Risk assessment method for vehicle-network interactive power information physical system

Abstract

The invention discloses a risk assessment method for a vehicle-network interactive power information physical system, and the method comprises the steps: designing the risk assessment of the vehicle-network interactive power information physical system under the false data injection attack through the detection of the false data injection attack of the vehicle-network interactive power information physical system based on the attack income, and can achieve the estimation of the current system state. And meanwhile, state prediction at the next moment is beneficial to system regulation and control, and vehicle network interactive power system management is promoted.

Description

Technical Field

[0001] The present invention relates to a technology in the field of information security, specifically a risk assessment method for a vehicle-grid interactive electric information-physical system subjected to false data injection attacks. Background Art

[0002] As a key infrastructure to ensure the normal operation of society, the stability and reliability of the power system are of vital importance. However, with the widespread access of smart devices, a large amount of data is pouring into the power system from various smart device interfaces. Among these data, the existence of false data is like a cancer hidden in the dark, posing a serious threat to the safe operation of the power system. False data exerts its destructive power by interfering with system status data. The control center of the power system relies on accurate system status data to make decisions, which are directly related to the dispatch and distribution of electricity. When false data is mixed in, the authenticity and accuracy of the system status data are destroyed, and the control center will issue erroneous instructions based on these contaminated data. The consequences of these erroneous instructions are disastrous, such as line overload. Line overload will cause the current in the line to exceed its designed carrying capacity, which will cause problems such as line heating and accelerated insulation aging. In severe cases, it may even cause line fires, leading to large-scale power outages. In addition, erroneous instructions may also cause load wear. This chain reaction caused by false data has brought great challenges to the stable operation of the power system. However, the limitations of existing results include the following: the current research on false data injection attacks is focused on the detection part, and it has not been combined with the risk assessment of the power physical information system; secondly, the current research on the detection of false data injection attacks in the power physical information system is focused on the existence, and its location has not been detected.

[0003] In the electric cyber-physical system of vehicle-grid interaction, the threat comes from false data injection attacks. Attackers can inject false data into the weak parts of the control center, communication system and actuators, causing decision-making failures and achieving the purpose of the attack. With the increase in the number of connected devices in the power grid network, the amount of data received in the data acquisition and monitoring control system has increased sharply, which can easily lead to the injection of false information, causing the command and control system to issue erroneous dispatch instructions, which will cause problems such as overload of distribution network lines in the vehicle-grid interaction system and damage to vehicle-grid interaction equipment such as charging piles. Summary of the invention

[0004] In view of the shortcomings of the prior art that false data injection attack detection and risk assessment are not combined and risk assessment of the system cannot be performed when a false data attack is detected, the present invention proposes a risk assessment method for a vehicle-grid interactive electric information-physical system. By detecting false data injection attacks on the vehicle-grid interactive electric information-physical system and designing a risk assessment of the vehicle-grid interactive electric information-physical system being attacked by false data injection based on the attack benefits, the current system state can be estimated, and the prediction of the state at the next moment is beneficial to system regulation and promotes the management of the vehicle-grid interactive electric system.

[0005] The present invention is achieved through the following technical solutions:

[0006] The present invention relates to a risk assessment method for a vehicle-grid interactive electric information-physical system, comprising:

[0007] Step 1: Collect the state parameters of the vehicle-grid interaction system and input them into the constructed system space state model, specifically: z t is the vehicle-grid interactive power cyber-physical system data (system active power, reactive power, etc.) collected by the control center at time t; t System status data (system voltage amplitude, phase angle); f(x t ) is x t At time t, the state transfer function is the change in the system state at the previous moment; h(x t ) is x t At time t, the measurement function is the coupling relationship between the measurement value and the state variable; φ t and v t are the state variables and measurement data errors at time t; the system state variables and the system state covariance K t is the Kalman filter gain at time t; P t is the predicted covariance at time t, and the measured state data changes after being attacked by false data injection State data change residual Attack constraints

