CN111222139B - GEP optimization-based smart power grid data anomaly effective identification method - Google Patents

GEP optimization-based smart power grid data anomaly effective identification method Download PDF

Info

Publication number
CN111222139B
CN111222139B CN202010112678.4A CN202010112678A CN111222139B CN 111222139 B CN111222139 B CN 111222139B CN 202010112678 A CN202010112678 A CN 202010112678A CN 111222139 B CN111222139 B CN 111222139B
Authority
CN
China
Prior art keywords
data
value
gep
attack
power grid
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.)
Active
Application number
CN202010112678.4A
Other languages
Chinese (zh)
Other versions
CN111222139A (en
Inventor
邓松
张建堂
祝展望
岳东
袁新雅
陈福林
董霞
蔡清媛
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.)
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing University of Posts and Telecommunications
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 Nanjing University of Posts and Telecommunications filed Critical Nanjing University of Posts and Telecommunications
Priority to CN202010112678.4A priority Critical patent/CN111222139B/en
Publication of CN111222139A publication Critical patent/CN111222139A/en
Application granted granted Critical
Publication of CN111222139B publication Critical patent/CN111222139B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/50—Monitoring users, programs or devices to maintain the integrity of platforms, e.g. of processors, firmware or operating systems
    • G06F21/55—Detecting local intrusion or implementing counter-measures
    • G06F21/56—Computer malware detection or handling, e.g. anti-virus arrangements
    • G06F21/562—Static detection
    • G06F21/565—Static detection by checking file integrity
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

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

Abstract

The invention discloses a GEP optimization-based intelligent power grid data anomaly effective identification method, which belongs to the field of electric power system information safety and aims to solve the problem of data anomaly identification under an intelligent power grid; power system state estimation is an important component of modern energy management systems, and its measurement data may contain bad data in addition to normal measurement noise. By using the method, abnormal data which cannot be completely detected by the traditional method can be made up, so that bad data in the power grid can be quickly and effectively identified. The method solves the problem of identifying data abnormity under the intelligent power grid by combining the idea of node energy based on the noise data processing capability of a supervisory machine learning model constructed by GEP, trains according to a large amount of data in the current active power distribution network, and effectively identifies the data abnormity under the power grid by comparing the data change modes before and after attack by using the reactive power optimization comprehensive index on the basis of residual detection, thereby ensuring the safe and reliable operation of the power grid.

