A dynamic defense method and system for malicious attacks in a DC microgrid system

Abstract

The present invention discloses a dynamic defense method and system for malicious attacks in a DC microgrid system, comprising the following steps: collecting input and output data from the DC microgrid system to form a historical data set, allowing attackers to invade the DC microgrid's communication network and launch FDIA and DoS attacks; establishing an equivalence relationship between the DC microgrid input and output in the historical data set, generating a residual term based on the equivalence relationship between the output and input, and optimizing the residual term to improve its sensitivity to attack vectors; when an FDIA attack is detected, calculating a mitigation vector to offset the false data vector; finally, adopting a dynamic security data deletion strategy to eliminate excessive historical data; collecting historical current data, assigning a weighting factor to the current in each sampling period, forming a weighted average to estimate the current value blocked by the DoS attack, and optimizing the weighting factor to estimate the current value. The present invention ensures the security of the DC microgrid.

Description

Technical Field

[0001] The present invention belongs to the technical field of DC microgrid system security, and in particular relates to a dynamic defense method and system for malicious attacks in a DC microgrid system. Background Art

[0002] A DC microgrid is a modern cyber-physical system that relies heavily on advanced information and communication technologies for control and communication. However, this reliance on communication technologies increases the risk of cyberattacks.

[0003] FDIA is primarily used to inject false data into DC microgrids, causing unstable deviations in voltage and current proportional sharing. Attackers typically target the link between sensors and controllers, as this area is often a vulnerable point with weak defenses. Denial of Service (DoS) attacks, on the other hand, disrupt information transmission, hindering the timely exchange of critical data between neighboring microgrid sensors or sensors. This interference can prevent controllers from making appropriate adjustments based on the corresponding information, significantly impacting the stability of the DC microgrid. DoS attacks often occur on communication lines between neighbors and can cause cascading failures in the DC microgrid cluster.

[0004] Therefore, it is urgent to solve the above problems. Summary of the Invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a dynamic defense method for malicious attacks in a DC microgrid system. The present invention can simultaneously detect and defend against FDIA and DoS attacks, reduce the computational burden, improve the computational efficiency of the DC microgrid, and ensure the security of the DC microgrid.

[0006] A second object of the present invention is to provide a dynamic defense system against malicious attacks in a DC microgrid system.

[0007] Technical solution: To achieve the above objectives, the present invention discloses a dynamic defense method for malicious attacks in a DC microgrid system, comprising the following steps:

[0008] (1) Collect input data and output data in the DC microgrid system to form a historical data set. The input data includes voltage, current data and disturbance items, and the output data is the control instruction.

[0009] (2) The attacker invades the communication network of the DC microgrid and launches FDIA and DoS attacks;

[0010] (3) Establish an equivalence relationship between the input and output of the DC microgrid in the historical data set, generate a residual term based on the equivalence relationship between the output and input, and improve the sensitivity of the residual term to the attack vector through optimization to ensure that the detected attack is not affected by the disturbance term; when an FDIA attack is detected, calculate the mitigation vector to offset the false data vector; finally, adopt a dynamic security data deletion strategy to eliminate excessive historical data;

[0011] (4) Collect historical current data, assign weight factors to the current in each sampling period, form a weighted average to estimate the current value blocked by the DoS attack, and optimize the weight factors to estimate the current value.

[0012] Optionally, the false data vector injected by FDIA in step (2) is expressed as:

[0013] V FDI (k)=V(k)+F(k)

[0014] Among them, V FDI (k) is the voltage at the kth sampling moment containing the false vector, V(k) is the actual voltage value, and F(k) is the false data vector of FDIA.

[0015] Optionally, the DoS attack set in step (2) is defined as:

[0016]

[0017] Among them, Φ1 represents the sampling time of DoS attack; Φ2 represents the normal sampling time, Φ = Φ1 ∪ Φ2 represents all sampling times of the microgrid; k d1 ,k d2 ,…,k dn is the discrete attack moment, k n1 ,k n2 ,...,k nn are discrete non-attack moments.

[0018] Optionally, the upper bound of the false data vector of FDIA in step (2) is The duration of the DoS attack is less than the sampling time of the DC microgrid, denoted as T D <T, the number of DoS attacks meets the condition: |n(k)|≤α+T D / τ D , where |n(k)| represents the number of DoS attacks within the sampling time, α≥0, τ D >0 is a constant.

[0019] Optionally, step (3) specifically includes the following steps:

[0020] Based on the equivalent relationship between the input and output of the DC microgrid system at m sampling moments in the historical data, FDIA is detected and prevented.

[0021] The equivalence relationship of the DC microgrid system in the past m steps can be written in the form of a vector matrix, expressed as:

[0022] Y(m)=Nx(0)+PU(m)+RD(m)+SX FDI (m)

[0023] Among them, Y(m) is the output of the past m sampling moments, U(m), D(m), X FDI (m) is the control input, noise input and false vector input of the past m sampling moments, x(0) is the state of the microgrid at the 0th sampling moment, N, P, R, S are the coefficient matrices of the microgrid;

[0024] The residual used to detect FDIA is defined as:

[0025] r(m)=r T (Y(m)-HU(m))

[0026] r(m)=r T (Nx(0)+(PH)U(m)+RD(m)+SX FDI (m))

[0027] r(m)=r T (VQ(m)+SX FDI (m))

[0028] Where r is the residual matrix; H is the Jacobian matrix of the DC microgrid, V = [N, (PH), R], Q(m) = [x(0), U(m), D(m)]; in order to make the residual term only sensitive to the attack vector, the residual matrix should be designed to be insensitive to noise or interference in the DC microgrid, so the optimization problem is expressed as:

[0029] min|r T VQ(m)|

[0030] Among them, there are two constraints in the optimization problem, which are expressed as:

[0031] L-δL≤L i ≤L+δL

[0032] C-δC≤C i ≤C+δC

[0033] Where L and C are the nominal values ​​of the inductor and capacitor of the DC buck converter, respectively, and δ is the percentage deviation, which is set to ±5%.

