A multi-source power terminal cross-domain attack detection method and system
By modeling and correcting the voltage phase angle, frequency and power of the power system, a cross-domain attack detection model is built, which solves the problem of low cross-domain attack detection accuracy in the power system, and achieves high-accuracy attack recognition and security Assure.
Patent Information
- Application Number
- CN202411105269.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-08-13
AI Technical Summary
The existing power system abnormality detection methods have low accuracy in cross-domain attack detection of power key information terminals, and cannot guarantee the safety of smart power systems.
By digitally modeling the voltage phase angle, frequency and power injection into the power grid, a multi-source power terminal cross-domain attack data detection model is built, and the loss function and neural network are used to identify attack parameters to achieve accurate detection of cross-domain attacks. .
It improves the accuracy of cross-domain attack detection of power systems, reduces the complexity of algorithms, can be deployed in a distributed manner at the terminal, with accurate detection results, and can identify corresponding attack parameters, ensuring the safe operation of intelligent power systems.
Smart Images

Figure CN118646599B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power system security protection, and in particular to a multi-source power terminal cross-domain attack detection method and system. Background Art
[0002] With the rapid development of computer technology, information technology and Internet of Things technology, edge computing is driving the development and deployment of IoT power systems to provide advanced services. More and more smart terminals are pouring into the power system, including electric vehicles, networked air conditioners, photovoltaic inverters, smart water heaters, etc., which are collectively referred to as power key information terminals. While power key information terminals facilitate people's daily lives, they also bring some security risks. Due to the many information-physical interaction links inside them, they are vulnerable to cross-domain attacks. These cross-domain attacks may affect multiple terminals and exist in the form of static attacks and dynamic attacks. Once a power key information terminal is invaded by a cross-domain attack, it may be maliciously exploited to cause a large disturbance to the power grid and affect its stability.
[0003] The existing methods for power system anomaly detection are mainly bad data detection, which determines whether the power system has been attacked by detecting some bad data in the master station. However, cross-domain attacks are not aimed at the SCADA system of the power grid, but directly at the power key information terminals in the power grid. Key information terminals are usually scattered everywhere and controlled by users. As a result, the traditional power system anomaly detection methods have low accuracy in cross-domain attack detection and cannot guarantee the security of the power system in the context of a large number of IoT terminals connected to the smart grid.
[0004] It can be seen that how to solve the problem that the existing cross-domain attack detection for key power information terminals has low detection accuracy and cannot ensure the security of the smart power system has become a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention
[0005] The present invention provides a multi-source power terminal cross-domain attack detection method and system to solve the technical problems of low detection accuracy and inability to ensure the security of smart power systems in the existing cross-domain attack detection of power key information terminals, thereby improving the accuracy of cross-domain attack detection of power key information terminals and ensuring the safe operation of smart power systems.
[0006] In a first aspect, the present invention provides a multi-source power terminal cross-domain attack detection method, the method comprising:
[0007] Digitally modeling the target power grid according to the voltage phase angle and frequency of each node of the target power grid and the power injected into the power grid to obtain a first power system model;
[0008] Based on the multi-source power terminal cross-domain attack existing in the target power grid, the first power system model is modified to obtain a modified second power system model;
[0009] Extracting the first equation of the second power system model, constructing a loss function using the first equation, a mean square error between an actual value and a predicted value, and a multi-source power terminal cross-domain attack parameter, and constructing a multi-source power terminal cross-domain attack data detection model based on the loss function;
[0010] In the actual detection process, the voltage phase angle and the frequency of each node obtained in real time are respectively input into the multi-source power terminal cross-domain attack data detection model to obtain the multi-source power terminal cross-domain attack data of the target power grid;
[0011] The multi-source power terminal cross-domain attack data is analyzed to obtain attack results with attack attributes.
[0012] Preferably, the target power grid is digitally modeled according to the voltage phase angle, frequency and power injected into the power grid at each node of the target power grid to obtain a first power system model, including:
[0013] Constructing a second equation for a power generation node in the target power grid and a third equation for a load node in the target power grid, wherein the second equation is defined to reflect the relationship between the voltage phase angle, frequency and power injected into the power grid of the power generation node, and the third equation is defined to reflect the relationship between the voltage phase angle, frequency and power injected into the power grid of the load node;
[0014] Performing nonlinear modeling on the power injected into the power grid corresponding to the second equation and the third equation respectively, to obtain a fourth equation and a fifth equation respectively;
[0015] According to the fourth equation and the fifth equation, a first power system model is constructed.