Description

GEP optimization-based smart power grid data anomaly effective identification method
Technical Field
The invention belongs to the field of information security of electric power systems, and particularly relates to an effective identification method for data abnormality of a smart power grid based on GEP optimization.
Background
With the continuous integration of network communication technology and power systems, the awareness of information network security of the power systems is strengthened, and establishment of a perfect information network security protection system becomes an important research subject of energy and power security. Especially, in 2015, the Ukran power grid attack event causes that a power data acquisition and monitoring System (SCADA) is seriously impacted, so that the economic loss is serious, a large amount of stored data is eliminated, and the recovery work of the SCADA server is also hindered in the later period of power failure. The rapid development of information technology brings more opportunities to the power grid, and undoubtedly brings challenges to the development of an electric power system towards an open intelligent direction. The method is particularly important for analyzing data leaks existing in the system and accurately detecting and identifying the power grid data abnormity.
The traditional effective identification methods include a residual search method, a non-quadratic criterion method, a zero-residual method and an estimation method. However, in 2009, the concept of power system state estimation false data attack is put forward for the first time, and the traditional identification method cannot meet the requirement of high precision under large data. The research shows that: traditional bad data detection can be avoided strategically by obtaining power system network parameters and topology. The state variable is modified to mislead the control center to make a wrong decision to attack the power grid, so that the safe and reliable operation of the power system is threatened, and the safe and stable development of the intelligent power grid is influenced. Therefore, the research on data information safety in an actual power system and the formulation of a corresponding detection data anomaly mechanism and defense measures are problems which cannot be ignored in the process of establishing a smart grid.
The GEP (gene expression programming) is gene expression programming, is provided by a grapevine evolution biologist Ferreira for referencing the genetic rule of biogenetic, and inherits the advantages of both genetic programming and genetic algorithm. The method can evolve computer programs with various forms such as mathematical expressions, neural networks, decision trees, polynomial construction, logic expressions and the like, and help researchers construct stable and accurate data models. In modern power system application, the data identification problem is researched based on a big data algorithm, so that the defect of traditional data abnormity identification can be effectively overcome, and the method is a feasible method for removing power grid data abnormity and improving data quality. The problem of identifying data anomalies generally falls into the research category of grid information safety.
The effective identification method for the data abnormity facing the smart power grid mainly needs to consider the problems in two aspects:
(1) how to extract the correlation among data from the mass data of state estimation, identify data abnormity through the data correlation and quickly detect the attacked node;
(2) how to guarantee that the characteristics and the integrality of original data are guaranteed to the maximum extent when the data abnormity under the power distribution network is identified.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to provide an effective identification method for data abnormity of a smart power grid based on GEP optimization, which is used for solving the problem of data abnormity identification under the smart power grid; the mechanism is a strategic method, abnormal data in the power grid can be rapidly and effectively identified by using the method, and the safe and stable operation of the power distribution network is ensured.
The technical scheme is as follows: in order to achieve the purpose, the invention provides the following technical scheme:
a GEP optimization-based intelligent power grid data anomaly effective identification method comprises the following steps:
1) obtaining a topological relation between a measurement value and a power system under an SCADA (supervisory control and data acquisition) system, executing a state estimation process, iteratively approaching a state initial value to a current operation state value by using minimum weighted multiplication for solving a minimum residual error, and solving a state solution of x under a direct current model;
2) an attack vector a is introduced in the state estimation process, when a is equal to Hc, the traditional residual error detection cannot completely identify data abnormality, and only preliminary index judgment can be carried out;
3) taking a reactive power network loss expected value and a voltage stability margin as comprehensive indexes by combining an energy function; performing n iterations by using a self-adaptive weight method to take the comprehensive index as a fitness function of the GEP algorithm, and solving the size of the comprehensive index at different times;
4) and (3) solving a comprehensive index value between n different points through a GEP algorithm, solving a distance difference by utilizing the comprehensive index of the same node before and after the attack according to the idea of a node energy function, and identifying that data abnormality occurs in state estimation at a certain time if the distance is too large and cannot meet the precision requirement.
Further, the step 1) includes the following steps:
step 1.1) firstly, acquiring a measurement value and a topological relation of an electric power system under an SCADA (supervisory control and data acquisition) system, and executing a state estimation process, wherein z is Hx + e, H is a Jacobi matrix, z is the measurement value, and x is a state value; e is an error; entering the step 2;
step 1.2) ideally, an estimated value of a state value for minimizing a residual error needs to be found
Figure GDA0003592273630000021
Usual objective function min J (x) ═ z-h (x)]TR-1[z-h(x)]Obtaining an extreme value; the least weighted two-multiplication iterative calculation is adopted to obtain the most approximate state solution meeting the convergence condition,
Figure GDA0003592273630000022
wherein R is an error variance matrix of the measured quantity;
Figure GDA0003592273630000023
is a state estimation value; step 2) is entered.
Further, the step 2) includes the following steps:
step 2.1) the data analyst introduces an attack vector a,at this time, the measured value is
Figure GDA0003592273630000031
