A vulnerability analysis method and device for abnormal load signals of electric vehicle power system

Abstract

A vulnerability analysis method and device for abnormal load signals of an electric vehicle power system, the method comprising: using a vector control strategy and an MTPA control strategy of a permanent magnet synchronous motor to perform simulink modeling on a single-wheel power system of an electric vehicle, establishing a load attack model in a basic square wave form, using a sliding window to perform calculation analysis on a torque value within an attack time, obtaining an index that can reflect the vulnerability of the electric vehicle power system under the current load signal attack, using a genetic algorithm to iteratively generate optimal load signal parameters, inputting a square wave load signal into a system through a load, performing quantitative analysis on the vulnerability of the electric vehicle power system, and using a parallel computing strategy to simultaneously run the simulink model on different cores and different computer clusters to achieve the purpose of acceleration, which has important application value.

Description

Technical Field

[0001] The present invention relates to the field of electric vehicle power system control and network physical security, and in particular to a vulnerability analysis method and device for abnormal load signals of an electric vehicle power system. Background Art

[0002] In recent years, the Internet of Things has developed rapidly, especially in the automotive field. With the increasing popularity of smart connected electric vehicles, their security issues have also received widespread attention, such as on-board network security, data security, and software security. The security of electric vehicle power systems has an additional layer of threat. Once an attacker destroys the control of the electric vehicle power system without being discovered, catastrophic damage is inevitable.

[0003] At present, the research on the security threats of electric vehicle power systems mainly focuses on motor drive and four-wheel drive vehicle systems. Three types of cyber-physical attacks are considered from the aspects of motor current feedback and four-wheel drive lateral stability control system parameters, namely replay attack, false signal injection attack and denial of service attack. The load has the most direct impact on the robustness of electric vehicle power systems, which has not been considered yet. For example, when a vehicle is driving on the road and encounters a speed bump, it will generate an excitation on the power motor, affecting the motor's operating performance. Once it is attacked by a network, the power system will be hijacked, and then a load state that is not conducive to the stability of the vehicle will be constructed, which will have unpredictable consequences.

[0004] Most electric vehicle drive systems use permanent magnet synchronous motors (PMSMs) because of their high power density, high torque density, and high efficiency, making them widely used in electric vehicles, household appliances, and other applications that require high efficiency. Field-Oriented Control (FOC) is mainly used, which is similar to the dual closed-loop control system of DC motors to achieve decoupling. Compared with surface-mounted permanent magnet synchronous motors, interior PMSMs (IPMSMs) have asymmetric cross-axis and direct-axis magnetic circuits, so permanent magnet torque and reluctance torque exist in the electromagnetic torque at the same time. In order to improve the working efficiency of the motor, it is necessary to make full use of its reluctance torque. The maximum torque per ampere (MTPA) control strategy is generally adopted because the MTPA working point is very close to the maximum efficiency point of the motor.

[0005] Random optimization algorithms are usually used to solve optimization problems with large search spaces and high complexity. These algorithms avoid falling into local optimal solutions by introducing randomness and explore a wider solution space during the search process. Common random optimization algorithms include: stochastic gradient descent, genetic algorithm, ant colony algorithm and particle swarm algorithm, etc. Each algorithm has its own unique advantages and limitations. For example, the stochastic gradient descent algorithm has low computational cost and is suitable for large-scale data and parameter space, but it is sensitive to learning rate and initial conditions; the genetic algorithm is suitable for multimodal optimization, complex constraints and high-dimensional space problems, and has strong global search capabilities, but is sensitive to parameter selection and encoding methods; the ant colony algorithm has good adaptability to discrete combinatorial optimization problems, can find better solutions and has certain robustness, but it performs poorly for optimization problems in continuous space, and the algorithm parameter settings and the choice of pheromone update strategy have a greater impact on the algorithm performance; the particle swarm algorithm is insensitive to initial conditions and converges quickly, which is suitable for continuous optimization and dynamic optimization problems, but the convergence speed may slow down in high-dimensional space. Summary of the invention

[0006] In order to overcome the above-mentioned shortcomings of the prior art and considering the new threat of network attacks to electric vehicle power systems in the era of the Internet of Things, the present invention proposes a vulnerability analysis method and device for abnormal load signals of electric vehicle power systems.