[0034] Based on the residual term r in the absence of attack T VQ(m) establishes the residual threshold, first collecting the past N groups of residual terms r r ={r r1 ,r r2 ,...,r rN}, the residual threshold is calculated by the average value and deviation term of the historical state, expressed as:

[0035]

[0036] σ r is the standard deviation of the residual terms of the past N groups, ζ σ is the adjustment parameter of the confidence interval;

[0037] When the residual value exceeds the threshold value r set by the DC microgrid A At t, the mitigation vector is calculated to offset the false data vector. The mitigation vector is expressed as:

[0038] Z(m)=S -1 (Nx(0)+PU(m)+RD(m)-Y(m))

[0039] Among them, Z(m)=[z(0),z(1),z(2),...,z(m)] T is the mitigation vector, and by injecting the mitigation vector into the DC microgrid, the output state y i (k) is expressed as:

[0040] y i (k) = C i x i (k)+C ai [x FDIi (k)+z i (k)]

[0041] C i is the measurement matrix of the DC microgrid system, x i (k) is the state vector of the DC microgrid system at the kth moment, C ai is the measurement matrix of the DC microgrid system under attack, x FDIi (k) is the state vector of the DC microgrid system at the kth moment under the attack condition, z i (k) is the relief vector of the DC microgrid system at the kth moment;

[0042] By setting an upper limit d on the residual test data capacity, if m>d, the residual items that do not exceed the residual threshold in the past m steps will be defined as safe data, and the safe data will not be used for the calculation of the residual test; define the safe data set s = [1,2,...,s], and the residual r(c) after removing the safe data is expressed as:

[0043] r(c)=r T [VQ(c)+SX FDI (c)],c≤d

[0044] where r T is the transposed matrix of the residual matrix, Q(c) is Q(m) after removing the security data, X FDI (c) is the false vector input after removing the security data.

[0045] Optionally, step (4) specifically includes the following steps:

[0046] Assuming that the DC microgrid suffers a DoS attack at step k, the current history data of the past km sampling cycles is first selected, and then a weight is assigned to the current of each cycle to form a weighted average estimated current;

[0047] The estimated current of the i-th DC microgrid in the k-th sampling period is expressed as:

[0048]

[0049] Where ω is the weight factor, satisfying ω k-m +ω k-m+1 +...+ω k-1 =1;

[0050] Based on the estimated current value, the estimated current proportional deviation is expressed as:

[0051]

[0052] Among them, I Ci is the parameter allocated according to the rated power of the i-th DC microgrid, I Cj is the parameter allocated according to the rated power of the jth DC microgrid, is the estimated current of the i-th DC microgrid in the k-th sampling period, is the estimated current of the jth DC microgrid in the kth sampling period; a ij is the weight coefficient between microgrids; under DoS attack, the estimated control output of the i-th DC microgrid is Expressed as:

[0053]

[0054] V i (k) is the actual control output of the i-th DC microgrid, and a is the control parameter;

[0055] The weight factor is optimized to minimize the error between the estimated voltage value and the DC microgrid reference voltage value; the optimization problem is expressed as:

[0056]

[0057] in, is the reference voltage value, and the constraint condition is 0<ω k-m ,ω k-m+1 ,...,ω k-1 <1.

[0058] Based on the same inventive concept, the present invention discloses a dynamic defense system for malicious attacks in a DC microgrid system, comprising:

[0059] A data collection unit is used to collect input data and output data in the DC microgrid system to form a historical data set. The input data includes voltage, current data and disturbance items, and the output data is control instructions.

[0060] An attack model unit is introduced to allow attackers to invade the communication network of the DC microgrid and launch FDIA and DoS attacks;

[0061] The FDIA defense unit establishes an equivalence relationship between the DC microgrid input and output in the historical data set. Based on the equivalence relationship between the output and input, a residual term is generated. The residual term is optimized to improve its sensitivity to attack vectors, ensuring that the detected attack is not affected by the disturbance term. When an FDIA attack is detected, a mitigation vector is calculated to offset the false data vector. Finally, a dynamic secure data deletion strategy is adopted to eliminate excessive historical data.

[0062] The DoS attack defense unit collects historical current data, assigns a weight factor to the current in each sampling period, forms a weighted average to estimate the current value blocked by the DoS attack, and optimizes the weight factor to estimate the current value.

[0063] Optionally, the false data vector injected by the FDIA in the data collection unit is expressed as:

[0064] V FDI (k)=V(k)+F(k)

[0065] Among them, V FDI (k) is the voltage at the kth sampling moment containing the false vector, V(k) is the actual voltage value, and F(k) is the false data vector of FDIA;

[0066] The DoS attack set is defined as:

[0067]

[0068] Among them, Φ1 represents the sampling time of DoS attack; Φ2 represents the normal sampling time, Φ = Φ1 ∪ Φ2 represents all sampling times of the microgrid; k d1 ,k d2,...,k dn is the discrete attack moment, k n1 ,k n2 ,...,k nn for discrete non-attack moments;

[0069] The upper bound of the false data vector of FDIA is The duration of the DoS attack is less than the sampling time of the DC microgrid, denoted as T D <T, the number of DoS attacks meets the condition: |n(k)|≤α+T D / τ D , where |n(k)| represents the number of DoS attacks within the sampling time, α≥0, τ D >0 is a constant.

