Distribution network information system risk assessment method and system based on graph convolutional network

Abstract

This invention discloses a risk assessment method and system for distribution network information systems based on graph convolutional networks (GCNs), belonging to the field of distribution network cyber-physical systems technology. It includes combining LSTM to capture the temporal dynamic characteristics during fault propagation to accurately assess the risk level of node voltage exceedance; using node voltage exceedance as the core evaluation criterion to comprehensively reflect the risk distribution under different attack scenarios; and learning the electrical characteristics of the system under non-fault conditions to improve assessment accuracy and complete the risk assessment of the distribution network information system. By introducing a combined model based on graph convolutional networks (GCNs) and long short-term memory (LSTMs), this invention can effectively capture the spatiotemporal characteristics of distribution network cyber-physical systems, accurately assess the risk level of node voltage exceedance, and significantly improve the accuracy and real-time performance of risk assessment. This invention can not only identify high-risk nodes but also provide reliable support for the safe operation of the power grid, thereby improving the security and resilience of the distribution network.

Description

Technical Field

[0001] This invention relates to the field of cyber-physical systems technology for power distribution networks, and in particular to a risk assessment method and system for power distribution network information systems based on graph convolutional networks. Background Technology

[0002] Currently, with the rapid development of distributed renewable energy and the in-depth application of information and communication technologies, the cyber-physical system of the distribution network has become an important part of the modern smart grid. In CPDS, the information system is highly coupled with the physical power system, which not only realizes real-time monitoring and automated control, but also exposes the system to more and more information security threats, such as attacks on the downlink communication link of control commands. Such attacks may cause controllable resources to go offline, and further cause node voltage to exceed the limit, seriously threatening the safe operation of the distribution network.

[0003] Currently, risk assessment methods for distribution networks mainly focus on static analysis and single-scenario simulation, lacking a comprehensive capture of the dynamic characteristics of the system. Traditional assessment methods rely on the static process of gradual repair after a fault, making it difficult to reflect the spatiotemporal correlation of information attacks on the operating status of the distribution network. This method has significant shortcomings in dealing with complex topologies and dynamic fault propagation, and cannot meet the needs of efficient assessment for modern smart distribution networks.

[0004] In recent years, risk assessment methods based on deep learning have gradually attracted attention, especially those that capture the complex dependencies between power grid operating states and fault characteristics through neural networks. For example, convolutional neural networks can extract the spatial features of the power grid, but their ability to capture temporal dynamic characteristics is limited. Long Short-Term Memory (LSTM) networks can handle time-series data well, but cannot directly reflect the power grid topology. Therefore, graph convolutional networks, as a type of neural network capable of processing graph-structured data, have been used to extract the spatial topological characteristics of the power grid. Combined with LSTM to capture temporal dynamic changes, this forms a risk assessment method that captures both spatial and temporal characteristics. By integrating these two networks, the global impact of information attacks on CPDS can be reflected more accurately. Summary of the Invention

[0005] In view of the existing risk assessment methods and problems in the distribution network information system based on graph convolutional networks, this invention is proposed.

[0006] Therefore, the problem to be solved by this invention is that while convolutional neural networks can extract the spatial features of the power grid, their ability to capture temporal dynamic characteristics is limited. Long short-term memory networks can handle time series data well, but cannot directly reflect the power grid topology. This invention adopts a risk assessment method for distribution network information systems based on graph convolutional networks to solve this problem.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, embodiments of the present invention provide a risk assessment method for a distribution network information system based on graph convolutional networks, which includes generating a distribution network cyber-physical system through simulation modeling, operating data under information attacks, extracting the complex topology and electrical characteristics of the system using GCN, and combining LSTM to capture the time dynamic characteristics during fault propagation to accurately assess the risk level of node voltage exceeding limits.

[0009] Design a simulation modeling method for controllable resource communication interruption under information attack. Generate power grid fault samples through Monte Carlo simulation, and define distribution network risk assessment indicators based on simulation results. Take node voltage over-limit as the core evaluation standard to comprehensively reflect the risk distribution under different attack scenarios.

[0010] The input data of the GCN-LSTM model is normalized to a uniform length, and the generalization ability of the model is enhanced by normalization processing technology. The model learns the electrical characteristics of the system under non-fault conditions, improves the evaluation accuracy, and completes the risk assessment of the distribution network information system.