[0016] Preferably, the second equation is:
[0017]
[0018] in, Represents a power generation node The voltage phase angle, express The derivative with respect to time, Represents a power generation node The frequency, express The derivative with respect to time, represents the set of power generation nodes, Represents a power generation node The inertia parameters of the generator, Represents a power generation node The damping parameters of the generator are Represents a power generation node The proportional coefficient of the generator proportional-integral control link, Represents a power generation node The integral coefficient of the generator proportional integral control link, Represents a power generation node Input power to the grid;
[0019] The third-party programs are:
[0020] ,
[0021] in, represents the equivalent damping coefficient of load node m, represents the time derivative of the voltage phase angle of load node m, Represents load node The actual power consumption, represents the power input from load node m to the grid, Represents a collection of load nodes;
[0022] The fourth equation is:
[0023] ,
[0024] Where N represents the set of power generation nodes and load nodes, Represents a power generation node and nodes The admittance parameter of the line between , represents the power generation node and power generation nodes The voltage phase angle deviation;
[0025] The fifth equation is:
[0026] ,
[0027] in, Represents load node and nodes The admittance parameter of the line between , represents the load node and load nodes The voltage phase angle deviation.
[0028] Preferably, the multi-source power terminal cross-domain attack based on the target power grid modifies the first power system model to obtain a modified second power system model, including:
[0029] Based on the protection levels of the multi-source power terminals, the multi-source power terminals are divided into secure terminals and vulnerable terminals;
[0030] Digitally modeling the power consumption of the vulnerable terminal to obtain a sixth equation of power of the vulnerable terminal, wherein the sixth equation is defined to reflect dynamic attacks and static attacks on the vulnerable terminal;
[0031] Obtaining the actual power consumption of the load node according to the sixth equation;
[0032] Substituting the actual power consumption of the load node into the fifth equation to obtain a seventh equation;
[0033] According to the seventh equation, the first power system model is corrected to obtain a corrected second power system model.
[0034] Preferably, the seventh equation is:
[0035] ,
[0036] in, represents the set of measurement nodes, , Indicates The frequency of the measurement nodes, Indicates The gain of the measurement node measuring the frequency, represents the static attack power of load node m, Indicates The power consumption of each safety terminal, represents the set of secure terminals, Represents the dynamic attack power of load node m.
[0037] Preferably, the static attack power is less than or equal to the power consumption of the corresponding vulnerable terminal;
[0038] The maximum value of the dynamic attack power is less than or equal to half of the difference between the power consumption of the corresponding vulnerable terminal and the static attack power.
[0039] Preferably, the loss function is constructed based on the first equation, the mean square error between the actual value and the predicted value, and the multi-source power terminal cross-domain attack parameter, including:
[0040] Constructing a first term of the loss function based on the mean square error between the actual value and the predicted value;
[0041] Constructing a second term of the loss function according to the first equation;
[0042] Constructing the third term of the loss function according to the multi-source power terminal cross-domain attack parameter;
[0043] The first term, the second term and the third term are combined to obtain the loss function.
[0044] Preferably, the first equation is:
[0045] ,
[0046]
[0047]
[0048] in, yes The predicted value of yes The predicted value of yes The predicted value of yes The predicted value of yes The predicted value of express The predicted value of express The predicted value of express The predicted value of express The predicted value of Indicates each The vector composed of Indicates each The vector composed of represents the first component of the first equation, represents the second component of the first equation, represents the third component of the first equation;
[0049] The first item is:
[0050] ,
[0051] in, express Power generation node at all times The voltage phase angle, Respectively Power generation node at all times The frequency, are the parameters of the neural network, express Power generation node at all times The voltage phase angle is The corresponding output prediction value under the parameter, express Power generation node at all times The frequency of The corresponding output prediction value under the parameter, where T represents the detection duration;
[0052] The second item is:
[0053] ,
[0054] in, Indicates each The vector composed of Indicates each The vector composed of
[0055] The third item is:
[0056]
[0057] in, is the expansion coefficient, express The one-norm of express The one-norm of express The predicted value of , a vector containing all dynamic attack parameters, express The predicted value of , a vector containing all static attack parameters;
[0058] The loss function is:
[0059] .
[0060] Preferably, the multi-source power terminal cross-domain attack data includes a vector composed of the gains of the measurement frequencies of each node under dynamic attack and a vector composed of the static attack powers of each node.
[0061] In a second aspect, the present invention further provides a multi-source power terminal cross-domain attack detection system, the system comprising: the system comprising: a first power system construction unit, a model correction unit, a detection model construction unit, a data detection unit and a data analysis unit;
[0062] The first power system construction unit is used to digitally model the target power grid according to the voltage phase angle and frequency of each node of the target power grid and the power injected into the power grid to obtain a first power system model;
[0063] The model correction unit is used to correct the first power system model based on the multi-source power terminal cross-domain attack existing in the target power grid to obtain a corrected second power system model;
[0064] The detection model construction unit is used to extract the first equation of the second power system model, respectively construct a loss function with the first equation, a mean square error between an actual value and a predicted value, and a multi-source power terminal cross-domain attack parameter, and construct a multi-source power terminal cross-domain attack data detection model based on the loss function;
[0065] The model detection unit is used to input the voltage phase angle and the frequency of each node obtained in real time into the multi-source power terminal cross-domain attack data detection model in an actual detection process to obtain the multi-source power terminal cross-domain attack data of the target power grid;
[0066] The data analysis unit is used to analyze the multi-source power terminal cross-domain attack data to obtain an attack result with attack attributes.
