CN117910516A - Network attack elasticity detection recovery method and system based on deep Jacobian - Google Patents

Abstract

The invention provides a network attack elasticity detection recovery method and system based on deep Jacobian. According to the invention, after an electric heat network model and an energy distribution diagram of a comprehensive energy system are established, DOS attack and FDI attack interference are considered, a neural network parallel CNN-BILSTM model is utilized to detect the attack type, and finally the attack influence is resisted by a deep Jacobian descent algorithm, so that energy control between the networks is realized. Compared with the existing network attack detection, the parallel CNN-BiLSTM network remarkably improves the calculation efficiency, accelerates the model training and reasoning process, and improves the prediction precision. Compared with the existing method, the invention provides a distributed elastic initialization-free Jacobian descent algorithm. The method is embedded with second-order information, and can exponentially converge to the global optimal solution of the studied problem, so that the convergence speed is increased.

Description

Network attack elasticity detection recovery method and system based on deep Jacobian

Technical Field

The invention relates to a detection recovery strategy of network attack in the field of energy control, in particular to a method and a system for detecting and recovering elasticity of network attack based on deep Jacobian.

Background

As energy resources become increasingly limited, it is vital to improve energy efficiency. An integrated energy management system (INTEGRATED ENERGY MANAGEMENT SYSTEM, IEMS) is a system that covers multiple aspects for monitoring, controlling, analyzing, and optimizing the use and allocation of energy resources. IEMS can help the organization monitor the energy consumption, identify the energy waste, and take measures to reduce the energy waste and improve the energy utilization efficiency. Through the integrated energy management system, renewable energy sources can be better managed, carbon emissions can be reduced, and sustainable energy use and environmental protection can be supported. IEMS is favorable for monitoring the state of the power grid in real time, can rapidly detect faults and take measures to reduce the power failure time and improve the stability of the power grid. IEMS provide data driven decision support for organizations by collecting and analyzing large amounts of energy data. This helps in developing more intelligent energy strategies and plans.

It is critical to protect these systems from network attacks, especially against denial of service attacks (DOS) and false data injection attacks (FDI). As they directly affect the quality and reliability of the power supply. IEMS relates to the collection and processing of large amounts of energy data, including energy supply, demand and load data. DOS attacks can lead to data loss or tampering, affecting the accuracy of energy decisions. The energy system needs real-time response to adjust the power supply strategy to accommodate the change in demand. DOS attacks can lead to delays and unavailability, affecting response times. IEMS rely on accurate data to make energy decisions. FDI attacks can cause the system to make false decisions using false data, compromising energy efficiency. An attacker may tamper with the data to disrupt the normal operation of the system. IEMS under the FDI attack may take an unreasonable energy allocation strategy, resulting in energy waste and performance degradation. It is important to formulate an emergency response plan, including backup energy data and system configuration, to quickly resume normal operation after an attack.

The comprehensive energy system considers the inherent mechanism, operation characteristics, various operation constraints, network constraints, the interconversion process and coupling characteristics of the multi-energy flows and the like of each energy device and each energy network in the energy production, transmission and consumption processes in the energy control construction requirements. The energy control problem is more complex and difficult in modeling, algorithm design, theoretical analysis, and the like because of strong coupling in different energy networks. For the response of the cooperative attack and the non-cooperative attack from the public network, the existing energy control method cannot accurately judge and quickly respond to the control among the network, and has the problems of low convergence speed, poor accuracy and the like.

Disclosure of Invention

In order to solve the defects in the prior art, the invention provides a network attack elasticity detection recovery method and system based on deep Jacobian. And considering system interference such as network attack and the like, detecting the attack type by utilizing a neural network parallel CNN-BILSTM model, and finally resisting the attack influence by a deep Jacobian descent algorithm to realize energy control among the network.

The invention adopts the following technical scheme.