[0007] The present invention adopts the vector control strategy and MTPA control strategy of the permanent magnet synchronous motor, models the single-wheel power system of the electric vehicle, establishes a load attack model, establishes indicators that can reflect the vulnerability of the electric vehicle power system, iterates the most aggressive load signal parameters through a random optimization algorithm, and analyzes the vulnerability of the electric vehicle power system, which has important application value.

[0008] In order to achieve the above object, the technical solution adopted by the present invention is:

[0009] The first aspect of the present invention relates to a vulnerability analysis method for abnormal load signals of an electric vehicle power system. The vulnerability analysis method for abnormal load signals of an electric vehicle power system adopts the MTPA control strategy of a permanent magnet synchronous motor for modeling, establishes a load attack model, generates abnormal load signals, and optimizes and iterates attack model parameters in combination with a random optimization algorithm; and comprises the following steps:

[0010] The first step is to establish the attacked model, adopt the vector control strategy and MTPA control strategy of the permanent magnet synchronous motor, and perform simulink modeling on the single-wheel power system of the electric vehicle;

[0011] The second step is to establish a suitable load attack model and conduct vulnerability analysis on the system;

[0012] The third step is to determine the evaluation criteria, use the sliding window to calculate and analyze the torque value within the attack time, and find out the index that can reflect the vulnerability of the electric vehicle power system under the current load signal attack;

[0013] The fourth step is to use the random signal optimization algorithm to iteratively optimize the parameters of the attack model. Through the evaluation indicators determined in the third step, the parameters of the attack model are continuously optimized to find the attack signal that can maximize the impact on the performance of the electric vehicle power system and analyze the vulnerability of the system.

[0014] The fifth step is to use parallel computing strategies to run the Simulink model on different cores and different computer clusters at the same time to achieve the purpose of acceleration.

[0015] The first step is to use the vector control strategy and MTPA control strategy of the permanent magnet synchronous motor to perform simulink modeling on the single-wheel power system of the electric vehicle, including:

[0016] The motor control adopts vector control, and tries to simulate the torque control law of DC motor on the permanent magnet synchronous motor. After coordinate transformation, the current vector is decomposed into the current component that generates magnetic flux and the current component that generates torque. The two components are perpendicular and independent of each other. They are adjusted separately, similar to the double closed-loop control system of DC motor, to achieve decoupling.

[0017] The whole process of vector control is: sampling the three-phase current of the motor to obtain the three-phase current i in the A, B, and C coordinate systems a 、i b 、i c , change i a 、i b 、i c After Clark transformation of the coordinate axis, the current i in the α and β coordinate systems is obtained α 、i β , change i α 、i β After Park transformation, the current i in the d and q coordinate systems is obtained d 、i q , calculate i d 、i q With its setting value i dref 、i qref The two errors are input into two PI controllers to obtain the output control voltage u d 、u q , perform an inverse Park transform to obtain u α 、u β , then input into SVPWM module for modulation, output PWM wave to control MOS tube switch of three-phase inverter and drive motor;

[0018] The MTPA control is realized by formula method. s Next, for i d 、i q The optimal current distribution can achieve the maximum electromagnetic output torque, use the minimum current to achieve the maximum output torque, reduce copper loss, and achieve the effect of high-efficiency control;

[0019] First, establish the steady-state voltage equations of the d and q axis motors:

[0020]

[0021] In the formula, u d 、u q 、i d 、i q are d-axis voltage and q-axis current respectively; L d , L q is the d and q axis inductance, R s is the stator resistance; ω e is the electrical angular velocity; is the flux amplitude of the permanent magnet;

[0022] The torque equation of permanent magnet synchronous motor is as follows:

[0023]

[0024] Where n p is the number of pole pairs of the motor;

[0025] From formula (2), we know that the electromagnetic torque of the motor consists of two parts. The first part is the permanent magnet torque, the second part is the reluctance torque; from the equation, we can see that when the motor quadrature axis inductance L q With direct axis inductance L d When they are not equal, the reluctance torque is not zero, but for i d 、i q Optimal current distribution can make full use of the reluctance torque and achieve maximum electromagnetic output torque;

[0026] dq axis current and i s The relationship is as follows:

[0027]

[0028] Where β is the current vector angle, i s is the current vector magnitude;

[0029] Substituting equation (3) into equation (2), we can obtain the electromagnetic torque T e , permanent magnet torque T r and reluctance torque T mWith the stator current i s The relationship between the current vector angle β is as follows:

[0030]

[0031] According to formula (4), the electromagnetic torque is a function of the current vector angle β. In the range of 90° to 180°, e (β) has only one extreme point, which is the maximum point, so T e The point where the β derivative is 0 is the MTPA point;

[0032] T e Derivatives about β yield:

[0033]

[0034] Solving for β MTPA The expression is:

[0035]

[0036] Substituting the obtained MTPA angle into equation (3), the optimal current distribution of the dq axis can be obtained, and the maximum electromagnetic torque can be achieved with the minimum current, thus realizing MTPA control. The above model strategy is finally modeled in Simulink.