[0011] As a preferred embodiment of the risk assessment method for distribution network information systems based on graph convolutional networks described in this invention, the step of generating a distribution network cyber-physical system through simulation modeling includes attacks on the downlink communication link of control commands, which is accomplished in the following ways;

[0012] In a CDLA attack, the state of the distribution network communication links is modified, specifically as follows:

[0013]

[0014] In the formula, l i-C These are the control commands issued by the master station to the controllable resources at node i before the attack. These are the control commands issued by the master station to the controllable resources at node i before the attack.

[0015] CDLA attacks can cause controllable resources such as wind and solar power connected to a node to go offline, leading to excessive node voltage.

[0016] As a preferred embodiment of the risk assessment method for distribution network information systems based on graph convolutional networks described in this invention, the step of generating power grid fault samples through Monte Carlo simulation includes simulating different types of faults occurring at different locations and times, and completing the fault analysis caused by information attacks through the following steps:

[0017] Initialize system parameters, set generator capacity, generator output, load and power system topology before CDLA attack, and randomly generate initial faults at different locations and time points using the MCS method, setting the initial time T=0;

[0018] Power flow calculation and data recording: Perform power flow calculation, update time T=T+1, record the system data of fault N at time T, and determine whether node voltage exceeds the limit. In order to simulate the fault repair process, the power flow calculation interval is 15 minutes, and the simulation duration is set to T=8.

[0019] Fault status judgment: If a voltage over-limit occurs, the simulation of the current fault N ends and enters fault N+1; if T<8 ​​and no voltage over-limit occurs, return to the power flow calculation step and continue the simulation.

[0020] As a preferred embodiment of the risk assessment method for distribution network information systems based on graph convolutional networks described in this invention, the definition of distribution network risk assessment indicators includes defining node voltage exceedance limits, expressed by the following formula:

[0021]

[0022] Where R i U is the risk assessment index for node i. i Let be the voltage at node i;

[0023] The method of enhancing the model's generalization ability by combining normalization processing techniques includes using data augmentation and normalization techniques to improve the model's generalization ability and ensure the uniformity of the input data. The specific steps are as follows:

[0024] The data augmentation includes inserting a non-fault feature map with a time length of 10-T into the beginning of the sequence when the fault sequence length is T, so that the time length of all fault map sequences is normalized to 10.

[0025] The normalization process involves standardizing all node data to a data format with a mean of 0 and a standard deviation of 1, thereby enhancing the model's adaptability to different data distributions.

[0026] As a preferred embodiment of the risk assessment method for distribution network information systems based on graph convolutional networks described in this invention, the method of learning the electrical characteristics of the system under non-fault conditions includes constructing a risk assessment framework by combining graph convolutional networks and long short-term memory networks, and completing data-driven risk prediction by constructing a GCN-LSTM model. This model consists of a graph convolutional network (GCN) and a long short-term memory network (LSTM). Taking the failure graph sequence generated by simulation as input, the GCN is used to extract the topology and electrical characteristics of the power system, thereby capturing the correlation between nodes. The failure graph sequence containing the topology and electrical characteristics is input into the LSTM layer. The dynamic changes in the fault propagation process are captured through information transmission between LSTM units, and the risk assessment results are output. Combined with the spatiotemporal characteristics of distribution network faults, the topology feature extraction model of the risk assessment framework is completed.

[0027] As a preferred embodiment of the risk assessment method for power distribution network information systems based on graph convolutional networks described in this invention, the topology feature extraction model includes constructing a power system topology feature extraction model using graph convolutional networks, obtaining the system's topology structure by learning the complex dependencies between electrical features, and the output of the GCN layer is represented as follows:

[0028]

[0029] In the formula, X and Y are the input and output matrices of the GCN layer, respectively. X represents the feature matrix of the power system network, with each row corresponding to the electrical characteristics on the bus. A is the adjacency matrix. Let the degree matrix be defined as:

[0030]

[0031] Where W is the trainable parameter matrix of the GCN layer, I is the identity matrix, and f σ (·) is the activation function, and the proposed method uses the active power P of the node. i reactive power Q i Voltage amplitude U i Voltage phase angle θ i Photovoltaic generator power P vi and wind turbine power P wi As a characteristic representing the state of a power system;

[0032] The topology of the acquisition system includes capturing the spatial correlation of the power grid, using a two-layer GCN model, the output of which is expressed as:

[0033]

[0034] Among them, X GCN and Y GCNW1 and W2 represent the input and output of the 2-layer GCN model, respectively, and represent the trainable parameter matrices of the first and second layers of the GCN, respectively.

