An emergency control method and medium for information-physical cross-space collaboration under false data injection attacks

Abstract

The present invention provides an emergency control method and medium for information-physical cross-space collaboration under false data injection attacks, which belongs to the field of smart grid technology, including: constructing an emergency control model according to information network attack characteristics and physical system operation status; the emergency control model includes: an information layer data recovery model that optimizes the accuracy and deviation rate of data recovery and a physical layer distribution network reconstruction model that optimizes the network topology; using a multi-task residual attention hypergraph neural network to solve the emergency control model, and obtaining the best data recovery plan and the best distribution network topology reconstruction strategy; performing data recovery according to the best data recovery plan, and realizing fast switching of distribution network topology according to the best distribution network topology reconstruction strategy. The present invention considers the impact of the accuracy and real-time performance of data recovery on the reconstruction of the distribution network, constructs an emergency control model based on the collaboration of data recovery and distribution network reconstruction, improves the emergency response capability of the distribution network, and ensures the safe operation of the distribution network.

Description

Technical Field

[0001] The invention relates to an emergency control method and medium for information-physical cross-space collaboration under false data injection attacks, and belongs to the technical field of smart grids. Background Art

[0002] At present, large-scale distributed power sources such as wind power and photovoltaics, as well as power electronic equipment, are widely connected to distribution networks. Their low interference resistance, weak support, and low inertia characteristics have brought severe challenges to the safety and reliability of distribution network operation.

[0003] As a cyber-physical system, the distribution network presents the characteristics of deep integration of information and physics, which provides an opportunity for malicious attacks that endanger the safe operation of the distribution network. Malicious network attacks such as false data injection attacks often use the weak parts of key links such as data acquisition and mobile communications as invasion channels, thereby causing grid dispatching decision errors or control system function failures. The accuracy and real-time performance of data recovery directly affect the performance of distribution reconstruction under emergency conditions. Among them, the accuracy of data recovery will restrict the optimal selection of reconstruction strategies, while the real-time performance of data recovery will hinder the timely deployment of emergency control, but these two are often difficult to take into account at the same time. Therefore, after suffering from false data injection attacks, it is necessary to collaboratively design physical layer distribution network reconstruction and information layer data recovery solutions to quickly restore the load / voltage distribution balance of the entire distribution system and ensure the safe operation of the distribution network.

[0004] In addition, in practical applications, due to factors such as communication problems or limitations of the equipment itself, the remote control of remote switches has unreliable problems such as refusal to operate and false operation. Existing studies often ignore the reliability of switches when considering the problem of distribution network reconstruction, which may increase the risk of power outages. In addition, differences in communication delays can cause the remote switch to operate asynchronously, which may cause a sudden drop in voltage and system instability. Therefore, the synchronization of remote switch switching during the distribution network reconstruction execution phase needs further research. Summary of the invention

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and provide an emergency control method and medium for information-physical cross-space collaboration under false data injection attacks, thereby solving the problem that it is difficult to simultaneously take into account the accuracy and real-time performance of existing data recovery.

[0006] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0007] In a first aspect, the present invention provides an emergency control method for information-physical cross-space collaboration under a false data injection attack, comprising:

[0008] Obtain an emergency control model pre-built according to the characteristics of information network attacks and the operating status of the physical system; wherein the emergency control model includes: an information layer data recovery model that optimizes the accuracy and deviation rate of data recovery and a physical layer distribution network reconstruction model that optimizes the network topology;

[0009] A pre-built multi-task residual attention hypergraph neural network is used to solve the emergency control model to obtain the best data recovery solution and the best distribution network topology reconstruction strategy;

[0010] Data recovery is performed according to the optimal data recovery solution, and corresponding action instructions are sent to the execution terminal according to the optimal distribution network topology reconstruction strategy to achieve rapid switching of the distribution network topology.

