阅读说明：本技术 智能电网中基于pmu部署的错误数据注入攻击防御方法 (PMU (phasor measurement unit) deployment-based error data injection attack defense method in smart power grid ) 是由 梁炜 裴超 *** 杨雨沱 刘帅 王恺 韩晓佳 肖杨 于 2019-03-20 设计创作，主要内容包括：本发明涉及智能电网安全技术,具体地说是智能电网中基于PMU部署的错误数据注入攻击防御方法。该方法包括新的低开销混合攻击机制、脆弱节点的PMU预部署以及全网基于PMU的贪心策略部署三个阶段。在新的低开销混合攻击机制阶段,攻击者可以用较低的开销攻击智能电网连接度较小的母线来引起错误状态估计结果。在脆弱节点的PMU预部署阶段,首先将最易受攻击的母线通过PMU的部署保护起来。在全网基于PMU贪心策略部署阶段,根据当前的攻击向量,部署的每轮在能保护最大数量测量值的母线部署一个PMU设备,直到整个电网完全可观。本发明提出的防御方法使得攻击者增加攻击开销,同时减小了部署迭代过程,在部署完成后,实现了对错误数据注入攻击的有效防御。(The invention relates to a security technology of a smart power grid, in particular to a defense method for error data injection attack based on PMU deployment in the smart power grid. The method comprises three stages of a new low-overhead hybrid attack mechanism, PMU pre-deployment of the fragile nodes and PMU-based greedy strategy deployment in the whole network. In a new low-overhead hybrid attack mechanism stage, an attacker can attack a bus with low smart grid connectivity with low overhead to cause an error state estimation result. In the PMU pre-deployment stage of the vulnerable node, the bus which is most vulnerable is firstly protected by the deployment of the PMU. In the deployment stage of the PMU greedy-based strategy in the whole network, one PMU device is deployed at a bus capable of protecting the maximum number of measured values in each deployed round according to the current attack vector until the whole power grid is completely observable. The defense method provided by the invention enables an attacker to increase attack overhead, reduces the deployment iteration process, and realizes effective defense on error data injection attack after deployment is completed.)

1. The method for defending the attack of the injection of the wrong data based on PMU deployment in the smart grid is characterized by comprising the following steps:

searching a fragile node according to a Jacobian matrix of the measured value of the power grid;

performing PMU pre-deployment on the fragile nodes;

and carrying out greedy strategy deployment on other nodes to realize defense against error data injection attack.

2. The method for defending against the injection of wrong data based on PMU deployment in the smart grid according to claim 1, characterized in that the weak nodes are obtained according to a Jacobian matrix of grid measurement values, comprising the following steps:

the method comprises the following steps: jacobian matrix for attacker receiving power grid measured value

step two: initializing an attack vector a₁、a₂，a₁＝a₂Zero (m,1), initializing tracking matrix Q I_m×mI is an identity matrix used for tracking column exchange of U; initializing a counter count zero zeros (1, n) to count the Jacobian matrix of the measured valuesThe number of non-zero elements in each column vector;

step three: jacobian matrix for measured values

step four: the attacker selects t minimum count (i) in the n power grid state variables, and for the t corresponding state variables, the attack vector satisfies a₁＝a₁+γ_ih_i；i∈[1，t]γ is a non-zero arbitrary constant, h_iIs the ith column vector in the Jacobian matrix;

step five: obtaining an attack vector a2 based on the elementary row transformation and the column exchange of the matrix U;

step six: attack vector a2 ═ Q (Q)^T)^-1(Ue_n) Wherein e is_nThe n-dimensional column vector with the last row of elements being 1 and the rest being 0 is used for selecting the last row of the matrix U; the final attack vector is a^*＝a₁+a₂。

3. The method for defending against the attack of the injection of the wrong data based on the PMU deployment in the smart grid according to claim 1, characterized in that the protected representation bus is provided with PMU devices, so that the measured values related to the bus, namely the bus active injection power and the branch active power flow power, are not attacked and tampered by an attacker.