New residual error is
Figure GDA0003592273630000032
Figure GDA0003592273630000033
In order to obtain a measured value,
Figure GDA0003592273630000034
is an estimated value, a is an attack vector; entering step 2.2);
step 2.2) can be known based on the inequality principle
Figure GDA0003592273630000035
c is the influence on the state value after attack;
Figure GDA0003592273630000036
residual errors after being influenced by bad data; entering step 2.3);
step 2.3) as seen from step 2.2), when a ═ Hc,
Figure GDA0003592273630000037
data attacks are undetectable; entering step 2.4);
step 2.4) setting a threshold value epsilon at the moment, judging whether bad data exists or not by comparing the difference value of residual errors before and after attack, wherein the judgment standard is
Figure GDA0003592273630000038
Entering step 2.5), otherwise, the rotation data is abnormal;
and 2.5) because data abnormity possibly exists when the data abnormity is less than or equal to epsilon and is not detected, and the influence of bad data on reactive power optimization and voltage stabilization of the power system is considered, evaluating by using a reactive power optimization comprehensive index as a fitness function in combination with a GEP algorithm, and entering the step 3).
Further, the step 3) includes the following steps:
step 3.1) introducing a GEP algorithm, coding variables in a reactive power optimization model, initializing population setting iteration times n, and entering step 3.2);
step 3.2) selecting a fitness function, wherein the reactive power optimization comprehensive index is maximized into a target function which is
Figure GDA0003592273630000039
ω1Is the reactive network loss weight coefficient and omega2For voltage stabilizing weight coefficient, respectively corresponding to λ1Reactive network loss expectation and lambda2The expected value of the voltage stability margin is obtained, and the step 3.3) is carried out;
and 3.3) because the two index dimensions are different, selecting a self-adaptive weight method, and carrying out normalization maximization treatment on the comprehensive index to obtain a form:
Figure GDA00035922736300000310
wherein
Figure GDA00035922736300000311
Corresponding to the desired maximum value of reactive network loss, lambda1,kCorresponding to the reactive network loss of the kth iteration,
Figure GDA00035922736300000312
corresponding to the minimum value of voltage stability, λ2,kCorresponding to the desired value of voltage stabilization for the k-th time,
Figure GDA00035922736300000313
for reactive power optimization of comprehensive indexes, the scheme corresponds to
Figure GDA00035922736300000314
The smaller the expected value of the reactive network loss is, the larger the voltage stability margin is, the more normal the data is, and the step 4) is carried out.
Further, the step 4) includes the following steps:
step 4.1) optimally retaining the reactive optimization comprehensive index according to the GEP 'preferred' idea, carrying out fitness function evaluation, judging whether a termination condition is met according to the iteration times, and entering step 4.2);
step 4.2) if the GEP algorithm end condition is not reached, the GEP algorithm carries out genetic operation and iteratively evaluates fitness functions of different nodes, otherwise, the step 4.3) is carried out;
step 4.3) if the termination condition is met, outputting the comprehensive indexes under different nodes, and taking the difference according to the comprehensive indexes of the nodes which are not attacked and the comprehensive indexes of the nodes after the attack, namely
Figure GDA0003592273630000041
Wherein
Figure GDA0003592273630000042
The training data of the attacked node of the table k,
Figure GDA0003592273630000043
table k data of nodes not attacked; entering step 4.4);
step 4.4) the comprehensive index difference before and after the attack
Figure GDA0003592273630000044
Judging with a threshold gamma, if the threshold gamma is smaller than the threshold gamma, meeting the precision requirement, and if the threshold gamma is not smaller than the threshold gamma, the data is normal, otherwise, entering the step 4.5);
and 4.5) identifying data abnormality and terminating.
Has the advantages that: compared with the prior art, the method for effectively identifying the data abnormity of the intelligent power grid based on GEP optimization is mainly used for solving the problem of effectively identifying the data abnormity under the intelligent power grid, and the method provided by the invention can be used for effectively identifying the data abnormity under the power grid by combining GEP and an energy function idea and utilizing a mode of comparing a comprehensive index distance with a threshold value according to a large amount of data in the current active power distribution grid, so that the safe and reliable operation of the power grid is well ensured.
Drawings
FIG. 1 is a block diagram of the efficient identification of data anomalies;
FIG. 2 is a reference architecture diagram;
FIG. 3 is a schematic flow diagram of the method of the present invention.
Detailed Description
For a better understanding of the contents of the present patent application, the technical solutions of the present invention will be further described below with reference to the accompanying drawings and specific examples.
As shown in fig. 1, a structure diagram of an effective identification method for data anomaly of a smart grid based on GEP optimization is provided, which mainly includes four parts: the device comprises a state estimator, a residual error detector, a sample trainer and a target recognizer. The state estimator performs state estimation on a large amount of power grid data acquired from the SCADA system; a residual detector: performing preliminary residual error data detection based on the current system operation state value; the sample trainer is used for carrying out iterative labeling by taking the reactive power optimization comprehensive index as a fitness function based on a GEP algorithm; the target recognizer is used for judging the distance difference and the threshold precision between the node data obtained by iteration and the node data which is not attacked, and then data abnormity is effectively recognized. Specific descriptions are given below:
a state estimator: the state estimator is essentially based on load flow calculations of the power system. Generating a nonlinear equation by constructing an objective function; and after the measurement value is obtained, iterative computation is carried out by adopting a weighted least square method to solve a solution vector of the current most approximate system operation state, so as to prepare for residual error identification.
A residual detector: due to the fact that errors exist in the acquired data or the physical equipment fails, bad data exist in the acquired data, and the result of state estimation is inaccurate. Therefore, preliminary detection needs to be performed on the basis of a state solution obtained by the state estimator by using a residual error, so as to preliminarily identify the abnormal data condition in the power grid.