[0070] Optionally, the FDIA prevention unit detects and prevents FDIA based on the equivalent relationship between the input and output of the DC microgrid system at m sampling moments in historical data.

[0071] The equivalence relationship of the DC microgrid system in the past m steps can be written in the form of a vector matrix, expressed as:

[0072] Y(m)=Nx(0)+PU(m)+RD(m)+SX FDI (m)

[0073] Among them, Y(m) is the output of the past m sampling moments, U(m), D(m), X FDI (m) is the control input, noise input and false vector input of the past m sampling moments, x(0) is the state of the microgrid at the 0th sampling moment, N, P, R, S are the coefficient matrices of the microgrid;

[0074] The residual used to detect FDIA is defined as:

[0075] r(m)=r T (Y(m)-HU(m))

[0076] r(m)=r T (Nx(0)+(PH)U(m)+RD(m)+SX FDI (m))

[0077] r(m)=r T (VQ(m)+SX FDI (m))

[0078] Where r is the residual matrix; H is the Jacobian matrix of the DC microgrid, V = [N, (PH), R], Q(m) = [x(0), U(m), D(m)]; in order to make the residual term only sensitive to the attack vector, the residual matrix should be designed to be insensitive to noise or interference in the DC microgrid, so the optimization problem is expressed as:

[0079] min|r T VQ(m)|

[0080] Among them, there are two constraints in the optimization problem, which are expressed as:

[0081] L-δL≤L i ≤L+δL

[0082] C-δC≤C i ≤C+δC

[0083] Where L and C are the nominal values ​​of the inductor and capacitor of the DC buck converter, respectively, and δ is the percentage deviation, which is set to ±5%.

[0084] Based on the residual term r in the absence of attack T VQ(m) establishes the residual threshold, first collecting the past N groups of residual terms r r ={r r1 ,r r2 ,...,r rN}, the residual threshold is calculated by the average value and deviation term of the historical state, expressed as:

[0085]

[0086] σ r is the standard deviation of the residual terms of the past N groups, ζ σ is the adjustment parameter of the confidence interval;

[0087] When the residual value exceeds the threshold value r set by the DC microgrid A At t, the mitigation vector is calculated to offset the false data vector. The mitigation vector is expressed as:

[0088] Z(m)=S -1 (Nx(0)+PU(m)+RD(m)-Y(m))

[0089] Among them, Z(m)=[z(0),z(1),z(2),...,z(m)] T is the mitigation vector, and by injecting the mitigation vector into the DC microgrid, the output state y i (k) is expressed as:

[0090] y i (k) = C i x i(k)+C ai [x FDIi (k)+z i (k)]

[0091] C i is the measurement matrix of the DC microgrid system, x i (k) is the state vector of the DC microgrid system at the kth moment, C ai is the measurement matrix of the DC microgrid system under attack, x FDIi (k) is the state vector of the DC microgrid system at the kth moment under the attack condition, z i (k) is the relief vector of the DC microgrid system at the kth moment;

[0092] By setting an upper limit d on the residual test data capacity, if m>d, the residual items that do not exceed the residual threshold in the past m steps will be defined as safe data, and the safe data will not be used for the calculation of the residual test; define the safe data set s = [1,2,...,s], and the residual r(c) after removing the safe data is expressed as:

[0093] r(c)=r T [VQ(c)+SX FDI (c)],c≤d

[0094] where r T is the transposed matrix of the residual matrix, Q(c) is Q(m) after removing the security data, X FDI (c) is the false vector input after removing the security data.

[0095] Optionally, the DoS attack defense unit assumes that the DC microgrid suffers a DoS attack in the kth step, firstly selects the current history data of the past km sampling cycles, and then assigns a weight to the current of each cycle to form a weighted average estimated current;

[0096] The estimated current of the i-th DC microgrid in the k-th sampling period is expressed as:

[0097]

[0098] Where ω is the weight factor, satisfying ω k-m +ω k-m+1 +...+ω k-1 =1;

[0099] Based on the estimated current value, the estimated current proportional deviation is expressed as:

[0100]

[0101] Among them, I Ciis the parameter allocated according to the rated power of the i-th DC microgrid, I Cj is the parameter allocated according to the rated power of the jth DC microgrid, is the estimated current of the i-th DC microgrid in the k-th sampling period, is the estimated current of the jth DC microgrid in the kth sampling period; a ij is the weight coefficient between microgrids; under DoS attack, the estimated control output of the i-th DC microgrid is Expressed as:

[0102]

[0103] V i (k) is the actual control output of the i-th DC microgrid, and a is the control parameter;

[0104] The weight factor is optimized to minimize the error between the estimated voltage value and the DC microgrid reference voltage value; the optimization problem is expressed as:

[0105]

[0106] in, is the reference voltage value, and the constraint condition is 0<ω k-m ,ω k-m+1 ,...,ω k-1 <1.