[0035] As a preferred embodiment of the risk assessment method for distribution network information systems based on graph convolutional networks described in this invention, the specific formula for the Long Short-Term Memory (LSTM) network is as follows:

[0036] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0037] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0038]

[0039] O t =σ(W o ·[h t-1 ,x t ]+b o )

[0040] h t =O t ⊙tanh(C t )

[0041] Among them, f t It is the output of the forget gate, σ is the sigmoid activation function, and W is the output of the forget gate. f This is the weight matrix of the forget gate, b f , b i , b c and b o It is the bias vector, h t-1 It is the hidden state of the previous time step, x t It is the input at the current time step, i t It is the output of the input gate, W i It is the weight matrix of the input gate. It is the state of the candidate memory cell, W c It is the weight matrix of the candidate memory units, C t It represents the cell state at the current time step, C. t-1 This represents the cell state at the previous time step. ⊙ indicates the Hadamard product, which is the element-wise multiplication. O t It is the output of the output gate, W o It is the weight matrix of the output gate, h tIt is the hidden state of the current time step.

[0042] Secondly, embodiments of the present invention provide a risk assessment system for distribution network information systems based on graph convolutional networks (GCNs), comprising: an extraction module, which generates operational data of the distribution network cyber-physical system under information attacks through simulation modeling, extracts the complex topology and electrical characteristics of the system using GCNs, and combines LSTMs to capture the temporal dynamic characteristics during fault propagation to accurately assess the risk level of node voltage exceeding limits; a design module, which designs a simulation modeling method for controllable resource communication interruption under information attacks, generates power grid fault samples through Monte Carlo simulation, and defines distribution network risk assessment indicators based on simulation results, with node voltage exceeding limits as the core evaluation standard to comprehensively reflect the risk distribution under different attack scenarios; and an assessment module, which normalizes the input data of the GCN-LSTM model to a uniform length, enhances the generalization ability of the model by combining normalization processing technology, learns the electrical characteristics of the system under non-fault conditions, improves the assessment accuracy, and completes the risk assessment of the distribution network information system.

[0043] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of the above-described risk assessment method for power distribution network information systems based on graph convolutional networks.

[0044] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the above-described risk assessment method for power distribution network information systems based on graph convolutional networks.

[0045] The beneficial effects of this invention are as follows: By introducing a combined model based on graph convolutional network (GCN) and long short-term memory network (LSTM), this invention can effectively capture the spatiotemporal characteristics of the distribution network cyber-physical system, accurately assess the risk level of node voltage exceeding limits, and significantly improve the accuracy and real-time performance of risk assessment. This invention can not only identify high-risk nodes, but also provide reliable support for the safe operation of the power grid, thereby improving the safety and resilience of the distribution network. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0047] Figure 1The present invention provides a flowchart of a risk assessment method and system for a power distribution network information system based on graph convolutional networks, which is an embodiment of the present invention.

[0048] Figure 2 The flowchart of the GCN-LSTM model of the risk assessment method and system for distribution network information system based on graph convolutional networks provided in one embodiment of the present invention is shown. Detailed Implementation

[0049] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0050] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0051] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0052] This invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of this invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not adhering to the usual scale. Furthermore, the schematic diagrams are merely examples and should not be construed as limiting the scope of protection of this invention. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0053] Furthermore, in the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0054] Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" in this invention should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0055] Example 1

[0056] Reference Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a risk assessment method for a power distribution network information system based on graph convolutional networks, including:

[0057] S1: By generating the cyber-physical system of the distribution network through simulation modeling, the operating data under information attack is used to extract the complex topology and electrical characteristics of the system using GCN, and combined with LSTM to capture the time dynamic characteristics during the fault propagation process, the risk level of node voltage exceeding the limit is accurately assessed.

[0058] Among them, generating the cyber-physical system of the power distribution network through simulation modeling includes attacks on the downlink communication link of control commands, which is accomplished in the following ways;

[0059] In a CDLA attack, the state of the distribution network communication links is modified, specifically as follows:

[0060]

[0061] In the formula, l i-C These are the control commands issued by the master station to the controllable resources at node i before the attack. These are the control commands issued by the master station to the controllable resources at node i before the attack.