4. The method for defending against the attack of the injection of the wrong data based on the PMU deployment in the smart grid according to claim 1, wherein the unprotected state means that a certain bus is not equipped with PMU equipment, so that the measured values related to the bus, namely the bus active injection power and the branch active power flow power, can be attacked and tampered by an attacker.

5. The method for defending against false data injection attack based on PMU deployment in smart grid according to claim 2, characterized in that the attack vector a is obtained based on matrix elementary row transformation and column exchange₂The method comprises the following steps:

1) performing primary row transformation on the matrix U to obtain a row simplest shape;

2) finding a row with the fewest non-zero elements in the simplest row shape, and placing the non-zero elements into the rearmost column in the row through column exchange;

3) the matrix Q is used to track changes in column switching;

4) the matrices U and Q are updated until the matrix U is no longer changed.

6. The method for defending against the injection of wrong data based on PMU deployment in the smart grid according to claim 5, characterized in that the matrix Q is used to track the change of column exchange as follows: in the simplest row form obtained by the initial row transformation of the matrix U, if some two rows of the row with the minimum number of non-zero elements are exchanged, two rows with the same row serial number in the unit matrix Q are correspondingly exchanged.

7. The method for defending against the PMU deployment-based fault data injection attack in the smart grid according to claim 1, wherein the PMU pre-deployment of the vulnerable nodes comprises the following steps:

respectively deploying PMU equipment on adjacent buses of buses corresponding to t state variables of the fragile node;

measured value Jacobian matrixThe row of measurements corresponding to the bus where the PMU is deployed and its respective adjacent bus is moved to H_SIn and updateAnd H_S；Representing a sub-matrix consisting of the row vectors corresponding to the unprotected measurements in the measurement jacobian matrix.

8. The method for defending against the injection of the wrong data based on PMU deployment in the smart grid according to claim 1, wherein the greedy strategy deployment for other nodes includes the following steps:

the method comprises the following steps: will matrixDivided into (n-t) sub-matrices

step two: according to the obtained attack vector a^*Obtaining a matrixIn which the submatrix is formed by row vectors corresponding to non-zero elements in the attack vector

step three: initializing a counter count (zeros (1, n-t)) and counting each submatrixAnd submatrices corresponding to non-zero elements in the attack vector

Where card is an operator for counting the number of elements in the set, L_targetA target bus corresponding to the maximum value of the counter is represented;

step four: in correspondence with L_targetA PMU device is arranged on the bus and the Jacobian moment drop of the measured value is measuredThe row of measurements corresponding to the bus where the PMU is deployed and its respective adjacent bus is moved to H_SPerforming the following steps;

step five: updating a matrix

9. The method for defending against false data injection attack based on PMU deployment in a smart grid according to claim 1, wherein the measured value includes at least one of bus voltage amplitude, bus active injection power and branch active tide.

10. The method for defending against erroneous data injection attack based on PMU deployment in the smart grid according to claim 1, wherein the state vector includes a magnitude and a phase angle of a bus voltage.

Technical Field

The invention relates to a security technology of a smart power grid, in particular to a PMU (phasor measurement unit) deployment-based error data injection attack defense method in the smart power grid.

Background

The smart grid is a modern power network system, and the smart grid enables the balance of supply and demand of electric energy to be more reliable, economic and sustainable through the fusion of an information technology and a communication technology. Compared with the traditional power network, the smart grid has the obvious characteristic that a large amount of data is generated due to bidirectional information interaction between the grid and users. The two-way information interaction between the consumer, the control center and the control device makes the supply and demand of energy more efficient. At the power supply side, the devices in the power system can be managed more intelligently, and the flexibility of power supply is greatly improved. On the user side, the user experience and charging system may be further enhanced. Although the development of smart grids brings many excellent characteristics, the strong coupling of network systems and physical systems also makes smart grids more vulnerable to a variety of network attacks. Generally, an electric power system belongs to a key infrastructure of a country, and the destruction or power failure of the electric power system can bring threats to the economy, safety and even life health of the country.