[0107] Beneficial effects: Compared with the existing technology, the present invention has the following significant advantages: the present invention significantly improves the stability and anti-attack capability of the DC microgrid by effectively defending against FDIA and DoS attacks, ensuring that the system can still operate normally under malicious attacks; the present invention adopts a strategy of dynamically removing security data, which not only reduces the computational burden but also improves the computational efficiency of the system; the weighted average estimation method ensures that the controller can still obtain accurate current estimation values ​​in the DoS attack scenario, thereby improving control accuracy; the present invention provides higher reliability for the DC microgrid, optimizes the utilization of renewable energy, thereby bringing economic and environmental benefits, and comprehensively guarantees the safe, stable and efficient operation of the microgrid. BRIEF DESCRIPTION OF THE DRAWINGS

[0108] Figure 1 Schematic diagram of the method flow of the present invention;

[0109] Figure 2 Schematic diagram of the DC microgrid model of the present invention;

[0110] Figure 3 This is a schematic diagram of the FDI malicious attack of the present invention;

[0111] Figure 4This is a schematic diagram of a DOS malicious attack of the present invention;

[0112] Figure 5 A schematic diagram of voltage changes under the defense method of the present invention;

[0113] Figure 6 Schematic diagram of current changes under the defense method of the present invention. DETAILED DESCRIPTION

[0114] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0115] Example 1, as Figure 1 and Figure 2 As shown, the present invention discloses a dynamic defense method for malicious attacks in a DC microgrid system, comprising the following steps:

[0116] (1) Collect input data and output data in the DC microgrid system to form a historical data set. The input data includes voltage, current data and disturbance items, and the output data is the control instruction.

[0117] (2) The attacker invades the communication network of the DC microgrid and launches FDIA and DoS attacks;

[0118] The false data vector injected by FDIA is expressed as:

[0119] V FDI (k)=V(k)+F(k)

[0120] Among them, V FDI (k) is the voltage at the kth sampling moment containing the false vector, V(k) is the actual voltage value, and F(k) is the false data vector of FDIA;

[0121] The control instructions for the next sampling moment of the DC microgrid require the data information from the previous sampling moment, which is transmitted by the sensor on the communication network. Under a DoS attack, the communication channel during sampling is blocked, resulting in the inability to transmit the information collected by the sensor about the current ratio sharing. The DoS attack set is defined as:

[0122]

[0123] Among them, Φ1 represents the sampling time of DoS attack; Φ2 represents the normal sampling time, Φ = Φ1 ∪ Φ2 represents all sampling times of the microgrid; k d1 ,k d2 ,...,k dn is the discrete attack moment, k n1 ,k n2 ,...,k nn for discrete non-attack moments;

[0124] Considering the limited resources of the attacker in the actual situation, the upper bound of the false data vector of FDIA is The duration of the DoS attack is less than the sampling time of the DC microgrid, denoted as T D <T, the number of DoS attacks meets the condition: |n(k)|≤α+T D / τ D , where |n(k)| represents the number of DoS attacks within the sampling time, α≥0, τ D >0 is a constant;

[0125] (3) Defending against FDIA: Establishing the equivalence relationship between the input and output of the DC microgrid in the historical data set, generating a residual term based on the equivalence relationship between the output and input, and improving the sensitivity of the residual term to the attack vector through optimization to ensure that the detected attack is not affected by the disturbance term; when FDIA is detected, calculating the mitigation vector to offset the false data vector and eliminate the negative impact of FDIA; finally, adopting a dynamic secure data deletion strategy to eliminate excessive historical data that may burden the computing and storage capabilities of the microgrid;

[0126] Based on the equivalent relationship between the DC microgrid input and output at m sampling moments in historical data, FDIA is detected and prevented.

[0127] The equivalence relation is expressed as:

[0128] k=0:y(0)=cx(0)+cx FDI (0)

[0129] k=1:y(1)=C[Ax(0)+Bu(0)+Md(0)]+Cx FDI (1)

[0130] k=2:y(2)=C[Ax(1)+Bu(1)+Md(1)]+Cx FDI (2)

[0131] =CA 2 x(0)+CABu(0)+CAMd(0)

[0132] +CBu(1)+CMd(1)+Cx FDI (2)

[0133] …

[0134]

[0135] The equivalence relation of the system in the past m steps can be written in the form of a vector matrix, expressed as:

[0136] Y(m)=Nx(0)+PU(m)+RD(m)+SXFDI (m)

[0137] Among them, Y(m) is the output of the past m sampling moments, U(m), D(m), X FDI (m) is the control input, noise input and false vector input of the past m sampling moments, x(0) is the state of the microgrid at the 0th sampling moment, N, P, R, S are the coefficient matrices of the microgrid;

[0138] The residual used to detect FDIA can be defined as:

[0139] r(m)=r T (Y(m)-HU(m))

[0140] r(m)=r T (Nx(0)+(PH)U(m)+RD(m)+SX FDI (m))

[0141] r(m)=r T (VQ(m)+SX FDI (m))

[0142] Where r is the residual matrix; H is the Jacobian matrix of the DC microgrid, V = [N, (PH), R], Q(m) = [x(0), U(m), D(m)]; in order to make the residual term only sensitive to the attack vector, the residual matrix should be designed to be insensitive to noise or interference in the DC microgrid, so the optimization problem can be expressed as:

[0143] min|r T VQ(m)|

[0144] Among them, there are two constraints in the optimization problem, which are expressed as:

[0145] L-δL≤L i ≤L+δL

[0146] C-δC≤C i ≤C+δC

[0147] Where L and C are the nominal values ​​of the inductor and capacitor of the DC buck converter, respectively, and δ is the percentage deviation. In a DC microgrid, the parameters of the LC filter in the DC buck converter may vary within a small range. Considering the impact of unknown variations on the residual term, the deviation value δ is set to ±5%. By taking different values ​​within this parameter range, the residual matrix that minimizes the unknown disturbance is obtained to achieve the optimization goal.