[0011] Furthermore, the information layer data recovery model aims to find the best trade-off between the real-time and accuracy of data recovery to adapt to the physical layer distribution network reconstruction after suffering a false data injection attack. Its expression is:

[0012] ;

[0013] in, To minimize, is the weight coefficient of time cost, is the weight coefficient of load loss cost, Indicates that the accuracy rate is , the data deviation rate is Optimal distribution network reconstruction strategy for data recovery, To take the accuracy rate , the data deviation rate is The data recovery time corresponding to the data recovery, Optimal distribution network reconstruction strategy The maximum communication delay is For bus The importance of the load, Optimal distribution network reconstruction strategy Lower bus The actual load loss, is the penalty coefficient for power flow exceeding the limit, is the total number of buses, is the power flow over-limit factor of the optimal distribution network reconstruction strategy under real operation data. When there is a power flow over-limit, ,otherwise .

[0014] Furthermore, the information layer data recovery model is constrained by the time and accuracy of data recovery and the actual load loss of each bus, and the expression is:

[0015] ;

[0016] ;

[0017] ;

[0018] ;

[0019] ;

[0020] ;

[0021] in, is the amount of data that is accurately recovered, is the amount of data recovered with deviation, is the amount of data that needs to be restored, is the amount of data that has not been tampered with and does not need to be restored. is the total amount of information data, The time required to accurately restore a piece of information data, The time required to recover a piece of information data with deviation, Reconstruction strategy The busbar estimated based on the restored data The load loss, The actual bus The maximum load that can be cut.

[0022] Furthermore, the physical layer distribution network reconstruction model aims to restore the normal operation of the distribution network under the condition of minimizing the loss of important loads, and its expression is:

[0023] ;

[0024] in, To minimize, is the weight coefficient of remote control reliability of the remote control switch, is the weight coefficient of load loss cost, For branch The initial state of the upper remote control switch, For branch The target state of the upper remote switch, is the action reliability coefficient of the remote control switch, For bus The importance of the load, To restore data based on Estimated busbar The load loss, is the set of all branches in the distribution network, is the total number of buses.

[0025] Furthermore, the physical layer distribution network reconstruction model takes the distribution network operation state as a constraint, including: power flow balance constraint, load loss constraint, topology radial constraint, node voltage constraint and line capacity constraint, and the expression is:

[0026] ;

[0027] ;

[0028] ;

[0029] ;

[0030] ;

[0031] ;

[0032] ;

[0033] in, For branch The active power, For branch The reactive power, To restore data based on Busbar The next active power generation, To restore data based on Busbar Reactive power generation, To restore data based on Busbar The active load demand is To restore data based on Busbar The reactive load demand, , Indicates busbar , The voltage amplitude at , Indicates busbar , The conductance and susceptance between the For bus , The phase difference between To restore data based on Busbar The maximum value of the load that can be cut is To restore data based on Busbar The total load, For branch The transmission capacity of For bus The minimum permissible voltage, For bus The maximum allowable voltage.

[0034] Furthermore, the multi-task residual attention hypergraph neural network introduces a hypergraph structure The data tampering characteristics and distribution network topology under false data injection attacks are described and modeled; is a set of nodes, each of which represents a distribution network bus. is a set of hyperedges, wherein the hyperedges represent network topology and data tampering characteristics; is the weight matrix of the hyperedge.

[0035] Furthermore, the multi-task residual attention hypergraph neural network includes: a hypergraph attention module;

[0036] The hypergraph attention module includes: a multi-head attention mechanism and a hypergraph convolution layer; wherein the hypergraph convolution expression is:

[0037] ;

[0038] in, For the The output of the layer, For the The output of the layer, For the The parameters of the layer, T is the transpose operation, is a nonlinear activation function, is the adjacency matrix, which is used to represent the relationship between hyperedges and nodes in the hypergraph structure. is the node degree of all nodes in the hypergraph structure The node degree matrix is ​​constructed. is the hyperedge degree of all hyperedges in the hypergraph structure The hyperedge degree matrix is ​​formed by

[0039] in:

[0040] ;

[0041] ;

[0042] ;

[0043] In the formula, For super edge The corresponding weight, For super edge and nodes The correlation relationship between them.

[0044] Furthermore, the multi-task residual attention hypergraph neural network takes high-order correlations on the hypergraph structure and normalized information-physical state data as input, adopts a multi-task learning architecture to describe data recovery and distribution network reconstruction as a multi-label classification problem, and uses a binary cross entropy function to calculate the error between the classification result and the actual label, and outputs the optimal data recovery strategy and the optimal distribution network topology reconstruction strategy while meeting the preset error requirements;

[0045] The information-physical state data include: bus flow data including active power and reactive power, operation reliability coefficient and communication delay of the remote control switch, and recovery time corresponding to the recovery deviation rate of each data.