[0008] Step 2: According to the false data injection attack detection model, when the system is not attacked, the Euclidean distance similarity is close to 1. When it is attacked by false data injection, its ratio will deviate sharply from 1, thereby achieving the purpose of detecting false data injection attacks. Specifically, the Euclidean distance similarity of the measurement state of the vehicle-grid interactive power information-physical system Among them: Euclidean similarity at the initial moment is the estimated value obtained based on the Kalman filter algorithm at time t. When attacked, its measured value changes to Corresponding to the Euclidean distance similarity at time t Setting detection thresholds based on historical data is to set the threshold; χ1 and χ2 are the lower and upper limits of the ratio under normal circumstances; If , it is attacked by injection of false data. Otherwise, it is normal, that is,

[0009] Step 3: Detect attacked nodes based on the false data injection attack model according to the state of the vehicle-grid interactive power-physical information system.

[0010] Step 4: Calculate the attack gain function of the false data injection attack Decision is made on the cost of false data injection attack on system nodes, where: R k For attack buff, C is the success probability of attacking point k under false injection attack; cnps The impact of false data injection attacks on vehicle-grid interactive power cyber-physical systems, re is the system recovery factor. The greater the system recovery cost is, the greater the impact of the attack is, and the greater the recovery factor is.

[0011] Step 5: Calculate the single-step benefit and success probability of the false data injection attack, including: overall attack benefit Among them: single step attack gain each step gain R k To generate revenue when attacking node k, C k is the vulnerability degree of node k that can be exploited by attackers; G k is the single-step attack benefit from node k to node k+1; the probability of a single node being successfully attacked when it is attacked by false data injection The node defense level when attacked by node k A k Inject attack vector for node k’s data; C k is the attack complexity when node k is attacked by data injection; the repair degree of node k vulnerability H k =1-exp(T k / 0.16) b and vulnerability exposure level E k =1-(0.05 / T k ) a ω1, ω2, ω3, and ω4 are weight coefficients of each parameter, which can be determined according to the situation of injecting false data into the attack node. a is the Pareto distribution parameter; the degree of exploitation of the vulnerability node by the attacker N is the number of system nodes that the attacker participates in the attack, h is the number of nodes that the attacker successfully attacks, and x k is the number of times node k is attacked,

[0012] Step 6: Calculate the risk loss of the vehicle-grid interactive power-physical information system, including: the instantaneous loss of charging station users connected to the distribution lines of the distribution network and the physical consequences of the long-term recovery consequences and the communication delay consequences.

[0013] Step 7: Calculate the attack gain function and evaluate the risk analysis of attack success on the attacked node in the vehicle-grid interactive power-physical cyber system. Technical Effects

[0014] The present invention detects false data injection attacks by estimating the state of the vehicle-grid interactive electric information-physical system and the Euclidean distance based on the Kerman filtering method; and conducts risk assessment of the vehicle-grid interactive electric information-physical system based on the attack gain, the physical loss of the false data injection attack and the communication consequences. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Schematic diagram of the vehicle-grid interactive power cyber-physical system;

[0016] Figure 2 It is a flow chart of the present invention;

[0017] Figure 3 Schematic diagram of the risk assessment module for the vehicle-grid interactive electric cyber-physical system. DETAILED DESCRIPTION

[0018] like Figure 1 As shown, this embodiment relates to a risk assessment method for a vehicle-grid interactive electric cyber-physical system, including:

[0019] Step 1) Detection of false information injection attacks, including:

[0020] 1.1 Establishing the system spatial state model Where: z t is the vehicle-grid interactive power cyber-physical system data (system active power, reactive power, etc.) collected by the control center at time t; t System status data (system voltage amplitude, phase angle); f(x t ) is x t At time t, the state transfer function is the change in the system state at the previous moment; h(x t ) is x t At time t, the measurement function is the coupling relationship between the measurement value and the state variable; φ t and v t are the state variable and measurement data error at time t respectively.

