Current transformer system network attack detection method and system based on electric signals

By calculating steady-state trusted boundaries and gray correlation filtering characteristics, training the support vector machine model, monitoring electrical signals in real time, and quickly determining that the converter system is subject to network attacks, solving the problem of insufficient detection lag and accuracy in the existing technology. , realizes fast and accurate detection and protection of cyberattacks.

CN118972141AActive Publication Date: 2024-11-15HUNAN UNIV

Patent Information

Application Number
CN202411161651.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-11-15
Estimated Expiration
2044-08-23

AI Technical Summary

Technical Problem

The prior art is difficult to take into account the real-time and accuracy of network attack detection inverter systems, and the existing methods have lag and unreliability in emergencies, which cannot effectively prevent system losses.

Method used

By obtaining historical steady-state operating conditions and network attack datasets, calculating steady-state trusted boundaries, filtering features using gray correlation method, training support vector machine models, monitoring electrical signals in real time, quickly determining whether they are subject to cyber attacks and determining the type of attack, Disconnect the server to reduce detection time and error.

Benefits of technology

It realizes rapid and accurate judgment of network attacks in the converter system, reduces system losses, takes into account real-time and accuracy of detection, and is suitable for emergency situations, reducing detection time and calculation amount.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118972141A_ABST
    Figure CN118972141A_ABST
Patent Text Reader

Abstract

The invention discloses a converter system network attack detection method and system based on an electric signal. The method comprises the following steps: acquiring a data set under a historical steady-state working condition and a historical network attack data set; obtaining a steady-state credible boundary according to the minimum value of the periodic regular change of the credible factor calculated by the data set under the steady-state working condition; screening a training data set according to the historical network attack data set and a grey relational degree method, and inputting the training data set into a machine learning model for training to obtain a network attack type detection model; and monitoring and collecting three-phase current signals and three-phase voltage signals of the converter system in real time, calculating credible factors, comparing the credible factors with the steady-state credible boundary to determine whether the converter system suffers from network attacks, if so, disconnecting the server to break away from an attack source, and judging the type of the network attacks in real time according to a network attack type detection model. According to the invention, the problem that the detection real-time performance and the detection accuracy are difficult to consider in the existing network attack detection is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of network attack detection, and in particular to a method and system for detecting network attacks on a converter system based on electrical signals. Background Art

[0002] Driven by Industry 4.0 and the Industrial Internet, cyber-physical information systems have developed rapidly, with various systems highly interconnected and sharing data. However, the increasing connectivity, expanding attack surface, and different infrastructures of these complex systems also bring cybersecurity challenges. Currently, cybersecurity has been given attention at the system level, but the security at the device level, especially the security of the converter system, has been largely ignored. Today's converters generally have communication functions and can be connected to the power grid or cloud services, which increases the risk of these devices being attacked. Therefore, it is of great significance to effectively and timely detect cyber attacks on converter systems.

[0003] Although some studies have explored the impact of cyber attacks on converter systems, most existing work focuses on fault detection and diagnosis. Converter systems play a role in the efficient control and transmission of electric energy, and cyber attacks should be detected in a timely and effective manner before equipment failure occurs to reduce losses. In order to address the vulnerabilities brought by cyber attacks, model-based and data-driven methods have been proposed. However, due to the inevitable parameter mismatch when building models and the complexity of smart power systems based on power electronics, model-based methods rely on accurate mathematical models of system health and are difficult to apply to actual projects. In contrast, data-driven methods have been widely used to detect cyber attacks using measurement data without explicit mathematical models, especially machine learning and deep learning models.