A sample trainer: due to the existence of the false data, the residual error cannot completely detect the data abnormity, but the false data influences the reactive power optimization of the power grid. Therefore, the trainer is mainly based on the idea of an energy function, adopts a self-adaptive weight method to take a reactive power network loss expected value and a voltage stability margin as a reactive power optimization comprehensive index, takes the reactive power optimization comprehensive index as a fitness function of a GEP algorithm, and obtains n times of weights to obtain a training set.
The target recognizer: the target recognizer mainly recognizes data which do not meet the precision requirement through a judgment process of comparing threshold values through double indexes based on residual detection and comprehensive indexes of the same nodes before and after attack, namely, normal data and abnormal data are distinguished, and data abnormity is identified.
First, method flow
1. State estimator
The power system state estimator mainly generates a nonlinear equation set by constructing a power system balance model. And after the measurement value is obtained, the estimation value of the current most approximate system running state value is solved through minimum weighting two-multiplication iterative computation. The relationship between the state values and the measured values is as follows:
z＝Hx+e；
wherein H is a Jacobi matrix, z is a measurement value, and x is a status value; and e is an error. Ideally, with respect to the measured value z, it is necessary to find the state value that minimizes the residual error, and establish the objective function:
min J(x)＝[z-h(x)]TR-1[z-h(x)]；
wherein R is an error variance matrix of the measurement, and an extreme value is solved for a nonlinear equation:
Figure GDA0003592273630000051
and (3) solving the most approximate state value meeting the convergence condition by adopting minimum weighted two-multiplication iterative computation:
Figure GDA0003592273630000052
2. residual error detector
Due to the fact that errors exist in the acquired data or the physical equipment fails, bad data exist in the acquired data, and the result of state estimation is inaccurate. It is therefore necessary to perform a preliminary detection with a residual based on the state solution solved by the state estimator. Wherein the measurement residual can be expressed as:
Figure GDA0003592273630000061
the data analyst introduces an attack vector a, where the measurement is
Figure GDA0003592273630000062
The measured new residual may be expressed as:
Figure GDA0003592273630000063
wherein the estimated value obtained by the state estimation is substituted
Figure GDA0003592273630000064
Measured value
Figure GDA0003592273630000065
The following were obtained:
Figure GDA0003592273630000066
based on the inequality principle
Figure GDA0003592273630000067
c is the influence on the state value after the attack, and obviously, when a is equal to Hc, the residual detection cannot meet the precision requirement.
3. Sample training device
Based on the GEP algorithm, the reactive power optimization problem influenced by data attack is used as a breakthrough, and a comprehensive index of a reactive power network loss expected value and a voltage stability margin of the reactive power optimization problem is used as a fitness function of a gene expression to obtain a training set by adopting a self-adaptive weight method. The specific process is as follows:
(1) encoding variables in a reactive power optimization model, and initializing a group to set a node iteration number n;
(2) and selecting the fitness function, and maximizing the reactive power optimization comprehensive index to be the target function. The comprehensive index mathematical model is as follows:
Figure GDA0003592273630000068
wherein the expected value of reactive network loss
Figure GDA0003592273630000069
Voltage stability margin value
Figure GDA00035922736300000610
Figure GDA00035922736300000611
Is a k-th order network loss, δkVoltage stability margin of order k, omega1Is the reactive network loss weight coefficient and omega2Is a voltage stabilization weight coefficient. Because the two indexes have different dimensions, a self-adaptive weight method is selected, and the comprehensive indexes are normalized to obtain a target function maximization form:
Figure GDA00035922736300000612
wherein λ is1,k，λ2,kAnd respectively obtaining the expected value of the network loss and the expected value of the voltage stability margin of the kth individual in the population. The proposal corresponds to the thought of the GEP algorithm of superiority and inferiority,
Figure GDA00035922736300000613
the larger the expected value of the k node network loss is, the smaller the expected value of the k node network loss is, the larger the voltage stability margin is, and the more normal the data is; in addition to this, the present invention is,
Figure GDA00035922736300000614
the training data of the attacked node of the table k,
Figure GDA00035922736300000615
table k data of nodes not attacked;
(3) evaluating the fitness, and judging whether a termination condition is met;
(4) genetic operation, turning to fitness function iteration;
4. target recognizer
The target recognizer mainly recognizes data which do not meet the precision requirement based on the processes of residual preliminary detection and dual index judgment of comparison between the comprehensive index distance of adjacent points and a threshold value, namely, normal data and abnormal data are distinguished.
If the residual error is detected
Figure GDA0003592273630000071
The detected data is abnormal; if the distance between the adjacent points of the composite index
Figure GDA0003592273630000072
The data at that point k is anomalous. Wherein epsilon is the threshold precision during residual error detection, and gamma is the threshold precision of the comprehensive index for detecting the same node energy before and after the attack.
As shown in fig. 2-3, the method for effectively identifying the data anomaly of the smart grid based on the GEP optimization of the present invention includes the following steps:
1. a GEP optimization-based intelligent power grid data anomaly effective identification method is characterized by comprising the following steps: comprises the following steps:
step 1: firstly, acquiring a measurement value and a topological relation of a power system under an SCADA (supervisory control and data acquisition) system, and executing a state estimation process, wherein z is Hx + e, H is a Jacobi matrix, z is the measurement value, and x is a state value; e is an error; entering the step 2;
step 2: ideally, it is necessary to find an estimated value of a state value for minimizing a residual error