[0037] The second step is to establish a suitable load attack model and conduct vulnerability analysis on the system, considering three types of network attacks, namely replay attack, false signal injection attack and denial of service attack;

[0038] Replay attack: refers to the attacker recording the original normal state data to conduct a network attack in order to deceive the operator into not taking action; in mathematical expression, replay attack is expressed as:

[0039]

[0040] In formula (7), Indicates the load data after the attack, T past Indicates the load data recorded when the system was in normal state in the past;

[0041] False signal injection attack: The attacker directly forges data or injects incorrect instructions into the system, which is expressed as:

[0042]

[0043] or

[0044]

[0045] In formula (8), T newrepresents the new attack data forged by the attacker. In formula (9), the data injection attack is designed by mixing the original value with the malicious factor. T(t) represents the original data, t a represents the attack time, ν and ε represent unknown signals generated by malicious modification, such as white noise, periodic function, periodic pulse injection, constant value, etc.;

[0046] Denial of service attack: refers to the attacker sending malicious messages or data at a very high frequency to destroy the original communication traffic. During the duration of the attack, the original correct signal cannot be transmitted normally, which can be expressed by the following formula:

[0047]

[0048] In formula (10), t start Indicates the moment the attack begins.

[0049] The third step is to determine the evaluation criteria, use the sliding window to calculate and analyze the torque value within the attack time, and find out the indicators that can reflect the vulnerability of the electric vehicle power system under the current load signal attack, including:

[0050] In order to obtain the transient process of the attack, the indicators are calculated in a sliding window. A sliding window is represented as The calculation formula of the vulnerability index fit is as follows:

[0051]

[0052]

[0053] Where T i Indicates the actual torque, T ave represents the average torque in a sliding window, t represents the simulation time, and t speed Indicates the end time of acceleration;

[0054] In the above formula, the variance s of the torque in a sliding window is first calculated, and the variance matrix S is obtained after the sliding ends. Then the elements of the variance matrix are summed, and then the square root is calculated. Finally, the fitness value fit is obtained by dividing it by the difference between the simulation time and the acceleration end time. This fitness value can reflect the fluctuation of the actual torque signal after the attack load signal is injected, and reflects the vulnerability of the system under the load network attack from a numerical perspective.

[0055] The fourth step uses a random signal optimization algorithm to iteratively optimize the parameters of the attack model, including:

[0056] The stochastic optimization algorithm is an algorithm for solving complex optimization problems, involving uncertainty and randomness; using randomness to explore the search space and gradually converge to the optimal solution; the algorithm uses the information obtained during the exploration process to guide subsequent searches to improve the efficiency of finding the best solution;

[0057] The following steps are included: first, a set of candidate solutions is randomly initialized, then the fitness value or objective function value of each candidate solution is calculated to evaluate the quality of the current solution, and then a part of the solutions are selected as candidate solutions according to the quality of the solution. The selection strategy can be proportional selection according to the size of the fitness value, or it can be selected based on probability; the selected candidate solutions are operated, including crossover, mutation, and neighborhood search operations; different algorithms have different operation modes, such as crossover and mutation operations in genetic algorithms, and temperature scheduling and state update in simulated annealing algorithms. According to the new solution obtained after the operation, the current solution set is updated, and the update method can be replacement, merging, or inserting a new solution; then determine whether the termination condition is met, usually including whether the maximum number of iterations is reached, the objective function value converges, and the change of the solution is less than a certain threshold. If the termination condition is not met, the evaluation, selection, operation, and update process will continue, and the next round of iteration will be performed. If the above conditions are met, the iteration ends; by introducing randomness through iterative updating of the solution set, the random optimization algorithm can search for better solutions in the solution space and avoid falling into the local optimal solution to a certain extent.

