Method for constructing large model of fractional order power system under network attack and security prediction

Abstract

The specification relates to the technical field of power system operation control, in particular to a method for constructing and safely predicting a fractional order power system large model under network attacks, comprising: constructing a fractional order first model of a power system according to characteristic sample data sets of each subsystem in the power system under a first state; determining a fractional order second model of the power system according to intermittent network attack modeling and characteristic sample data sets of each subsystem under a second state; constructing a fractional order multivariate regression model of the power system under the second state according to the characteristic sample data sets of each subsystem under the second state and the fractional order second model of the power system; determining the sensitivity of each subsystem in the power system to parameter perturbation and intermittent network attacks; and constructing a prediction model according to the sensitivity, error indicators of an initial prediction model of intermittent network attacks and an initial prediction model of parameter perturbation. The application improves the estimation accuracy of intermittent network attacks and parameter perturbation suffered by a fractional order power system.

Description

Technical Field

[0001] This manual relates to the field of power system operation and control technology, especially to the construction of large-scale fractional-order power system models and security prediction methods under network attacks. Background Technology

[0002] Currently, deterministic networking, as an emerging network communication architecture, ensures server quality through technologies such as resource reservation, service guarantees, and explicit routing. Therefore, it places stricter requirements on data transmission in deterministic networks (e.g., low packet loss rate and real-time accuracy of information between ends). However, power systems are susceptible to intermittent network attacks and parameter perturbations during operation, leading to low prediction accuracy and consequently affecting their safe and reliable operation.

[0003] In existing technologies, Chinese invention patent application number 202010801108.6 employs a deep learning architecture to establish a nonlinear mapping relationship between time-series features and frequency security, achieving end-to-end frequency security assessment, and improving assessment accuracy by optimizing key parameters of the power system frequency security assessment model. However, deep learning models often require a large amount of training data and are prone to overfitting, especially when data is insufficient or of poor quality. In such cases, the model's generalization ability may be poor, failing to comprehensively and accurately describe system characteristics, thus limiting the accuracy of the assessment results. Chinese invention patent application number 202210272522.1 estimates attack signals using a distributed sliding mode observer. While this method is effective in certain scenarios, its performance is highly dependent on parameter settings. Inappropriate parameter settings can weaken the observer's detection capability, leading to false or missed detections of attack signals, reducing prediction accuracy, and exacerbating controller chattering under network attacks. Furthermore, the above methods do not consider the prediction of system parameter perturbations, ignoring their impact on system state, thus limiting their application and effectiveness in complex power systems. Summary of the Invention

[0004] To address the shortcomings of existing technologies in comprehensively and accurately describing system characteristics and in predicting low accuracy when facing intermittent network attacks and system parameter perturbations, this specification provides a method for constructing a large-scale model of a fractional-order power system and making security predictions under network attacks.

[0005] This specification provides an embodiment of a method for constructing a fractional-order power system large model under network attacks. The method includes: constructing a first fractional-order power system model based on feature sample datasets of each subsystem in a first state; determining a second fractional-order power system model based on intermittent network attack modeling, feature sample datasets of each subsystem in a second state, and the first fractional-order power system model, where the second state represents being subjected to parameter perturbations and intermittent network attacks; constructing a second fractional-order power system multivariate regression model based on feature sample datasets of each subsystem in the second state and the second fractional-order power system model; determining the sensitivity of each subsystem in the power system to parameter perturbations and intermittent network attacks based on the second fractional-order power system model; constructing an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbations based on the sensitivity; iteratively updating the weight parameters in the initial prediction models for intermittent network attacks and parameter perturbations based on the error indices of the initial prediction models for intermittent network attacks and parameter perturbations to construct an intermittent network attack prediction model and a parameter perturbation prediction model, which constitute a fractional-order power system large model under network attacks.

[0006] According to one aspect of the embodiments of this specification, constructing a first fractional-order model of the power system based on a feature sample dataset of each subsystem in a first state includes:

[0007] A fractional-order model of the power system is constructed using the following formula:

[0008] Among them, D α Let x(t) represent the fractional differential operator, and let x(t) represent the power system state matrix, x(t) = [ΔP] AE ,ΔP FC ,ΔP DEG ,ΔP BESS ΔP c ,Δf] T ΔP AE ΔP represents the change in power generation from the hydroelectrolyzer. FC ΔP represents the change in power generation from the fuel cell. DEG ΔP represents the change in power generation from the diesel generator. BESS ΔP represents the change in power generation of the battery energy storage system. c Δf represents the change in the output of the integral controller, and Δf represents the change in the power system frequency. T represents the state moments of the power system. AE T represents the time constant of the water electrolyzer. FC T represents the fuel cell time constant. DEG T represents the diesel engine time constant;BESS K represents the time constant of the battery energy storage system. I K represents the gain of the integral controller. p T represents the frequency response coefficient of the power system. p Represents the time constant of the power system. Represents the power system control matrix; u(t) represents the power system control input, B u =[K AE / T AE ,K FC / T FC ,K DEG / T DEG ,K BESS / T BESS [0,0],K AE K represents the gain of the water electrolyzer. FC K represents the fuel cell gain. DEG K represents the diesel engine gain. BESS Indicates the gain of the battery energy storage system, ΔA and ΔB. u Let represent the parameter perturbation matrices of the power system state matrix and control matrix, respectively; y(t) represents the power system output; and C represents the power system output matrix.

[0009] Where f(x) is a function, and Γ(·) is a gamma function. x Let α be a variable, and let α be a fractional order. n Let ξ be an integer greater than α, and let ξ be the integral variable.

[0010] According to one aspect of the embodiments of this specification, determining the fractional-order second model of the power system based on intermittent network attack modeling, the feature sample dataset of each subsystem in the second state, and the first fractional-order model of the power system includes: constructing the second fractional-order second model of the power system using the following formula:

[0011] Among them, δ(x,t)=ΔAx(t)+ΔB u u(t) represents the fractional-order power system parameter perturbation, and η(t) represents the intermittent network attack related to the system state, which is determined by the intermittent network attack model.

[0012] According to one aspect of the embodiments of this specification, determining the sensitivity of each subsystem in a power system to parameter perturbations and intermittent network attacks based on the fractional-order multiple regression model of the power system includes: estimating the regression coefficient matrix of multiple subsystems of the power system using the least squares method; determining the sensitivity of the subsystem to intermittent network attacks by summing the second and fourth elements in each regression coefficient matrix; and determining the sensitivity of the subsystem to parameter perturbations by summing the third and fifth elements in each regression coefficient matrix.

[0013] According to one aspect of the embodiments of this specification, constructing an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbation based on the sensitivity includes: constructing an intermittent network attack prediction weight matrix based on the sensitivity of the subsystem to intermittent network attacks; constructing a parameter perturbation prediction weight matrix based on the sensitivity of the subsystem to parameter perturbation; determining the output of the hidden layer neurons of the intermittent network attack prediction model based on the intermittent network attack prediction weight matrix; determining the output of the hidden layer neurons of the parameter perturbation prediction model based on the parameter perturbation prediction weight matrix; determining the initial prediction model for intermittent network attacks based on the output of the hidden layer neurons of the intermittent network attack prediction model and the output weights of the prediction model; and determining the initial prediction model for parameter perturbation based on the output of the hidden layer neurons of the parameter perturbation prediction model and the output weights of the prediction model.