The invention provides a network attack elasticity detection recovery method based on deep Jacobian, which comprises the following steps:

s1, performing energy control on multiple targets in each energy body network in a comprehensive energy management system, establishing an objective function with the minimum cost function of the comprehensive energy management system, and establishing constraint conditions of an electric heating network;

s2, under the established constraint condition, the signal acquisition module acquires related information of an Energy Body (EB) and an Energy Router (ER), and the objective function and the constraint condition are simplified;

S3, building an energy distribution model among energy bodies, defining three energy distribution diagrams, and corresponding to different physical links of power, heat and gas quantity among the energy bodies;

s4, establishing DOS and FDI network attack models;

S5, performing network attack detection by using a parallel CNN-BILSTM model, respectively performing tensor modeling on characteristic data and power data of a power grid, an air network and a heat supply network in the comprehensive energy management system, and then connecting the characteristics into a final characteristic vector, wherein the characteristic vector is used for judging attack types;

S6, setting an elastic initial free Jacobian descent algorithm, and dynamically updating by utilizing real-time information to resist DOS attacks and FDI attacks in an untrusted communication network;

And S7, when each energy network meets constraint conditions and meets an objective function, judging that the energy network converges to an optimal state, and controlling energy generation and transmission of the energy body according to the data by each energy body when the energy bodies reach optimal energy exchange quantity.

Preferably, the step S1 includes:

S1, establishing a cost function for each device in a comprehensive energy management system, and establishing an objective function with the minimum cost function of the comprehensive energy management system, wherein the constraints of supply and demand balance and the constraints of output of each device are considered, and respective constraint conditions are set for unbalanced power, heat and gas of an electric heating network;

s1.1: the cost function required in the calculation is proposed,

The cost function of the fuel generator and the fuel heating device and the cost function of the cogeneration are respectively obtained by the fuel cost:

wherein, Is a non-negative cost coefficient for a cost function of the fuel generator; /(I)Is a non-negative cost coefficient for the cost function of the fuel heating apparatus,/>Is a non-negative cost coefficient of a cost function for cogeneration,/>The time step distance t and the power and heat energy of cogeneration of the ith energy body are respectively;

the cost functions of the renewable power generator and the renewable heating device are respectively:

wherein, And/>A non-negative cost coefficient that is a cost function for the renewable power generator; /(I)Penalty coefficients for cost functions for renewable power generators; /(I)And/>A non-negative cost coefficient that is a cost function for the renewable heating device; /(I)Penalty coefficients for cost functions for renewable heating devices;

the cost functions of electrical and thermal storage are:

wherein, And/>Is a cost coefficient for the cost function of the electrical storage; /(I)And/>Is a cost coefficient for a cost function of electrical and thermal storage;

The cost function of the gas supply is:

wherein, And/>A non-negative cost coefficient for the gas supply cost function; Is convex within the constraint;

The energy load utility function considering the demand response is:

wherein, And/>Is the non-negative utility coefficient of the energy load effect function of the demand response,/>And/>Respectively represent controllable power, thermal energy and gas load;

revenue/cost of selling/purchasing energy to other energy bodies is:

wherein, And/>Is a market settlement price; /(I)A non-negative coefficient of value for selling or purchasing energy; /(I)And/>Respectively representing power, heat energy and gas exchanged in the energy network;

The total cost of the comprehensive energy management system is as follows:

s1.2, in the comprehensive energy management system, the power is subjected to global supply and demand balance constraint, and unbalanced power, heat and gas constraint conditions of an ith energy body are as follows:

(1) The unbalanced power of the ith energy body at time step t is estimated by the following equation:

wherein, Representing the output of power,/>Load representing power,/>Representing the schedulable power of the ith energy volume at time stride t,/>Respectively represents the output power of the fuel generator, the renewable generator, the cogeneration and the electricity storage of the ith energy body at the time step distance t,Representing the power load that must be operated and the power load that is scheduled to be operated, respectively; beta ₂、β₂ is the non-negative coefficient of cogeneration power, electricity storage,/>Setting the charge as positive and the discharge as negative;