[0021] 1.2 The unscented Kalman filter method is used to estimate the state of the system and obtain its state variables and covariance, which are specifically: Where: K t is the Kalman filter gain at time t; P t is the forecast covariance at time t;

[0022] 1.3 In the complete vehicle-grid interactive electric cyber-physical system, the residual is calculated based on the change of the measured state data after being attacked by false data injection, specifically: in: When the vehicle-grid interactive electric information-physical system is attacked, the actual measurement values ​​obtained will quickly deviate. However, when the Kalman filter algorithm estimates the system state, due to the inherent characteristics of the Kalman filter algorithm, the state estimate obtained will not mutate due to the introduction of the attack. Therefore, in order to identify false data injection attacks, the present invention uses Euclidean distance similarity to describe the similarity between the actual measurement value and the predicted value. At the same time, the ratio at different times is used to describe whether it is under attack, and the appropriate threshold is set according to the system historical data.

[0023] 1.4 The purpose of the false data injection attack is to overload the lines of the vehicle-grid interactive power cyber-physical system, so the attack constraints are: Where: P l and Q l is the active and reactive power of the target line l; ΔP l att and is the increase of active and reactive power when the target line l is attacked; is the maximum apparent power of the target line l.

[0024] Step 2) False data intrusion detection: The Euclidean distance method is used to detect the similarity of adjacent moments. When false data is injected into the system, its actual measurement value will change suddenly, but it will converge to the normal value at the next moment under the influence of the Kalman filter algorithm. Therefore, the historical data under the normal operation of the vehicle-grid interactive electric information-physical system is used as the basis to set the appropriate threshold, including:

[0025] 2.1 Attack Detection Algorithm: When the system is in normal operation and has not been attacked by false data injection, its Euclidean similarity ratio with other reference data will remain at a stable level. However, once the system is invaded by false data injection attack, this ratio will change significantly or mutate. It is based on this characteristic that it can effectively determine whether the system is Suffered from an attack involving the injection of false data.

[0026] The Euclidean distance similarity in: is the distance similarity at time t

[0027] When the system is not attacked, the Euclidean distance similarity is close to 1. When it is attacked by false data injection, its ratio will be It deviates sharply from 1, thus achieving the purpose of detecting false data injection attacks.

[0028] 2.2 Set the detection threshold through historical data, specifically: in: is to set the threshold; χ1 and χ2 are the lower and upper limits of the ratio under normal circumstances; If it is, it is attacked by injection of false data. Otherwise, it is normal.

[0029] Step 3) Risk assessment of the vehicle-grid interactive power cyber-physical system: Based on the false data injection attack detection results introduced in Section 1, the system risk assessment is based on the results of the general information vulnerability scoring system to obtain the risk of being affected by false data. The probability of attack success during an injection attack is used to evaluate the system attack risk by calculating the attack benefit, including:

[0030] 3.1 Attack gain calculation: Where: R k For attack buff, C is the success probability of attacking point k under false injection attack; cnps The impact of false data injection attacks on vehicle-grid interactive power cyber-physical systems, re is the system recovery factor. The greater the system recovery cost is, the greater the impact of the attack is, and the greater the recovery factor is.

[0031] 3.2 Single-step attack gain and attack success probability: In the process of false data injection attack, the success of the attack is related to the attacker’s attack capability on the one hand, and to the defense of the vehicle-grid interactive power information-physical system on the other hand. Therefore, the probability of a single node being successfully attacked when it is attacked by false data injection is as follows: in: is the degree of vulnerability exploited by the attacker when attacking node k; is the node defense level when node k is attacked; Among them: A k Inject attack vector for node k’s data; C k H is the attack complexity when node k is attacked by data injection; k and E kare the degree of repair and exposure of the vulnerability of node k respectively; ω1, ω2, ω3, ω4 are the weight coefficients of each parameter respectively, which can be determined according to the situation of false data injection attack node; H k =1-exp(T k / 0.16) b , where: T k is the vulnerability discovery time of node k, b is the Weibull distribution parameter; E k =1-(0.05 / T k ) a , where a is the Pareto distribution parameter. The attacker's degree of exploitation of the vulnerability node Where: N is the number of system nodes that the attacker participates in the attack, h is the number of nodes that the attacker successfully attacks, and x k is the number of times node k is attacked. Single-step attack gain Each step gain Where: R k To generate revenue when attacking node k, C k is the vulnerability degree of node k that can be exploited by attackers; G k is the single-step attack benefit from node k to node k+1, then the overall attack benefit