[0004] In the detection problem based on time series, the use of machine learning or deep learning models often requires coordination between detection real-time and detection accuracy. Using data from fewer time sampling points means fewer effective features, which affects the accuracy. As the attack time increases, the correlation between features and different types of network attacks increases. In order to obtain strongly correlated features and improve the detection accuracy, it is necessary to select data from more time sampling points, resulting in weak detection real-time. In practical applications, network attack detection of converter systems should include two steps: 1) determine whether a network attack has occurred; 2) detect the specific type of network attack. The existing methods all complete the two steps through an algorithm model, that is, when a specific network attack is detected, it is also determined that the network attack has occurred. At this point, the algorithm model has consumed a lot of time to detect accurate network attacks, and the converter system may have suffered a failure and caused inevitable losses. Therefore, it is necessary to determine whether the system has encountered a network attack in the shortest time within step 1. If it is determined that the system has encountered an attack, the server's network connection is immediately disconnected to disconnect the attack source to ensure that the system operation stability is not further damaged. Secondly, it is necessary to detect the specific type of network attack and regulate the unstable operating conditions of the system caused by different attack types.

[0005] In summary, the existing technologies for converter system network attack detection methods can be roughly divided into the following three categories, and each of them has the following defects:

[0006] (1) Model-driven approach: Mathematical models are usually based on some assumptions that may not fully match the real-world situation, leading to model errors, especially when dealing with nonlinear, multivariate, and high-dimensional problems. At the same time, mathematical models are usually difficult to deal with noise and uncertainty, which may lead to model instability.

[0007] (2) Data-driven methods: When applying data-driven methods to time series-based detection problems, the real-time nature of detection and the accuracy of detection should be coordinated. Some studies ignore the high computing power and time consumption of complex models in order to obtain high accuracy, which is not applicable to engineering problems in emergency situations.

[0008] (3) Network attack detection strategy: The current existing methods generally complete the two detection steps (detecting whether a network attack has occurred and detecting the specific type of network attack) through a machine learning or deep learning algorithm; if the algorithm model is relatively complex or requires the input of a large amount of real-time detection data, there will be a lag in obtaining the detection results, which may have caused difficult-to-repair vulnerabilities in system security, and is not suitable for emergency situations and critical infrastructure. The more data different machine learning algorithms input, the more complex the model structure, and the longer the calculation time. It is worth noting that in the existing inventions, only the detection accuracy is focused on, without considering the impact of detection time. Summary of the invention

[0009] 1. Technical issues to be resolved

[0010] Based on the above problems, the present invention provides a method and system for detecting network attacks on a converter system based on electrical signals, so as to solve the problem that it is difficult to balance the real-time detection and the detection accuracy in the existing network attack detection.

[0011] (II) Technical solution

[0012] Based on the above technical problems, the present invention provides a method for detecting network attacks on a converter system based on electrical signals, comprising:

[0013] S1. Obtaining a historical steady-state data set and a historical network attack data set: the historical steady-state data set includes a three-phase current signal and a three-phase voltage signal of a converter system under a steady-state condition, and the historical network attack data set includes a three-phase current signal and a three-phase voltage signal of a converter system, and corresponding network attack types;

[0014] S2. Determine a steady-state credible boundary: calculate an influence factor according to the three-phase current signal and the three-phase voltage signal, calculate a credible factor according to the influence factor, and obtain a steady-state credible boundary based on the minimum value of the periodic regular change of the credible factor calculated according to the data set under the steady-state working condition;

[0015] S3. Train to obtain a network attack type detection model: take each sampling point of the time series of the three-phase voltage signal and the three-phase current signal of the historical network attack data set as a feature, select effective features to form a training data set through the grey correlation method, input the training data set into the machine learning model for training, and obtain a network attack type detection model;

[0016] S4. Real-time monitoring of network attacks and determination of network attack types: real-time monitoring and collection of the three-phase current signal and the three-phase voltage signal of the converter system, real-time calculation of the trust factor, when the trust factor calculated in real time meets the network attack threshold requirement set according to the steady-state trust boundary, it is determined to be under network attack, the server is disconnected from the attack source, and the type of network attack is determined in real time according to the network attack type detection model.