[0014] According to one aspect of the embodiments of this specification, constructing an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbations based on the sensitivity includes:

[0015] The feature datasets of each subsystem collected in the second state are input into the intermittent network attack prediction model to determine the output of the hidden layer neurons of the intermittent network attack prediction model:

[0016] h j,m For the intermittent network attack prediction model of a superspiral fractional-order constrained RBF neural network, ρ is the output of the m-th hidden layer neuron. j c represents the weight matrix for predicting intermittent network attacks or the weight matrix for predicting parameter perturbations. j,m b represents the center point vector value of the m-th hidden layer neuron in the intermittent network attack prediction model of a superspiral fractional-order constrained RBF neural network. j,m Let b represent the width of the Gaussian function of the m-th hidden layer neuron in the intermittent network attack prediction model of the superspiral fractional-order constrained RBF neural network, and b j,m >0, subscripts m = 1, 2, ..., M represent the m-th hidden layer neuron of the superspiral fractional-order constrained RBF neural network intermittent network attack prediction model, and M is the total number of hidden layer neurons in the superspiral fractional-order constrained RBF neural network intermittent network attack prediction model.

[0017] According to one aspect of the embodiments of this specification, the error indices of the intermittent network attack initial prediction model and the parameter perturbation initial prediction model are determined by the following formula:

[0018] Where E1(t) represents the evaluation index of intermittent network attack prediction error, ν1 represents the weight of intermittent network attack prediction error, η(t) represents the system's intermittent network attack, E2(t) represents the evaluation index of parameter perturbation prediction error, ν2 represents the weight of parameter perturbation error, and δ(x,t) represents the system parameter perturbation. The values ​​are the observations of η(t) and δ(x,t) for the initial prediction model of intermittent network attacks and the initial prediction model of parameter perturbation, respectively.

[0019] According to one aspect of the embodiments of this specification, iteratively updating the weight parameters in the intermittent network attack initial prediction model and the parameter perturbation initial prediction model includes: constructing a superspiral path using a superspiral algorithm to guide the weights in the intermittent network attack initial prediction model and the parameter perturbation initial prediction model to converge along the optimal direction; and constraining the outputs of the intermittent network attack initial prediction model and the parameter perturbation initial prediction model using an exponentially dependent barrier function.

[0020] This specification also provides a security prediction method, including: acquiring feature data of each subsystem in the power system under a second state; inputting the feature data of each subsystem in the power system under the second state into an intermittent network attack prediction model and a parameter perturbation prediction model to obtain prediction results for each subsystem.

[0021] This specification also provides an embodiment of a fractional-order power system large model construction device under network attacks. The device includes: a first construction unit for constructing a first fractional-order power system model based on feature sample datasets of each subsystem in a first state; a second construction unit for determining a second fractional-order power system model based on intermittent network attack modeling, feature sample datasets of each subsystem in a second state, and the first fractional-order power system model, wherein the second state represents the presence of parameter perturbations and intermittent network attacks; and a third construction unit for constructing a second fractional-order power system multivariate regression model in the second state based on feature sample datasets of each subsystem in the second state and the second fractional-order power system model. The system comprises five subsystems: a first subsystem, a second subsystem, and a third subsystem, a fourth subsystem, which are configured to determine the sensitivity of each subsystem in the power system to parameter perturbations and intermittent network attacks based on the sensitivity; a fifth subsystem, which is configured to iteratively update the weight parameters in the intermittent network attack prediction model and the parameter perturbation prediction model based on the error indices of the intermittent network attack prediction model and the parameter perturbation prediction model, thereby constructing the intermittent network attack prediction model and the parameter perturbation prediction model, which together constitute a large-scale model of the power system under network attacks.

[0022] This specification also provides a security prediction device, which includes: an acquisition unit for acquiring feature data of each subsystem in a power system under a second state; and a prediction unit for inputting the feature data of each subsystem in the power system under the second state into an intermittent network attack prediction model and a parameter perturbation prediction model to obtain prediction results for each subsystem.

[0023] This specification also provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the method for constructing a large fractional-order power system model and predicting security under network attacks.

[0024] This specification also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for constructing a large-scale model of a fractional-order power system under network attacks and for security prediction.

[0025] This invention improves the prediction accuracy of intermittent network attacks and parameter perturbations suffered by fractional-order power systems, effectively making up for the shortcomings of traditional methods in that they cannot fully and accurately describe system characteristics. In particular, it addresses the problem of low prediction accuracy when facing intermittent network attacks and system parameter perturbations, significantly improving the accuracy and robustness of security assessment.

[0026] This specification utilizes the least squares method to estimate the regression coefficients of intermittent network attacks and parameter perturbations, analyzes the sensitivity of fractional-order power system states to intermittent network attacks and parameter perturbations, designs prediction weights for intermittent network attacks and parameter perturbations, and compensates the neural network prediction model. The neural network prediction model employs a superspiral algorithm to construct a superspiral path, guiding the neural network weights to converge rapidly along the optimal direction. An exponentially dependent barrier function is used to constrain the neural network output, effectively preventing large fluctuations during training, thereby improving the prediction accuracy for intermittent network attacks and system parameter perturbations. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 The diagram shown is a flowchart of a method for constructing a large-scale model of a fractional-order power system under network attacks, according to an embodiment of this specification.

[0029] Figure 2 The diagram shown is a flowchart of a method for determining the sensitivity of each subsystem to parameter perturbations and intermittent network attacks according to an embodiment of this specification.

[0030] Figure 3 The diagram shown is a flowchart of a method for constructing an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbations according to an embodiment of this specification.

[0031] Figure 4 The diagram shown is a flowchart of a method for iteratively updating an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbations, according to an embodiment of this specification.

[0032] Figure 5 The flowchart shown in this specification illustrates an embodiment of intermittent network attack and parameter perturbation prediction for each subsystem.

[0033] Figure 6 The diagram shown is a structural schematic of a device for constructing a large-scale model of a fractional-order power system under network attacks, according to an embodiment of this specification.

[0034] Figure 7 The diagram shown is a structural schematic of a safety prediction device according to an embodiment of this specification.

[0035] Figure 8 The diagram shown is a structural schematic of a computer device according to an embodiment of this specification.