[0046] Furthermore, the execution terminal sets a residence time to ensure that all execution terminals that need to act at the same time in the optimal distribution network topology reconstruction strategy synchronously execute action instructions at the corresponding time;

[0047] The residence time of any execution terminal is the difference between the longest communication delay of all execution terminals that need to act at the same time as the execution terminal in the optimal distribution network topology reconstruction strategy and the communication delay of the execution terminal.

[0048] In a second aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the emergency control method for information-physical cross-space collaboration under a false data injection attack as described in the first aspect.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] (1) The information-physical cross-space collaborative emergency control method under false data injection attacks provided by the present invention considers the impact of the accuracy and real-time performance of information layer data recovery on the physical layer distribution network reconstruction, and constructs an emergency control model based on the collaboration of information layer data recovery and physical layer distribution network reconstruction to improve the emergency response capability of the distribution network and ensure the safe operation of the distribution network.

[0051] (2) The information-physical cross-space collaborative emergency control method under false data injection attacks provided by the present invention adopts a multi-task residual attention hypergraph neural network to solve the constructed emergency control model in real time, obtain the optimal data recovery strategy and the optimal distribution network reconstruction strategy; and set the residence time in the execution terminal to ensure that the execution terminal synchronously executes the action instruction at the predetermined time, effectively ensuring the effect of emergency control. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of the emergency control method for information-physical cross-space collaboration under false data injection attacks provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0053] The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0054] Embodiment 1, the present invention provides an emergency control method for information-physical cross-space collaboration under false data injection attack, such as Figure 1 As shown, the method comprises the following steps:

[0055] Step 1: Obtain the emergency control model that is pre-built based on the characteristics of information network attacks and the operating status of the physical system.

[0056] Step 2: Use a pre-built multi-task residual attention hypergraph neural network to solve the emergency control model and obtain the optimal data recovery solution and the optimal distribution network topology reconstruction strategy.

[0057] Step 3: Perform data recovery according to the optimal data recovery solution, and send corresponding action instructions to the execution terminal according to the optimal distribution network topology reconstruction strategy to achieve rapid switching of the distribution network topology.

[0058] Specifically, the distribution network control center builds an emergency control model based on cross-space collaboration between information layer data recovery and physical layer distribution network reconstruction based on the characteristics of information network attacks and the operating status of the physical system.

[0059] The emergency control model includes: an information layer data recovery model that optimizes the accuracy and deviation rate of data recovery and a physical layer distribution network reconstruction model that optimizes the network topology.

[0060] In some specific embodiments, the information layer data recovery model aims to find the best trade-off between the real-time and accuracy of data recovery to adapt to the physical layer distribution network reconstruction after suffering a false data injection attack, and its expression is:

[0061] ;

[0062] in, To minimize, is the weight coefficient of time cost, is the weight coefficient of load loss cost, Indicates that the accuracy rate is , the data deviation rate is Optimal distribution network reconstruction strategy for data recovery, To take the accuracy rate , the data deviation rate is The data recovery time corresponding to the data recovery, Optimal distribution network reconstruction strategy The maximum communication delay is For bus The importance of the load, Optimal distribution network reconstruction strategy Lower bus The actual load loss, is the penalty coefficient for power flow exceeding the limit, is the total number of buses, is the power flow over-limit factor of the optimal distribution network reconstruction strategy under real operation data. When there is a power flow over-limit, ,otherwise .

[0063] The constraints of recovery time and accuracy of information layer data recovery must be met, and the expression is:

[0064] ;

[0065] ;

[0066] ;

[0067] ;

[0068] ;

[0069] The constraint that the actual load loss of each busbar does not exceed the maximum value of the actual shelvable load of each busbar under actual working conditions should also be met, and the expression is:

[0070] ;

[0071] in, is the amount of data that is accurately recovered, is the amount of data recovered with deviation, is the amount of data that needs to be restored, is the amount of data that has not been tampered with and does not need to be restored. is the total amount of information data, The time required to accurately restore a piece of information data, The time required to recover a piece of information data with deviation, Reconstruction strategy The busbar estimated based on the restored data The load loss, The actual bus The maximum load that can be cut.