[0017] Further, in S2, the calculating of the influence factor according to the three-phase current signal and the three-phase voltage signal, and the calculating of the credibility factor according to the influence factor include:

[0018] S21, calculating the influence factors of the three-phase voltage according to the three-phase voltage signal, including the influence factor of the A-phase voltage, the influence factor of the B-phase voltage, the influence factor of the C-phase voltage and the average influence factor of the three-phase voltage, that is, V af 、V bf 、Vcf and V sf :

[0019]

[0020] In the formula, i represents the current moment, i-1 represents the previous moment; Va i , Vb i and Vc i Respectively represent the current phase A voltage, phase B voltage and phase C voltage; Va i-1 , Vb i-1 and Vc i-1 They represent the A-phase voltage, B-phase voltage and C-phase voltage at the previous moment respectively; α is the weight coefficient of the voltage fluctuation amplitude; β is the weight coefficient of the voltage distortion degree.

[0021] S22, calculating the influence factors of the three-phase current according to the three-phase current signal, including the influence factor of the A-phase current, the influence factor of the B-phase current, the influence factor of the C-phase current and the average influence factor of the three-phase current, that is, I af ,I bf ,I cf and I sf :

[0022]

[0023] In the formula, Ia i , Ib i and Ic i Respectively represent the current A phase current, B phase current and C phase current at the current moment; Ia i-1 , Ib i-1 and Ic i-1 Respectively represent the A phase current, B phase current and C phase current at the previous moment; I af ,I bf ,I cf and I sf They represent the influence factor of phase A current, phase B current, phase C current and the average influence factor of the three-phase current respectively; α is also the weight coefficient of the current fluctuation amplitude; β is also the weight coefficient of the current distortion degree.

[0024] S23, introduce the sigmod function, and calculate the credible factor Y_re according to the three-phase voltage average influence factor and the three-phase current average influence factor:

[0025]

[0026] Furthermore, the α is 0.6, and the β is 0.4.

[0027] Furthermore, S3 includes:

[0028] S31, taking each sampling point of the time series of the three-phase voltage signal and the three-phase current signal of the historical network attack data set as a feature, analyzing the correlation coefficient between each feature and different network attack types by using the grey correlation method;

[0029] S32, forming a training data set from features corresponding to correlation coefficients greater than the average correlation coefficient;

[0030] S33. Input the training data set into the machine learning model for training to obtain a network attack type detection model.

[0031] Furthermore, the grey correlation method is used to analyze the correlation coefficient between each feature and different network attack types, including: taking the label data corresponding to different attack situations as the reference sequence Y, selecting the electrical signal time series within a part of the network attack time as the comparison sequence X m , calculate the correlation coefficient between the reference sequence and the comparison sequence:

[0032]

[0033] Where j = 1, 2, ... n g , n g Indicates the total number of data for each factor; m g is the number of factors; min (j) and ζ max (j) is the minimum and maximum difference between sequences; γ represents X m is the correlation coefficient with Y; ρ is the resolution coefficient.

[0034] Furthermore, the ρ is 0.5.

[0035] Furthermore, the machine learning model includes a support vector machine model.

[0036] Furthermore, in S4, when the trust factor calculated in real time meets the network attack threshold requirement set according to the steady-state trust boundary, it is: when the number of times that the trust factor calculated in real time is lower than the steady-state trust boundary exceeds the set threshold.

[0037] The present invention also discloses a converter system network attack detection system based on electrical signals, comprising:

[0038] at least one processor; and at least one memory in communication with the processor, wherein:

[0039] The memory stores program instructions that can be executed by the processor, and the processor can execute the method by calling the program instructions.

[0040] The present invention also discloses a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, wherein the computer instructions enable the computer to execute the method.