[0062] CDLA attacks can cause controllable resources such as wind and solar power connected to a node to go offline, leading to excessive node voltage.

[0063] S2: Design a simulation modeling method for controllable resource communication interruption under information attack. Generate power grid fault samples through Monte Carlo simulation, and define distribution network risk assessment indicators based on simulation results. The core evaluation standard is node voltage over-limit situation, which comprehensively reflects the risk distribution under different attack scenarios.

[0064] The generation of power grid fault samples through Monte Carlo simulation includes simulations of different types of faults occurring at different locations and times. The analysis of faults caused by information attacks is completed through the following steps:

[0065] Initialize system parameters, set generator capacity, generator output, load and power system topology before CDLA attack, and randomly generate initial faults at different locations and time points using the MCS method, setting the initial time T=0;

[0066] Power flow calculation and data recording: Perform power flow calculation, update time T=T+1, record the system data of fault N at time T, and determine whether node voltage exceeds the limit. In order to simulate the fault repair process, the power flow calculation interval is 15 minutes, and the simulation duration is set to T=8.

[0067] Fault status judgment: If a voltage over-limit occurs, the simulation of the current fault N ends and enters fault N+1; if T<8 ​​and no voltage over-limit occurs, return to the power flow calculation step and continue the simulation.

[0068] S2.1: Define the risk assessment indicators for the distribution network, including defining node voltage exceedances, expressed by the following formula:

[0069]

[0070] Where R i U is the risk assessment index for node i. i Let be the voltage at node i;

[0071] Enhancing the generalization ability of a model by combining normalization techniques includes employing data augmentation and normalization methods to improve the model's generalization ability and ensure the uniformity of the input data. The specific steps are as follows:

[0072] Data augmentation involves inserting a non-fault feature map with a time length of 10-T into the beginning of the fault map sequence, given a fault sequence length of T, so that the time length of all fault map sequences is standardized to 10.

[0073] Normalization involves standardizing all node data to a data format with a mean of 0 and a standard deviation of 1, thereby enhancing the model's adaptability to different data distributions.

[0074] S3: Normalize the input data of the GCN-LSTM model to a uniform length, and combine normalization processing technology to enhance the generalization ability of the model. Learn the electrical characteristics of the system under non-fault conditions, improve the evaluation accuracy, and complete the risk assessment of the distribution network information system.

[0075] Among them, learning the electrical characteristics of the system under non-fault conditions includes constructing a risk assessment framework by combining graph convolutional networks and long short-term memory networks. Data-driven risk prediction is completed by constructing a GCN-LSTM model. This model consists of a graph convolutional network (GCN) and a long short-term memory network (LSTM). Taking the failure graph sequence generated by simulation as input, the GCN is used to extract the topology and electrical characteristics of the power system, thereby capturing the correlation between nodes. The failure graph sequence containing the topology and electrical characteristics is input into the LSTM layer. The dynamic changes in the fault propagation process are captured through information transmission between LSTM units, and the risk assessment results are output. Combined with the spatiotemporal characteristics of distribution network faults, the topology feature extraction model of the risk assessment framework is completed.

[0076] S3.1: The topology feature extraction model includes constructing a topology feature extraction model for the power system using graph convolutional networks (GCNs). It obtains the system's topology by learning the complex dependencies between electrical features. The output of the GCN layer is represented as:

[0077]

[0078] In the formula, X and Y are the input and output matrices of the GCN layer, respectively. X represents the feature matrix of the power system network, with each row corresponding to the electrical characteristics on the bus. A is the adjacency matrix. Let the degree matrix be defined as:

[0079]

[0080] Where W is the trainable parameter matrix of the GCN layer, I is the identity matrix, and f σ (·) is the activation function, and the proposed method uses the active power P of the node. i reactive power Q i Voltage amplitude U i Voltage phase angle θ i Photovoltaic generator power P vi and wind turbine power P wi As a characteristic representing the state of a power system;

[0081] Obtaining the system's topology includes capturing the spatial correlation of the power grid. A two-layer GCN model is used, and its output is represented as follows:

[0082]

[0083] Among them, X GCN and Y GCN W1 and W2 represent the input and output of the 2-layer GCN model, respectively, and represent the trainable parameter matrices of the first and second layers of the GCN, respectively.