[0058] The parallel computing strategy described in step 5 is to run the Simulink model on different cores and different computer clusters at the same time to achieve the purpose of acceleration, including:

[0059] In order to solve the problem of long Simulink running time and slow speed during the iteration of the random optimization algorithm, make full use of multi-core processors and computing clusters to run multiple Simulink simulations in parallel; use the Parallel Computing Toolbox toolbox to execute applications on the worker (MATLAB computing engine) running locally to use the full processing power of the multi-core desktop, and the MATLAB Parallel Server toolbox can expand Simulink to run on clusters and clouds; note that when setting up the environment, you need to keep the MATLAB version of multiple computers consistent.

[0060] A second aspect of the present invention relates to a vulnerability analysis device for abnormal load signals of an electric vehicle power system, comprising a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement a vulnerability analysis method for abnormal load signals of an electric vehicle power system of the present invention.

[0061] A third aspect of the present invention relates to a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements a vulnerability analysis method for abnormal load signals of an electric vehicle power system according to the present invention.

[0062] The technical concept of the present invention is: in order to consider the new threat of network attacks to electric vehicle power systems in the era of the Internet of Things, the present invention proposes a vulnerability analysis method for abnormal load signals of electric vehicle power systems, adopts the vector control strategy and MTPA control strategy of permanent magnet synchronous motors, models the single-wheel power system of electric vehicles, establishes a load attack model, establishes indicators that can reflect the vulnerability of electric vehicle power systems, and iterates the most aggressive load signal parameters through a random optimization algorithm to analyze the vulnerability of electric vehicle power systems. The method is simple and efficient, has a novel perspective, and has important application value. Finally, the simulink parallel computing strategy is used to make full use of multi-core processors and computing clusters to run multiple simulink simulations in parallel, fully utilize computing resources, and greatly improve computing speed.

[0063] The beneficial effects of the present invention are mainly manifested in:

[0064] (1) Improving the safety of electric vehicle power systems. The present invention proposes a vulnerability analysis method for abnormal load signals of electric vehicle power systems, examining the threat of signal intrusion from the perspective of load. This method is novel and can effectively simulate cyber-physical attack scenarios that electric vehicles may encounter in reality, conduct vulnerability analysis, and help discover and repair possible security vulnerabilities in the design and control strategies of electric vehicle power systems.

[0065] (2) A square wave signal attack model was designed, and the indicators for evaluating load attacks were calculated in a sliding window. The parameters of the square wave signal attack model were iterated using a random optimization algorithm to achieve vulnerability analysis of the electric vehicle power system under optimal load attacks.

[0066] (3) Improve the efficiency of automobile companies in detecting road conditions, expand the detection range, and reduce detection costs. The present invention only needs to simulate real load signals to complete the test. For example, square waves are used to simulate the road conditions of sudden braking and immediate starting in reality. There is no need to build actual road conditions for vehicle testing, thereby reducing the detection costs of automobile companies. At the same time, the present invention can generate other load signals that do not exist in reality to simulate network load signal attacks, expand the detection range, and the method is simple and efficient, which improves detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 The present invention shows a method for analyzing the vulnerability of an abnormal load signal of an electric vehicle power system.

[0068] Figure 2 A structural block diagram of a vulnerability analysis method for abnormal load signals of an electric vehicle power system according to the present invention is shown.

[0069] Figure 3 The structural block diagram of the permanent magnet synchronous motor vector control and MTPA control of the present invention is shown.

[0070] Figure 4 The flowchart of the random optimization algorithm (genetic algorithm) used in the present invention is shown.

[0071] Figure 5 It shows the change of the fitness value corresponding to the attack signal of the present invention with the number of iterations.

[0072] Figure 6 The load attack signal and the actual torque signal diagram of the present invention are shown.

[0073] Figure 7 Multiple simulink parallel computing interface diagrams in the present invention are shown. DETAILED DESCRIPTION

[0074] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings.

[0075] Example 1

[0076] Reference Figure 1 to Figure 7 , a vulnerability analysis method for abnormal load signals of electric vehicle power systems,simulink modeling of single-wheel power systems of electric vehicles,establishing load attack models in the basic form of square waves,establishing indicators that can reflect the vulnerability of electric vehicle power systems under square wave load attacks,using genetic algorithms to iteratively generate optimal load signal parameters,inputting square wave load signals into the system,quantitatively analyzing the vulnerability of electric vehicle power systems,and finally using simulink parallel computing strategies to accelerate the operation process.