[0041] (III) Beneficial effects

[0042] The above technical solution of the present invention has the following advantages:

[0043] (1) The present invention first determines the steady-state credible boundary based on the data set under the historical steady-state working conditions, screens the effective features based on the historical network attack data set and the grey correlation method, and trains to obtain a network attack type detection model; then calculates the credible factor in real time through the real-time monitored electrical signal, quickly determines whether a network attack has occurred through the steady-state credible boundary, and quickly and accurately determines the type of network attack based on the network attack type detection model, so as to perform regulation and repair; the detection of whether a network attack has occurred is separated from the detection of the specific network attack type, thereby simplifying the model, taking into account both the real-time detection and the detection accuracy, and being applicable to power electronic systems equipped with converters, and having generalization to network attacks;

[0044] (2) When the present invention detects a network attack, it disconnects the server from the attack source to avoid deep damage to the converter system;

[0045] (3) The network attack type sample size of the present invention is small, and the sample set is high-dimensional data based on time series. The support vector machine model is used to solve the problem of small sample data, and the kernel function method is used to solve the problem of high dimensionality of data, taking into account the detection accuracy while reducing the detection time;

[0046] (4) The present invention effectively screens effective features through the grey correlation method, reduces feature redundancy, can effectively reduce the calculation amount of the model, and further reduce the detection time. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:

[0048] Figure 1 Flow chart of a method for detecting network attacks on a converter system based on electrical signals according to an embodiment of the present invention;

[0049] Figure 2 A flowchart of determining a steady-state trust boundary according to an embodiment of the present invention;

[0050] Figure 3 A three-phase voltage signal diagram, a three-phase current signal diagram and a credible factor diagram according to an embodiment of the present invention;

[0051] Figure 4A flow chart of a network attack type detection model obtained through training according to an embodiment of the present invention;

[0052] Figure 5 A schematic diagram of feature extraction of voltage signals and current signals according to an embodiment of the present invention;

[0053] Figure 6 The present invention is a flowchart of an embodiment of the present invention for real-time monitoring of network attacks and determining the type of network attacks. DETAILED DESCRIPTION

[0054] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0055] The overall process of the converter system network attack detection method based on electrical signals of the present invention is shown in FIG. Figure 1 As shown, the specific steps are as follows:

[0056] S1. Obtain historical steady-state data sets and historical network attack data sets;

[0057] The data set under the historical steady-state working condition includes the three-phase current signal and the three-phase voltage signal of the converter system under the steady-state working condition, and the historical network attack data set includes the three-phase current signal and the three-phase voltage signal of the converter system, and the corresponding network attack type;

[0058] The data set under steady-state conditions is used to design the steady-state boundary for judging network attacks. At the same time, the historical network attack data set will be recorded for training the detection model to detect the type of network attacks. There is no requirement for the order of S2 and S3, which can be performed simultaneously or not.

[0059] S2. Determine a steady-state credible boundary: calculate an influence factor according to the three-phase current signal and the three-phase voltage signal, calculate a credible factor according to the influence factor, and obtain a steady-state credible boundary based on the minimum value of the periodic regular change of the credible factor calculated according to the data set under the steady-state working condition;

[0060] The data set under steady-state conditions is used to design a steady-state boundary to judge network attacks. Under stable operating conditions, a steady-state trust boundary can be obtained. The steady-state trust boundary calculation process is as follows: Figure 2 As shown, including:

[0061] S21, calculating the influence factors of the three-phase voltage according to the three-phase voltage signal, including the influence factor of the A-phase voltage, the influence factor of the B-phase voltage, the influence factor of the C-phase voltage and the average influence factor of the three-phase voltage;

[0062] The influence factors of the three-phase voltage are calculated, and the proportion of the voltage difference between the current moment and the previous moment in the voltage at the previous moment is represented as the voltage fluctuation amplitude, and the proportion of the voltage difference between the current moment and the steady-state mean voltage in the steady-state mean voltage is represented as the voltage distortion degree, as shown in equations (1)-(4):

[0063]

[0064] In the formula, i represents the current moment, i-1 represents the previous moment; Va i , Vb i and Vc i Respectively represent the current phase A voltage, phase B voltage and phase C voltage; Va i-1 , Vb i-1 and Vc i-1 Respectively represent the A phase voltage, B phase voltage and C phase voltage at the previous moment; V af 、V bf 、V cf and V sf They represent the A-phase voltage influence factor, B-phase voltage influence factor, C-phase voltage influence factor and the three-phase voltage average influence factor respectively; α is the weight coefficient of the voltage fluctuation amplitude; β is the weight coefficient of the voltage distortion degree.