(2) The unbalanced thermal energy of the ith energy at time step t is estimated by:

wherein, Representing the output of heat energy,/>Representing the load of thermal energy,/>Representing the schedulable thermal energy of the ith energy volume at time step t,/>Respectively representing the fuel heating device, the renewable heating device, the cogeneration and the heat energy output of the heat storage of the ith energy body at the time step distance t,Representing the thermal energy load that must be operated and the thermal energy load that is scheduled to be operated, respectively; beta ₂、β₂ is the non-negative coefficient of cogeneration heat energy, heat storage;

(3) The imbalance of the ith energy cell at time step t is estimated by:

wherein, Representing the amount of dispatchable gas of the ith energy volume at time step t; /(I)Representing the gas output of the ith energy body at time step t; /(I)The necessary operating gas load for the ith energy body at time step t; /(I)For a planned operating gas load of the ith energy body at time step t;

the energy load of each energy body comprises an electric load, a thermal load and a gas load, and each load can be divided into an equivalent energy load which is required to be operated and a schedulable energy load;

S1.3, each participant in the energy body is limited, a decision is made according to a group of local constraints, and constraint conditions of an electric heat network in the comprehensive energy management system are as follows:

(1) For electrical energy, the power energy constraints and ramp rate limits of the fuel generator are respectively:

Wherein, the method comprises the steps of, wherein, Slope rate representing two successive time periods of the fuel generator,/> Maximum and minimum values of fuel generator power representing the ith energy body, respectively;

the trade-off constraints for optimality and omnipotency of renewable generators are as follows:

Where b is the positive parameter of the renewable generator limit equation, And/>Representing the maximum and minimum values of renewable generator power of the ith energy volume respectively,

Constraints on electrical energy storage are:

wherein, And/>Respectively represent the maximum charge-discharge rate,/>Representing charge-discharge efficiency,/>Representing the state of charge of the electrical energy stored in the device, R (·) being a function of the SOC range;

the limitations of the schedulable power load are:

wherein, In order to be a maximum electrical power load,

The ratio of schedulable electrical power loads is:

Wherein, psi represents the conversion ratio from SCM/h to MW, SCM/h is the unit for gas delivery, and this value is set to 1/84; gamma _i,g→p、γ_i,g→h、γ_i,h→g represents the power ratio of electric power to cogeneration and gas load, the power ratio of heat to cogeneration and gas load, the power ratio of gas load and heat to cogeneration, respectively;

(2) For thermal energy, the constraints of the fuel heating apparatus are as follows:

Where a is a positive parameter of the fuel heating apparatus restriction equation, Maximum and minimum values of heat energy of fuel heating device representing ith energy body, respectively,/>A ramp rate representing two successive time periods of the fuel heating apparatus;

The trade-off constraints between optimality and likelihood of renewable heating devices are as follows:

Where c is the positive parameter of the constraint equation for the renewable heating device, Maximum and minimum values of renewable heating device power representing the ith energy body, respectively;

The constraints of thermal energy storage are:

Wherein, the method comprises the steps of, wherein, And/>Respectively represent the maximum charge-discharge rate,/>Representing charge-discharge efficiency,/>Representing the state of charge of thermal energy stored in the device, R (·) is related to/>A range function;

The limit of the schedulable thermal load is:

Wherein, the method comprises the steps of, wherein, Is the maximum thermal energy load;

The ratio of schedulable thermal loads is:

Wherein γ _i,g→h represents the ratio of heat to combined heat and gas load;

(3) In a gas system, the relevant constraints are as follows;

Q (·) is about Is a monotonically increasing constraint function of (1), the constraints of the gas supplier are:

wherein, A minimum value and a maximum value of gas output representing the ith energy body, respectively;

The gas load limits are:

Wherein, the method comprises the steps of, wherein, For maximum gas load of the ith energy body at time step t,

The ratio of the gas loads is:

Wherein, the method comprises the steps of, wherein, Representing the minimum and maximum values of the ratio of electric power to combined power and gas load, respectively,/>Representing the minimum and maximum values of the ratio of heat to combined heat to gas load, respectively;

in cogeneration, while in the electric and thermal systems, its local operating constraints are:

wherein, And/>Is the coefficient of the linear inequality constraint determined by the feasible operating region of cogeneration,/>Is the ramp rate between two consecutive time steps of the cogeneration generator of the ith energy.