[0077] Reference Figure 1 , showing the overall workflow of the present invention.

[0078] S1. The vector control strategy and MTPA control strategy of permanent magnet synchronous motor are used to build simulink model of single-wheel power system of electric vehicle, which is the basis of simulation experiment.

[0079] S2. Establish a suitable load attack model and conduct vulnerability analysis on the system.

[0080] S3. Use the sliding window to calculate and analyze the torque value within the attack time, and find the index that can reflect the vulnerability of the electric vehicle power system under the current load signal attack.

[0081] S4. Use the random signal optimization algorithm to iteratively optimize the parameters of the attack model.

[0082] S5. Use parallel computing strategies to run Simulink models on different cores and different computer clusters at the same time to speed up the computing process.

[0083] Reference Figure 2 , shows the overall idea diagram of the present invention, which is mainly divided into two modules, namely the genetic algorithm code written in Matlab, which is responsible for generating the parameters of the square wave load signal, and the second module, the detected model built by simulink. After the execution of the second module, it will return the fitness value to the genetic algorithm module for the next iteration.

[0084] Reference Figure 3 , describes the implementation structure of the vector control and maximum torque current ratio control strategy of the permanent magnet synchronous motor. This control strategy optimizes the performance of the motor and ensures that the motor can achieve the highest efficiency under different working conditions. It is the control model currently used in electric vehicle power systems and is close to reality. It includes multiple modules such as Clark transformation, Park transformation, SVPWM modulation, MTPA, etc., and samples the three-phase current of the motor to obtain the current i in the A, B, and C coordinate systems. a 、i b 、i c , change i a 、i b 、i c The current i in the α and β coordinate systems is obtained through the coordinate axis Clark transformation module α 、i β , change i α 、i β The current i in the d and q coordinate systems is obtained through the Park transformation module d 、i q , calculate i d 、i q With its setting value i dref 、i qref The two errors are input into two PI controller modules to obtain the output control voltage u d 、u q , and perform the inverse Park transform module to obtain u α 、u β , and then input into the SVPWM module for modulation, and output PWM wave to control the MOS tube switch of the three-phase inverter to drive the motor.

[0085] The MTPA module first establishes the steady-state voltage equations of the d-axis and q-axis motors:

[0086]

[0087] In the formula, u d 、u q 、i d 、i q are d-axis voltage and q-axis current respectively; L d , L q is the d and q axis inductance, R s is the stator resistance; ω e is the electrical angular velocity; is the flux amplitude of the permanent magnet.

[0088] The torque equation of permanent magnet synchronous motor is as follows:

[0089]

[0090] In formula (14), n p is the number of pole pairs of the motor.

[0091] From formula (14), we know that the electromagnetic torque of the motor consists of two parts. The first part is the permanent magnet torque, the second part is the reluctance torque. From the equation, we can see that when the motor quadrature axis inductance L q With direct axis inductance L d When they are not equal, the reluctance torque is not zero, but for i d 、i q Optimal current distribution can make full use of the reluctance torque and achieve maximum electromagnetic output torque.

[0092] dq axis current and i s The relationship is as follows:

[0093]

[0094] Where β is the current vector angle, i s is the current vector magnitude.

[0095] Substituting equation (15) into equation (14), we can obtain the electromagnetic torque T e , permanent magnet torque T r and reluctance torque T m With the stator current i s and the relationship between the current vector angle β.

[0096]

[0097] According to formula (16), the electromagnetic torque is a function of the current vector angle β. In the range of 90° to 180°, e (β) has only one extreme point, which is the maximum point, so T e The point where the β derivative is 0 is the MTPA point.

[0098] T e Derivatives about β yield:

[0099]

[0100] Solving for β MTPA The expression is:

[0101]

[0102] Substituting the obtained MTPA angle into equation (15), the optimal current distribution of the dq axis can be obtained, and the maximum electromagnetic torque can be achieved with the minimum current, thus realizing MTPA control.

[0103] Reference Figure 4 , shows in detail the workflow of the random optimization algorithm. This implementation uses the genetic algorithm as the optimization algorithm for iteration, including initializing the population, evaluating fitness, selection, crossover, mutation, and the formation of a new generation of population. This flowchart provides a clear guide for understanding how to optimize the parameters of the square wave load signal through the genetic algorithm.