[0067] The present invention provides a multi-source power terminal cross-domain attack detection method and system. Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following:
[0068] (1) High detection accuracy. Combined with the physical constraints of the power system, the algorithm complexity is greatly reduced, so it can be deployed on the terminal to run the detection in a distributed manner, and the detection results are highly accurate.
[0069] (2) Ability to identify corresponding attack parameters. BRIEF DESCRIPTION OF THE DRAWINGS
[0070] Figure 1 It is a schematic diagram of the steps of a multi-source power terminal cross-domain attack detection method provided by a preferred embodiment of the present invention;
[0071] Figure 2 It is a schematic diagram of the steps of a first power system model building method provided by a preferred embodiment of the present invention;
[0072] Figure 3 It is a schematic diagram of the steps of a method for correcting a first power system model provided by a preferred embodiment of the present invention;
[0073] Figure 4 It is a schematic diagram of the steps of a method for constructing a loss function provided by a preferred embodiment of the present invention;
[0074] Figure 5 It is a schematic diagram of a multi-source power terminal cross-domain attack detection system provided by a preferred embodiment of the present invention. DETAILED DESCRIPTION
[0075] The following is a detailed explanation of the implementation mode of the present invention in conjunction with the accompanying drawings. The embodiments are provided only for illustrative purposes and cannot be understood as limiting the present invention. The accompanying drawings are only for reference and illustration purposes and do not constitute a limitation on the scope of patent protection of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In the description of the present invention, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, the meaning of "multiple" is two or more.
[0076] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for illustrative purposes, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0077] In the description of the present invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood by specific circumstances.
[0078] See also Figure 1 In an embodiment of the present invention, a multi-source power terminal cross-domain attack detection method is provided, the method comprising:
[0079] S1. Digitally modeling the target power grid according to the voltage phase angle and frequency of each node of the target power grid and the power injected into the power grid to obtain a first power system model.
[0080] S2. Based on the cross-domain attack of multiple power terminals existing in the target power grid, the first power system model is modified to obtain a modified second power system model.
[0081] S3. Extract the first equation of the second power system model, construct a loss function with the first equation shown, the mean square error between the actual value and the predicted value, and the multi-source power terminal cross-domain attack parameter, and construct a multi-source power terminal cross-domain attack data detection model based on the loss function.
[0082] S4. During the actual detection process, the voltage phase angle and the frequency of each node acquired in real time are respectively input into the multi-source power terminal cross-domain attack data detection model to obtain the multi-source power terminal cross-domain attack data of the target power grid.
[0083] S5. Analyze the multi-source power terminal cross-domain attack data to obtain attack results with attack attributes.
[0084] In a preferred embodiment of the present invention, a target power grid with multiple power terminals is digitally modeled according to the voltage phase angle, frequency and power injected into the power grid at each node of the target power grid to obtain a first power system model, such as Figure 2 As shown, including the following:
[0085] S101. Construct a second equation for the power generation node in the target power grid, and a third equation for the load node in the target power grid, wherein the second equation is defined to reflect the relationship between the voltage phase angle, frequency and power injected into the power grid of the power generation node, and the third equation is defined to reflect the relationship between the voltage phase angle, frequency and power injected into the power grid of the load node.
[0086] S102, performing nonlinear modeling on the power injected into the power grid corresponding to the second equation and the third equation, respectively, to obtain a fourth equation and a fifth equation, respectively.
[0087] S103: Construct a first power system model according to the fourth equation and the fifth equation.
[0088] The nodes connected to the generators in the target power grid are defined as power generation nodes, which generate electrical energy. The nodes connected to the electrical appliances in the target power grid are defined as load nodes, which consume electrical energy. For the power generation nodes, the second equation of the power generation nodes is constructed. The second equation is defined to reflect the relationship between the voltage phase angle, frequency and power injected into the power grid of the power generation nodes. According to the physical relationship in the operation of the power system, the second equation of the power generation node is obtained as follows:
[0089]
[0090] in, Represents a power generation node The voltage phase angle, express The derivative with respect to time, Represents a power generation node The frequency, express The derivative with respect to time, represents the set of power generation nodes, Represents a power generation node The inertia parameters of the generator, Represents a power generation node The damping parameters of the generator are Represents a power generation node The proportional coefficient of the generator proportional-integral control link, Represents a power generation node The integral coefficient of the generator proportional integral control link, Represents a power generation node Input power to the grid.