[0084] S3.2: The specific formula for constructing a Long Short-Term Memory (LSTM) network is as follows:

[0085] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0086] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0087]

[0088] O t =σ(W o ·[h t-1 ,x t ]+b o )

[0089] h t =O t ⊙tanh(C t )

[0090] Among them, f t It is the output of the forget gate, σ is the sigmoid activation function, and W is the output of the forget gate. f This is the weight matrix of the forget gate, b f b i b c and b o It is the bias vector, h t-1 It is the hidden state of the previous time step, x t It is the input at the current time step, i t It is the output of the input gate, W i It is the weight matrix of the input gate. It is the state of the candidate memory cell, W c It is the weight matrix of the candidate memory units, C t It represents the cell state at the current time step, C. t-1 This represents the cell state at the previous time step. ⊙ indicates the Hadamard product, which is the element-wise multiplication. O t It is the output of the output gate, W o It is the weight matrix of the output gate, h t It is the hidden state of the current time step.

[0091] In a preferred embodiment, a risk assessment system for a distribution network information system based on graph convolutional networks (GCNs) includes an extraction module that generates operational data of the distribution network cyber-physical system under information attacks through simulation modeling. It utilizes GCNs to extract the complex topology and electrical characteristics of the system and combines LSTMs to capture the temporal dynamics of fault propagation, accurately assessing the risk level of node voltage exceedances. A design module designs a simulation modeling method for controllable resource communication interruptions under information attacks. It generates power grid fault samples through Monte Carlo simulation and defines distribution network risk assessment indicators based on the simulation results, using node voltage exceedances as the core evaluation criterion to comprehensively reflect the risk distribution under different attack scenarios. An assessment module normalizes the input data of the GCN-LSTM model to a uniform length and enhances the model's generalization ability through normalization processing techniques. It learns the electrical characteristics of the system under non-fault conditions, improving assessment accuracy and completing the risk assessment of the distribution network information system.

[0092] The above-mentioned unit modules can be embedded in the processor of the computer device in hardware form or independent of it, or they can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0093] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0094] In summary, by introducing a combined model based on Graph Convolutional Network (GCN) and Long Short-Term Memory (LSTM), this invention can effectively capture the spatiotemporal characteristics of the distribution network cyber-physical system, accurately assess the risk level of node voltage exceeding limits, and significantly improve the accuracy and real-time performance of risk assessment. This invention can not only identify high-risk nodes but also provide reliable support for the safe operation of the power grid, thereby improving the safety and resilience of the distribution network.

[0095] Example 2

[0096] Reference Figure 1 and Figure 2 This is the second embodiment of the present invention, which provides a risk assessment method for a power distribution network information system based on graph convolutional networks. To verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0097] A simulated power distribution network model with 14 nodes was constructed. This model can generate operational data under different information attack scenarios. For attacks on the downlink communication link of control commands, the state of the power distribution network communication link is modified to simulate the changes in the master station's control commands to controllable resources at the nodes before and after the attack. Specifically, the attack caused controllable resources such as wind power and photovoltaics to go offline, which in turn caused the node voltage to exceed the limit. Using the Monte Carlo simulation method, we generated 10,000 sets of power grid fault samples. These samples cover different types of faults that occur at different locations and at different times. In each simulation cycle, the system parameters were initialized, including a generator capacity of 50MW, an output power of 30MW, a load of 40MW, and the topology of the power system.

[0098] Power flow calculations are performed at 15-minute intervals, updating time T = T+1, recording system data for fault N at time T, and determining whether node voltage exceedance has occurred. The simulation duration is set to an upper limit of 8 hours. For each fault scenario, if node voltage exceedance occurs at any time point, defined as voltage exceeding ±10% of the rated value, the simulation of the current fault N is immediately terminated and the simulation of the next fault N+1 begins. If no voltage exceedance occurs within 8 hours, the simulation continues until the upper limit of the duration is reached.