[0065] S22, calculating the influence factors of the three-phase current according to the three-phase current signal, including the influence factor of the A-phase current, the influence factor of the B-phase current, the influence factor of the C-phase current and the average influence factor of the three-phase current;

[0066] Similarly, the influencing factors of the three-phase current are calculated, and the proportion of the current difference between the current moment and the previous moment in the current value at the previous moment is represented as the current fluctuation amplitude, and the proportion of the difference between the current moment and the steady-state mean current in the steady-state mean current is represented as the current distortion degree, as shown in equations (5)-(8):

[0067]

[0068]

[0069] In the formula, Ia i , Ib i and Ic i Respectively represent the current A phase current, B phase current and C phase current at the current moment; Ia i-1 , Ib i-1 and Ic i-1 Respectively represent the A phase current, B phase current and C phase current at the previous moment; I af ,I bf ,I cf and I sfThey represent the influence factor of phase A current, phase B current, phase C current and the average influence factor of the three-phase current respectively; α is also the weight coefficient of the current fluctuation amplitude; β is also the weight coefficient of the current distortion degree.

[0070] S23, introducing a sigmod function, and calculating a credible factor according to the three-phase voltage average influence factor and the three-phase current average influence factor;

[0071] In order to obtain a stable boundary value and quickly detect network attacks, the sigmoid function is introduced to avoid jumping output values, as shown in formula (9):

[0072]

[0073] In the formula, Y_re is the credible factor.

[0074] S24, obtaining a steady-state credible boundary according to a minimum value of a periodic regular change of the credible factor in a steady state;

[0075] Figure 3 (a) Graph of three-phase voltage signal of Example 1, Figure 3 Figure (b) shows the three-phase current signal diagram of Example 1. Figure 3 Figure (c) shows the trust factor graph of Example 1 in steady state and during network attack. It can be seen from Example 1 that in the simulated converter system, even in steady state, the calculated trust factor is not fixed, but changes periodically and has a minimum value, which is 0.89 in this experiment. Therefore, this value can be used as the steady-state trust boundary. In Example 1, a false data injection attack is carried out on the converter system at the first second, and the weight coefficient α of the current and current fluctuation amplitude is 0.6; the weight coefficient β of the voltage and current distortion degree is 0.4. In Example 1, at 1.0124s, the trust factor calculated in real time mutates, and the trust factor at this time is 0.83, which is less than the steady-state trust boundary, and remains below the steady-state trust boundary value for a period of time.

[0076] S3. Train to obtain a network attack type detection model: take each sampling point of the time series of the three-phase voltage signal and the three-phase current signal of the historical network attack data set as a feature, select effective features to form a training data set through the grey correlation method, input the training data set into the machine learning model for training, and obtain a network attack type detection model; Figure 4 As shown, specifically including:

[0077] S31, taking each sampling point of the time series of the three-phase voltage signal and the three-phase current signal of the historical network attack data set as a feature, analyzing the correlation coefficient between each feature and different network attack types by using the grey correlation method;

[0078] The present invention selects each sampling point of the time series of the three-phase voltage signal and the three-phase current signal during the network attack as a feature, and analyzes the degree of association between each feature and different types of network attacks by the grey correlation (GRA) method. GRA is a method that uses grey correlation to describe the degree of influence between factors. The feature sequence in the present invention is a time-based electrical signal sequence, which is affected by objective factors such as equipment operating conditions and subjective factors of hacker network attacks, presents partially unknown grey characteristics, and is suitable for the use of GRA. Therefore, the label data corresponding to different attack situations is used as the reference sequence Y, and the electrical signal time series within part of the network attack time is selected as the comparison sequence X m , calculate the correlation coefficient between the reference sequence and the comparison sequence and its average value Select Characterized by Y and X m Relevance:

[0079]

[0080] Where j = 1, 2, ... n g , n g Indicates the total number of data for each factor; m g is the number of factors; min (j) and ζ max (j) is the minimum and maximum difference between sequences; γ represents X m is the correlation coefficient with Y; ρ is the resolution coefficient.