[0104] Genetic algorithm is a kind of random search algorithm that draws on the natural selection and natural genetic mechanism of the biological world. First, a group of initial populations are randomly generated, and a suitable fitness function is selected as the criterion for judging the quality of the population. Then, a new generation of candidate populations is generated through selection, crossover, and mutation. This process is repeated and iterated continuously to finally generate a solution with the optimal fitness value. In the present invention, the initial population is the duty cycle and period of ten square waves, and 42 groups of populations are initially set.

[0105] The fitness function is set according to the operating performance of the permanent magnet synchronous motor of the electric vehicle single-wheel power system. In order to obtain the transient process of the attack, all indicators are calculated within the sliding window and expressed as The fitness function fit calculation formula is as follows:

[0106]

[0107]

[0108] Where T i Indicates the actual torque, T ave represents the average torque in a sliding window, t represents the simulation time, and t speed Indicates the time when acceleration ends.

[0109] In the above formula, the variance s of the torque in a sliding window is calculated first, and the variance matrix S is obtained after the sliding is completed. Then the elements of the variance matrix are summed, and then the square root is calculated. Finally, the fitness value fit is obtained by dividing the difference between the simulation time and the acceleration end time. This fitness value can reflect the fluctuation of the actual torque signal after the attack load signal is injected.

[0110] The genetic algorithm process is as follows: first, encoding is performed to represent the solution data in the solution space as genotype string structure data in the genetic space. Different combinations of these string structure data constitute different solutions. Then, 42 groups of initial populations are generated in the currently encoded solution space. The genetic algorithm starts to evolve with this group of populations as the starting point, assigns the initial population to the attack load signal parameter, and then performs simulink simulation. After the simulation ends, the fitness value is returned. Then, according to the fitness of individuals in the population, individuals with high fitness are selected from the current population through a roulette wheel, that is, the individual fitness is mapped to probability for replication. Individuals with high fitness have a greater probability of replication and more copies.

[0111] Calculate the corresponding fitness f(x i ), then the probability that the individual is copied is p i It is expressed as:

[0112]

[0113] Crossover means crossing over between two individuals in order according to probability, and mutation means mutating each individual according to probability.

[0114] Reference Figure 5 , shows how the fitness value corresponding to the attack signal changes with the number of iterations during the iteration of the genetic algorithm, and the final convergence process. The horizontal axis is the number of iterations, and the vertical axis is the fitness value. The upper trajectory represents the average fitness value of this iteration, and the point below represents the maximum absolute value of the fitness of this iteration.

[0115] Reference Figure 6 , showing the actual torque response of the permanent magnet synchronous motor drive system after the load signal is input. This implementation case uses a square wave as the load signal injection detection, and generates the duty cycle and period of the square wave signal through the genetic algorithm described above. The square wave in the figure is the load attack signal. Finally, by comparing the attack load signal and the actual motor torque signal, it can be intuitively seen that the actual motor torque signal can exceed the load attack signal by up to 70%, reflecting a large degree of load fluctuation.

[0116] Reference Figure 7, showing the interface of using Simulink for parallel computing. Through the parallel computing strategy, forty simulation models can be run on different cores and computing nodes at the same time, which significantly speeds up the iteration and simulation process and improves research efficiency. As can be seen from the figure, it only takes seven minutes and thirty seconds to run forty simulations in one iteration. The steps to build a cluster environment are as follows: first create a cluster, then configure the scheduler (MATLAB Job Scheduler, MJS), start MJS, start the worker, and select the scheduler for the worker. Note that when building the environment, you need to keep the MATLAB version of multiple computers consistent.

[0117] Example 2

[0118] The present embodiment relates to a vulnerability analysis device for abnormal load signals of an electric vehicle power system, comprising a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement a vulnerability analysis method for abnormal load signals of an electric vehicle power system of embodiment 1.

[0119] Example 3

[0120] This embodiment relates to a computer-readable storage medium on which a program is stored. When the program is executed by a processor, a vulnerability analysis method for abnormal load signals of an electric vehicle power system according to embodiment 1 is implemented.