[0099] To train and test the GCN-LSTM model, all fault samples were divided into a training set (70%), a validation set (15%), and a test set (15%). The model input data was normalized to a uniform length, meaning the time length of all fault sequences was standardized to 10 time steps. Non-fault feature maps were inserted at the beginning of the sequences to pad to 10 time steps. Furthermore, all node data were normalized to a form with a mean of 0 and a standard deviation of 1 to enhance the model's generalization ability. Ultimately, the GCN-LSTM model achieved an accuracy of 92.3% on the test set, indicating that the model can effectively predict the risk level of node voltage exceedance in the distribution network under information attacks, thus providing important decision support for the safe and stable operation of the power grid. Experimental data are shown in Table 1 below.

[0100] Table 1 Experimental Data

[0101]

[0102] Table 1 shows how a simulated power distribution network model was constructed to simulate power grid operation under information attacks, generating a large number of fault samples. Risk assessment was then performed through power flow calculations and voltage limit detection. After data standardization, the GCN-LSTM model was used for training and testing, ultimately verifying the model's effectiveness in predicting node voltage limit risks and providing decision support for the safe and stable operation of the power grid. A comparison of this invention with existing technologies is shown in Table 2 below.

[0103] Table 2 Comparison of the present invention with the prior art

[0104]

[0105] Table 2 shows significant advantages in terms of the accuracy of risk assessment, automation of data processing, capture of topology, consideration of time dynamics, and generalization ability of the model.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A risk assessment method for power distribution network information systems based on graph convolutional networks, characterized in that: include, By generating a cyber-physical system of the distribution network through simulation modeling, and using operational data under information attack, the complex topology and electrical characteristics of the system are extracted using GCN, and the time dynamic characteristics of the fault propagation process are captured by LSTM to accurately assess the risk level of node voltage exceeding the limit. Design a simulation modeling method for controllable resource communication interruption under information attack. Generate power grid fault samples through Monte Carlo simulation, and define distribution network risk assessment indicators based on simulation results. Take node voltage over-limit as the core evaluation standard to comprehensively reflect the risk distribution under different attack scenarios. The input data of the GCN-LSTM model is normalized to a uniform length, and the generalization ability of the model is enhanced by normalization processing technology. The model learns the electrical characteristics of the system under non-fault conditions, improves the evaluation accuracy, and completes the risk assessment of the distribution network information system.

2. The risk assessment method for distribution network information systems based on graph convolutional networks as described in claim 1, characterized in that: The generation of the distribution network cyber-physical system through simulation modeling includes attacks on the downlink communication link of control commands, which are accomplished in the following ways; In a CDLA attack, the state of the distribution network communication links is modified, specifically as follows: In the formula, l i-C These are the control commands issued by the master station to the controllable resources at node i before the attack. These are the control commands issued by the master station to the controllable resources at node i before the attack. CDLA attacks can cause controllable wind and solar power resources connected to a node to go offline, leading to node voltage exceeding limits.

3. The risk assessment method for distribution network information systems based on graph convolutional networks as described in claim 2, characterized in that: The generation of power grid fault samples through Monte Carlo simulation includes simulating different types of faults occurring at different locations and times, and the analysis of faults caused by information attacks is completed through the following steps: Initialize system parameters, set generator capacity, generator output, load and power system topology before CDLA attack, and randomly generate initial faults at different locations and time points using the MCS method, setting the initial time T=0; Power flow calculation and data recording: Perform power flow calculation, update time T=T+1, record the system data of fault N at time T, and determine whether node voltage exceeds the limit. In order to simulate the fault repair process, the power flow calculation interval is 15 minutes, and the simulation duration upper limit is set to T=8. Fault status judgment: If a voltage over-limit occurs, the simulation of the current fault N ends and enters fault N+1; if T<8 ​​and no voltage over-limit occurs, return to the power flow calculation step and continue the simulation.

4. The risk assessment method for distribution network information systems based on graph convolutional networks as described in claim 3, characterized in that: The defined distribution network risk assessment indicators include defining node voltage exceedances, expressed by the following formula: In the formula, R i U is the risk assessment index for node i. i Let be the voltage at node i; The method of enhancing the model's generalization ability by combining normalization processing techniques includes using data augmentation and normalization techniques to improve the model's generalization ability and ensure the uniformity of the input data. The specific steps are as follows: The data augmentation includes inserting a non-fault feature map with a time length of 10-T into the beginning of the sequence when the fault sequence length is T, so that the time length of all fault map sequences is normalized to 10. The normalization process involves standardizing all node data to a data format with a mean of 0 and a standard deviation of 1, thereby enhancing the model's adaptability to different data distributions.