In order to ensure the normal operation of the smart grid, the power network needs to be continuously monitored and controlled by a SCADA (data acquisition and monitoring control system) system and an EMS (energy management system). It is generally not possible to measure directly all the state variables of the power system, especially the phase angle for the individual busbars, with sensors. Therefore, obtaining accurate state estimates by the state estimator is crucial for subsequent control and analysis. One important factor affecting the accuracy of state estimation is the introduction of bad data. Bad data is typically caused by non-malicious failures or malicious network attacks. Detection techniques for bad data are typically based on: erroneous measurements will result in larger normalized residuals. However, the error data injection attack proposed in 2009 can avoid the conventional detection based on the normalized measurement value residual error, and cause the wrong state estimation, which brings great threat to the subsequent control and decision of the power grid.

In order to defend against the attack of error data injection occurring in the smart grid, an effective way is to add an additional security mechanism. Since PMUs are advanced measurement units, accurate real-time passing phase information can be provided, and since it can be synchronized with GPS, it is generally difficult for an attacker to tamper with the PMU-protected measurements. While the existing defense algorithm based on PMU deployment usually deploys PMUs by means of a greedy strategy, only one PMU device is deployed on a bus capable of protecting the maximum number of measured values according to the greedy strategy in the PMU deployment process, but when wrong data injection attacks on the bus with low connectivity in a power grid in a strategic manner in the PMU deployment process, the algorithms based on the greedy strategy are often insufficient and cannot effectively defend the attacks. It is necessary to redesign a new PMU deployment method to protect against the occurrence of erroneous data injection attacks.

Disclosure of Invention

The invention provides a method for defending an error data injection attack deployed based on a greedy strategy PMU in a smart grid, aiming at the problem that the existing method for defending the error data injection attack deployed based on the greedy strategy PMU in the smart grid fails when a new mixed error data injection attack is faced. The method comprises three steps of a new low-overhead hybrid attack mechanism, PMU pre-deployment of the fragile nodes and PMU-based greedy strategy deployment in the whole network.

In order to solve the technical problems, the invention adopts the technical scheme that: a PMU deployment-based error data injection attack defense method in a smart power grid comprises the following steps:

searching a fragile node according to a Jacobian matrix of the measured value of the power grid;

performing PMU pre-deployment on the fragile nodes;

and carrying out greedy strategy deployment on other nodes to realize defense against error data injection attack.

The method for obtaining the fragile node according to the Jacobian matrix of the power grid measurement value comprises the following steps:

the method comprises the following steps: jacobian matrix for attacker receiving power grid measured value

And obtaining a measured value Jacobian matrixThe number of rows m and columns n, is noted

Wherein the content of the first and second substances,representing a sub-matrix consisting of row vectors corresponding to unprotected measurement values in a measurement value Jacobian matrix; at the same time order

Namely U is the transposition of the Jacobian matrix of the measured values;

step two: initializing an attack vector a₁、a₂，a₁＝a₂Zero (m,1), initializing tracking matrix Q I_M×MI is an identity matrix used for tracking column exchange of U; initializing a counter count zero zeros (1, n) to count the Jacobian matrix of the measured values

The number of non-zero elements in each column vector;

step three: jacobian matrix for measured valuesSearching to obtain the number of non-zero elements in each column vector; the node corresponding to the column vector with the minimum number of the nonzero elements is a fragile node;

step four: the attacker selects t minimum count (i) in the n power grid state variables, and for the t corresponding state variables, the attack vector satisfies a₁＝a₁+γ_Ih_I；i∈[1,t]γ is a non-zero arbitrary constant, h_IIs the ith column vector in the Jacobian matrix;

step five: attack vector a is obtained based on matrix U elementary row transformation and column exchange₂；

Step six: attack vector a₂＝(Q^t)^-1(Ue_N) Wherein e is_NThe n-dimensional column vector with the last row of elements being 1 and the rest being 0 is used for selecting the last row of the matrix U; the final attack vector is a^*＝a₁+a₂。

The protected representation bus is provided with PMU equipment, so that measured values related to the bus, namely bus active injection power and branch active tidal power, are not attacked and tampered by attackers.