[0121] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms described in the embodiments. The protection scope of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A vulnerability analysis method for abnormal load signals of electric vehicle power systems, characterized in that: The MTPA control strategy of the permanent magnet synchronous motor is used for modeling, a load attack model is established, an abnormal load signal is generated, and the attack model parameters are optimized and iterated in combination with the random optimization algorithm; the following steps are included: The first step is to establish the attacked model, adopt the vector control strategy and MTPA control strategy of the permanent magnet synchronous motor, and perform simulink modeling on the single-wheel power system of the electric vehicle; The second step is to establish a suitable load attack model and conduct vulnerability analysis on the system; The third step is to determine the evaluation criteria, use the sliding window to calculate and analyze the torque value within the attack time, and find out the index that can reflect the vulnerability of the electric vehicle power system under the current load signal attack; specifically, it includes: In order to obtain the transient process of the attack, the indicators are calculated in a sliding window. The total amount of data in a sliding window is expressed as The calculation formula of the vulnerability index fit is as follows: Where T i Indicates the actual torque, T ave represents the average torque in a sliding window, t represents the simulation time, and t speed Indicates the end time of acceleration; First, the variance s of the torque in a sliding window is calculated, and the variance matrix S is obtained after the sliding is completed. Then, the elements of the variance matrix are summed, and then the square root is calculated. Finally, the vulnerability index fit is obtained by dividing the difference between the simulation time and the acceleration end time. This vulnerability index can reflect the fluctuation of the actual torque signal after the attack load signal is injected, and reflects the vulnerability of the system under the load network attack from a numerical perspective. The fourth step is to use the random signal optimization algorithm to iteratively optimize the parameters of the load attack model. Through the evaluation indicators determined in the third step, the parameters of the load attack model are continuously optimized to find the attack signal that can maximize the impact on the performance of the electric vehicle power system and analyze the vulnerability of the system. The fifth step is to use parallel computing strategies to run the Simulink model on different cores and different computer clusters at the same time to achieve the purpose of acceleration.

2. A vulnerability analysis method for abnormal load signals of electric vehicle power system according to claim 1, characterized in that: The first step is to use the vector control strategy and MTPA control strategy of the permanent magnet synchronous motor to perform simulink modeling on the single-wheel power system of the electric vehicle, including: The motor control adopts vector control, and tries to simulate the torque control law of DC motor on the permanent magnet synchronous motor. After coordinate transformation, the current vector is decomposed into the current component that generates magnetic flux and the current component that generates torque. The two components are perpendicular and independent to each other. They are adjusted separately to achieve decoupling. The whole process of vector control is: sampling the three-phase current of the motor to obtain the three-phase current i in the A, B, and C coordinate systems a 、i b 、i c , change i a 、i b 、i c After Clark transformation of the coordinate axis, the current i in the α and β coordinate systems is obtained α 、i β , change i α 、i β After Park transformation, the current i in the d and q coordinate systems is obtained d 、i q , calculate i d 、i q With its set value i dref 、i qref The two errors are input into two PI controllers to obtain the output control voltage u d 、u q , perform an inverse Park transform to obtain u α 、u β , then input into SVPWM module for modulation, output PWM wave to control MOS tube switch of three-phase inverter and drive motor; The MTPA control is realized by formula method. s Next, for i d 、i q The optimal current distribution can achieve the maximum electromagnetic output torque, use the minimum current to achieve the maximum output torque, reduce copper loss, and achieve the effect of high-efficiency control; First, establish the steady-state voltage equations of the d and q axis motors: In the formula, u d 、u q 、i d 、i q are d-axis voltage and q-axis current respectively; L d , L q is the d and q axis inductance, R s is the stator resistance; ω e is the electrical angular velocity; ψ f is the flux amplitude of the permanent magnet; The torque equation of permanent magnet synchronous motor is as follows: Where n p is the number of pole pairs of the motor; From formula (2), we know that the electromagnetic torque of the motor consists of two parts. The first part is the permanent magnet torque, the second part is the reluctance torque; from the equation, we can see that when the motor quadrature axis inductance L q With direct axis inductance L d When they are not equal, the reluctance torque is not zero, but for i d 、i q Optimal current distribution can make full use of the reluctance torque and achieve maximum electromagnetic output torque; dq axis current and i s The relationship is as follows: Where β is the current vector angle, i s is the current vector magnitude; Substituting equation (3) into equation (2), we can obtain the electromagnetic torque T e , permanent magnet torque T r and reluctance torque T m With the stator current i s The relationship between the current vector angle β is as follows: According to formula (4), the electromagnetic torque is a function of the current vector angle β. In the range of 90° to 180°, e (β) has only one extreme point, which is the maximum point, so T e The point where the β derivative is 0 is the MTPA point; T e Derivatives about β yield: Solving for β MTPA The expression is: Substituting the obtained MTPA angle into equation (3), the optimal current distribution of the dq axis can be obtained, and the maximum electromagnetic torque can be achieved with the minimum current, thus realizing MTPA control; the MTPA control strategy is finally modeled in Simulink.