5. The risk assessment method for distribution network information systems based on graph convolutional networks as described in claim 4, characterized in that: The method for learning the electrical characteristics of the system under non-fault conditions includes constructing a risk assessment framework by combining graph convolutional networks and long short-term memory networks. A data-driven risk prediction is achieved by building a GCN-LSTM model, which consists of a graph convolutional network (GCN) and a long short-term memory network (LSTM). Taking the simulated failure graph sequence as input, the GCN extracts the topology and electrical characteristics of the power system, thereby capturing the relationships between nodes. The failure graph sequence containing the topology and electrical characteristics is then input into the LSTM layer. Information transmission between LSTM units captures the dynamic changes during fault propagation, outputting risk assessment results. Combined with the spatiotemporal characteristics of distribution network faults, the topology feature extraction model of the risk assessment framework is completed.

6. The risk assessment method for distribution network information systems based on graph convolutional networks as described in claim 5, characterized in that: The topology feature extraction model includes constructing a power system topology feature extraction model using a graph convolutional network (GCN). It obtains the system's topology by learning the complex dependencies between electrical features. The output of the GCN layer is represented as follows: In the formula, X and Y are the input and output matrices of the GCN layer, respectively. X represents the feature matrix of the power system network, with each row corresponding to the electrical characteristics on the bus. A is the adjacency matrix. Let the degree matrix be defined as: Where W is the trainable parameter matrix of the GCN layer, I is the identity matrix, and f σ (·) is the activation function, and the proposed method uses the active power P of the node. i reactive power Q i Voltage amplitude U i Voltage phase angle θ i Photovoltaic generator power P vi and wind turbine power P wi As a characteristic representing the state of a power system; The topology of the acquisition system includes capturing the spatial correlation of the power grid, using a two-layer GCN model, the output of which is expressed as: Among them, X GCN and Y GCN W1 and W2 represent the input and output of the 2-layer GCN model, respectively, and represent the trainable parameter matrices of the first and second layers of the GCN, respectively.

7. The risk assessment method for distribution network information systems based on graph convolutional networks as described in claim 6, characterized in that: The specific formula for the Long Short-Term Memory (LSTM) network is as follows: f t =σ(W f ·[h t-1 ,x t ]+b f ) i t =σ(W i ·[h t-1 ,x t ]+b i ) The t =σ(W o ·[h t-1 ,x t ]+b o ) h t =O t ⊙tanh(C t ) Among them, f t It is the output of the forget gate, σ is the sigmoid activation function, and W is the output of the forget gate. f This is the weight matrix of the forget gate, b f b i b c and b o It is the bias vector, h t-1 It is the hidden state of the previous time step, x t It is the input at the current time step, i t It is the output of the input gate, W i It is the weight matrix of the input gate. It is the state of the candidate memory cell, W c It is the weight matrix of the candidate memory units, C t It represents the cell state at the current time step, C. t-1 This represents the cell state at the previous time step. ⊙ indicates the Hadamard product, which is the element-wise multiplication. O t It is the output of the output gate, W o It is the weight matrix of the output gate, h t It is the hidden state of the current time step.

8. A risk assessment system for distribution network information systems based on graph convolutional networks, based on the risk assessment method for distribution network information systems based on graph convolutional networks as described in any one of claims 1 to 7, characterized in that: include, The extraction module generates the cyber-physical system of the distribution network through simulation modeling, and extracts the operating data of the system under information attack using GCN. It also uses LSTM to capture the time dynamic characteristics of the fault propagation process and accurately assesses the risk level of node voltage exceeding the limit. The design module designs a simulation modeling method for controllable resource communication interruption under information attack. It generates power grid fault samples through Monte Carlo simulation and defines distribution network risk assessment indicators based on simulation results. The core evaluation standard is node voltage over-limit situation, which comprehensively reflects the risk distribution under different attack scenarios. The evaluation module normalizes the input data of the GCN-LSTM model to a uniform length and enhances the model's generalization ability by combining normalization processing technology. It learns the electrical characteristics of the system under non-fault conditions, improves the evaluation accuracy, and completes the risk assessment of the distribution network information system.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the risk assessment method for power distribution information systems based on graph convolutional networks as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the risk assessment method for power distribution network information system based on graph convolutional networks as described in any one of claims 1 to 7.

Images