[0148] In order to more accurately detect FDIA in DC microgrids, the residual term r is used in the absence of attack. T VQ(m) establishes the residual threshold; first, collect the past N groups of residual terms rr ={r r1 ,r r2 ,...,r rN}, the residual threshold can be calculated by the average value and deviation term of the historical state, expressed as:

[0149]

[0150] σ r is the standard deviation of the residual terms of the past N groups, ζ σ is the adjustment parameter of the confidence interval;

[0151] When the residual value exceeds the threshold value r set by the DC microgrid A At t, the mitigation vector is calculated to offset the false data vector. The mitigation vector is expressed as:

[0152] Z(m)=S -1 (Nx(0)+PU(m)+RD(m)-Y(m))

[0153] Among them, Z(m)=[z(0),z(1),z(2),...,z(m)] T is the mitigation vector, and by injecting the mitigation vector into the DC microgrid, the output state y i (k) is expressed as:

[0154] y i (k) = C i x i (k)+C ai [x FDIi (k)+z i (k)]

[0155] C i is the measurement matrix of the DC microgrid system, x i (k) is the state vector of the DC microgrid system at the kth moment, C ai is the measurement matrix of the DC microgrid system under attack, x FDIi (k) is the state vector of the DC microgrid system at the kth moment under the attack condition, z i (k) is the relief vector of the DC microgrid system at the kth moment;

[0156] Therefore, the negative impact of false data is eliminated, and mitigation and resistance to FDIA are achieved through residual testing and mitigation vectors;

[0157] In addition, considering the storage capacity and computational burden of DC microgrids, a dynamic safety data deletion strategy is proposed. By setting an upper limit d on the residual test data capacity, if m>d, the residual items that do not exceed the residual threshold in the past m steps will be defined as safety data, and the safety data will not be used for the calculation of the residual test. The dynamic data deletion method not only reduces the storage and computational burden of DC microgrids, but also can accurately detect FDIA. The safety data set s=[1,2,...,s] is defined, and the residual r(c) after removing the safety data is expressed as:

[0158] r(c)=r T [VQ(c)+SX FDI (c)],c≤d

[0159] r T is the transposed matrix of the residual matrix, Q(c) is Q(m) after removing the security data, X FDI (c) is the false vector input after removing the security data;

[0160] (4) Defense against DoS attacks: Collect historical current data, assign weight factors to the current in each sampling period, form a weighted average to estimate the current value blocked by the DoS attack, and optimize the weight factors to make the estimated current value more accurate.

[0161] Assume that the DC microgrid suffers a DoS attack at step k. The DoS attack mainly targets the proportional current sharing channel in the DC microgrid. The attacker consumes the resources of the target DC microgrid by sending a large number of invalid or malicious requests, resulting in the blocking of the proportional current sharing channel. As a result, the controller cannot receive proportional current sharing information from the local and neighboring DC microgrids.

[0162] First, the current history data of the past km sampling cycles are selected, and then a weight is assigned to the current of each cycle to form a weighted average estimated current;

[0163] The estimated current of the i-th DC microgrid in the k-th sampling period can be expressed as:

[0164]

[0165] Where ω is the weight factor, satisfying ω k-m +ω k-m+1 +...+ω k-1 =1;

[0166] Based on the estimated current value, the estimated current proportional deviation can be expressed as:

[0167]

[0168] Among them, I Ciis the parameter allocated according to the rated power of the i-th DC microgrid, I Cj is the parameter allocated according to the rated power of the jth DC microgrid, is the estimated current of the i-th DC microgrid in the k-th sampling period, is the estimated current of the jth DC microgrid in the kth sampling period; a ij is the weight coefficient between microgrids; under DoS attack, the estimated control output of the i-th DC microgrid is It can be expressed as:

[0169]

[0170] V i (k) is the actual control output of the i-th DC microgrid, and a is the control parameter;

[0171] In order to make the estimated control output more accurate, the weight factor needs to be optimized to minimize the error between the estimated voltage value and the DC microgrid reference voltage value; this optimization problem can be expressed as:

[0172]

[0173] in, is the reference voltage value, and the constraint condition is 0<ω k-m ,ω k-m+1 ,...,ω k-1 <1.

[0174] In order to verify the effectiveness of the method of the present invention, a simulation experiment was carried out; Figure 2 As shown in FIG, four interconnected DC microgrids were used for testing, and the system parameters included nominal voltage (volts), buck converter resistance (ohms), rated current (amperes), line resistance (ohms), line inductance (henry), impedance load (ohms), and current load (amperes). The value of nominal voltage (volts) was 6, the value of buck converter resistance (ohms) was 50, the value of line resistance (ohms) was 3, the value of line resistance (ohms) was 0.02, the value of line inductance (henry) was 0.02, the value of impedance load (ohms) was 1, and the value of current load (amperes) was 0.1.

[0175] The initiation and duration of malicious attacks in the experiment, e.g. Figure 3 and Figure 4 As shown in the figure, the defense effect is as follows Figure 5 and Figure 6As shown in the figure, the method of the present invention has a good defense effect in the presence of FDIA and DoS attacks, and the voltage and current values ​​of the DC microgrid are restored to the nominal state. In summary, the present invention can effectively defend against malicious attacks, significantly improve the stability and anti-attack capability of the DC microgrid, ensure that the system can still operate normally under malicious attacks, and can be used in practical engineering applications.

[0176] Example 2

[0177] Based on the same inventive concept, the present invention discloses a dynamic defense system for malicious attacks in a DC microgrid system, comprising:

[0178] A data collection unit is used to collect input data and output data in the DC microgrid system to form a historical data set. The input data includes voltage, current data and disturbance items, and the output data is control instructions.