[0036] Explanation of symbols in the attached drawings:

[0037] 601. First building block;

[0038] 602. Second building block;

[0039] 603. Third building block;

[0040] 604. Determine the unit;

[0041] 605. Fourth building block;

[0042] 606. The fifth building block;

[0043] 701. First Acquisition Unit;

[0044] 702. Prediction Result Determination Unit;

[0045] 802. Computer equipment;

[0046] 804, Processor;

[0047] 806. Memory;

[0048] 808. Drive mechanism;

[0049] 810. Input / Output Module;

[0050] 812. Input devices;

[0051] 814. Output devices;

[0052] 816. Presentation equipment;

[0053] 818. Graphical User Interface;

[0054] 820. Network interface;

[0055] 822. Communication link;

[0056] 824. Communication bus. Detailed Implementation

[0057] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0058] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0059] This specification provides the operational steps of the methods described in the embodiments or flowcharts, but based on conventional or non-inventive labor, more or fewer operational steps may be included. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only possible execution order. In actual system or device products, the methods shown in the embodiments or drawings can be executed sequentially or in parallel.

[0060] It should be noted that the methods, apparatus, and computer equipment described in this specification can be used in the field of power system operation and control technology. This specification does not limit the application areas of the method for constructing a large fractional-order power system model and predicting security under network attacks.

[0061] Figure 1 The diagram shown is a flowchart of a method for constructing a large-scale model of a fractional-order power system under network attacks, according to an embodiment of this specification. The method specifically includes the following steps:

[0062] Step 101: Construct a first-order fractional model of the power system based on the feature sample dataset of each subsystem in the power system under the first state.

[0063] In this step, the first state represents the state of the power system during normal operation. In the first state, characteristic sample data of each subsystem in the power system are collected to construct a fractional-order first model of the power system. In the embodiments of this specification, the power system includes multiple types of subsystems, specifically including: a water electrolyzer system, a fuel cell system, a diesel generator system, and a battery energy storage system. Each subsystem generates different characteristic data when operating in the first state. For example, the characteristic data in the water electrolyzer system is the change in power generation from the water electrolyzer; the characteristic data in the fuel cell system is the change in power generation from the fuel cell; and the characteristic data in the diesel generator system is the change in power generation from the diesel generator.

[0064] This step aims to construct a fractional-order first-order model of the power system. It involves collecting feature data from each subsystem within the power system during its first-state operation, creating a feature sample dataset. This feature sample dataset is then used to further construct the fractional-order first-order model of the power system, specifically including:

[0065] Based on the feature sample dataset of each subsystem in the power system under the first state, a first-state matrix of the power system, a second-state matrix of the power system describing the system response speed, and a power system control matrix are constructed. Specifically, the power system state matrix describing the system response speed is determined based on the time constants of each subsystem; the parameter perturbation matrix is ​​determined based on the ratio of the gain of each subsystem to its time constant. The gain of each system is obtained through experimental testing.

[0066] Furthermore, based on the feature sample dataset of each subsystem in the power system under the first state, a power system state matrix is ​​constructed.

[0067] The first-order fractional model of the power system is constructed using the following formula:

[0068] Among them, D α Let denote a fractional-order differential operator, and let x(t) denote the power system state matrix. Represents the state matrix of the power system. The control matrix of the power system is represented by u(t); the control inputs of the power system are ΔA and ΔB. u Let y(t) represent the parameter perturbation matrices of the power system state matrix and control matrix, respectively, and let C represent the power system output matrix. Then x(t) = [ΔP] AE ,ΔP FC ,ΔP DEG ,ΔP BESS ΔP c ,Δf] T ΔP AE ΔP represents the change in power generation from the hydroelectrolyzer. FC ΔP represents the change in power generation from the fuel cell. DEG ΔP represents the change in power generation from the diesel generator. BESS ΔP represents the change in power generation of the battery energy storage system. c B represents the change in the output of the integral controller, Δf represents the change in the power system frequency, and B represents the change in the output of the integral controller. u =[K AE / T AE ,K FC / T FC ,K DEG / T DEG ,K BESS / T BESS [0,0],K AE K represents the gain of the water electrolyzer. FC K represents the fuel cell gain. DEG K represents the diesel engine gain. BESS This indicates the gain of the battery energy storage system. The various gains described in this manual can be obtained through experimental testing.

[0069] in, T AE T represents the time constant of the water electrolyzer, used to describe the system response speed; FC T represents the fuel cell time constant. DEG T represents the diesel engine time constant; BESS K represents the time constant of the battery energy storage system. I K represents the gain of the integral controller. p T represents the frequency response coefficient of the power system. p This represents the time constant of the power system.

[0070] Before constructing the first fractional-order model of the power system, we first construct a fractional-order differential operator to obtain the fractional-order differential of the function: Where f(x) is a function, and Γ(·) is a gamma function. x Let α represent the fractional order, and let α represent the variable. nLet ξ represent an integer greater than α, and let ξ represent the integral variable.

[0071] Step 102: Based on the intermittent network attack modeling, the feature sample dataset of each subsystem in the second state, and the first fractional-order power system model, determine the second fractional-order power system model. The second state represents the state of being subjected to parameter perturbations and intermittent network attacks.

[0072] In the embodiments of this specification, the second state represents the power system suffering from parameter perturbations and indirect network attacks. In the second state, it is necessary to construct a second fractional-order model of the power system based on the characteristic sample data of each subsystem in the power system and the first fractional-order model of the power system constructed in step 101.

[0073] Specifically, based on the first fractional-order model of the power system, we add parameters related to power system perturbation and intermittent network attacks related to system state.

[0074] Specifically, the fractional-order second model of the power system is constructed using the following formula:

[0075] Among them, δ(x,t)=ΔAx(t)+ΔB u u(t) represents the fractional-order power system parameter perturbation, and η(t) represents the intermittent network attack related to the system state, determined by the intermittent network attack model. Where δ(x,t) = ΔAx(t) + ΔB u u(t) represents the fractional-order power system parameter perturbation; It is a positive number.

[0076] Step 103: Based on the feature sample dataset of each subsystem in the second state and the second fractional-order power system model, construct the fractional-order multiple regression model of the power system in the second state.

[0077] In this step, intermittent network attacks cause system instability and increase the risk of failure by intermittent, short-term interference. Attackers do not need to continuously consume large amounts of resources. Therefore, this specification first models intermittent network attacks:

[0078] Where η(t) represents the intermittent network attack related to the system state, ρ(t) represents the gain coefficient of the intermittent network attack, and T DoS (t) represents the weight of an intermittent network attack, t represents the current time t, n∈N is the number of cycles, and T represents the duration of one cycle of the power system. on (t) represents the start time of the intermittent network attack, T off (t) represents the duration of an intermittent network attack, T on(t) and T off (t) changes randomly and satisfies T on (t)+T off (t)≤T.

[0079] Based on the intermittent network attack modeling and the fractional-order second model of the power system, in the case of intermittent network attacks and parameter perturbations on the power system described in step 102, the characteristic data and interference datasets of each subsystem are collected [η]. p (t),δ p (x,t)].

[0080] The feature dataset of the p-th subsystem is shown below: [ΔP AE,p ,ΔP FC,p ,ΔP DEG,p ,ΔP BESS,p ΔP c,p ,Δf p ] T Subscripts p = 1, 2, ..., n. Where ΔP AE ΔP represents the change in power generation from the hydroelectrolyzer. FC ΔP represents the change in power generation from the fuel cell. DEG ΔP represents the change in power generation from the diesel generator. BESS ΔP represents the change in power generation of the battery energy storage system. c This represents the change in the output of the integral controller. Here, η(t) in the disturbance dataset represents intermittent network attacks related to the system state, and δ... p (x,t) represents the fractional-order power system parameter perturbation.