[0081] S32, forming a training data set from features corresponding to correlation coefficients greater than the average correlation coefficient;

[0082] Figure 5 Figure (a) shows the feature extraction of voltage signal. Figure 5 Figure (b) shows the feature extraction of the current signal. In Example 2, the resolution coefficient ρ is taken as 0.5, and the effective features within 0.2s of the voltage signal and the current signal are extracted respectively. In Figure (a), the mean correlation degree calculated for the voltage signal is 0.68, and the features greater than the mean are screened out as the data set for SVM model training. 88 features are selected from 200 features. In Figure (b), the mean correlation degree calculated for the current signal is 0.69, and 93 features greater than the mean are screened out as the training data set. Therefore, the gray correlation method proposed in the present invention can effectively screen effective features, reduce feature redundancy, and effectively reduce the amount of calculation of the model.

[0083] S33, inputting the training data set into a machine learning model for training to obtain a network attack type detection model;

[0084] In this embodiment, the machine learning model is a support vector machine model (SVM). Considering that data on network attack types is not easy to collect, in order to obtain sufficiently high accuracy in a small sample data set, a support vector machine model (SVM) is constructed as a detection model. The SVM model can now be applied to solve multi-classification problems. The SVM multi-classification method constructs N(N-1) / 2 hyperplanes for N types of problems, where each hyperplane is trained using only two types of data sets, that is, solving the following quadratic programming problem:

[0085]

[0086] In the formula, is the normal vector of the hyperplane; ε is the slack variable; τ is the penalty factor, which determines the classification accuracy; x is the feature vector; y is the classification result; real number is the threshold; φ s represents a nonlinear spatial mapping; the subscript t represents the index of the sample in the union of class i and class j, and t = 1, 2...sn, sn is the spatial dimension. The decision function of the two-class SVM is:

[0087]

[0088] Where λ is the Lagrange multiplier and 0≤λ t ≤τ;K h is the kernel function; the subscript new represents new data. In high-dimensional feature space, the radial basis kernel function can maintain sensitivity to local features of the data, and the value of the kernel function changes smoothly between different samples, which can provide a relatively consistent feature mapping. Therefore, the K used in the present invention h is the radial basis kernel function:

[0089]

[0090] Where ω is the radial basis width. The complete calculation process of this process is as follows Figure 5 shown.

[0091] Compared with other machine learning algorithms, the random sampling and feature selection of the random forest algorithm lead to unstable recognition results, the BP neural network algorithm has a slow learning speed, the recognition process is not real-time, and the recognition accuracy is easily affected by the training samples. Considering that the sample size of network attack types is small and the sample set is high-dimensional data based on time series, the SVM classifier is suitable for small sample data and has strong generalization ability. At the same time, it solves the problem of high dimensionality of data by using the kernel function method. In addition, the time complexity of the support vector machine depends on the type and number of features of the network attack type. Under limited network attack types, the detection time can be effectively reduced by reducing the input of feature data. Other machine learning models can also be used, but the effect of using SVM is better.

[0092] S4, real-time monitoring of network attacks and determination of network attack types: real-time monitoring and collection of the three-phase current signal and three-phase voltage signal of the converter system, real-time calculation of the trust factor, when the trust factor calculated in real time meets the network attack threshold requirement set according to the steady-state trust boundary, it is determined to be a network attack, the server is disconnected from the attack source, and the type of network attack is determined in real time according to the network attack type detection model. Figure 6 shown.

[0093] The phasor measurement unit (PMU) is used to monitor and collect the three-phase current and three-phase voltage signals of the converter system in real time, with a sampling frequency of 10k~20k.