[0091] According to the physical relationship in the operation of the power system, the third-party equation of the load node is obtained. The third-party equation is defined as reflecting the relationship between the voltage phase angle, frequency and power injected into the power grid at the load node. The third-party equation is as follows:
[0092] ,
[0093] in, represents the equivalent damping coefficient of load node m, represents the time derivative of the voltage phase angle of load node m, Represents load node The actual power consumption, represents the power input from load node m to the grid, Represents a collection of load nodes.
[0094] In the third-party program, since the load node consumes power, the power injected into the grid by the load node is a negative value.
[0095] The power injected into the grid in the second equation and third-party programs to inject power into the grid , nonlinear modeling is used to obtain the fourth equation and the fifth equation respectively. The fourth equation is as follows:
[0096] ,
[0097] Where N represents the set of power generation nodes and load nodes, Represents a power generation node and nodes The admittance parameter of the line between , represents the power generation node and nodes The voltage phase angle deviation.
[0098] The fifth equation is as follows:
[0099] ,
[0100] in, Represents load node and nodes The admittance parameter of the line between , represents the load node and nodes The voltage phase angle deviation.
[0101] In a preferred embodiment of the invention, based on the first power system model, a cross-domain attack on power terminals existing in the target power grid is digitally modeled to obtain a second power system model, such as Figure 3 As shown, the following steps are included:
[0102] S201. Based on the protection level of the power terminal, divide the power terminal into a secure terminal and a vulnerable terminal.
[0103] S202: Digitally model the power consumption of the vulnerable terminal to obtain a sixth equation of the power consumption of the vulnerable terminal, wherein the sixth equation is defined to reflect dynamic attacks and static attacks on the vulnerable terminal.
[0104] S203. Obtain the actual power consumption of the load node according to the sixth equation.
[0105] S204. Substitute the actual power consumption of the load node into the fourth equation and the fifth equation to obtain a seventh equation and an eighth equation respectively.
[0106] S205. According to the seventh equation and the eighth equation, the first power system model is corrected to obtain a corrected second power system model.
[0107] In a preferred embodiment of the present invention, based on the protection level of the terminal, the terminal is classified into secure terminals and vulnerable terminals, wherein the secure terminal is a terminal with a high protection level and cannot be attacked, and the vulnerable terminal is a terminal with a low protection level and may be attacked. and Represents the power consumption of the secure terminal and the vulnerable terminal. For each load node, the relationship between the power consumption of the secure terminal and the vulnerable terminal is as follows:
[0108]
[0109] Furthermore, for vulnerable nodes Digital modeling is performed to obtain the sixth equation of vulnerable terminal power. The sixth equation is defined to reflect the dynamic attack and static attack of the vulnerable terminal, as shown below:
[0110] ,
[0111] in, represents the set of measurement nodes, , Indicates The frequency of the measurement nodes, Indicates The gain of the measurement node measuring the frequency, Represents the static attack power of load node m.
[0112] In the sixth equation, the first term It represents the modeling of dynamic attacks on vulnerable terminals. For dynamic attacks, attackers need to sense some nodes and attack according to the frequency information of the sensed nodes. Only by sensing the frequency information of the nodes can dynamic attacks be launched and measured. Therefore, the measurement node is the node whose frequency information can be sensed by the attacker. Indicates The gain of the measurement frequency of each measurement node is set in advance by the attacker and its value is greater than 0. The static attack power represented by represents a step signal.
[0113] Substituting the sixth equation into the relationship between the power consumption of the secure terminal and the vulnerable terminal, the actual power consumption of the load node is obtained as follows:
[0114]
[0115] in, Indicates The power consumption of each safety terminal, Represents a collection of secure endpoints.
[0116] In the embodiment of the present invention, considering the actual situation, dynamic attacks and static attacks are constrained. For the dynamic attack constraint, considering that the dynamic threat will increase and decrease the load at the same time, the maximum value of the dynamic attack power is less than or equal to half of the difference between the power consumption of the corresponding vulnerable terminal and the static attack power, which is specifically expressed as follows:
[0117] .
[0118] For static attack constraints, the static attack power is less than or equal to the power consumption of the corresponding vulnerable terminal, which is specifically expressed as follows:
[0119] .
[0120] Furthermore, the actual power consumption of the load node is substituted into the fifth equation to obtain the seventh equation.
[0121] The seventh equation is as follows:
[0122] .
[0123] According to the seventh equation, the first power system model is corrected to obtain a corrected second power system model. Then, the equations corresponding to the second power system model include the fourth equation and the seventh equation, as shown below:
[0124] ,
[0125] ,
[0126] .
[0127] Furthermore, in an embodiment of the present invention, the first equation of the second power system model is extracted, and a loss function is constructed using the first equation, the mean square error between the actual value and the predicted value, and the multi-source power terminal cross-domain attack parameter.