[0032] 3.3 System risk analysis under false data injection attack: In the electric information-physical system of vehicle-grid interaction, the impact on the communication system, lines and charging piles when attacked; false data injection causes the system to obtain wrong data, and the control system may make wrong instructions based on the wrong data during operation, causing losses to the system. The impact includes the physical consequences of instantaneous losses to users of charging stations connected to the distribution lines of the distribution network and the long-term consequences of recovery, as well as the consequences of communication delays.

[0033] The physical consequences described include:

[0034] 1. Instantaneous loss: Establish a fault load loss model and a fault user loss model that considers the user's power consumption level. Specifically: Fault load loss in: is the number of load losses after node k is attacked, is the user cluster that includes charging pile users on the attacked node k; is the level factor of user m, L m is the load carried by user m. Number of users lost due to failure Among them: U m is the total number of users in the affected user cluster m.

[0035] 2. Long-term loss, specifically: number of hours of load loss due to fault Where: T(f,θ) is the number of hours the fault lasts, where θ is the location of the fault point. Since the indicators of various physical consequences have different degrees of influence on the power grid, dimensionless processing is performed through the benchmark values ​​of each indicator, and the weights are assigned according to the actual fault conditions, specifically: in: are the benchmark values ​​of each indicator, Among them: σ1, σ2, σ3, σ4 are the weight coefficients of each indicator.

[0036] The communication consequences mentioned above refer to: after being attacked by false data injection, the communication line will also be affected, causing the impact of communication delay, specifically: Where: C TX For communication consequences, DP avg,i is the average communication delay of the ith data packet, T min,delay is the minimum acceptable communication delay; the overall impact of the false data injection attack is C cnps =C Ph +C TX .

[0037] Compared with the prior art, the present invention adopts the Kerman filtering method to estimate the state of the vehicle-grid interactive power information-physical system and the Euclidean distance to detect false data injection attacks; then the system attack position is evaluated by the attack gain; compared with a single risk assessment, the present invention can not only provide global fault warning, but also predict the location where the fault will occur.