[0094] According to the method of S21-S23, a trust factor calculated in real time is obtained. When a network attack occurs, the oscillating changes of the three-phase voltage and the three-phase current make the trust factor calculated in real time less than the steady-state trust boundary. In this case, it is considered that a network attack may occur. In order to avoid fluctuations caused by interference, the trust factor value can be continuously detected, and the number of times it is lower than the boundary value is accumulated. When the set threshold is reached, it is determined to be a network attack. That is, when the number of times the trust factor calculated in real time is lower than the steady-state trust boundary exceeds the set threshold, it is determined to be a network attack, the server is disconnected from the attack source, and the specific type of network attack is determined in real time according to the network attack type detection model.

[0095] Finally, it should be noted that the above method can be converted into software program instructions, which can be implemented by using a system including a processor and a memory, or by computer instructions stored in a non-transitory computer-readable storage medium. The above integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above software functional unit is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

[0096] In summary, the above-mentioned method and system for detecting network attacks on a converter system based on electrical signals have the following beneficial effects:

[0097] (1) The present invention first determines the steady-state credible boundary based on the data set under the historical steady-state working conditions, screens the effective features based on the historical network attack data set and the grey correlation method, and trains to obtain a network attack type detection model; then calculates the credible factor in real time through the real-time monitored electrical signal, quickly determines whether a network attack has occurred through the steady-state credible boundary, and quickly and accurately determines the type of network attack based on the network attack type detection model, so as to perform regulation and repair; the detection of whether a network attack has occurred is separated from the detection of the specific network attack type, thereby simplifying the model, taking into account both the real-time detection and the detection accuracy, and being applicable to power electronic systems equipped with converters, and having generalization to network attacks;

[0098] (2) When the present invention detects a network attack, it disconnects the server from the attack source to avoid deep damage to the converter system;

[0099] (3) The network attack type sample size of the present invention is small, and the sample set is high-dimensional data based on time series. The support vector machine model is used to solve the problem of small sample data, and the kernel function method is used to solve the problem of high dimensionality of data, taking into account the detection accuracy while reducing the detection time;

[0100] (4) The present invention effectively screens effective features through the grey correlation method, reduces feature redundancy, can effectively reduce the calculation amount of the model, and further reduce the detection time.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Although the embodiments of the present invention are described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations shall fall within the scope defined by the appended claims.

Claims

1. A method for detecting network attacks on a converter system based on electrical signals, characterized in that: include: S1. Obtaining a historical steady-state data set and a historical network attack data set: the historical steady-state data set includes a three-phase current signal and a three-phase voltage signal of a converter system under a steady-state condition, and the historical network attack data set includes a three-phase current signal and a three-phase voltage signal of a converter system, and corresponding network attack types; S2. Determine a steady-state credible boundary: calculate an influence factor according to the three-phase current signal and the three-phase voltage signal, calculate a credible factor according to the influence factor, and obtain a steady-state credible boundary based on the minimum value of the periodic regular change of the credible factor calculated according to the data set under the steady-state working condition; S3. Train to obtain a network attack type detection model: take each sampling point of the time series of the three-phase voltage signal and the three-phase current signal of the historical network attack data set as a feature, select effective features to form a training data set through the grey correlation method, input the training data set into the machine learning model for training, and obtain a network attack type detection model; S4. Real-time monitoring of network attacks and determination of network attack types: real-time monitoring and collection of the three-phase current signal and the three-phase voltage signal of the converter system, calculation of the trust factor, and when the trust factor calculated in real time meets the network attack threshold requirement set according to the steady-state trust boundary, it is determined to be a network attack, the server is disconnected from the attack source, and the type of network attack is determined in real time according to the network attack type detection model.