[0128] The first equation extracted from the second power system model is:
[0129] ,
[0130]
[0131]
[0132] in, yes The predicted value of yes The predicted value of yes The predicted value of yes The predicted value of yes The predicted value of express The predicted value of express The predicted value of express The predicted value of express The predicted value of Indicates each The vector composed of Indicates each The vector composed of represents the first component of the first equation, represents the second component of the first equation, represents the third component of the first equation.
[0133] The predicted values involved in the embodiments of the present invention are all outputs of the multi-source power terminal cross-domain attack data detection model for the corresponding quantities.
[0134] In a preferred embodiment of the present invention, the first equation, the mean square error between the actual value and the predicted value, and the multi-source power terminal cross-domain attack parameter construct a loss function, such as Figure 4 As shown, the following steps are included:
[0135] S301. Construct a first term of a loss function according to a mean square error between the actual value and the predicted value.
[0136] S302. Construct a second term of the loss function according to the first equation.
[0137] S303. Construct the third term of the loss function according to the multi-source power terminal cross-domain attack parameters.
[0138] S304: Combine the first item, the second item and the third item to obtain the loss function
[0139] In a preferred embodiment of the present invention, a neural network is used to construct a multi-source power terminal cross-domain attack data detection model, and a loss function of the neural network is designed. The loss function includes three parts, namely the first term of the loss function, the second term of the loss function, and the third term of the loss function.
[0140] Among them, the first term of the loss function is used to reduce the deviation between the predicted value and the actual value. The first term is expressed as follows:
[0141]
[0142] in, express Power generation node at all times The voltage phase angle, Respectively Power generation node at all times The frequency, are the parameters of the neural network, express Power generation node at all times The voltage phase angle is The corresponding output prediction value under the parameter, express Power generation node at all times The frequency of The corresponding output prediction value under the parameter, T represents the detection duration.
[0143] Among them, the second term of the loss function is constrained by the first equation, so that the neural network can be optimized according to the actual situation and reduce the complexity of training. The second term of the loss function is expressed as follows:
[0144]
[0145] in, Indicates each The vector composed of Indicates each Vector composed of.
[0146] The third term of the loss function is to generate a sparse solution, and this term depends on and , the third term of the loss function is expressed as follows:
[0147]
[0148] in, is the expansion coefficient, express The one-norm of express The one-norm of express The predicted value of , a vector containing all dynamic attack parameters, express The predicted value of contains the vector of all static attack parameters. Specifically, the gain of the measured frequency is the dynamic attack parameter, and the static attack power is the static attack parameter.
[0149] The third term of the loss function is Corresponding to dynamic attack, Corresponding to static attack.
[0150] Therefore, the loss function of the multi-source power terminal cross-domain attack data detection model is:
[0151]
[0152] Furthermore, the basic parameters of the neural network are set, and the multi-source power terminal cross-domain attack data detection model is built. Based on the simulation, the training data is obtained. For the target power grid, a specific node is selected as the attack node. In order to be universal, several different types of attack node sets should be selected, including single-point threat and multi-point threat, and then the parameters of the dynamic attack are set. , parameters of static attack , by solving the second power system model, the training data can be obtained. The dynamic attack parameters and static attack parameters of the node position that is not selected as the attack node are 0. At the same time, considering the interference problem in the actual physical world, a layer of noise is added to the voltage phase angle and frequency. The noise uses Gaussian noise, and the standard deviation is set to 0.01pu. Pu is a standardized unit of measurement.
[0153] In the actual detection process, the voltage phase angle and frequency of the target power grid are monitored, and the voltage phase angle and frequency of each node obtained in real time are input into the multi-source power terminal cross-domain attack data detection model to obtain the multi-source power terminal cross-domain attack data of the target power grid. The output multi-source power terminal cross-domain attack data includes a vector composed of the gain of the measured frequency of each node of the dynamic attack and a vector composed of the static attack power of each node.
[0154] According to the vector composed of the gain of the measurement frequency of each node and the vector composed of the static attack power of each node, it is judged whether each node is under malicious attack or normal environmental disturbance. First, the gain of the measurement frequency corresponding to each node is judged. If the gain of the measurement frequency is 0, the static attack frequency of the node is further judged. If the static attack frequency of the node is greater than the preset static attack threshold, it indicates that the target power grid is under static attack; if the gain of the measurement frequency is greater than 0, it is compared with the preset dynamic attack threshold. If the dynamic attack frequency of the node is greater than the preset dynamic attack threshold, it indicates that the target power grid is under dynamic attack.
[0155] The multi-source power terminal cross-domain attack detection method provided by the present invention can detect the gain parameters of the measurement frequency and the static attack power of the dynamic attack, so as to judge the attack according to the gain parameters of the measurement frequency and the static attack power, thereby improving the accuracy of attack detection.