[0072] In some specific embodiments, the physical layer distribution network reconstruction model aims to restore the normal operation of the distribution network under the condition of minimizing the loss of important loads, and its expression is:

[0073] ;

[0074] in, To minimize, is the weight coefficient of remote control reliability of the remote control switch, is the weight coefficient of load loss cost, For branch The initial state of the upper remote control switch, For branch The target state of the upper remote switch, is the action reliability coefficient of the remote control switch, For bus The importance of the load, To restore data based on Estimated busbar The load loss, is the set of all branches in the distribution network, is the total number of buses.

[0075] In actual engineering applications, due to environmental factors, communication problems and equipment problems, remote control failures often occur, mainly manifested as false operation and refusal to operate. Description such as:

[0076] ;

[0077] in, is the rejection rate of the switch, is the malfunction rate of the switch.

[0078] The physical layer distribution network reconstruction model takes the distribution network operation status as the constraint, including: power flow balance constraint, load loss constraint, topology radial constraint, node voltage constraint and line capacity constraint, and the expression is:

[0079] ;

[0080] ;

[0081] ;

[0082] ;

[0083] ;

[0084] ;

[0085] ;

[0086] in, For branch The active power, For branch The reactive power, To restore data based on Busbar The next active power generation, To restore data based on Busbar Reactive power generation, To restore data based on Busbar The active load demand is To restore data based on Busbar The reactive load demand, , Indicates busbar , The voltage amplitude at , Indicates busbar , The conductance and susceptance between the For bus , The phase difference between To restore data based on Busbar The maximum value of the load that can be cut is To restore data based on Busbar The total load, For branch The transmission capacity of For bus The minimum permissible voltage, For bus The maximum allowable voltage.

[0087] The present invention constructs an emergency control model for the problem of information-physical collaborative recovery, wherein the information layer is a data recovery model that optimizes the accuracy and deviation rate of data recovery, and the physical layer is a distribution network reconstruction model that optimizes the network topology, and realizes collaborative recovery of the information layer and the physical layer by exchanging coupling variables.

[0088] Example 2. Based on Example 1, this example provides the specific content of the multi-task residual attention hypergraph neural network.

[0089] Multi-task residual attention hypergraph neural network introduces hypergraph structure The data tampering characteristics and distribution network topology under false data injection attacks are described and modeled; is a set of nodes, each of which represents a distribution network bus. is a set of hyperedges, wherein the hyperedges represent network topology and data tampering characteristics; is the weight matrix of the hyperedge.

[0090] The multi-task residual attention hypergraph neural network takes the high-order correlation on the hypergraph structure and the normalized information-physical state data as input, adopts the multi-task learning architecture to describe the data recovery and distribution network reconstruction as a multi-label classification problem, and uses the binary cross entropy function to calculate the error between the classification result and the actual label, which is expressed as:

[0091] ;

[0092] in, represents the number of training samples, Indicates the number of labels corresponding to each sample, For the Sample No. The predicted probability of the labels, For the Sample No. The actual situation of a label.

[0093] Output the optimal data recovery strategy and the optimal distribution network reconstruction strategy while meeting the preset error requirements; wherein the information-physical state data includes: bus flow data including active power and reactive power, the action reliability coefficient and communication delay of the remote control switch, and the recovery time corresponding to each data recovery deviation rate.

[0094] The multi-task residual attention hypergraph neural network that introduces the hypergraph structure includes: a hypergraph attention module.

[0095] In some specific embodiments, the hypergraph attention module includes: a multi-head attention mechanism and a hypergraph convolution layer; the hypergraph convolution expression in the hypergraph convolution layer is:

[0096] ;

[0097] in, For the The output of the layer, For the The output of the layer, For the The parameters of the layer, T is the transpose operation, is a nonlinear activation function, is the adjacency matrix, which is used to represent the relationship between hyperedges and nodes in the hypergraph structure. is the node degree of all nodes in the hypergraph structure The node degree matrix is ​​constructed. is the hyperedge degree of all hyperedges in the hypergraph structure The hyperedge degree matrix is ​​formed by

[0098] in:

[0099] ;

[0100] ;

[0101] ;

[0102] In the formula, For super edge The corresponding weight, For super edge and nodes The correlation relationship between them.