3. A vulnerability analysis method for abnormal load signals of electric vehicle power system according to claim 1, characterized in that: The second step is to establish a suitable load attack model and conduct vulnerability analysis on the system, considering three types of network attacks, namely replay attack, false signal injection attack and denial of service attack; Replay attack: refers to the attacker recording the original normal state data to conduct a network attack in order to deceive the operator into not taking action; in mathematical expression, replay attack is expressed as: In formula (7), Indicates the load data after the attack, T past Indicates the load data recorded when the system was in normal state in the past; False signal injection attack: The attacker directly forges data or injects incorrect instructions into the system, which is expressed as: or In formula (8), T new represents the new attack data forged by the attacker. In formula (9), the data injection attack is designed by mixing the original value with the malicious factor. T(t) represents the original data, t a represents the attack time, ν and ε represent unknown signals generated by malicious modification; Denial of service attack: refers to the attacker sending malicious messages or data at a very high frequency to destroy the traffic conditions of the original communication; during the duration of the attack, the original correct signal cannot be transmitted normally, which can be expressed by the following formula: In formula (10), t start Indicates the moment the attack begins.

4. The vulnerability analysis method for abnormal load signal of electric vehicle power system according to claim 1, characterized in that: In the fourth step, the random signal optimization algorithm is used to iteratively optimize the parameters of the load attack model, including: The stochastic optimization algorithm is an algorithm for solving complex optimization problems, involving uncertainty and randomness; using randomness to explore the search space and gradually converge to the optimal solution; the algorithm uses the information obtained during the exploration process to guide subsequent searches to improve the efficiency of finding the best solution; The method includes the following steps: first, a group of candidate solutions are randomly initialized, then the fitness value or objective function value of each candidate solution is calculated to evaluate the quality of the current solution, and then a part of the solutions are selected as candidate solutions according to the quality of the solutions. The selection strategy is to select proportionally according to the size of the fitness value, or to select based on probability; the selected candidate solutions are operated, including crossover, mutation, and neighborhood search operations; different algorithms have different operation methods; according to the new solution obtained after the operation, the current solution set is updated, and the update method is to replace, merge or insert the new solution; then it is determined whether the termination condition is met, including whether the maximum number of iterations is reached, the objective function value converges, and the change of the solution is less than a certain threshold; if the termination condition is not met, the evaluation, selection, operation and update process continues, and the next round of iteration is performed. If the termination condition is met, the iteration ends; randomness is introduced by iteratively updating the solution set, and the random optimization algorithm can search for better solutions in the solution space and avoid falling into local optimal solutions to a certain extent.

5. The vulnerability analysis method for abnormal load signal of electric vehicle power system according to claim 1, characterized in that: The parallel computing strategy described in step 5 is to run the Simulink model on different cores and different computer clusters at the same time to achieve the purpose of acceleration, including: In order to solve the problem of long simulink running time and slow speed during the iteration of the random optimization algorithm, make full use of multi-core processors and computing clusters to run multiple simulink simulations in parallel; use the Parallel Computing Toolbox toolbox to execute applications on the worker running locally to use the full processing power of the multi-core desktop, and the MATLAB Parallel Server toolbox to extend Simulink to run on clusters and clouds; note that when setting up the environment, the MATLAB versions of multiple computers need to be kept consistent; the worker is the MATLAB computing engine.

6. A vulnerability analysis device for abnormal load signals of electric vehicle power systems, characterized in that: It comprises a memory and one or more processors, wherein the memory stores executable codes, and when the one or more processors execute the executable codes, they are used to implement a vulnerability analysis method for abnormal load signals of an electric vehicle power system as described in any one of claims 1-5.

7. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, a vulnerability analysis method for abnormal load signals of an electric vehicle power system as described in any one of claims 1-5 is implemented.

Images