[0156] In a preferred embodiment of the present invention, the target power grid is digitally modeled according to the voltage phase angle, frequency and power injected into the power grid of each node of the target power grid to obtain a first power system model; based on the cross-domain attack of multi-source power terminals in the target power grid, the first power system model is corrected to obtain a corrected second power system model; the first equation of the second power system model is extracted, and the loss function is constructed by the first equation, the mean square error between the actual value and the predicted value, and the cross-domain attack parameter of the multi-source power terminal, and a multi-source power terminal cross-domain attack data detection model is constructed based on the loss function; in the actual detection process, the voltage phase angle and frequency of each node obtained in real time are respectively input into the cross-domain attack data detection model of the multi-source power terminal to obtain the cross-domain attack data of the multi-source power terminal of the target power grid, and the cross-domain attack data of the multi-source power terminal is analyzed to obtain an attack result with attack attributes. The multi-source power terminal cross-domain attack detection method of the present invention greatly reduces the algorithm complexity by combining the physical constraints of the power system, so it can be deployed on the terminal to run the detection in a distributed manner, and the detection result is highly accurate and can identify the corresponding attack parameters.
[0157] Accordingly, if Figure 5 As shown, based on a multi-source power terminal cross-domain attack detection method, an embodiment of the present invention also provides a multi-source power terminal cross-domain attack detection system, the system comprising: a first power system construction unit 1, a model correction unit 2, a detection model construction unit 3, a model detection unit 4 and a data analysis unit 5;
[0158] The first power system construction unit 1 is used to digitally model the target power grid according to the voltage phase angle and frequency of each node of the target power grid and the power injected into the power grid to obtain a first power system model.
[0159] The model correction unit 2 is used to correct the first power system model based on the multi-source power terminal cross-domain attack existing in the target power grid to obtain a corrected second power system model.
[0160] The detection model construction unit 3 is used to extract the first equation of the second power system model, and construct a loss function with the first equation, the mean square error between the actual value and the predicted value, and the multi-source power terminal cross-domain attack parameter, and construct a multi-source power terminal cross-domain attack data detection model based on the loss function.
[0161] The model detection unit 4 is used to input the voltage phase angle and the frequency of each node obtained in real time into the multi-source power terminal cross-domain attack data detection model during the actual detection process, so as to obtain the multi-source power terminal cross-domain attack data of the target power grid.
[0162] The data analysis unit 5 is used to analyze the multi-source power terminal cross-domain attack data to obtain an attack result with attack attributes.
[0163] For the specific definition of a multi-source power terminal cross-domain attack detection system, please refer to the above-mentioned definition of a multi-source power terminal cross-domain attack detection method, which will not be repeated here. A person of ordinary skill in the art can appreciate that the various modules and steps described in conjunction with the embodiments disclosed in the present invention can be implemented in hardware, software, or a combination of both. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.
[0164] A multi-source power terminal cross-domain attack detection method and system provided in this embodiment is aimed at the technical problem that the detection accuracy of cross-domain attack detection of power key information terminals is low and the security of the power system under the background of a large number of Internet of Things terminals accessing the smart grid cannot be guaranteed. The present invention digitally models the target power grid according to the voltage phase angle, frequency and power injected into the power grid of each node of the target power grid to obtain a first power system model; based on the multi-source power terminal cross-domain attack existing in the target power grid, the first power system model is corrected to obtain a corrected second power system model; the first equation of the second power system model is extracted, and the loss function is constructed by the first equation, the mean square error between the actual value and the predicted value, and the multi-source power terminal cross-domain attack parameter, and a multi-source power terminal cross-domain attack data detection model is constructed based on the loss function; in the actual detection process, the voltage phase angle and frequency of each node obtained in real time are respectively input into the multi-source power terminal cross-domain attack data detection model to obtain the multi-source power terminal cross-domain attack data of the target power grid, and the multi-source power terminal cross-domain attack data is analyzed to obtain an attack result with attack attributes. The multi-source power terminal cross-domain attack detection method of the present invention, combined with the physical constraints of the power system, greatly reduces the algorithm complexity, so it can be deployed on the terminal to run the detection in a distributed manner. The detection results are highly accurate and can identify the corresponding attack parameters.
[0165] The above-mentioned embodiments only express several preferred implementation modes of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this technical field, several improvements and substitutions can be made without departing from the technical principle of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the protection scope of the claims.