[0103] In some more specific embodiments, the multi-task residual attention hypergraph neural network consists of an input layer, a residual block containing 3 hypergraph attention modules, 3 fully connected layers and an output layer.

[0104] Embodiment 3: Based on Embodiment 2, this embodiment provides a resident mechanism of a remote control terminal.

[0105] Based on the emergency control model constructed in Example 1 and the solution method in Example 2, an optimal distribution network topology reconstruction strategy is generated, and the distribution network control center sends corresponding action instructions, i.e., opening / closing instructions, to the remote control execution terminal according to the optimal distribution network topology reconstruction strategy.

[0106] The execution terminal sets a residence mechanism and introduces a residence time to ensure that all execution terminals that need to act at the same time in the optimal distribution network reconstruction strategy synchronously execute action instructions at the corresponding time.

[0107] The residence time of each execution terminal is defined as the difference between the longest communication delay of all execution terminals that need to act at the same time as the execution terminal in the optimal distribution network reconstruction strategy and the communication delay of the execution terminal.

[0108] After receiving the action command, the execution terminals wait for each other and control the remote control switches to act synchronously at the specified time to complete the rapid switching of the distribution network topology.

[0109] The present invention can balance rapidity and accuracy when solving the problem of power distribution system fault recovery under false data injection attacks, thereby improving the operational safety and reliability of the power distribution network under cyber-physical fusion.

[0110] Example 4. This embodiment provides a computer storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements the emergency control method of information-physical cross-space collaboration under the false data injection attack as described in any of Examples 1 to 3.

[0111] The storage medium may include, for example, a storage component of a tablet computer, a hard disk of a computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer storage medium may be any combination of one or more computer storage media.

[0112] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An emergency control method for information-physical cross-space collaboration under false data injection attack, characterized in that: include: Obtain an emergency control model pre-built according to the characteristics of information network attacks and the operating status of the physical system; wherein the emergency control model includes: an information layer data recovery model that optimizes the accuracy and deviation rate of data recovery and a physical layer distribution network reconstruction model that optimizes the network topology; A pre-built multi-task residual attention hypergraph neural network is used to solve the emergency control model to obtain the best data recovery solution and the best distribution network topology reconstruction strategy; Performing data recovery according to the optimal data recovery solution, and sending corresponding action instructions to the execution terminal according to the optimal distribution network topology reconstruction strategy to achieve rapid switching of the distribution network topology; The information layer data recovery model aims to find the best trade-off between the real-time and accuracy of data recovery to adapt to the physical layer distribution network reconstruction after being attacked by false data injection. Its expression is: Among them, min is minimized, r T is the weight coefficient of time cost, r λ is the weight coefficient of load loss cost, S λ,δ represents the optimal distribution network reconstruction strategy when the accuracy rate is λ and the data deviation rate is δ for data recovery, RT is the data recovery time corresponding to the data recovery with the accuracy rate of λ and the data deviation rate of δ, and ET is the optimal distribution network reconstruction strategy S λ,δ The maximum communication delay is is the importance of the load under bus i, is the optimal distribution network reconstruction strategy S λ,δ The actual load loss of the lower busbar i, is the power flow over-limit penalty coefficient, N is the total number of buses, is the power flow over-limit factor of the optimal distribution network reconstruction strategy under real operation data. When there is a power flow over-limit, otherwise The physical layer distribution network reconstruction model aims to restore the normal operation of the distribution network under the condition of minimizing the loss of important loads. Its expression is: Among them, min is minimization, μ1 is the weight coefficient of remote control reliability of remote control switch, μ2 is the weight coefficient of load loss cost, is the initial state of the remote switch on branch ij, x ij is the target state of the remote switch on branch ij, c ij is the action reliability coefficient of the remote control switch, is the importance of the load under bus i, Based on the recovery data D R The estimated load loss of bus i; ε is the set of all branches in the distribution network.