Figure GDA0003592273630000073
Usual objective function min J (x) ═ z-h (x)]TR-1[z-h(x)]Obtaining an extreme value; the least weighted two-multiplication iterative calculation is adopted to obtain the most approximate state solution meeting the convergence condition,
Figure GDA0003592273630000074
wherein R is an error variance matrix of the measurement;
Figure GDA0003592273630000075
is a state estimation value; entering the step 3;
and step 3: the data analyst introduces an attack vector a, in which case the measurement value is
Figure GDA0003592273630000076
New residual error is
Figure GDA0003592273630000077
Figure GDA0003592273630000078
A is a measurement value in an attack state, and a is an attack vector; entering the step 4;
and 4, step 4: based on the inequality principle
Figure GDA0003592273630000079
c is the influence on the state value after attack;
Figure GDA00035922736300000710
the residual error is the residual error influenced by bad data; entering the step 5;
and 5: as can be seen from step 4, when a ═ Hc,
Figure GDA00035922736300000711
data attacks are undetectable; entering step 6;
step 6: at the moment, a threshold value epsilon is set, whether bad data exists is judged by comparing the difference value of residual errors before and after attack, and the judgment standard is
Figure GDA0003592273630000081
Entering step 7, otherwise, the data is abnormal;
and 7: if the data abnormality is less than or equal to epsilon, the data abnormality is not detected, and the influence of bad data on reactive power optimization and voltage stabilization of the power system is considered, so that the reactive power optimization comprehensive index is used as a fitness function for evaluation by combining a GEP algorithm, and the step 8 is carried out;
and 8: introducing a GEP algorithm, coding variables in the reactive power optimization model, initializing population and setting iteration times n, and entering step 9;
and step 9: the fitness function is selected by maximizing the reactive power optimization comprehensive index into an objective function which is
Figure GDA0003592273630000082
ω1Is the reactive network loss weight coefficient and omega2For voltage stabilizing weight coefficient, respectively corresponding to λ1Reactive network loss expectation and lambda2The expected value of the voltage stability margin enters the step 10;
step 10: because the two indexes have different dimensions, a self-adaptive weight method is selected, and the comprehensive index is subjected to normalization and maximization treatment to obtain a form:
Figure GDA0003592273630000083
wherein
Figure GDA0003592273630000084
Corresponding to the desired maximum value of reactive network loss, lambda1,kCorresponding to the reactive network loss of the kth iteration,
Figure GDA0003592273630000085
corresponding to the minimum value of voltage stability, λ2,kCorresponding to the expected value of the voltage stability at the k time
Figure GDA0003592273630000086
The larger the expected value of the reactive power network loss is, the larger the voltage stability margin is, the more normal the data is, and the step 11 is carried out;
step 11: optimally reserving the reactive power optimization comprehensive index according to the GEP (GEP) idea, evaluating a fitness function, judging whether a termination condition is met according to the iteration times, and entering step 12;
step 12: if the GEP algorithm end condition is not met, the GEP algorithm carries out genetic operation and iteratively evaluates fitness functions of different nodes, otherwise, the step 13 is carried out;
step 13: if the termination condition is met, outputting the comprehensive indexes under different nodes according to the condition that the node is not receivedThe integrated index of the attacked node is subtracted from the integrated index of the attacked node, i.e.
Figure GDA0003592273630000087
Wherein
Figure GDA0003592273630000088
The node comprehensive index data representing the attack of the k nodes,
Figure GDA0003592273630000089
representing the node comprehensive index data of the k nodes which are not attacked; (ii) a Entering step 14;
step 14: the difference of the comprehensive indexes before and after attack
Figure GDA00035922736300000810
Judging with a threshold gamma, if the threshold gamma is smaller than the threshold gamma, meeting the precision requirement, and if the threshold gamma is not smaller than the threshold gamma, the data is normal, otherwise, entering the step 15;
step 15: and identifying data abnormality and terminating.
Examples
Data abnormity and data attack in the power grid can not only cause unstable voltage of the power system, but also cause huge economic loss. Assuming that an estimation vector a is introduced in the state estimation process, the dummy data cannot be completely identified in the residual detection. Based on the influence of false data on the reactive power optimization of the power grid, the idea of an energy function is combined, a reactive power grid loss expected value and a voltage stability margin are used as comprehensive indexes, the comprehensive indexes are used as fitness functions in a GEP algorithm, and n iterations are carried out. And finally, comparing the precision of the distance threshold of the comprehensive index, and identifying data abnormality.
The specific implementation scheme is as follows:
(1) firstly, acquiring a topological relation between a measurement value and a power system under an SCADA system, executing a state estimation process, iteratively approaching a state initial value to a current running state value by using minimum weighted double multiplication for solving a minimum residual error, and solving a state solution of x under a direct current model.
(2) An attack vector a is introduced in the state estimation process, and when a is equal to Hc, the traditional residual detection cannot completely identify data abnormality and only can make preliminary index judgment.
(3) Bad data cannot be completely detected based on traditional residual error detection, and the data influences reactive power optimization of a power grid. Therefore, the reactive power grid loss expectation value and the voltage stability margin are used as comprehensive indexes by combining the idea of an energy function. And (3) performing n iterations by using the comprehensive index as a fitness function of the GEP algorithm by using a self-adaptive weight method, and solving the size of the comprehensive index at different times. K is the number of nodes, n is the number of nodes, where the relationship is 0< ═ K < ═ n
(4) The comprehensive index value between different points of n times can be obtained through the GEP algorithm, the distance difference is obtained by utilizing the comprehensive index of the same node before and after attack according to the node energy function idea, and if the distance is too large to meet the precision requirement, the data abnormality in the state estimation of a certain time can be identified.