Claims
1. A multi-source power terminal cross-domain attack detection method, characterized in that: The method comprises: Digitally modeling the target power grid according to the voltage phase angle and frequency of each node of the target power grid and the power injected into the power grid to obtain a first power system model; Based on the multi-source power terminal cross-domain attack existing in the target power grid, the first power system model is modified to obtain a modified second power system model; Extracting the first equation of the second power system model, constructing a loss function using the first equation, a mean square error between an actual value and a predicted value, and a multi-source power terminal cross-domain attack parameter, and constructing a multi-source power terminal cross-domain attack data detection model based on the loss function; In the actual detection process, the voltage phase angle and the frequency of each node obtained in real time are respectively input into the multi-source power terminal cross-domain attack data detection model to obtain the multi-source power terminal cross-domain attack data of the target power grid; Analyzing the multi-source power terminal cross-domain attack data to obtain attack results with attack attributes; The target power grid is digitally modeled according to the voltage phase angle, frequency and power injected into the power grid at each node of the target power grid to obtain a first power system model, including: Constructing a second equation for a power generation node in the target power grid and a third equation for a load node in the target power grid, wherein the second equation is defined to reflect the relationship between the voltage phase angle, frequency and power injected into the power grid of the power generation node, and the third equation is defined to reflect the relationship between the voltage phase angle, frequency and power injected into the power grid of the load node; Performing nonlinear modeling on the power injected into the power grid corresponding to the second equation and the third equation respectively, to obtain a fourth equation and a fifth equation respectively; Constructing a first power system model according to the fourth equation and the fifth equation; The first equation is: , in, yes The predicted value of yes The predicted value of yes The predicted value of yes The predicted value of yes The predicted value of express The predicted value of express The predicted value of express The predicted value of express The predicted value of Indicates each The vector composed of Indicates The gain of the measurement node measuring the frequency, Indicates each The vector composed of represents the first component of the first equation, represents the second component of the first equation, represents the third component of the first equation, Represents a power generation node The voltage phase angle, express The derivative with respect to time, Represents a power generation node The frequency, express The derivative with respect to time, express The predicted value of Indicates The frequency of the measurement nodes, represents the set of power generation nodes, Represents a power generation node The inertia parameters of the generator, Represents a power generation node The damping parameters of the generator are Represents a power generation node The proportional coefficient of the generator proportional-integral control link, Represents a power generation node The integral coefficient of the generator proportional integral control link, represents the equivalent damping coefficient of load node m, represents the time derivative of the voltage phase angle of load node m, represents the set of load nodes, N represents the set of power generation nodes and load nodes, represents the set of measurement nodes, , represents the set of secure terminals, Represents load node and nodes The admittance parameter of the line between , represents the load node and nodes The voltage phase angle deviation, Indicates The power consumption of each safety terminal; The loss function is constructed based on the first equation, the mean square error between the actual value and the predicted value, and the multi-source power terminal cross-domain attack parameter, including: Constructing a first term of the loss function based on the mean square error between the actual value and the predicted value; Constructing a second term of the loss function according to the first equation; Constructing the third term of the loss function according to the multi-source power terminal cross-domain attack parameter; Combining the first term, the second term and the third term to obtain the loss function; The first item is , in, express Power generation node at all times The voltage phase angle, Respectively Power generation node at all times The frequency, are the parameters of the neural network, express Power generation node at all times The voltage phase angle is The corresponding output prediction value under the parameter, express Power generation node at all times The frequency of The corresponding output prediction value under the parameter, where T represents the detection duration; The second item is: , in, Indicates each The vector composed of represents the static attack power of load node m; The third item is: in, is the expansion coefficient, express The one-norm of express The one-norm of express The predicted value of , a vector containing all dynamic attack parameters, express The predicted value of , a vector containing all static attack parameters; The loss function is: 。 2. The multi-source power terminal cross-domain attack detection method according to claim 1, characterized in that: The second equation is: in, Represents a power generation node The voltage phase angle, express The derivative with respect to time, Represents a power generation node The frequency, express The derivative with respect to time, represents the set of power generation nodes, Represents a power generation node The inertia parameters of the generator, Represents a power generation node The damping parameters of the generator are Represents a power generation node The proportional coefficient of the generator proportional-integral control link, Represents a power generation node The integral coefficient of the generator proportional integral control link, Represents a power generation node Input power to the grid; The third-party programs are: , in, represents the equivalent damping coefficient of load node m, represents the time derivative of the voltage phase angle of load node m, Represents load node The actual power consumption, represents the power input from load node m to the grid, represents a collection of load nodes; The fourth equation is: , Where N represents the set of power generation nodes and load nodes, Represents a power generation node and nodes The admittance parameter of the line between , represents the power generation node and nodes The voltage phase angle deviation; The fifth equation is: , in, Represents load node and nodes The admittance parameter of the line between , represents the load node and nodes The voltage phase angle deviation.
3. The multi-source power terminal cross-domain attack detection method according to claim 2, characterized in that: The method of modifying the first power system model based on the multi-source power terminal cross-domain attack existing in the target power grid to obtain a modified second power system model includes: Based on the protection levels of the multi-source power terminals, the multi-source power terminals are divided into secure terminals and vulnerable terminals; Digitally modeling the power consumption of the vulnerable terminal to obtain a sixth equation of power of the vulnerable terminal, wherein the sixth equation is defined to reflect dynamic attacks and static attacks on the vulnerable terminal; Obtaining the actual power consumption of the load node according to the sixth equation; Substituting the actual power consumption of the load node into the fifth equation to obtain a seventh equation; According to the seventh equation, the first power system model is corrected to obtain a corrected second power system model.