[0179] An attack model unit is introduced to allow attackers to invade the communication network of the DC microgrid and launch FDIA and DoS attacks;

[0180] The false data vector injected by FDIA in the data collection unit is expressed as:

[0181] V FDI (k)=V(k)+F(k)

[0182] Among them, V FDI (k) is the voltage at the kth sampling moment containing the false vector, V(k) is the actual voltage value, and F(k) is the false data vector of FDIA;

[0183] The DoS attack set is defined as:

[0184]

[0185] Among them, Φ1 represents the sampling time of DoS attack; Φ2 represents the normal sampling time, Φ = Φ1 ∪ Φ2 represents all sampling times of the microgrid; k d1 ,k d2 ,...,k dn is the discrete attack moment, k n1 ,k n2 ,...,k nn for discrete non-attack moments;

[0186] The upper bound of the false data vector of FDIA is The duration of the DoS attack is less than the sampling time of the DC microgrid, denoted as T D <T, the number of DoS attacks meets the condition: |n(k)|≤α+T D / τ D, where |n(k)| represents the number of DoS attacks within the sampling time, α≥0, τ D >0 is a constant.

[0187] The FDIA defense unit establishes an equivalence relationship between the DC microgrid input and output in the historical data set. Based on the equivalence relationship between the output and input, a residual term is generated. The residual term is optimized to improve its sensitivity to attack vectors, ensuring that the detected attack is not affected by the disturbance term. When an FDIA attack is detected, a mitigation vector is calculated to offset the false data vector. Finally, a dynamic secure data deletion strategy is adopted to eliminate excessive historical data.

[0188] In the FDIA defense unit, the equivalent relationship between the input and output of the DC microgrid system at m sampling moments in the historical data is used to detect and defend against FDIA.

[0189] The equivalence relationship of the DC microgrid system in the past m steps can be written in the form of a vector matrix, expressed as:

[0190] Y(m)=Nx(0)+PU(m)+RD(m)+SX FDI (m)

[0191] Among them, Y(m) is the output of the past m sampling moments, U(m), D(m), X FDI (m) is the control input, noise input and false vector input of the past m sampling moments, x(0) is the state of the microgrid at the 0th sampling moment, N, P, R, S are the coefficient matrices of the microgrid;

[0192] The residual used to detect FDIA is defined as:

[0193] r(m)=r T (Y(m)-HU(m))

[0194] r(m)=r T (Nx(0)+(PH)U(m)+RD(m)+SX FDI (m))

[0195] r(m)=r T (VQ(m)+SX FDI (m))

[0196] Where r is the residual matrix; H is the Jacobian matrix of the DC microgrid, V = [N, (PH), R], Q(m) = [x(0), U(m), D(m)]; in order to make the residual term only sensitive to the attack vector, the residual matrix should be designed to be insensitive to noise or interference in the DC microgrid, so the optimization problem is expressed as:

[0197] min|r T VQ(m)

[0198] Among them, there are two constraints in the optimization problem, which are expressed as:

[0199] L-δL≤L i ≤L+δL

[0200] C-δC≤C i ≤C+δC

[0201] Where L and C are the nominal values ​​of the inductor and capacitor of the DC buck converter, respectively, and δ is the percentage deviation, which is set to ±5%.

[0202] Based on the residual term r in the absence of attack T VQ(m) establishes the residual threshold, first collecting the past N groups of residual terms r r ={r r1 ,r r2 ,...,r rN}, the residual threshold is calculated by the average value and deviation term of the historical state, expressed as:

[0203]

[0204] σ r is the standard deviation of the residual terms of the past N groups, ζ σ is the adjustment parameter of the confidence interval;

[0205] When the residual value exceeds the threshold value r set by the DC microgrid A At t, the mitigation vector is calculated to offset the false data vector. The mitigation vector is expressed as:

[0206] Z(m)=S -1 (Nx(0)+PU(m)+RD(m)-Y(m))

[0207] Among them, Z(m)=[z(0),z(1),z(2),...,z(m)] T is the mitigation vector, and by injecting the mitigation vector into the DC microgrid, the output state y i (k) is expressed as:

[0208] y i (k) = C i x i (k)+C ai [x FDIi (k)+z i (k)]

[0209] C i is the measurement matrix of the DC microgrid system, x i (k) is the state vector of the DC microgrid system at the kth moment, C aiis the measurement matrix of the DC microgrid system under attack, x FDIi (k) is the state vector of the DC microgrid system at the kth moment under the attack condition, z i (k) is the relief vector of the DC microgrid system at the kth moment;

[0210] By setting an upper limit d on the residual test data capacity, if m>d, the residual items that do not exceed the residual threshold in the past m steps will be defined as safe data, and the safe data will not be used for the calculation of the residual test; define the safe data set s = [1,2,...,s], and the residual r(c) after removing the safe data is expressed as:

[0211] r(c)=r T [VQ(c)+SX FDI (c)],c≤d

[0212] where r T is the transposed matrix of the residual matrix, Q(c) is Q(m) after removing the security data, X FDI (c) is the false vector input after removing the security data.

[0213] The DoS attack defense unit collects historical current data, assigns a weight factor to the current in each sampling period, forms a weighted average to estimate the current value blocked by the DoS attack, and optimizes the weight factor to estimate the current value.