The unprotected state indicates that a certain bus is not provided with PMU equipment, so that the measured values related to the bus, namely the bus active injection power and the branch active tidal current power, can be attacked and tampered by an attacker.

Obtaining an attack vector a based on matrix elementary row transformation and column exchange₂The method comprises the following steps:

1) performing primary row transformation on the matrix U to obtain a row simplest shape;

2) finding a row with the fewest non-zero elements in the simplest row shape, and placing the non-zero elements into the rearmost column in the row through column exchange;

3) the matrix Q is used to track changes in column switching;

4) the matrices U and Q are updated until the matrix U is no longer changed.

The matrix Q is used to track the change in column switching as follows: in the simplest row form obtained by the initial row transformation of the matrix U, if some two rows of the row with the minimum number of non-zero elements are exchanged, two rows with the same row serial number in the unit matrix Q are correspondingly exchanged.

The PMU pre-deployment of the fragile node comprises the following steps:

respectively deploying PMU equipment on adjacent buses of buses corresponding to t state variables of the fragile node;

measured value Jacobian matrix

The row of measurements corresponding to the bus where the PMU is deployed and its respective adjacent bus is moved to H_SIn and updateAnd H_S；

Representing a sub-matrix consisting of the row vectors corresponding to the unprotected measurements in the measurement jacobian matrix.

The greedy strategy deployment for other nodes comprises the following steps:

the method comprises the following steps: will matrix

Divided into (n-t) sub-matrices

Wherein L is_I∈[1,n-t]，Representation matrix

A sub-matrix formed by column vectors corresponding to the remaining unprotected state variables;

step two: according to the obtained attack vector a^*Obtaining a matrixIn which the submatrix is formed by row vectors corresponding to non-zero elements in the attack vector

Whereink∈[1,Ω]Omega is the number of nonzero elements in the attack vector;

step three: initializing a counter count (zeros (1, n-t)) and counting each submatrixAnd submatrices corresponding to non-zero elements in the attack vector

The number of the same elements in the solution and the maximum value obtained are recorded as

Where card is an operator for counting the number of elements in the set, L_targetA target bus corresponding to the maximum value of the counter is represented;

step four: in correspondence with L_targetA PMU device is arranged on the bus, and a Jacobian matrix of measured values is obtained

The row of measurements corresponding to the bus where the PMU is deployed and its respective adjacent bus is moved to H_SPerforming the following steps;

step five: updating a matrix

H_SAnd attack vector a^*(ii) a Return to step until rank (H)_S) N, i.e. matrix H_SAnd if the rank is full, the whole power grid is completely observable, and the wrong data injection attack cannot happen again.

The measurements include at least one of bus voltage amplitude, bus active injected power, and branch active tide.

The state vector includes a magnitude and a phase angle of the bus voltage.

The invention provides a PMU deployment-based error data injection attack defense method in a smart power grid. The method comprehensively considers the concealable characteristic of error data injection attack in the smart grid and the problem that the traditional PMU deployment strategy based on a greedy method fails to inject new mixed error data, and further provides a new method for defending error data injection attack based on PMU deployment, so that an attacker is prevented from attacking a bus with low connectivity of the smart grid with low cost to cause error state estimation. The concrete points are as follows:

1. the invention provides a novel low-overhead mixed error data injection attack mechanism, an attacker can attack a bus with low connectivity of a smart grid by using low overhead to cause an error state estimation result, and a traditional greedy method-based PMU deployment strategy fails to work on the mixed attack mechanism.

2. The invention provides PMU pre-deployment of a fragile node, when a bus with low connectivity is protected first, corresponding state variables in a power grid are also protected, and meanwhile, attack cost required by an attacker for generating an attack vector in the PMU deployment process is increased.

3. The invention provides PMU-based greedy strategy deployment in the whole network, and after the PMU is pre-deployed on a fragile node, one PMU device is deployed on a bus capable of protecting the maximum number of measured values in each deployed round according to the current attack vector by using the idea based on the greedy strategy until the whole power grid is completely observable. After deployment is completed, effective defense against error data injection attacks is achieved.

