data corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
the establishing module is used for establishing a false data identification model based on the actual measurement value and the false data corresponding to the actual measurement value;
and the identification module is used for inputting the data to be detected into the false data identification model and identifying to obtain corresponding false data in the data to be detected.
7. An apparatus for identifying false data according to claim 6, wherein said computing module comprises:
the variable submodule is used for calculating to obtain a state variable corresponding to the actual measurement value according to the preset state estimation model and the actual measurement value;
the attack submodule is used for calculating to obtain an attack vector corresponding to the actual measurement value based on the state variable and the actual measurement value;
and the false submodule is used for calculating the sum of the attack vector and the actual measurement value to obtain false data corresponding to the actual measurement value.
8. An apparatus for identifying false data according to claim 7, wherein said variable submodule includes:
the generating unit is used for generating a pseudo-attack vector corresponding to the actual measurement value and different attack strengths;
the calculation unit is used for calculating to-be-measured data according to the to-be-attacked vector and the actual measurement value;
and the variable unit is used for inputting the data to be measured into the preset state estimation model to obtain the state variable corresponding to the actual measurement value.
9. An apparatus for identifying false data according to claim 6, wherein the establishing module comprises:
the establishing submodule is used for establishing a preliminary neural network model based on the actual measuring value and the false data corresponding to the actual measuring value;
the dividing submodule is used for dividing the actual measuring value and the false data corresponding to the actual measuring value into training data and verification data;
and the training submodule is used for training and verifying the preliminary neural network model based on the training data and the verification data to obtain the false data identification model.
10. A false data recognition apparatus according to claim 9, wherein the training submodule includes:
an initial unit, configured to initialize parameters of the preliminary neural network model;
a training unit, configured to train the preliminary neural network model based on the training data and the parameters, to obtain a trained neural network model;
and the verification unit is used for verifying the trained false data identification model based on the training data to obtain the false data identification model.
CN202210942357.6A 2022-08-08 2022-08-08 False data identification method and device Pending CN114997346A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210942357.6A CN114997346A (en) 2022-08-08 2022-08-08 False data identification method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210942357.6A CN114997346A (en) 2022-08-08 2022-08-08 False data identification method and device

Publications (1)

Publication Number Publication Date
CN114997346A true CN114997346A (en) 2022-09-02

Family

ID=83023169

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210942357.6A Pending CN114997346A (en) 2022-08-08 2022-08-08 False data identification method and device

Country Status (1)

Country Link
CN (1) CN114997346A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107808105A (en) * 2017-10-18 2018-03-16 南京邮电大学 False data detection method based on prediction in a kind of intelligent grid
CN110995761A (en) * 2019-12-19 2020-04-10 长沙理工大学 Method and device for detecting false data injection attack and readable storage medium
CN112929381A (en) * 2021-02-26 2021-06-08 南方电网科学研究院有限责任公司 Detection method, device and storage medium for false injection data
CN113612733A (en) * 2021-07-07 2021-11-05 浙江工业大学 Twin network-based few-sample false data injection attack detection method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107808105A (en) * 2017-10-18 2018-03-16 南京邮电大学 False data detection method based on prediction in a kind of intelligent grid
CN110995761A (en) * 2019-12-19 2020-04-10 长沙理工大学 Method and device for detecting false data injection attack and readable storage medium
CN112929381A (en) * 2021-02-26 2021-06-08 南方电网科学研究院有限责任公司 Detection method, device and storage medium for false injection data
CN113612733A (en) * 2021-07-07 2021-11-05 浙江工业大学 Twin network-based few-sample false data injection attack detection method

Similar Documents

Publication Publication Date Title
Gao et al. A physically inspired data-driven model for electricity theft detection with smart meter data
Dorr et al. Detection, isolation, and identification of sensor faults in nuclear power plants
Alblawi Fault diagnosis of an industrial gas turbine based on the thermodynamic model coupled with a multi feedforward artificial neural networks
CN110542819A (en) transformer fault type diagnosis method based on semi-supervised DBNC
CN111780933B (en) Method and system for diagnosing leakage fault of high-pressure heater based on neural network and thermodynamic modeling
CN109214708A (en) Study of Risk Evaluation Analysis for Power System method based on cross entropy theoretical optimization support vector machines
CN104536290A (en) Soft measuring method and system based on kernel principal component analysis and radial basis function neural network
Wu et al. Application of Bayesian regularization back propagation neural network in sensorless measurement of pump operational state
Makonin Investigating the switch continuity principle assumed in Non-Intrusive Load Monitoring (NILM)
CN114239725B (en) Electric larceny detection method for data poisoning attack
CN112329276A (en) Safety state analysis method and device based on comprehensive energy system
Luo et al. Novel pattern recognition-enhanced sensor fault detection and diagnosis for chiller plant
CN101813920B (en) Virtual redundancy method for temperature sensor of power station turboset
Huang et al. Outlier detection for process monitoring in industrial cyber-physical systems
Wang et al. Deep-learning modeling and control optimization framework for intelligent thermal power plants: A practice on superheated steam temperature
CN114997346A (en) False data identification method and device
Bettocchi et al. Set up of a robust neural network for gas turbine simulation
CN109638892B (en) Photovoltaic power station equivalent modeling method based on improved fuzzy clustering algorithm
US20180276322A1 (en) Failure diagnosis apparatus, monitoring apparatus, failure diagnosis method and recording medium
Chen et al. Stacked autoencoder framework of false data injection attack detection in smart grid
CN106529041A (en) Refined modeling and identifying method for steam turbine
CN106815641A (en) A kind of Gas path fault diagnosis method based on model and Fuzzy Pattern Recognition
CN113468797A (en) State estimation method and system for electricity-gas comprehensive energy system
CN113465930B (en) Gas turbine multi-sensor fault detection method based on hybrid method
Brahma Using Long Short-Term Memory (LSTM) networks with the toy model concept for compressible pulsatile flow metering

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: 20220902