[0214] In the DoS attack defense unit, it is assumed that the DC microgrid suffers a DoS attack at the kth step. First, the current historical data of the past km sampling cycles are selected, and then a weight is assigned to the current of each cycle to form a weighted average estimated current;

[0215] The estimated current of the i-th DC microgrid in the k-th sampling period is expressed as:

[0216]

[0217] Where ω is the weight factor, satisfying ω k-m +ω k-m+1 +...+ω k-1 =1;

[0218] Based on the estimated current value, the estimated current proportional deviation is expressed as:

[0219]

[0220] Among them, I Ci is the parameter allocated according to the rated power of the i-th DC microgrid, I Cj is the parameter allocated according to the rated power of the jth DC microgrid, is the estimated current of the i-th DC microgrid in the k-th sampling period, is the estimated current of the jth DC microgrid in the kth sampling period; a ij is the weight coefficient between microgrids; under DoS attack, the estimated control output of the i-th DC microgrid is Expressed as:

[0221]

[0222] V i (k) is the actual control output of the i-th DC microgrid, and a is the control parameter;

[0223] The weight factor is optimized to minimize the error between the estimated voltage value and the DC microgrid reference voltage value; the optimization problem is expressed as:

[0224]

[0225] in, is the reference voltage value, and the constraint condition is 0<ω k-m ,ω k-m+1 ,...,ω k-1 <1.

Claims

1. A dynamic defense method for malicious attacks in a DC microgrid system, characterized in that: The steps include: (1) Collect input data and output data in the DC microgrid system to form a historical data set. The input data includes voltage, current data and disturbance items, and the output data is the control instruction. (2) The attacker invades the communication network of the DC microgrid and launches FDIA and DoS attacks; (3) Establish an equivalence relationship between the input and output of the DC microgrid in the historical data set, generate a residual term based on the equivalence relationship between the output and input, and improve the sensitivity of the residual term to the attack vector by designing an optimization problem; when an FDIA attack is detected, calculate the mitigation vector to offset the false data vector; finally, adopt a dynamic security data deletion strategy to eliminate excessive historical data; (4) Collect historical current data, assign weight factors to the current in each sampling period, estimate the current value blocked by the DoS attack through the weighted average estimation method, optimize the weight factors, and estimate the current value.

2. The method for dynamic defense against malicious attacks in a DC microgrid system according to claim 1, characterized in that: The false data vector injected by FDIA is expressed as: , in, is the voltage at the kth sampling moment containing the false vector, is the actual voltage value, is the false data vector of FDIA.

3. The method for dynamic defense against malicious attacks in a DC microgrid system according to claim 1, characterized in that: The DoS attack set is defined as: ， in, Represents the sampling time of DoS attack; represents the normal sampling time, represents all sampling times of the microgrid; is the discrete attack moment, are discrete non-attack moments.

4. The method for dynamic defense against malicious attacks in a DC microgrid system according to claim 1, characterized in that: The upper bound of the false data vector of FDIA is , the duration of the DoS attack is less than the sampling time of the DC microgrid, which is expressed as , the number of DoS attacks meets the following conditions: ,in, Represents the number of DoS attacks during the sampling time, , is a constant.

5. The method for dynamic defense against malicious attacks in a DC microgrid system according to claim 1, characterized in that: The step (3) specifically includes the following steps: Based on the equivalent relationship between the input and output of the DC microgrid system at m sampling moments in the historical data, FDIA is detected and prevented. The equivalence relationship of the DC microgrid system history m steps can be written in the form of a vector matrix, expressed as: ， in, is the output of the historical m sampling moments, is the control input, noise input and false vector input at the historical m sampling moments, is the microgrid state at the 0th sampling moment, is the coefficient matrix of the microgrid; The residual used to detect FDIA is defined as: ， ， ， in, is the residual matrix; is the Jacobian matrix of the DC microgrid, , ; In order to make the residual term only sensitive to the attack vector, the residual matrix should be designed to be insensitive to noise or interference in the DC microgrid, so the optimization problem is based on the residual term in the absence of attack Minimize, expressed as: ， Among them, there are two constraints in the optimization problem, which are expressed as: ， in, and are the nominal values ​​of the inductor and capacitor of the DC step-down converter, is the percentage deviation, the deviation value Set to ±5%; Based on the residual term in the absence of attack To establish the residual threshold, first collect N groups of historical residual items , the residual threshold is calculated by the average value and deviation term of the historical state, expressed as: ， is the standard deviation of the residual term of the historical N groups, is the adjustment parameter of the confidence interval; When the residual value exceeds the threshold set by the DC microgrid When , the mitigation vector is calculated to offset the false data vector, and the mitigation vector is expressed as: ， in, is the mitigation vector, and by injecting the mitigation vector into the DC microgrid, the output state Expressed as: ， is the measurement matrix of the DC microgrid system, is the state vector of the DC microgrid system at the kth moment, is the measurement matrix of the DC microgrid system under attack conditions, is the state vector of the DC microgrid system at the kth moment under attack, is the relief vector of the DC microgrid system at the kth moment; By setting an upper limit d on the residual test data capacity, if , then the residual items that do not exceed the residual threshold in the historical m steps will be defined as safe data, and the safe data will not be used for the calculation of the residual test; define the safe data set , the residual after removing the safety data Expressed as: ， in is the transposed matrix of the residual matrix, After removing the security data , is the false vector input after removing the security data.

6. The method for dynamic defense against malicious attacks in a DC microgrid system according to claim 1, characterized in that: The step (4) specifically includes the following steps: Assuming that the DC microgrid suffers a DoS attack at step k, the current history data of the historical k-m sampling cycles is first selected, and then a weight is assigned to the current of each cycle to form a weighted average estimated current; The estimated current of the i-th DC microgrid in the k-th sampling period is expressed as: ， in, is a weight factor that satisfies ; Based on the estimated current value, the estimated current proportional deviation is expressed as: ， in, is the parameter allocated according to the rated power of the i-th DC microgrid, is the parameter allocated according to the rated power of the jth DC microgrid, is the estimated current of the i-th DC microgrid in the k-th sampling period, is the estimated current of the j-th DC microgrid in the k-th sampling period; is the weight coefficient between microgrids; under DoS attack, the estimated control output of the i-th DC microgrid is Expressed as: ， is the actual control output of the i-th DC microgrid, is the control parameter; The weight factor is optimized to minimize the error between the estimated voltage value and the DC microgrid reference voltage value. The optimization problem is to minimize the Euclidean distance between the estimated voltage value and the DC microgrid reference voltage value, which is expressed as: ， in, is the reference voltage value, and the constraints are .