Claims (1)

1. A GEP optimization-based intelligent power grid data anomaly effective identification method is characterized by comprising the following steps: comprises the following steps:
1) obtaining a topological relation between a measurement value and a power system under an SCADA (supervisory control and data acquisition) system, executing a state estimation process, iteratively approaching a state initial value to a current running state value by using minimum weighted double multiplication for solving a minimum residual error, and solving a state solution of x under a direct current model;
the method comprises the following steps: step 1.1) firstly, acquiring a measurement value and a topological relation of an electric power system under an SCADA (supervisory control and data acquisition) system, and executing a state estimation process, wherein z is Hx + e, H is a Jacobi matrix, z is the measurement value, and x is a state value; e is an error; entering step 1.2);
step 1.2) ideally, an estimated value of a state value for minimizing a residual error needs to be found
Figure FDA00035922736200000110
Usual objective function minJ (x) ═ z-h (x)]TR-1[z-h(x)]Obtaining an extreme value; the least weighted two-multiplication iterative calculation is adopted to obtain the most approximate state solution meeting the convergence condition,
Figure FDA0003592273620000011
wherein R is an error variance matrix of the measurement;
Figure FDA0003592273620000012
is a state estimation value; entering step 2);
2) an attack vector a is introduced in the state estimation process, when a is equal to Hc, the traditional residual error detection cannot completely identify data abnormality, and only preliminary index judgment can be carried out; the method specifically comprises the following steps:
step 2.1) data analyst introduces an attack vector a, in which case the measured value is
Figure FDA0003592273620000013
New residual error is
Figure FDA0003592273620000014
Figure FDA0003592273620000015
A is a measurement value in an attack state, and a is an attack vector; entering step 2.2);
step 2.2) can be known based on the inequality principle
Figure FDA0003592273620000016
c is the influence on the state value after attack;
Figure FDA0003592273620000017
the residual error is the residual error influenced by bad data; entering step 2.3);
step 2.3) as seen from step 2.2), when a ═ Hc,
Figure FDA0003592273620000018
data attacks are undetectable; entering step 2.4);
step 2.4) setting a threshold value epsilon at the moment, judging whether bad data exists or not by comparing the difference value of residual errors before and after attack, wherein the judgment standard is
Figure FDA0003592273620000019
Entering step 2.5), otherwise, the rotation data is abnormal;
step 2.5) because data abnormity possibly exists when the data abnormity is less than or equal to epsilon and is not detected, and the influence of bad data on reactive power optimization and voltage stabilization of the power system is considered, the reactive power optimization comprehensive index is used as a fitness function for evaluation in combination with a GEP algorithm, and the step 3) is carried out;
3) taking a reactive power network loss expected value and a voltage stability margin as comprehensive indexes by combining an energy function; performing n iterations by using a self-adaptive weight method to take the comprehensive index as a fitness function of the GEP algorithm, and solving the size of the comprehensive index at different times; the method specifically comprises the following steps:
step 3.1) introducing a GEP algorithm, coding variables in a reactive power optimization model, initializing population setting iteration times n, and entering step 3.2);
step 3.2) selecting a fitness function, wherein the reactive power optimization comprehensive index is maximized into a target function which is
Figure FDA0003592273620000021
ω1Is the reactive network loss weight coefficient and omega2For voltage stabilizing weight coefficient, respectively corresponding to λ1Reactive network loss expectation and lambda2The expected value of the voltage stability margin is obtained, and the step 3.3) is carried out;
and 3.3) selecting a self-adaptive weight method to carry out normalization and maximization treatment on the comprehensive indexes to obtain a form:
Figure FDA0003592273620000022
wherein
Figure FDA0003592273620000023
Corresponding to the desired maximum value of reactive network loss, lambda1,kCorresponding to the reactive network loss of the kth iteration,
Figure FDA0003592273620000024
corresponding to the minimum value of voltage stability, λ2,kCorresponding to the expected value of the voltage stability at the k time
Figure FDA0003592273620000025
The larger the expected value of the reactive power network loss is, the larger the voltage stability margin is, the more normal the data is, and the step 4) is carried out;
4) the method comprises the steps of solving a comprehensive index value between different nodes for n times through a GEP algorithm, solving a distance difference by utilizing the comprehensive index of the same node before and after attack according to a node energy function idea, and identifying that data abnormality occurs in state estimation at a certain time if the distance is too large and cannot meet the precision requirement; the method specifically comprises the following steps:
step 4.1) optimally retaining the reactive power optimization comprehensive index according to the GEP 'preferred' idea, carrying out fitness function evaluation, judging whether a termination condition is met according to iteration times, and entering step 4.2);
step 4.2) if the GEP algorithm end condition is not reached, the GEP algorithm carries out genetic operation and iteratively evaluates fitness functions of different nodes, otherwise, the step 4.3) is carried out;
step 4.3) if the termination condition is met, outputting the comprehensive indexes under different nodes, and taking the difference according to the comprehensive indexes of the nodes which are not attacked and the comprehensive indexes of the nodes after the attack, namely
Figure FDA0003592273620000026
Wherein
Figure FDA0003592273620000027
The node comprehensive index data representing the attack of the k nodes,
Figure FDA0003592273620000028
representing the node comprehensive index data of the k nodes which are not attacked; entering step 4.4);
step 4.4) the comprehensive index difference before and after the attack
Figure FDA0003592273620000029
Judging with a threshold gamma, if the threshold gamma is smaller than the threshold gamma, meeting the precision requirement, and if the threshold gamma is not smaller than the threshold gamma, the data is normal, otherwise, entering the step 4.5);
and 4.5) identifying data abnormality and terminating.
CN202010112678.4A 2020-02-24 2020-02-24 GEP optimization-based smart power grid data anomaly effective identification method Active CN111222139B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010112678.4A CN111222139B (en) 2020-02-24 2020-02-24 GEP optimization-based smart power grid data anomaly effective identification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010112678.4A CN111222139B (en) 2020-02-24 2020-02-24 GEP optimization-based smart power grid data anomaly effective identification method