Drawings

FIG. 1 is a diagram illustrating state estimation and error data attack of a smart grid;

FIG. 2 is a schematic diagram of a PMU deployment for an IEEE14 bus system;

fig. 3 is a schematic diagram of a PMU deployment of an IEEE30 bus system.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings.

The invention provides a PMU (phasor measurement unit) deployment-based error data injection attack defense method in a smart power grid. In a new low-overhead hybrid attack mechanism stage, an attacker can attack a bus with a small connection degree of a smart grid by using low overhead to cause an error state estimation result, and a traditional PMU (phasor measurement Unit) deployment strategy based on a greedy method fails to work on the hybrid attack mechanism. To defend against this new error data injection attack, the bus that is most vulnerable is first protected by the deployment of the PMU during the PMU pre-deployment phase of the vulnerable node. In the deployment stage of the PMU greedy-based strategy in the whole network, one PMU device is deployed at a bus capable of protecting the maximum number of measured values in each deployed round according to the current attack vector until the whole power grid is completely observable. The defense method provided by the invention enables an attacker to increase attack overhead, reduces the deployment iteration process, and realizes effective defense on error data injection attack after deployment is completed.

1. The method is suitable for state estimation and error data injection attacks of the smart grid.

The method considers a steady state lossless power transmission network as shown in fig. 1, and smart grid oriented state estimation, where remote terminal units such as smart meters, smart sensors and actuators are used to monitor real-time measurements of the grid. The collected data is transmitted via a communication network to a SCADA server in the control center, where a state estimator then estimates the state variables of the grid using these received redundant measurements and other available information such as topology information. The state variables are typically voltage magnitude and phase angle in an ac transmission network and voltage phase angle in a dc transmission network. These state variables must be accurately estimated so that other functions of the subsequent energy management system, such as optimal power flow analysis, automatic generation control, economic dispatch and contingency analysis, can be reliably controlled to balance the supply and demand of the electric energy. The initiation of the fault data injection attack may be by various means, such as by tampering with the readings of the smart sensor or smart meter, by intervening transmission links that disturb the data, or by directly damaging the database in the control center.

Suppose that the grid includes n buses and m measurement meters, generally m > n. The direct current measurement equation formed between the state variable and the measured value is expressed as

z_k＝Hx_k+v_k

Wherein k represents time, z_kRepresenting the vector of the measurement values obtained by the measuring meter, typically including the bus voltage amplitude, the bus active injection power and the active power flow of the branch, x_kRepresenting the state vector of the power grid, the bus voltage phase angle in the direct current measurement equation, H represents the measured Jacobian matrix, v_kRepresenting the measurement noise, usually white gaussian noise assumed to be zero mean, and the covariance matrix of the measurement noise is denoted R. When the error data injection attack is at k₀When the time happens, the measurement equation becomes

WhereinRepresenting the vector of measurements after attack, a_k＝[a_1,k,a_2,k,…,a_m,k,]Error data representing an attacker injected at time k, k₀Indicating the moment at which the attack begins to occur. The widely used weighted least square state estimation method optimally estimates the state variable by minimizing the weighted least square error, and the estimated state variable in the direct current model is

The corresponding estimated measurement vector is

Conventional bad data detectionThe mechanism is typically a detector based on a residual, defined as the residual between the observed measurement vector and the estimated measurement vector, i.e. the residual is

Wherein I is a unit array. If the residual error r_kAbove a predefined threshold, poor data is indicated, whereas the measurement vector is considered normal.

However, when an attack occurs, the injection of erroneous data may alter the result of the state estimation, which may be expressed as

WhereinRepresenting the estimated state variable after the injection of the error data into the attack, and c representing the deviation of the state variable caused by the occurrence of the attack. Meanwhile, the measurement residual at the time of attack occurrence is represented as

So that the data injected when the error data injection attack satisfies a_kWhen Hc, measure residualWhere τ is the threshold of detection. This means that a fault data injection attack can circumvent the traditional residual-based detection mechanism, making it undetectable.