7. A dynamic defense system for malicious attacks in a DC microgrid system, characterized in that: include: A data collection unit is used to collect input data and output data in the DC microgrid system to form a historical data set. The input data includes voltage, current data and disturbance items, and the output data is control instructions. An attack model unit is introduced to allow attackers to invade the communication network of the DC microgrid and launch FDIA and DoS attacks; The FDIA defense unit establishes an equivalence relationship between the DC microgrid input and output in the historical data set, generates a residual term based on the equivalence relationship between the output and input, and improves the sensitivity of the residual term to attack vectors by designing an optimization problem. When an FDIA attack is detected, a mitigation vector is calculated to offset the false data vector. Finally, a dynamic secure data deletion strategy is adopted to eliminate excessive historical data. The DoS attack defense unit collects historical current data, assigns a weight factor to the current in each sampling period, estimates the current value blocked by the DoS attack through a weighted average estimation method, optimizes the weight factor, and estimates the current value.

8. The dynamic defense system for malicious attacks in a DC microgrid system according to claim 7, characterized in that: The false data vector injected by FDIA is expressed as: , in, is the voltage at the kth sampling moment containing the false vector, is the actual voltage value, is the false data vector of FDIA; The DoS attack set is defined as: ， in, Represents the sampling time of DoS attack; represents the normal sampling time, represents all sampling times of the microgrid; is the discrete attack moment, for discrete non-attack moments; The upper bound of the false data vector of FDIA is , the duration of the DoS attack is less than the sampling time of the DC microgrid, which is expressed as , the number of DoS attacks meets the following conditions: ,in, Represents the number of DoS attacks during the sampling time, , is a constant.

9. The dynamic defense system for malicious attacks in a DC microgrid system according to claim 7, characterized in that: The FDIA defense unit detects and defends FDIA based on the equivalent relationship between the input and output of the DC microgrid system at m sampling moments in the historical data. The equivalence relationship of the DC microgrid system history m steps can be written in the form of a vector matrix, expressed as: ， in, is the output of the historical m sampling moments, is the control input, noise input and false vector input at the historical m sampling moments, is the microgrid state at the 0th sampling moment, is the coefficient matrix of the microgrid; The residual used to detect FDIA is defined as: ， ， ， in, is the residual matrix; is the Jacobian matrix of the DC microgrid, , ; In order to make the residual term only sensitive to the attack vector, the residual matrix should be designed to be insensitive to noise or interference in the DC microgrid, so the optimization problem is based on the residual term in the absence of attack Minimize, expressed as: ， Among them, there are two constraints in the optimization problem, which are expressed as: ， in, and are the nominal values ​​of the inductor and capacitor of the DC step-down converter, is the percentage deviation, the deviation value Set to ±5%; Based on the residual term in the absence of attack To establish the residual threshold, first collect N groups of historical residual items , the residual threshold is calculated by the average value and deviation term of the historical state, expressed as: ， is the standard deviation of the residual term of the historical N groups, is the adjustment parameter of the confidence interval; When the residual value exceeds the threshold set by the DC microgrid When , the mitigation vector is calculated to offset the false data vector, and the mitigation vector is expressed as: ， in, is the mitigation vector, and by injecting the mitigation vector into the DC microgrid, the output state Expressed as: ， is the measurement matrix of the DC microgrid system, is the state vector of the DC microgrid system at the kth moment, is the measurement matrix of the DC microgrid system under attack conditions, is the state vector of the DC microgrid system at the kth moment under attack, is the relief vector of the DC microgrid system at the kth moment; By setting an upper limit d on the residual test data capacity, if , then the residual items that do not exceed the residual threshold in the historical m steps will be defined as safe data, and the safe data will not be used for the calculation of the residual test; define the safe data set , the residual after removing the safety data Expressed as: ， in is the transposed matrix of the residual matrix, After removing the security data , is the false vector input after removing the security data.

10. The dynamic defense system for malicious attacks in a DC microgrid system according to claim 7, characterized in that: In the DoS attack defense unit, it is assumed that the DC microgrid suffers a DoS attack at step k. First, the current history data of the historical k-m sampling cycles are selected, and then a weight is assigned to the current of each cycle to form a weighted average estimated current. The estimated current of the i-th DC microgrid in the k-th sampling period is expressed as: ， in, is a weight factor that satisfies ; Based on the estimated current value, the estimated current proportional deviation is expressed as: ， in, is the parameter allocated according to the rated power of the i-th DC microgrid, is the parameter allocated according to the rated power of the jth DC microgrid, is the estimated current of the i-th DC microgrid in the k-th sampling period, is the estimated current of the j-th DC microgrid in the k-th sampling period; is the weight coefficient between microgrids; under DoS attack, the estimated control output of the i-th DC microgrid is Expressed as: ， is the actual control output of the i-th DC microgrid, is the control parameter; The weight factor is optimized to minimize the error between the estimated voltage value and the DC microgrid reference voltage value. The optimization problem is to minimize the Euclidean distance between the estimated voltage value and the DC microgrid reference voltage value, which is expressed as: ， in, is the reference voltage value, and the constraints are .

Images