2. The emergency control method according to claim 1, characterized in that: The information layer data recovery model is constrained by the time and accuracy of data recovery and the actual load loss of each bus, and the expression is: RT=N p t c +N d t u ; N p +N d =N att ; N n +N att =N all ; N att t u ≤RT≤N att t c ; Among them, N p is the amount of data accurately recovered, N d is the amount of data recovered with deviation, N att is the amount of data that needs to be restored, N n N is the amount of data that has not been tampered with and does not need to be restored. all is the total amount of information data, t c The time required to accurately restore a piece of information data, t u The time required to recover a piece of information data with deviation, For the reconstruction strategy S λ,δ The load loss of bus i estimated based on the restoration data is: is the maximum shearable load of the actual busbar i.

3. The emergency control method according to claim 1, characterized in that: The physical layer distribution network reconstruction model takes the distribution network operation state as the constraint, including: power flow balance constraint, load loss constraint, topology radial constraint, node voltage constraint and line capacity constraint, and the expression is: Among them, P ij is the active power of branch ij, Q ij is the reactive power of branch ij, Based on the recovery data D R Active power generation under bus i, Based on the recovery data D R Reactive power generation under bus i, Based on the recovery data D R Active load demand under bus i, Based on the recovery data D R The reactive load demand under bus i, U i , U j Represents the voltage amplitude at busbar i, j, G ij , B ij represents the conductance and susceptance between busbars i and j, θ ij is the phase angle difference between busbars i and j, Based on the recovery data D R The maximum value of the shelving load of busbar i is Based on the recovery data D R The total load of busbar i, is the transmission capacity of branch ij, is the minimum allowable voltage of busbar i, is the maximum allowable voltage of bus i.

4. The emergency control method according to claim 1, characterized in that: The multi-task residual attention hypergraph neural network introduces a hypergraph structure G=(V, E, W) to describe and model the data tampering characteristics and distribution network topology under false data injection attacks; wherein V is a set of nodes, each of which represents a distribution network bus, E is a set of hyperedges, each of which represents the network topology and data tampering characteristics; and W is a weight matrix of the hyperedges.

5. The emergency control method according to claim 4, characterized in that: The multi-task residual attention hypergraph neural network includes: a hypergraph attention module; The hypergraph attention module includes: a multi-head attention mechanism and a hypergraph convolution layer; wherein the hypergraph convolution expression is: in, is the output of the l+1th layer, is the output of the lth layer, Θ l is the parameter of the lth layer, T is the transposition operation, σ(·) is the nonlinear activation function, H hyper is the adjacency matrix, which represents the relationship between hyperedges and nodes in the hypergraph structure, D V is the node degree matrix composed of the node degrees d(v) of all nodes in the hypergraph structure, D E is the hyperedge degree matrix composed of the hyperedge degrees d(e) of all hyperedges in the hypergraph structure, in: Where W(e) is the weight corresponding to the hyperedge e, H hyper (v,e) is the association relationship between hyperedge e and node v.

6. The emergency control method according to claim 5, characterized in that: The multi-task residual attention hypergraph neural network takes high-order correlations on the hypergraph structure and normalized information-physical state data as input, adopts a multi-task learning architecture to describe data recovery and distribution network reconstruction as a multi-label classification problem, and uses a binary cross entropy function to calculate the error between the classification result and the actual label, and outputs the optimal data recovery strategy and the optimal distribution network topology reconstruction strategy while meeting the preset error requirements; The information-physical state data include: bus flow data including active power and reactive power, operation reliability coefficient and communication delay of the remote control switch, and recovery time corresponding to the recovery deviation rate of each data.

7. The emergency control method according to claim 1, characterized in that: The execution terminal sets a residence time to ensure that all execution terminals that need to act at the same time in the optimal distribution network topology reconstruction strategy synchronously execute action instructions at the corresponding time; The residence time of any execution terminal is the difference between the longest communication delay of all execution terminals that need to act at the same time as the execution terminal in the optimal distribution network topology reconstruction strategy and the communication delay of the execution terminal.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the emergency control method for information-physical cross-space collaboration under a false data injection attack as described in any one of claims 1 to 7.

Images