[0081] Furthermore, based on the characteristic data of each subsystem under parameter perturbations and intermittent network attacks, as well as the regression coefficient matrix of the characteristic data of each subsystem, the error matrix of the change in the characteristic data of each subsystem, and the polynomial characteristic matrix, a fractional-order multiple regression model of the power system in the second state is constructed. For a detailed description of the fractional-order multiple regression model of the power system, please refer to [link to relevant documentation]. Figure 2 describe.

[0082] Step 104: Based on the fractional-order multiple regression model of the power system, determine the sensitivity of each subsystem in the power system to parameter perturbations and intermittent network attacks.

[0083] After constructing the fractional-order multiple regression model of the power system, the least squares method is used to estimate the regression coefficients of intermittent network attacks and parameter perturbations, thereby analyzing the sensitivity of the fractional-order power system state to intermittent network attacks and parameter perturbations.

[0084] Step 105: Based on the sensitivity, construct an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbation, respectively.

[0085] In this step, after determining the sensitivity of each subsystem in the power system to parameter perturbations and intermittent network attacks, the sensitivity of all subsystems to intermittent network attacks is combined to form the intermittent network attack prediction weight matrix for the initial prediction model of intermittent network attacks; similarly, the sensitivity of all subsystems to parameter perturbations is combined to form the parameter perturbation prediction weight matrix for the initial prediction model of parameter perturbations. By introducing the intermittent network attack prediction weight matrix and the parameter perturbation prediction weight matrix into the initial prediction model, the initial prediction model pays more attention to system state variables with higher sensitivity to intermittent network attacks or parameter perturbations.

[0086] Step 106: Based on the error indices of the initial prediction model for intermittent network attacks and the initial prediction model for parameter perturbation, iteratively update the weight parameters in the initial prediction model for intermittent network attacks and the initial prediction model for parameter perturbation to construct the intermittent network attack prediction model and the parameter perturbation prediction model. The intermittent network attack prediction model and the parameter perturbation prediction model constitute a large fractional-order power system model under network attacks.

[0087] This step employs the superspiral algorithm to construct a superspiral path, guiding the weights of each layer of the neural network to converge rapidly. An exponentially dependent barrier function is used to constrain the neural network's output, preventing significant fluctuations during training. This improves the prediction accuracy for intermittent network attacks and parameter perturbations affecting fractional-order power systems, ultimately constructing a large-scale model of a fractional-order power system under network attacks. For detailed descriptions of training and constructing the intermittent network attack prediction model and the parameter perturbation prediction model, please refer to [link to relevant documentation]. Figure 4 describe.

[0088] Figure 2 The diagram shown is a flowchart of a method for determining the sensitivity of each subsystem to parameter perturbations and intermittent network attacks according to an embodiment of this specification, specifically including the following steps:

[0089] Step 201: Use the least squares method to estimate the regression coefficient matrix of multiple subsystems of the power system in fractional order.

[0090] In this step, before determining the regression coefficient matrix of each subsystem, characteristic sample data of each subsystem in the power system under the second state are first collected to construct a fractional-order multiple regression model of the power system. The formula of the multiple regression model is as follows:

[0091] Where, ΔP AE This represents the dataset of changes in power generation from the hydroelectrolyzer under the second state, ΔP. AE =[ΔP AE,1 ΔPAE,2 … ΔP AE,n ] T ΔP FC This represents the dataset of changes in fuel cell power generation in the second state, ΔP. FC =[ΔP FC,1 ΔP FC,2 … ΔP FC,n ] T ΔP DEG This represents the dataset of changes in diesel engine power generation under the second state, ΔP. DEG =[ΔP DEG,1 ΔP DEG,2 … ΔP DEG,n ] T ΔP BESS This represents the dataset of changes in power generation of the battery energy storage system under the second state, ΔP. BESS =[ΔP BESS,1 ΔP BESS,2 … ΔP BESS,n ] T ΔP represents the dataset of changes in the output of the integral controller, ΔP = [ΔP1 ΔP2 … ΔP] n ] T Δf represents the frequency change of the power system. The dataset Δf = [Δf1 Δf2 … Δf n ] T β1 represents the regression coefficient matrix of the change in power generation of the hydroelectrolyzer under the second state: β1=[β 1,1 ,β 2,1 ,β 3,1 ,β 4,1 ,β 5,1 ,β 6,1 ] T β2 represents the regression coefficient matrix of the change in fuel cell power generation in the second state: β2=[β 1,2 ,β 2,2 ,β 3,2 ,β 4,2 ,β 5,2 ,β 6,2 ] T β3 represents the regression coefficient matrix of the change in diesel engine power generation in the second state: β3=[β 1,3 ,β 2,3 ,β 3,3 ,β 4,3 ,β 5,3 ,β 6,3 ] T β4 represents the regression coefficient matrix of the change in power generation of the battery energy storage system in the second state: β4=[β 1,4 ,β 2,4 ,β 3,4 ,β 4,4,β 5,4 ,β 6,4 ] T β5 represents the regression coefficient matrix of the change in the integral controller output in the second state: β5 = [β 1,5 ,β 2,5 ,β 3,5 ,β 4,5 ,β 5,5 ,β 6,5 ] T β6 represents the regression coefficient matrix of the power system frequency change in the second state: β6=[β 1,6 ,β 2,6 ,β 3,6 ,β 4,6 ,β 5,6 ,β 6,6 ] T ε1 represents the error matrix of the change in power generation of the water electrolyzer in the second state: ε1=[ε 1,1 ,ε 2,1 ,ε 3,1 ,ε 4,1 ,ε 5,1 ,ε 6,1 ] T ε2 represents the error matrix of the change in fuel cell power generation in the second state: ε2=[ε 1,2 ,ε 2,2 ,ε 3,2 ,ε 4,2 ,ε 5,2 ,ε 6,2 ] T ε3 represents the error matrix of the diesel engine power generation change in the second state: ε3=[ε 1,3 ,ε 2,3 ,ε 3,3 ,ε 4,3 ,ε 5,3 ,ε 6,3 ] T ε4 represents the error matrix of the change in power generation of the battery energy storage system in the second state: ε4=[ε 1,4 ,ε 2,4 ,ε 3,4 ,ε 4,4 ,ε 5,4 ,ε 6,4 ] T ε5 represents the error matrix of the integral controller output change in the second state: ε5=[ε 1,5 ,ε 2,5 ,ε 3,5 ,ε 4,5 ,ε 5,5 ,ε 6,5 ] T ε6 represents the error matrix of the power system frequency change in the second state: ε6=[ε1,6 ,ε 2,6 ,ε 3,6 ,ε 4,6 ,ε 5,6 ,ε 6,6 ] T X poly Represents the characteristic matrix of a polynomial

[0092] After constructing the fractional-order multiple regression model of the power system, the fractional-order regression coefficient matrix of each subsystem in the fractional-order multiple regression model of the power system is estimated by the least squares method. The form of the power system fractional regression coefficient matrix for each subsystem is shown below:

[0093] in, This represents the regression coefficient matrix of the hydroelectric power generation system. This represents the regression coefficient matrix of the fuel cell power generation system. This represents the regression coefficient matrix of the diesel generator system. This represents the regression coefficient matrix of the battery energy storage system. This represents the regression coefficient matrix of the integral controller system. X represents the regression coefficient matrix of the power system frequency system. polt X is the multinomial characteristic matrix in the multiple regression model. poly .