2. The method for detecting network attacks on a converter system based on electrical signals according to claim 1, characterized in that: In S2, the calculating of the influence factor according to the three-phase current signal and the three-phase voltage signal, and the calculating of the credibility factor according to the influence factor include: S21, calculating the influence factors of the three-phase voltage according to the three-phase voltage signal, including the influence factor of the A-phase voltage, the influence factor of the B-phase voltage, the influence factor of the C-phase voltage and the average influence factor of the three-phase voltage, that is, V af 、V bf 、V cf and V sf : In the formula, i represents the current moment, i-1 represents the previous moment; Va i , Vb i and Vc i Respectively represent the current phase A voltage, phase B voltage and phase C voltage; Va i-1 , Vb i-1 and Vc i-1 They represent the A-phase voltage, B-phase voltage and C-phase voltage at the previous moment respectively; α is the weight coefficient of the voltage fluctuation amplitude; β is the weight coefficient of the voltage distortion degree. S22, calculating the influence factors of the three-phase current according to the three-phase current signal, including the influence factor of the A-phase current, the influence factor of the B-phase current, the influence factor of the C-phase current and the average influence factor of the three-phase current, that is, I af ,I bf ,I cf and I sf : In the formula, Ia i , Ib i and Ic i Respectively represent the current A phase current, B phase current and C phase current at the current moment; Ia i-1 , Ib i-1 and Ic i-1 Respectively represent the A phase current, B phase current and C phase current at the previous moment; I af ,I bf ,I cf and I sf They represent the influence factor of phase A current, phase B current, phase C current and the average influence factor of the three-phase current respectively; α is also the weight coefficient of the current fluctuation amplitude; β is also the weight coefficient of the current distortion degree. S23, introduce the sigmod function, and calculate the credible factor Y_re according to the three-phase voltage average influence factor and the three-phase current average influence factor:

3. The method for detecting network attacks on a converter system based on electrical signals according to claim 2, characterized in that: The α is 0.6, and the β is 0.

4.

4. The method for detecting network attacks on a converter system based on electrical signals according to claim 1, characterized in that: The S3 includes: S31, taking each sampling point of the time series of the three-phase voltage signal and the three-phase current signal of the historical network attack data set as a feature, analyzing the correlation coefficient between each feature and different network attack types by using the grey correlation method; S32, forming a training data set from features corresponding to correlation coefficients greater than the average correlation coefficient; S33. Input the training data set into the machine learning model for training to obtain a network attack type detection model.

5. The method for detecting network attacks on a converter system based on electrical signals according to claim 4, characterized in that: The grey correlation method is used to analyze the correlation coefficient between each feature and different network attack types, including: taking the label data corresponding to different attack situations as the reference sequence Y, selecting the electrical signal time series within a part of the network attack time as the comparison sequence X m , calculate the correlation coefficient between the reference sequence and the comparison sequence: Where j = 1, 2, ..., n g Indicates the total number of data for each factor; m g is the number of factors; min (j) and ζ max (j) is the minimum and maximum difference between sequences; γ represents X m is the correlation coefficient with Y; ρ is the resolution coefficient.

6. The method for detecting network attacks on a converter system based on electrical signals according to claim 5, characterized in that: The ρ is 0.

5.

7. The method for detecting network attacks on a converter system based on electrical signals according to claim 4, characterized in that: The machine learning model includes a support vector machine model.

8. The method for detecting network attacks on a converter system based on electrical signals according to claim 1, characterized in that: In S4, when the trust factor calculated in real time meets the network attack threshold requirement set according to the steady-state trust boundary, it is: when the number of times that the trust factor calculated in real time is lower than the steady-state trust boundary exceeds the set threshold.

9. A converter system network attack detection system based on electrical signals, characterized in that: include: at least one processor; and at least one memory in communication with the processor, wherein: The memory stores program instructions executable by the processor, and the processor can execute the method according to any one of claims 1 to 8 by calling the program instructions.

10. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions that cause the computer to execute the method according to any one of claims 1 to 8.

Citation Information

Patent Citations

  • Attack recognition method based on random forest algorithm and energy storage coordination control device

    CN111107092A

  • Efficiency-optimized PMSM current prediction control method and system

    CN111211716A

  • LDoS attack detection method based on PSO-K algorithm

    CN112261000A

  • VSG transient stability enhancement method for frequency stability improvement

    CN117411030A

  • Power distribution network security assessment method and system considering network attack influence

    CN118041655A

Cited By

  • DC voltage control link network attack detection method, DC voltage control link network attack defending method and DC voltage control link network attack detection device

    CN121309036A