Publications (2)

Publication Number Publication Date
CN111222139A CN111222139A (en) 2020-06-02
CN111222139B true CN111222139B (en) 2022-06-03

Family

ID=70811941

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010112678.4A Active CN111222139B (en) 2020-02-24 2020-02-24 GEP optimization-based smart power grid data anomaly effective identification method

Country Status (1)

Country Link
CN (1) CN111222139B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112803402B (en) * 2021-02-23 2023-03-31 重庆大学 Radiation network forward-push back substitution robust state estimation method containing bad data preprocessing
CN113449060B (en) * 2021-06-29 2022-07-01 金陵科技学院 Geographic big data security risk assessment method based on mixed gene expression programming
CN113488980B (en) * 2021-07-07 2023-12-19 南京邮电大学 Attack tolerance control method of direct-current micro-grid under denial of service attack
CN115935284A (en) * 2022-11-04 2023-04-07 贵州电网有限责任公司信息中心 Power grid abnormal voltage detection method, device, equipment and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102915423A (en) * 2012-09-11 2013-02-06 中国电力科学研究院 System and method for filtering electric power business data on basis of rough sets and gene expressions
CN104980440A (en) * 2015-06-23 2015-10-14 南京邮电大学 Active power distribution network big data transmission method based on content filtering and multi-Agent cooperation
CN110297911A (en) * 2018-03-21 2019-10-01 国际商业机器公司 Internet of Things (IOT) calculates the method and system that cognition data are managed and protected in environment