[0094] Step 202: The sum of the second and fourth elements in each regression coefficient matrix is ​​determined as the sensitivity of the subsystem to intermittent network attacks.

[0095] In the embodiments of this specification, the regression coefficient matrix β1 of the water electrolyzer power generation system, the regression coefficient matrix β2 of the fuel cell power generation system, the regression coefficient matrix β3 of the diesel engine power generation system, the regression coefficient matrix β4 of the battery energy storage system power generation system, the regression coefficient matrix β5 of the integral controller system, and the regression coefficient matrix β6 of the power system frequency system, each of these six subsystem regression coefficient matrices includes 6 elements, which are respectively referred to as the first element, the second element, the third element, the fourth element, the fifth element, and the sixth element.

[0096] In this specification, based on the formula of the fractional-order multiple regression model for power systems, it can be determined that the characteristic dataset of each subsystem in the power system can be expressed as the product of a polynomial characteristic matrix and the regression coefficient matrix of each subsystem. For example, the dataset ΔP of the change in power generation from the hydroelectrolyzer in the second state... AE Unfold, and you get:

[0097] Among them, the η1×β2 term and These two terms are related to intermittent cyberattacks suffered by hydroelectric power generation systems. Therefore, the coefficients β2 and β4 before these two terms are determined to be coefficients related to the sensitivity of fractional-order power systems to intermittent cyberattacks.

[0098] Specifically, the sensitivity of the fractional-order power system to intermittent network attacks is determined by adding the second and fourth elements of the regression coefficient matrix of each subsystem. The sensitivity of a fractional-order power system to an intermittent network attack η(t) is considered, and thus the sensitivity of a hydroelectric power generation system, a fuel cell power generation system, a diesel engine power generation system, a battery energy storage system, an integral controller system, and a power system frequency system to intermittent network attacks are determined respectively.

[0099] After determining the sensitivity of each subsystem to intermittent network attacks, the regression coefficient matrices of each subsystem are further sorted according to their sensitivity to intermittent network attacks. A higher sensitivity indicates a greater impact of intermittent network attacks on the state of the fractional-order power system; a lower sensitivity indicates a smaller impact.

[0100] Step 203: The sum of the third and fifth elements in each regression coefficient matrix is ​​determined as the sensitivity of the subsystem to parameter perturbations.

[0101] In this embodiment of the specification, similar to the determination of the terminology subjected to intermittent network attacks in step 202, β3 and β5 are determined in this step to be coefficients related to the sensitivity of the fractional-order power system to parameter perturbations.

[0102] The sensitivity of the fractional-order power system to intermittent network attacks is determined by adding the third and fifth elements of the regression coefficient matrix of each subsystem. The sensitivity of a fractional-order power system to parameter perturbation δ(x,t) is considered. Based on this, the sensitivities of the hydroelectric power generation system, fuel cell power generation system, diesel engine power generation system, battery energy storage system, integral controller system, and power system frequency system to parameter perturbation are determined. The regression coefficient matrices of each subsystem are then sorted by their sensitivity to parameter perturbation in ascending order. A higher sensitivity indicates a greater impact of the parameter perturbation on the state of the fractional-order power system; a lower sensitivity indicates a smaller impact.

[0103] Figure 3 The diagram shown is a flowchart of a method for constructing an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbations according to an embodiment of this specification. The method specifically includes the following steps:

[0104] Step 301: Construct an intermittent network attack prediction weight matrix based on the subsystem's sensitivity to intermittent network attacks.

[0105] This step is based on Figure 2 The sensitivity calculated in the above formula is used to calculate the intermittent network attack prediction weight matrix and the parameter perturbation prediction weight matrix, as shown in the following formula:

[0106] Wherein, ρ1 represents the intermittent network attack prediction weight matrix, and the six elements in the intermittent network attack prediction weight matrix are the sensitivity of each of the six subsystems to intermittent network attacks and the sum of the sensitivity of all subsystems to intermittent network attacks.

[0107] Step 302: Construct the parameter perturbation prediction weight matrix based on the subsystem's sensitivity to parameter perturbations.

[0108] Where ρ2 represents the parameter perturbation prediction weight matrix. The six elements in the parameter perturbation prediction weight matrix are the sensitivity of each of the six subsystems to intermittent network attacks and the sum of the sensitivities of all subsystems to intermittent network attacks.

[0109] Step 303: Determine the output of the hidden layer neurons of the intermittent network attack initial prediction model based on the intermittent network attack prediction weight matrix.

[0110] In this step, a superspiral fractional-order algorithm is used to construct an intermittent network attack prediction model. To improve the accuracy of the model's prediction, the intermittent network attack prediction weight matrix established in step 302 is introduced, making the intermittent network attack prediction model pay more attention to system state variables that are highly sensitive to intermittent network attacks and parameter perturbations.

[0111] In this step, the feature datasets of each subsystem collected in the second state will be used: [ΔP AE,p ,ΔP FC,p ,ΔP DEG,p ,ΔP BESS,p ΔP c,p ,Δf p ] T The input is fed into the intermittent network attack prediction model to determine the output of the hidden layer neurons. The hidden layer neurons are shown below: Among them, h j,m ρ represents the output of the m-th hidden layer neuron in the intermittent network attack prediction model. j c represents the weight matrix for predicting intermittent network attacks or the weight matrix for predicting parameter perturbations. j,m =[c j,m,1 ,cj,m,2 ,c j,m,3 ,c j,m,4 ,c j,m,5 ,c j,m,6 ] T c j,m b represents the center point vector value of the m-th hidden layer neuron in the intermittent network attack prediction model. j,m Let b represent the width of the Gaussian function of the m-th hidden layer neuron in the intermittent network attack prediction model. j,m >0, subscripts m = 1, 2, ..., M represent the m-th hidden layer neuron of the intermittent network attack prediction model, and M represents the total number of hidden layer neurons of the intermittent network attack prediction model.