Preferably, the step S2 includes:

S2.1: setting an integrated energy management system containing n energy bodies, wherein each energy body has m _i participants, finding out the optimal operation with the maximum social benefit, the optimal energy distribution and the minimum delivery cost among the energy bodies, and defining the optimal targets as follows:

wherein, Is the maximum benefit value,/>For minimum transport costs, obj is the optimization objective;

in the case of electrical energy which is to be used, The following formula is shown:

as to the thermal energy of the heat energy, The following formula is shown:

s2.2: definition x _ij∈R³ is a three-dimensional vector of energy body participants' power, heat and gas quantity, where i represents the energy body node, i is the i energy body node, definition Is a three-dimensional vector consisting of power, heat and gas quantity of the energy load which must be operated j times of the ith energy body, and converts the variable of each participant into a form of x _ij as shown in the following formula;

Wherein, if x _ij represents a controllable energy load, then B _ij＝-I₃; otherwise, B _ij＝I₃; wherein I ₃ is a three-dimensional identity matrix; omega _ij is a local closed convex set constrained by a local inequality;

Setting the cost function corresponding to the zero variable as any type of strong convex function to estimate the value of the exchange energy in the distributed energy network And/>A virtual variable x _ij is allocated to the energy router in the energy body, let j=0, a strong convex cost function is assigned to x _i0 and set to zero, and then let W (x _ij) represent the corresponding cost function or negative utility function as shown in the following formula:

S2.3: the objective function and its constraints are converted into the following form:

wherein, And/>Is a locally closed convex set determined by local inequality constraints.

Preferably, the participants in the energy mass include renewable power generators, fuel power generators, renewable heating devices, cogeneration, electricity storage, gas supply, and dispatchable energy loads.

Preferably, the step S2 further includes:

The related information collected by the signal collection module comprises: the power, heat energy and gas output, load of each energy body participant and the power, heat energy and gas exchange quantity between energy body networks, and the unbalanced power, heat energy and gas exchange quantity of the ith energy body at the time step t.

Preferably, the step S3 includes:

the distributed communication network is represented as a graph Wherein/>Representing a set of nodes,/>Representing edge sets, each edge set referring to a communication link,/>Representing an adjacency matrix, when node i can receive information from node j, node j is referred to as node i's neighbor, and node i's neighbor set is defined asIf/>Then a _ij = 1; otherwise, a _ij = 0;

Definition of the definition For the diagram/>Laplace matrix of (2), wherein/>Setting upIs connected when no attack occurs, in which case all eigenvalues of L are non-negative; meanwhile, only one zero eigenvalue exists, and the sequence of all eigenvalues of L is 0=λ ₁＜λ₂≤,…,≤λ_n;

If an attack is detected, the lost data is replaced by the latest received data, and for a computing unit which is not attacked by the network, each energy-body participant dynamically exchanges information with its neighbor, then the energy router dynamically exchanges information with its neighbor energy router, and the power, heat and gas output of each energy-body participant and the power, heat and gas exchange of the energy router are updated.