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10896378B2 (en) * 2018-01-02 2021-01-19 International Business Machines Corporation Fast detection of energy consumption anomalies in buildings

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102915423A (en) * 2012-09-11 2013-02-06 中国电力科学研究院 System and method for filtering electric power business data on basis of rough sets and gene expressions
CN104980440A (en) * 2015-06-23 2015-10-14 南京邮电大学 Active power distribution network big data transmission method based on content filtering and multi-Agent cooperation
CN110297911A (en) * 2018-03-21 2019-10-01 国际商业机器公司 Internet of Things (IOT) calculates the method and system that cognition data are managed and protected in environment

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
Application of gene expression programming (GEP) in power transformers fault diagnosis using DGA;Hasmat Malik 等;《2014 6th IEEE Power India International Conference (PIICON)》;20150604;全文 *
Detecting False Data Injection Attacks on Power Grid by Sparse Optimization;Liu Lanchao 等;《IEEE Transactions on Smart Grid ( Volume: 5, Issue: 2, March 2014)》;20140331;全文 *
Security risk assessment of cyber physical power system based on rough set and gene expression programming;Song Deng 等;《IEEE/CAA JOURNAL OF AUTOMATICA SINICA》;20151031;全文 *
基于混合基因表达式编程的入侵检测算法;邓松 等;《计算机与现代化》;20110930;全文 *

Also Published As

Publication number Publication date
CN111222139A (en) 2020-06-02

Similar Documents

Publication Publication Date Title
CN111222139B (en) GEP optimization-based smart power grid data anomaly effective identification method
CN111598179B (en) Power monitoring system user abnormal behavior analysis method, storage medium and equipment
CN111585948B (en) Intelligent network security situation prediction method based on power grid big data
CN110909811A (en) OCSVM (online charging management system) -based power grid abnormal behavior detection and analysis method and system
CN112333194B (en) GRU-CNN-based comprehensive energy network security attack detection method
CN111582542B (en) Power load prediction method and system based on anomaly repair
CN109902740B (en) Re-learning industrial control intrusion detection method based on multi-algorithm fusion parallelism
CN111711608B (en) Method and system for detecting abnormal flow of power data network and electronic equipment
CN110689069A (en) Transformer fault type diagnosis method based on semi-supervised BP network
CN112433518B (en) Industrial control system intrusion detection method based on recurrent neural network
CN107220540A (en) Intrusion detection method based on intensified learning
Wang et al. An accurate false data detection in smart grid based on residual recurrent neural network and adaptive threshold
CN114760098A (en) CNN-GRU-based power grid false data injection detection method and device
CN111882041A (en) Power grid attack detection method and device based on improved RNN (neural network)
CN114167284B (en) Lithium battery RUL prediction method and equipment based on BMS big data and integrated learning
CN112215279B (en) Power grid fault diagnosis method based on immune RBF neural network
CN107102223A (en) NPC photovoltaic DC-to-AC converter method for diagnosing faults based on improved hidden Markov model GHMM
CN117580046A (en) Deep learning-based 5G network dynamic security capability scheduling method
CN116545764B (en) Abnormal data detection method, system and equipment of industrial Internet
CN110061986B (en) Network intrusion anomaly detection method based on combination of genetic algorithm and ANFIS
CN112422546A (en) Network anomaly detection method based on variable neighborhood algorithm and fuzzy clustering
CN112333147A (en) Nuclear power plant DCS platform network operation situation sensing method and system
CN115145790A (en) False data injection attack detection method and system for smart power grid
Tan et al. Using hidden markov models to evaluate the real-time risks of network
CN114169395A (en) Construction method and application of dominant instability mode recognition model of power system

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
CB02 Change of applicant information

Address after: 210003 Gulou District, Jiangsu, Nanjing new model road, No. 66

Applicant after: NANJING University OF POSTS AND TELECOMMUNICATIONS

Address before: Yuen Road Qixia District of Nanjing City, Jiangsu Province, No. 9 210023

Applicant before: NANJING University OF POSTS AND TELECOMMUNICATIONS

CB02 Change of applicant information
GR01 Patent grant
GR01 Patent grant