[0112] When j equals 1, the output of the hidden layer neurons in the intermittent network attack prediction model can be obtained as follows:

[0113] Among them, h 1,m c represents the output of the m-th hidden layer neuron in the intermittent network attack prediction model, which is constructed from the intermittent network attack prediction weight matrix. 1,m c represents the vector value of the center point of the m-th hidden layer neuron in the intermittent network attack prediction model. 1,m =[c 1,m,1 ,c 1,m,2 ,c 1,m,3 ,c 1,m,4 ,c 1,m,5 ,c 1,m,6 ] T b 1,m Let b represent the width of the Gaussian function of the m-th hidden layer neuron in the intermittent network attack prediction model. 1,m >0, subscripts m = 1, 2, ..., M represent the m-th hidden layer neuron of the intermittent network attack prediction model, and M represents the total number of hidden layer neurons of the intermittent network attack prediction model.

[0114] Step 304: Determine the output of the hidden layer neurons in the parameter perturbation initial prediction model based on the parameter perturbation prediction weight matrix. This step is similar to step 303; when j equals 2, h... 2,m c represents the output of the m-th hidden layer neuron in the parametric perturbation prediction model, which is constructed from the parametric perturbation prediction weight matrix. 2,m c represents the vector value of the center point of the m-th hidden layer neuron in the parameter perturbation prediction model. 2,m =[c 2,m,1 ,c 2,m,2 ,c 2,m,3 ,c 2,m,4 ,c 2,m,5 ,c2,m,6 ] T b 2,m Let b represent the width of the Gaussian function of the m-th hidden layer neuron in the parameter perturbation prediction model. 2,m >0, the subscripts m = 1, 2, ..., M represent the m-th hidden layer neuron of the parameter perturbation prediction model, and M represents the total number of hidden layer neurons of the parameter perturbation prediction model.

[0115] Step 305: Determine the initial prediction model for intermittent network attacks based on the outputs of the hidden layer neurons of the initial prediction model for intermittent network attacks and the output weights of the prediction model; determine the initial prediction model for parameter perturbation based on the outputs of the hidden layer neurons of the initial prediction model for parameter perturbation and the output weights of the prediction model.

[0116] In this step, the output of the neural network prediction model is:

[0117] Where y1 and y2 represent the observations of the intermittent network attack η(t) and the system parameter perturbation δ(x,t) by the initial prediction model of the intermittent network attack and the initial prediction model of the parameter perturbation, respectively, and the observations are expressed as follows: w i,m This represents the output weight of the i-th initial prediction model at layer m. The subscripts i = 1 and 2 represent the prediction output weights of the intermittent network attack prediction model and the parameter perturbation prediction model, respectively. h i This represents the output of the hidden layer neuron in the i-th initial prediction model.

[0118] The outputs of the hidden layer neurons of the intermittent network attack initial prediction model and the parameter perturbation initial prediction model determined in steps 303 and 304 are substituted into the neural network prediction model formula to construct the intermittent network attack initial prediction model and the parameter perturbation initial prediction model, respectively.

[0119] After constructing the initial prediction models for intermittent network attacks and parameter perturbations, this manual further calculates the performance of these models to fine-tune them. Specifically, the error indices of the initial prediction models for intermittent network attacks and parameter perturbations are calculated using the following formulas:

[0120] Where E1(t) and ν1 represent the error evaluation index and prediction error weight of the initial prediction model for intermittent network attacks, respectively, and E2(t) and ν2 represent the error evaluation index and prediction error weight of the initial prediction model for parameter perturbation. This represents the observed value of the intermittent network attack prediction model for the system's intermittent network attack η(t), and can also be called the predicted value; This represents the observed value of the system parameter perturbation δ(x,t) by the parameter perturbation prediction model, and can also be called the predicted value.

[0121] Figure 4 The diagram shown is a flowchart of a method for iteratively updating an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbations, according to an embodiment of this specification. The method specifically includes the following steps:

[0122] Step 401: Construct a superspiral path using the superspiral algorithm to guide the weights in the initial prediction model for intermittent network attacks and the initial prediction model for parameter perturbation to converge along the optimal direction.

[0123] Specifically, the formula for the superhelical algorithm is as follows:

[0124] χ i (t)=-λ|D α x(t)| ζ tanh(D α-1 x(t))+χ 1,i (t);

[0125]

[0126] Where, χ i (t) represents the superhelical global error adjustment factor, χ 1,i (t) represents the superhelical local error suppression factor, where λ and ζ are constants.

[0127] In this step, the superspiral algorithm is used to guide the weights of each layer in the initial prediction model for intermittent network attacks and the initial prediction model for parameter perturbation to converge rapidly along the optimal direction, thereby improving the training efficiency of the two prediction models.

[0128] Step 402: Use an exponentially dependent barrier function to constrain the outputs of the initial prediction model for intermittent network attacks and the initial prediction model for parameter perturbation.

[0129] In this step, the superspiral algorithm calculation formula determined in step 401 is used to construct a superspiral fractional-order constraint optimization formula, and then an exponentially dependent barrier function is constructed according to the following formula:

[0130]

[0131] k(t)=k a e -rt +k b ;

[0132] Where Λ(t) represents the superspiral fractional-order constrained optimization function, k(t) represents the output error constraint function, and E i (t) represents the error evaluation index of the i-th prediction model, h i k represents the output of the hidden layer neuron of the i-th prediction model; a k b r, Ψ1, and Ψ2 are constants. In this specification, the exponentially dependent barrier function is used to limit the error evaluation index E. i (t) Significant fluctuations occur during training, improving the prediction accuracy of intermittent network attacks and parameter perturbations suffered by fractional-order power systems.

[0133] Specifically, the formula for constructing a superhelical path using the superhelical algorithm is shown below:

[0134]

[0135] The superspiral algorithm is used to construct the weights w of the superspiral path-guided neural network. i It converges rapidly along the optimal direction and employs an exponentially dependent barrier function on the neural network's output. (Formula meaning explained above) Figure 3 (As explained in the text) Constraints are applied to prevent large fluctuations in the neural network during training, thereby improving the prediction accuracy of intermittent network attacks and parameter perturbations suffered by fractional-order power systems. When the error evaluation indicators of the intermittent network attack prediction model and the parameter perturbation prediction model converge to preset thresholds (e.g., 0) during the training process, the training of the intermittent network attack prediction model and the parameter perturbation prediction model is considered complete, and the intermittent network attack prediction model and the parameter perturbation prediction model are constructed, which is the large-scale model of the fractional-order power system under network attacks described in this specification.

[0136] Figure 5 The diagram shown is a flowchart of a method for intermittent network attacks and parameter perturbation prediction of various subsystems according to an embodiment of this specification, which specifically includes the following steps:

[0137] Step 501: Obtain the characteristic data of each subsystem in the power system under the second state.

[0138] In this step, characteristic data of each subsystem in the power system that is under intermittent network attacks and parameter perturbations are collected, including: power generation data of the water electrolyzer system, power generation data of the fuel cell system, power generation data of the diesel generator system, power generation data of the battery energy storage system, changes in the integral controller, and changes in the power system frequency.

[0139] Step 502: Input the characteristic data of each subsystem in the power system under the second state into the intermittent network attack prediction model and the parameter perturbation prediction model to obtain the prediction results for each subsystem.