Preferably, the step S4 includes:

S4.1: the time interval for the kth attack is defined as t _k ^Φ,t_k ^Φ+Δ_k ^Φ, where t _k ^Φ is the moment of attack initiation, a _k ^Φ is the duration ,Ξ_a(t^Φ)＝∪[t_k ^Φ,t_k ^Φ+Δ_k ^Φ)∩[t₀ ^Φ,t^Φ] of the respective attack, the union ,Ξ_s(t^Φ)＝∪[t_k ^Φ+Δ_k ^Φ,t_k+1 ^Φ)∩[t₀ ^Φ,t^Φ] of communication moments subject to DoS attack represents the union of communication moments not subject to attack, where t ^Φ ₀ is the initial time, Total duration of attack for [ t0, t ];

It is defined that if any communication channel of agent i is attacked, agent i is the attacked node, the set of attacked nodes being denoted An;

S4.2: sensor and actuator data subject to FDI attacks are:

Representing superimposed malicious data of integrated energy management system control input,/> Malicious data representing superimposed output values corresponding to comprehensive energy management system,/>Respectively representing whether the executor channel and the sensor channel are subjected to FDI attack, wherein the attack is 1, and the attack is not 0;

setting attack vector Are bounded in a collection and haveHere/>Is a known positive scalar.

Preferably, the step S5 includes:

S5.1: tensor modeling is respectively carried out on characteristic data and power data of a power grid, a gas network and a heat supply network in the comprehensive energy management system, and as X _Λ and P _Λ, attack types judged by a neural network model are defined as

X_Λ＝[X_ΛV,X_ΛH,X_ΛP]^T，

P_Λ＝[P_ΛD,P_ΛQ,P_ΛR]^T，

X_Λi＝[x_Λ(i-L),x_Λ(i-L+1),…,x_Λ(i-2),x_Λ(i-1)]^T，

Wherein the method comprises the steps ofInput characteristic data required for attack detection at the ith sampling time point are represented, wherein the characteristic data comprise voltage of a power grid, heat of a heat supply network and pressure of an air network, X _ΛVi represents input power grid voltage required for attack detection at the ith sampling time point, X _ΛHi represents input heat of the heat supply network required for attack detection at the ith sampling time point, X _ΛPi represents input air network pressure required for attack detection at the ith sampling time point, L represents time steps of input data of a parallel CNN-BiLSTM model, and X _Λi comprises time sequence data;

S5.2: after normalization, X _Λi is considered a grayscale image and as input to Conv2D, X _Λi is flattened and used as input to BiLSTM;

X _Λi represents the characteristic data acquired at the ith sampling time point; p _Λi represents the comprehensive energy management system power of the ith sampling time point, P _ΛDi represents the grid system power of the ith sampling time point, P _ΛQi represents the air grid system power of the ith sampling time point, and P _ΛRi represents the heat grid system power of the ith sampling time point; representing attack types detected by a parallel CNN-BiLSTM model;

S5.3: the judgment of the attack type is carried out,

If X _Λ＝[X_ΛV,X_ΛH,X_ΛP]^T is zero, the communication is interrupted, the attack type is DOS attack, and then S5.2 is returned;

If X _Λ＝[X_ΛV,X_ΛH,X_ΛP]^T is changed drastically and fluctuates greatly, the FDI attack is temporarily judged and then S5.2 is returned;

If X _Λ＝[X_ΛV,X_ΛH,X_ΛP]^T has no change or small amplitude fluctuation, judging that the attack is not generated, and returning to S5.2;

S5.4: outputting the network attack detection result and outputting For DOS attack, output/>For FDI attack, output/>In order to avoid attack, local information is output, and the characteristic data comprise relevant quantities such as voltage of a power grid, heat of a heat supply network, pressure of an air network and the like and power data of the power grid, the air network and the heat supply network as input of a free Jacobian descent algorithm.