[0038] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A risk assessment method for a vehicle-grid interactive electric cyber-physical system, characterized in that: include: Step 1: Collect the state parameters of the vehicle-grid interaction system and input them into the constructed system space state model, specifically: z t is the vehicle-grid interactive electric cyber-physical system data collected by the control center at time t, x t System status data, f(x t ) is x t At time t, the state transfer function is the change in the system state at the previous moment, h(x t ) is x t At time t, the measurement function is the coupling relationship between the measurement value and the state variable, φ t and v t are the state variables and measurement data errors at time t, and the system state variables are and the system state covariance K t is the Kalman filter gain at time t, P t is the predicted covariance at time t, and the measured state data changes after being attacked by false data injection State data change residual Attack constraints Step 2: According to the false data injection attack detection model, when the system is not attacked, the Euclidean distance similarity is close to 1. When it is attacked by false data injection, its ratio will deviate sharply from 1, thereby achieving the purpose of detecting false data injection attacks. Specifically, the Euclidean distance similarity of the measurement state of the vehicle-grid interactive power information-physical system Among them: Euclidean similarity at the initial moment is the estimated value obtained based on the Kalman filter algorithm at time t. When attacked, its measured value changes to Corresponding to the Euclidean distance similarity at time t Setting detection thresholds based on historical data To set the threshold, χ1 and χ2 are the lower and upper limits of the ratio under normal circumstances. , it is attacked by injection of false data, otherwise, it is normal, that is, Step 3: Detect the attacked node according to the false data injection attack model based on the state of the vehicle-grid interactive power-physical information system; Step 4: Calculate the attack gain function of the false data injection attack Decision is made on the cost of false data injection attack on system nodes, where: R k For attack buff, is the success probability of attacking point k under false injection attack, C cnps The impact of false data injection attacks on vehicle-grid interactive power cyber-physical systems, re is the system recovery factor; Step 5: Calculate the single-step benefit and success probability of the false data injection attack, including: overall attack benefit Among them: single step attack gain each step gain R k To generate revenue when attacking node k, C k is the vulnerability level of node k that can be exploited by an attacker, G k is the single-step attack benefit from node k to node k+1, and the probability of a single node being successfully attacked when it is attacked by false data injection The node defense level when attacked by node k A k The data injection attack vector for node k, C k is the attack complexity when node k is attacked by data injection, and the degree of repair of the vulnerability of node k H k =1-exp(T k / 0.16) b and vulnerability exposure level E k =1-(0.05 / T k ) a ω1, ω2, ω3, and ω4 are weight coefficients of each parameter, which can be determined according to the situation of injecting false data into the attack node. a is the Pareto distribution parameter, which indicates the degree of exploitation of the vulnerability node by the attacker. N is the number of system nodes that the attacker participates in the attack, h is the number of nodes that the attacker successfully attacks, and x k is the number of times node k is attacked; Step 6: Calculate the risk loss of the vehicle-grid interactive power-physical information system, including: the physical consequences of the instantaneous loss of the charging station users connected to the distribution lines of the distribution network and the long-term consequences of recovery and the communication delay consequences; Step 7: Calculate the attack gain function and evaluate the risk analysis of attack success on the attacked node in the vehicle-grid interactive power-physical cyber system.

2. The risk assessment method for the vehicle-grid interactive electric information-physical system according to claim 1 is characterized in that: The step 1 is implemented in the following way: 1.1 Establishing the system spatial state model Where: z t is the vehicle-grid interactive electric cyber-physical system data collected by the control center at time t, x t System status data, f(x t ) is x t At time t, the state transfer function is the change in the system state at the previous moment, h(x t ) is x t At time t, the measurement function is the coupling relationship between the measurement value and the state variable, φ t and v t are the state variables and measurement data errors at time t respectively; 1.2 The unscented Kalman filter method is used to estimate the state of the system and obtain its state variables and covariance, which are specifically: Where: K t is the Kalman filter gain at time t, P t is the forecast covariance at time t; 1.3 In the complete vehicle-grid interactive electric cyber-physical system, the residual is calculated based on the change of the measured state data after being attacked by false data injection, specifically: in: The Euclidean distance similarity is used to describe the similarity between the actual measured value and the predicted value. At the same time, the ratio at different times is used to describe whether it is under attack, and the appropriate threshold is set according to the system's historical data; 1.4 The purpose of the false data injection attack is to overload the lines of the vehicle-grid interactive power cyber-physical system, so the attack constraints are: Where: P l and Q l are the active and reactive power of the target line l, and is the increase in active and reactive power when the target line l is attacked, is the maximum apparent power of the target line l.

3. The risk assessment method for the vehicle-grid interactive electric information-physical system according to claim 1 is characterized in that: The threshold in step 2 is obtained by: 2.1 Attack detection algorithm: When the system is in normal operation and has not been attacked by false data injection, its Euclidean similarity ratio with other reference data will remain at a stable level. However, once the system is invaded by false data injection attack, this ratio will change significantly or mutate. It is based on this characteristic that it is effective to judge whether the system has been attacked by false data injection; 2.2 Set the detection threshold through historical data, specifically: in: To set the threshold, χ1 and χ2 are the lower and upper limits of the ratio under normal circumstances. , it is attacked by injection of false data, otherwise, it is normal.