[0140] In this step, the data collected in step 501 is input into the pre-trained intermittent network attack prediction model and parameter perturbation prediction model to obtain the model's prediction results on whether each subsystem is subject to intermittent network attacks and parameter perturbations. This specification improves the prediction accuracy of intermittent network attacks and parameter perturbations suffered by fractional-order power systems.

[0141] like Figure 6 The diagram shown is a structural schematic of a device for constructing a large-scale fractional-order power system model under network attacks, according to an embodiment of this specification. The basic structure of this device is illustrated in the diagram. The functional units and modules can be implemented in software, or using general-purpose or specific chips to construct the large-scale fractional-order power system model under network attacks. Specifically, the device includes:

[0142] The first construction unit 601 is used to construct a first fractional-order model of the power system based on the feature sample dataset of each subsystem in the power system under the first state.

[0143] The second construction unit 602 is used to determine the second fractional-order power system model based on the intermittent network attack modeling, the feature sample dataset of each subsystem in the second state and the first fractional-order power system model, wherein the second state represents the suffering of parameter perturbation and intermittent network attacks.

[0144] The third construction unit 603 is used to construct a fractional-order multiple regression model of the power system in the second state based on the feature sample dataset of each subsystem in the second state and the second fractional-order power system model.

[0145] The determining unit 604 is used to determine the sensitivity of each subsystem in the power system to parameter perturbations and intermittent network attacks based on the fractional-order multiple regression model of the power system.

[0146] The fourth construction unit 605 is used to construct an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbation based on the sensitivity, respectively.

[0147] The fifth construction unit 606 is used to iteratively update the weight parameters in the intermittent network attack initial prediction model and the parameter perturbation initial prediction model according to the error evaluation index of the intermittent network attack initial prediction model and the parameter perturbation initial prediction model, and construct the intermittent network attack prediction model and the parameter perturbation prediction model. The intermittent network attack prediction model and the parameter perturbation prediction model constitute a fractional-order power system large model under network attack.

[0148] Figure 7 The diagram shown is a structural schematic of a security prediction device according to an embodiment of this specification. The basic structure of the security prediction device is illustrated in this diagram. The functional units and modules can be implemented using software to achieve security prediction based on a large fractional-order power system model under network attacks, or they can be implemented using general-purpose chips or specific chips. The device specifically includes:

[0149] The first acquisition unit 701 is used to acquire the characteristic data of each subsystem in the power system in the second state;

[0150] The prediction result determination unit 702 is used to input the characteristic data of each subsystem in the power system under the second state into the intermittent network attack prediction model and the parameter perturbation prediction model to obtain the prediction results for each subsystem.

[0151] Figure 8 The diagram illustrates a computer device according to an embodiment of this specification. The method for constructing a large-scale model of a fractional-order power system under network attacks and for security prediction described in this application can be applied to the computer device. The computer device 802 may include one or more processors 804, such as one or more central processing units (CPUs), each of which can implement one or more hardware threads. The computer device 802 may also include any memory 806 for storing any kind of information such as code, settings, data, etc. Non-limitingly, for example, the memory 806 may include any type of RAM, any type of ROM, flash memory, hard disk, optical disk, etc. More generally, any memory can use any technology to store information. Further, any memory can provide volatile or non-volatile retention of information. Further, any memory can represent a fixed or removable component of the computer device 802. In one case, when the processor 804 executes associated instructions stored in any memory or combination of memories, the computer device 802 can perform any operation of the associated instructions. The computer device 802 also includes one or more drive mechanisms 808 for interacting with any memory, such as a hard disk drive mechanism, an optical disk drive mechanism, etc.

[0152] Computer device 802 may also include an input / output module 810 (I / O) for receiving various inputs (via input device 812) and providing various outputs (via output device 814). A specific output mechanism may include a presentation device 816 and an associated graphical user interface (GUI) 818. In other embodiments, the input / output module 810 (I / O), input device 812, and output device 814 may be omitted, and the device may function solely as a computer device within a network. Computer device 802 may also include one or more network interfaces 820 for exchanging data with other devices via one or more communication links 822. One or more communication buses 824 couple the components described above together.

[0153] Communication link 822 can be implemented in any way, such as via a local area network, a wide area network (e.g., the Internet), a point-to-point connection, or any combination thereof. Communication link 822 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0154] Corresponding to Figures 1 to 5 In addition to the method shown, embodiments of this specification also provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the above-described method.

[0155] This specification also provides computer-readable instructions, wherein when a processor executes the instructions, the program therein causes the processor to perform the following... Figures 1 to 5 The method shown.

[0156] It should be understood that in the various embodiments of this specification, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this specification.

[0157] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this specification generally indicates that the preceding and following related objects have an "or" relationship.

[0158] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this specification can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this specification.

[0159] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0160] In the several embodiments provided in this specification, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or units, or they may be electrical, mechanical, or other forms of connection.

[0161] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments described in this specification, depending on actual needs.

[0162] Furthermore, the functional units in the various embodiments of this specification can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0163] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this specification, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this specification. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0164] This specification uses specific embodiments to illustrate the principles and implementation methods of this specification. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this specification. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this specification. Therefore, the content of this specification should not be construed as a limitation of this specification.

Claims

1. A method for constructing a large-scale model of a fractional-order power system under network attacks, characterized in that, The method includes: Based on the feature sample dataset of each subsystem in the power system under the first state, construct the first fractional-order power system model. Based on the intermittent network attack modeling, the feature sample dataset of each subsystem in the second state, and the first fractional-order power system model, the second fractional-order power system model is determined, where the second state represents the occurrence of parameter perturbations and intermittent network attacks. Based on the feature sample dataset of each subsystem in the second state and the second fractional-order power system model, a fractional-order multiple regression model of the power system in the second state is constructed. Based on the aforementioned fractional-order multiple regression model of the power system, the sensitivity of each subsystem in the power system to parameter perturbations and intermittent network attacks is determined; Based on the aforementioned sensitivity, an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbation are constructed respectively. Based on the error indices of the initial prediction model for intermittent network attacks and the initial prediction model for parameter perturbation, the weight parameters in the initial prediction model for intermittent network attacks and the initial prediction model for parameter perturbation are iteratively updated to construct the intermittent network attack prediction model and the parameter perturbation prediction model. The intermittent network attack prediction model and the parameter perturbation prediction model constitute a large fractional-order power system model under network attacks.

2. The method according to claim 1, characterized in that, Based on the feature sample dataset of each subsystem in the power system under the first state, the first-order fractional model of the power system is constructed as follows: A fractional-order model of the power system is constructed using the following formula: ;in, Describes a fractional differential operator. Represents the state matrix of the power system. , This indicates the change in power generation from the hydroelectrolysis cell. This indicates the change in power generation from the fuel cell. This indicates the change in power generation from the diesel generator. This indicates the change in power generation from the battery energy storage system. This indicates the change in the output of the integral controller. This represents the frequency variation of the power system. Represents the state moments of the power system. This represents the time constant of the water electrolyzer. Represents the time constant of a fuel cell. Represents the diesel engine time constant; This represents the time constant of a battery energy storage system. Indicates the gain of the integral controller. Represents the frequency response coefficient of the power system. Represents the time constant of the power system. Represents the power system control matrix; Indicates the control input of the power system. , Indicates the gain of the water electrolyzer. Indicates fuel cell gain. Indicates the diesel engine gain. Indicates the gain of the battery energy storage system. , These represent the parameter perturbation matrices of the power system state matrix and control matrix, respectively. Indicates the output of the power system. Represents the power system output matrix; ;in, For function, For gamma function, For variables, For fractional order, For a higher integers, It is the integral variable.