Preferably, the step S6 includes:

A distributed elastic non-initialized Jacobian descent (DRIFJD) algorithm is proposed to resist attacks in an untrusted communication network, defining three control variables, expressed respectively as And/>Each control variable needs to be adjusted during the attack, and the DRIFJD algorithm is expressed as follows:

Using combined measurements

And design control variables

Wherein the method comprises the steps ofAnd/>Is an auxiliary variable; /(I)And/>Is within attack interval/>And/>Is determined by the estimation of (a); /(I)And/>The latest received or measured state that is successful for the kth attack, blocking network communications for the attacking node during [ t _k ^Φ,t_k ^Φ+Δ_k ^Φ); /(I)

S6.1, inputting local information, includingAnd/> And/>The characteristic data comprises related quantities such as voltage of a power grid, heat of a heat supply network, pressure of an air network and the like, and power data of the power grid, the air network and the heat supply network;

S6.2: each energy network attempts to exchange with its neighbors And/>Information of (2);

S6.3: if the communication between the energy sources i and j is successful, the energy sources i and j are judged not to be attacked, and the control variable is selected as

And/> For updating/>Equation/>And/>For auxiliary variables/>And/>Is updated according to the update of (a); for/> Is/are of the dual variable of (1)And/>The main function of (a) is to collectively converge all z _i to the same value by distributed computing;

s6.4: if the communication between the energy sources i and j is unsuccessful, judging that the node is attacked, judging whether the node can communicate with two adjacent energy sources, and if the node cannot communicate, judging that the node is attacked; if communication is possible, it is determined that i is not a node under attack, and the control variable is selected as And/> For updating/>Equation(s)And/>For auxiliary variables/>And/>Is updated according to the update of (a);

S6.5: if the communication between the energy source i and the energy source j is unsuccessful, judging that the energy source i is attacked, judging whether the energy source i is a attacked node, and if the energy source i cannot communicate, judging that the energy source i is a attacked node; if communication is possible, determining that i is not the node under attack; if yes, the attacked node i can not receive the neighbor information during the attack period And/> With most recently received dataAnd/>Calculating/>, instead of lost dataAnd/>These two data are further used as estimates to calculate the evolution of the system, selecting the control variable as/>And For updating/>Equation(s)And/>For auxiliary variables/>And/>Is updated according to the update of (a);

s6.6: judging variable If the iteration is converged, ending the iteration if the iteration is not converged, returning to the step S6.1 to continue the iteration until the convergence of the result is ended.

The invention also provides a network attack elasticity detection and recovery system based on the depth Jacobian, which is used by the network attack elasticity detection and recovery method based on the depth Jacobian, and comprises the following steps:

The signal acquisition module is used for acquiring the related information of the energy body and the energy router;

the attack detection module is used for detecting network attacks and judging attack types;

and the recovery module is used for resisting the network attack by utilizing the dynamic update of the real-time information.

Compared with the prior art, the invention provides the detection method and the recovery strategy for resisting the DoS attack and the FDI attack. An analysis model of DoS attacks and FDI attacks is established.

Compared with the existing network attack detection, the parallel CNN-BiLSTM network can process different parts of input data at the same time and fully utilize the capability of parallel computation. This significantly improves computational efficiency and speeds up the model training and reasoning process. Furthermore, the parallel CNN and BiLSTM layers facilitate the simultaneous extraction and integration of data features. This enables us to capture information about various aspects of the data and provide an efficient model representation, thereby improving prediction accuracy.

Compared with the existing method, the invention provides a distributed elastic initialization-free Jacobian descent algorithm. The algorithm designs three switching control protocols, so that the algorithm can reasonably use the estimated value to replace lost information when an attack occurs. The method embeds second-order information, thereby accelerating convergence speed. The method has the advantages of high convergence speed, no special initialization condition, asynchronous communication and the like. Meanwhile, the convergence and the optimality can be maintained in the duration of denial of service attack, and the algorithm can exponentially converge to the global optimal solution of the studied problem.

Drawings

Fig. 1 is a schematic structural diagram of a distributed communication network graph provided by an embodiment of the present invention, where fig. a is an original communication topology, fig. b is a communication topology destroyed in DOS attack mode, and fig. c is a communication topology destroyed in FDI attack mode;

FIG. 2 is a diagram of a parallel CNN-BILSTM network attack detection model according to an embodiment of the present invention;

FIG. 3 is a flowchart of a deep Jacobian descent algorithm provided by an embodiment of the present invention;

Fig. 4 is a general flowchart of a network attack elasticity detection recovery method and system based on deep jacobian according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. The described embodiments of the application are only some, but not all, embodiments of the application. All other embodiments, which can be made by those skilled in the art without making any inventive effort, are within the scope of the present application.