2. The new low-overhead hybrid attack mechanism comprises the following steps:

the method comprises the following steps: the attacker updates the received measured value Jacobian matrix according to the power grid topological structure informationAnd obtaining a measurementValue Jacobian matrixNumber of rows and columns, notedWhere S represents the set of indices to which the measurements are protected due to the deployment of PMUs,is the complement of S, accordinglyRepresenting a sub-matrix consisting of the row vectors corresponding to the unprotected measurements in the measurement jacobian matrix. Protected means that the measurements (bus active injection power and branch active power flow power) related to the deployed bus are not attacked and tampered by an attacker due to the deployment of the PMU devices, and unprotected means that the measurements (bus active injection power and branch active power flow power) related to the deployed bus are attacked and tampered by the attacker due to the absence of the deployment of the PMU devices. At the same time orderNamely, U is the transpose of the Jacobian matrix of the measured values and is represented by corresponding row vectors;

step two: initializing an attack vector a₁And a₂The two attack vectors are column vectors, m represents the dimension of the attack vector and is equal to the dimension of the measurement value vector; initializing matrix Q ═ I_m×mAn m-dimensional identity matrix, which is used for tracking the column exchange of the matrix U; a counter count (1, n) (representing an n-dimensional zero vector) is initialized to count the measured value Jacobian matrix

The number of non-zero elements in each column vector;

standing at the angle of an attacker, the attacker can only capture the attack due to the restriction of own resources of the attackerAcquiring and tampering a limited number of smart meters; and due to the hidden condition of the error data injection attack satisfied by the attacker, namely a_kBy Hc, it is essentially meant that the attack vector belongs to the column space of the jacobian matrix, i.e.Symbol

Representing the column space of the matrix, the error data injection attack can be constructed as a function of:

wherein | | c | purple light_∞The maximum value of the deviation of the state variable after the attack is represented and is a set threshold value to restrict | | c | calculation_∞The method is used for ensuring that the attack can generate certain influence on the power grid.

Step three: jacobian matrix for measured valuesThe number count (i) of non-zero elements in each column vector is found and obtained corresponding to each bus of the bus system in practice, and then an attacker selects t minimum count values in all n state variables; and the node corresponding to the column vector with the minimum number of the nonzero elements is the fragile node.

Step four: for the corresponding t state variables, making the attack variable satisfy a₁＝a₁+γ_ih_i，i∈[1,t]γ is a non-zero arbitrary constant;

the attack vector provided by the invention is composed of the ith column vector h in the Jacobian matrix_iThe main idea is that: if all the measured values corresponding to a certain state variable are simultaneously tampered, the attack vector can still avoid the detection of the system, and an alarm cannot be triggered. These state variables correspond to column vectors with the smallest number of non-zero elements that are easily tampered with by attackers,since there are few measurements that an attacker needs to tamper with these state variables, these buses are also a bottleneck to the security of the entire system.

Step five: attack vector a is obtained based on matrix elementary row transformation and column exchange₂The method comprises the following specific steps:

1) performing primary row transformation on the matrix U to obtain a row simplest shape;

2) finding a row with the fewest non-zero elements in the simplest row shape, and placing the non-zero elements into the rearmost column in the row through column exchange;

3) the matrix Q is used to track changes in column switching: in the simplest row shape obtained by the initial row transformation of the matrix U, if some two rows of the rows with the minimum number of non-zero elements are exchanged, two rows with the same row serial number in the unit matrix Q are correspondingly exchanged;

4) updating the matrix U and the matrix Q until the matrix U is not changed;

the invention provides an attack vector based on matrix elementary row transformation and column exchange, which has the main idea that: a linear transformation on a matrix does not change the solution space of the matrix. Namely, the following conditions are satisfied:

wherein [. ]]_aLine simplest form indicating that eventually no change occurs, P denotes a pairAnd performing primary line transformation.