3. The method according to claim 2, characterized in that, Based on the intermittent network attack modeling, the feature sample dataset of each subsystem in the second state, and the first fractional-order power system model, the second fractional-order power system model is determined to include: The fractional-order second model of the power system is constructed using the following formula: ;in, This represents the perturbation of fractional-order power system parameters. This represents intermittent network attacks related to the system state, determined by intermittent network attack modeling.

4. The method according to claim 1, characterized in that, Based on the aforementioned fractional-order multiple regression model of the power system, the sensitivity of each subsystem in the power system to parameter perturbations and intermittent network attacks is determined, including: The least squares method is used to estimate the regression coefficient matrix of multiple subsystems of the power system of fractional order; The sum of the second and fourth elements in each regression coefficient matrix is ​​used to determine the sensitivity of the subsystem to intermittent network attacks. The sum of the third and fifth elements in each regression coefficient matrix is ​​used to determine the sensitivity of the subsystem to parameter perturbations.

5. The method according to claim 4, characterized in that, Based on the aforementioned sensitivity, the initial prediction model for intermittent network attacks and the initial prediction model for parameter perturbations are constructed respectively, including: Based on the subsystem's sensitivity to intermittent network attacks, construct an intermittent network attack prediction weight matrix; Based on the subsystem's sensitivity to parameter perturbations, construct the parameter perturbation prediction weight matrix; Based on the intermittent network attack prediction weight matrix, determine the output of the hidden layer neurons of the initial prediction model for intermittent network attacks; Based on the parameter perturbation prediction weight matrix, determine the output of the hidden layer neurons of the parameter perturbation initial prediction model; The initial prediction model for intermittent network attacks is determined based on the outputs of the hidden layer neurons of the initial prediction model for intermittent network attacks and the output weights of the prediction model; the initial prediction model for parameter perturbation is determined based on the outputs of the hidden layer neurons of the initial prediction model for parameter perturbation and the output weights of the prediction model.

6. The method according to claim 1, characterized in that, Based on the aforementioned sensitivity, the initial prediction model for intermittent network attacks and the initial prediction model for parameter perturbations are constructed respectively, including: The feature sample datasets of each subsystem collected in the second state are input into the initial prediction model for intermittent network attacks to determine the output of the hidden layer neurons of the intermittent network attack prediction model: ; For the prediction model of intermittent network attacks of superspiral fractional-order constrained RBF neural networks, the first... m The output of each hidden layer neuron This represents the weight matrix for predicting intermittent network attacks or the weight matrix for predicting parameter perturbations. The superspiral fractional-order constrained RBF neural network intermittent network attack prediction model represents the first... m The center point vector values ​​of each hidden layer neuron The superspiral fractional-order constrained RBF neural network intermittent network attack prediction model represents the first... m The width of the Gaussian function of each hidden layer neuron, and subscript These represent the first, second, and third generations of the intermittent network attack prediction model for the superspiral fractional-order constrained RBF neural network. m One hidden layer neuron, The total number of hidden layer neurons in the intermittent network attack prediction model of the superspiral fractional-order constrained RBF neural network. j Indicates the first j An intermittent network attack prediction model, in Indicates input, This represents the input feature dataset of all hidden layer neurons in a superspiral fractional-order constrained RBF neural network intermittent network attack prediction model.

7. The method according to claim 6, characterized in that, The error indices of the initial prediction models for intermittent network attacks and parameter perturbation are determined by the following formulas: ;in, This represents an evaluation index for the prediction error of intermittent network attacks. This represents the weight of the prediction error for intermittent network attacks. This indicates an intermittent network attack on the system. Indicators for evaluating parameter perturbation prediction error Indicates the parameter perturbation error weight; Indicates system parameter perturbation, , These represent the initial prediction models for intermittent network attacks and parameter perturbation, respectively. , The observed values.

8. The method according to claim 1, characterized in that, Iterative updates to the weight parameters in the intermittent network attack initial prediction model and the parameter perturbation initial prediction model include: The superhelical algorithm is used to construct a superhelical path to guide the weights in the initial prediction model of intermittent network attacks and the initial prediction model of parameter perturbation to converge along the optimal direction; An exponentially dependent barrier function is used to constrain the outputs of the initial prediction models for intermittent network attacks and parameter perturbation.

9. A security prediction method, characterized in that, The method is applied to a large-scale fractional-order power system model under network attacks as described in any one of claims 1 to 8, including: Obtain characteristic data of each subsystem in the power system under the second state; The characteristic data of each subsystem in the power system under the second state are input into the intermittent network attack prediction model and the parameter perturbation prediction model to obtain the prediction results for each subsystem.

10. A device for constructing a large-scale model of a fractional-order power system under network attacks, characterized in that, The device includes: The first building unit is used to build a first-order fractional model of the power system based on the feature sample dataset of each subsystem in the power system under the first state. The second construction unit is used to determine the second fractional-order power system model based on the intermittent network attack modeling, the feature sample dataset of each subsystem in the second state, and the first fractional-order power system model. The second state represents the occurrence of parameter perturbations and intermittent network attacks. The third construction unit is used to construct a fractional-order multiple regression model of the power system in the second state based on the feature sample dataset of each subsystem in the second state and the second fractional-order power system model. The determining unit is used to determine the sensitivity of each subsystem in the power system to parameter perturbations and intermittent network attacks based on the fractional-order multiple regression model of the power system. The fourth construction unit is used to construct an initial prediction model for intermittent network attacks and an initial prediction model for parameter perturbation based on the sensitivity, respectively. The fifth construction unit is used to iteratively update the weight parameters in the intermittent network attack initial prediction model and the parameter perturbation initial prediction model based on the error index of the intermittent network attack initial prediction model and the parameter perturbation initial prediction model, thereby constructing the intermittent network attack prediction model and the parameter perturbation prediction model. The intermittent network attack prediction model and the parameter perturbation prediction model constitute a fractional-order power system large model under network attack.

11. A safety prediction device, characterized in that, The apparatus, when applied to the intermittent network attack prediction model and parameter perturbation prediction model in any one of claims 1 to 8, and when applied to the method of claim 9, includes: The first acquisition unit is used to acquire characteristic data of each subsystem in the power system under the second state; The prediction result determination unit is used to input the characteristic data of each subsystem in the power system under the second state into the intermittent network attack prediction model and the parameter perturbation prediction model to obtain the prediction results for each subsystem.

12. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 9.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 9.

Images