4. The risk assessment method for the vehicle-grid interactive electric information-physical system according to claim 3 is characterized in that: The Euclidean distance similarity in: is the estimated value at time t, which is obtained based on the Kalman filter algorithm, and its Euclidean similarity at the initial time When under attack, its measured value will change significantly: It can be inferred that the Euclidean distance similarity at this time Euclidean distance similarity at time t When the system is not attacked, the Euclidean distance similarity is close to 1. When attacked by false data injection, its ratio will deviate sharply from 1, thus achieving the purpose of detecting false data injection attacks.

5. The risk assessment method for the vehicle-grid interactive electric cyber-physical system according to claim 1 is characterized in that: The attack success probability is obtained by: 3.1 Attack gain calculation: Where: R k For attack buff, is the success probability of attacking point k under false injection attack, C cnps The impact of false data injection attacks on vehicle-grid interactive power cyber-physical systems, re is the system recovery factor; 3.2 Single-step attack gain and attack success probability: In the process of false data injection attack, the success of the attack is related to the attacker’s attack capability on the one hand, and to the defense of the vehicle-grid interactive power information-physical system on the other hand. Therefore, the probability of a single node being successfully attacked when it is attacked by false data injection is as follows: in: is the degree of vulnerability exploited by the attacker when attacking node k, is the node defense level when node k is attacked, Among them: A k The data injection attack vector for node k, C k is the attack complexity when node k is attacked by data injection, H k and E k are the degree of repair and exposure of the vulnerability of node k, ω1, ω2, ω3, and ω4 are the weight coefficients of each parameter, which can be determined according to the situation of false data injection attacking the node. k =1-exp(T k / 0.16) b , where: T k is the time it takes to discover the vulnerability of node k, b is the Weibull distribution parameter, and E k =1-(0.05 / T k ) a , where: a is the Pareto distribution parameter, and the attacker's degree of exploitation of the vulnerability node Where: N is the number of system nodes that the attacker participates in the attack, h is the number of nodes that the attacker successfully attacks, and x k is the number of times node k is attacked, the single-step attack gain is the gain per step Where: R k To generate revenue when attacking node k, C k is the vulnerability level of node k that can be exploited by an attacker, G k is the single-step attack benefit from node k to node k+1, then the overall attack benefit 3.3 System risk analysis under false data injection attack: In the electric information-physical system of vehicle-grid interaction, the attack affects the communication system, lines and charging piles. False data injection causes the system to obtain erroneous data. During operation, the control system may make erroneous instructions based on the erroneous data, causing losses to the system. The impact includes the physical consequences of instantaneous losses to users of charging stations connected to the distribution lines of the distribution network and the long-term consequences of recovery, as well as the consequences of communication delays.

6. The risk assessment method for the vehicle-grid interactive electric cyber-physical system according to claim 1 or 5 is characterized in that: The physical consequences described include: 4.

1. Instantaneous loss: Establish a fault load loss model and a fault user loss model that considers the user's power consumption level. Specifically: Fault load loss in: is the number of load losses after node k is attacked, is the user cluster that includes charging pile users on the attacked node k, is the level factor of user m, L m is the load carried by user m, and the number of users lost due to the failure Among them: U m is the total number of users in the affected user cluster m; 4.

2. Long-term loss, specifically: number of hours of load loss due to fault Where: T(f,θ) is the number of hours the fault lasts, θ is the location of the fault point, and the number of hours the fault causes the loss of users Since the indicators of various physical consequences have different degrees of influence on the power grid, dimensionless processing is performed through the benchmark values ​​of each indicator, and the weights are assigned according to the actual fault conditions, specifically: in: are the benchmark values ​​of each indicator, Among them: σ1, σ2, σ3, σ4 are the weight coefficients of each indicator.

7. The risk assessment method for the vehicle-grid interactive electric cyber-physical system according to claim 1 or 5 is characterized in that: The communication consequences mentioned above refer to: after being attacked by false data injection, the communication line will also be affected, causing the impact of communication delay, specifically: Where: C TX For communication consequences, DP avg,i is the average communication delay of the ith data packet, T min,delay is the minimum acceptable communication delay, and the overall impact of the false data injection attack is C cnps =C Ph +C TX .

Images