4. The multi-source power terminal cross-domain attack detection method according to claim 3, characterized in that: The seventh equation is: , in, Represents the dynamic attack power of load node m.
5. The multi-source power terminal cross-domain attack detection method according to claim 4, characterized in that: The static attack power is less than or equal to the power consumption of the corresponding vulnerable terminal; The maximum value of the dynamic attack power is less than or equal to half of the difference between the power consumption of the corresponding vulnerable terminal and the static attack power.
6. The multi-source power terminal cross-domain attack detection method according to claim 1, characterized in that: The multi-source power terminal cross-domain attack data includes a vector composed of the gain of each node measurement frequency of the dynamic attack and a vector composed of the static attack power of each node.
7. A multi-source power terminal cross-domain attack detection system, characterized in that: The system comprises: a first power system construction unit, a model correction unit, a detection model construction unit, a data detection unit and a data analysis unit; The first power system construction unit is used to digitally model the target power grid according to the voltage phase angle and frequency of each node of the target power grid and the power injected into the power grid to obtain a first power system model; The model correction unit is used to correct the first power system model based on the multi-source power terminal cross-domain attack existing in the target power grid to obtain a corrected second power system model; The detection model construction unit is used to extract the first equation of the second power system model, respectively construct a loss function with the first equation, a mean square error between an actual value and a predicted value, and a multi-source power terminal cross-domain attack parameter, and construct a multi-source power terminal cross-domain attack data detection model based on the loss function; The model detection unit is used to input the voltage phase angle and the frequency of each node obtained in real time into the multi-source power terminal cross-domain attack data detection model in an actual detection process to obtain the multi-source power terminal cross-domain attack data of the target power grid; The data analysis unit is used to analyze the multi-source power terminal cross-domain attack data to obtain an attack result with attack attributes; The target power grid is digitally modeled according to the voltage phase angle, frequency and power injected into the power grid at each node of the target power grid to obtain a first power system model, including: Constructing a second equation for a power generation node in the target power grid and a third equation for a load node in the target power grid, wherein the second equation is defined to reflect the relationship between the voltage phase angle, frequency and power injected into the power grid of the power generation node, and the third equation is defined to reflect the relationship between the voltage phase angle, frequency and power injected into the power grid of the load node; Performing nonlinear modeling on the power injected into the power grid corresponding to the second equation and the third equation respectively, to obtain a fourth equation and a fifth equation respectively; Constructing a first power system model according to the fourth equation and the fifth equation; The first equation is: , in, yes The predicted value of yes The predicted value of yes The predicted value of yes The predicted value of yes The predicted value of express The predicted value of express The predicted value of express The predicted value of express The predicted value of Indicates each The vector composed of Indicates The gain of the measurement node measuring the frequency, Indicates each The vector composed of represents the first component of the first equation, represents the second component of the first equation, represents the third component of the first equation, Represents a power generation node The voltage phase angle, express The derivative with respect to time, Represents a power generation node The frequency, express The derivative with respect to time, express The predicted value of Indicates The frequency of the measurement nodes, represents the set of power generation nodes, Represents a power generation node The inertia parameters of the generator, Represents a power generation node The damping parameters of the generator are Represents a power generation node The proportional coefficient of the generator proportional-integral control link, Represents a power generation node The integral coefficient of the generator proportional integral control link, represents the equivalent damping coefficient of load node m, represents the time derivative of the voltage phase angle of load node m, represents the set of load nodes, N represents the set of power generation nodes and load nodes, represents the set of measurement nodes, , represents the set of secure terminals, Represents load node and nodes The admittance parameter of the line between , represents the load node and nodes The voltage phase angle deviation, Indicates The power consumption of each safety terminal; The loss function is constructed based on the first equation, the mean square error between the actual value and the predicted value, and the multi-source power terminal cross-domain attack parameter, including: Constructing a first term of the loss function based on the mean square error between the actual value and the predicted value; Constructing a second term of the loss function according to the first equation; Constructing the third term of the loss function according to the multi-source power terminal cross-domain attack parameter; Combining the first term, the second term and the third term to obtain the loss function; The first item is , in, express Power generation node at all times The voltage phase angle, Respectively Power generation node at all times The frequency, are the parameters of the neural network, express Power generation node at all times The voltage phase angle is The corresponding output prediction value under the parameter, express Power generation node at all times The frequency of The corresponding output prediction value under the parameter, where T represents the detection duration; The second item is: , in, Indicates each The vector composed of represents the static attack power of load node m; The third item is: in, is the expansion coefficient, express The one-norm of express The one-norm of express The predicted value of , a vector containing all dynamic attack parameters, express The predicted value of , a vector containing all static attack parameters; The loss function is: 。