Step six: attack vector a₂＝(Q^T)^-1(Te_n) Wherein e is_nThe last row of n-dimensional column vectors with elements of 1 and the rest of 0 are used for selecting the last row of the matrix U, and the final attack vector is a^*＝a₁+a₂。

3. The PMU pre-deployment of the vulnerable bus comprises the following steps:

the method comprises the following steps: the control center firstly analyzes the Jacobian matrix of the measured values and the topological structure of the power gridThe most vulnerable bus is identified and determined. I.e. for the measured value Jacobian matrix

Searching to obtain the number of non-zero elements in each column vector;

step two: selecting t corresponding state variables corresponding to t minimum counter values, and deploying a PMU device on an adjacent bus corresponding to the t state variable buses;

as shown in FIG. 2, the measured value Jacobian matrix of the IEEE14 bus system is foundIf the number of non-zero elements in the column vector corresponding to the middle bus 8 is the minimum, a PMU device is disposed in advance on the adjacent bus 7, and the PMU device is represented by diagonal filling. This is because the measurements related to the edge bus 8 are rare and it is easy for an attacker to locate this bus from the topology information; and once a PMU is deployed on a specific bus, the voltage phase angle of the bus and all branch power flow measurement values connected with the bus can be directly measured, and the phase angles of other buses adjacent to the bus are protected. Namely, the state variables corresponding to the bus 4, the bus 7, the bus 8 and the bus 9 are also protected. Similarly, as shown in FIG. 3, in the IEEE30 bus system, the Jacobian matrix is measured

If the number of non-zero elements in the column vector corresponding to the middle buses 11, 13 and 26 is the minimum, a PMU device is disposed in advance on each of the adjacent buses 9, 12 and 25, and is indicated by diagonal filling.

Step three: measured value Jacobian matrixThe row vector corresponding to the measurements of the bus where the PMU is deployed and its corresponding adjacent bus is moved to H_SIn and updateAnd H_S。

4. The PMU-based greedy strategy deployment in the whole network comprises the following steps:

the method comprises the following steps: will matrixDivided into (n-t) sub-matricesWherein L is_i∈[1,n-t]，

Representation matrixEach column vector corresponding to the remaining unprotected state variables;

step two: according to the obtained attack vector a^*To obtainSubmatrix corresponding to non-zero elements of attack vector

Whereink∈[1,Ω]Omega is the number of nonzero elements in the attack vector;

step three: initializing a counter count (zeros (1, n-t)) and counting each submatrix

And submatrices corresponding to non-zero elements in the attack vector

The number of the same elements in the solution and the maximum value obtained are recorded as

Where card is an operator for the number of elements in the set. L is_targetA target bus corresponding to the maximum value of the counter is represented; arg denotes the value of the variable at which the objective function is maximized, and zeros denotes that the elements in the defined matrix are all zero.

Step four: in correspondence with L_targetA PMU is deployed on the bus and the Jacobian matrix of the measured values is obtained

The row vector corresponding to the measurements of the bus where the PMU is deployed and its corresponding adjacent bus is moved to H_SPerforming the following steps;

after pre-deployment of PMU devices on edge buses in a power grid, vulnerable buses in the network are first protected. And then, adopting the idea of a greedy strategy, and deploying a PMU device on a bus capable of protecting the maximum measured value in each PMU deployment round. As shown in fig. 2, a PMU is deployed on the bus 12 according to the attack vector generated in the first round after pre-deployment. In fig. 3, a PMU is first deployed on the bus 23 according to the attack vector generated in the first round after pre-deployment.

Step five: updating a matrixH_SAnd attack vector a^*；

Step six: repeating the second to fifth steps until rank (H)_S) N, i.e. matrix H_SAnd if the transmission system is full, the whole power bus transmission system is completely considerable, and the wrong data injection attack cannot happen again.

As shown in fig. 2, PMU processes deployed by a greedy strategy are shown as bus nodes filled with dots and arrows in the figure, and the PMU is deployed in the order of bus 12, bus 10, bus 14, bus 3, and bus 2. In fig. 3, the PMUs deployed by the greedy strategy are in order bus 23, bus 29, bus 17, bus 8, bus 7, bus 3, bus 21, bus 20, bus 18, and bus 2. The attack defense dynamic game of the attacker and the power grid control center is finished when all state variables are completely observable, all the state variables are protected due to the deployment of PMUs, and the smart power grid cannot be attacked again by error data injection attack.