The schematic structural diagram of the graph theory of the distributed communication network as shown in fig. 1, the distributed communication network is represented as a graphWherein/>Representing a set of nodes. /(I)Representing edge sets, each edge set referring to a communication link. /(I)Representing the adjacency matrix. If node i can receive information from node j, node j is referred to as node i's neighbor. The neighbor set of node i is defined as/>If/>Then a _ij = 1; otherwise, a _ij =0. Furthermore, we define/>For the diagram/>Of (c) whereinHypothesis/>Is connected when no attack occurs. In this case, all eigenvalues of L are non-negative; at the same time, there is only one zero eigenvalue. All eigenvalues of L are in the order 0=λ ₁＜λ₂≤,…,≤λ_n.

As shown in fig. 2, the parallel CNN-BILSTM model respectively performs tensor modeling on the characteristic data and the power data of the power grid, the air grid and the heat supply network in the integrated energy management system, and uses the tensor modeling as X _Λ and P _Λ. Meanwhile, the attack type judged by the neural network model is defined as

As shown in the flow chart of the deep jacobian descent algorithm of fig. 3, in one aspect, the system operates during normal periods. In this case, each agent can control updates using real-time information. On the other hand, the system is subject to attack. If node i is not the node under attack, it will still be updated dynamically using real-time information. If node i belongs to the attacked node, switching to dynamic updating with the estimation. Meanwhile, a switching strategy is adopted to defend attacks.

As shown in fig. 4, the specific method is as follows.

S1, establishing a cost function for each device in the comprehensive energy management system, establishing an objective function with the minimum cost function of the comprehensive energy management system, and establishing respective constraint conditions for unbalanced power, heat and gas of an electric heating network by considering supply and demand balance constraint and output limit of each device.

S1.1: the cost function required in the calculation is presented, as follows.

The cost function of the fuel generator and the fuel heating device and the cost function of the cogeneration are respectively obtained by the fuel cost:

wherein, Is a non-negative cost coefficient for a cost function of the fuel generator; /(I)Is a non-negative cost coefficient for the cost function of the fuel heating apparatus,/>Is a non-negative cost coefficient of a cost function for cogeneration,/>The time step distance t and the power and heat energy of cogeneration of the ith energy body are respectively.

The cost functions of the renewable power generator and the renewable heating device are respectively:

wherein, And/>A non-negative cost coefficient that is a cost function for the renewable power generator; /(I)Penalty coefficients for cost functions for renewable power generators; /(I)And/>A non-negative cost coefficient that is a cost function for the renewable heating device; /(I)Is a penalty factor for the cost function of the renewable heating device.

The cost functions of electrical and thermal storage are:

wherein, And/>Is a cost coefficient for the cost function of the electrical storage; /(I)And/>Is a cost factor for the cost function of electrical and thermal storage.

The cost function of the gas supply is:

wherein, And/>A non-negative cost coefficient for the gas supply cost function; Is convex within the constraint.

The energy load utility function considering the demand response is:

wherein, And/>Is the non-negative utility coefficient of the energy load effect function of the demand response,/>And/>Representing controllable power, thermal energy and gas loads, respectively.

Revenue/cost of selling/purchasing energy to other energy bodies is:

wherein, And/>Is a market settlement price; /(I)A non-negative coefficient of value for selling or purchasing energy; /(I)And/>Representing the power, thermal energy and gas exchanged in the energy network, respectively.

The total cost of the comprehensive energy management system is as follows:

S1.2, in the comprehensive energy management system, the power can receive global supply and demand balance constraint, and unbalanced power, heat and gas constraint conditions of the ith energy body are as follows:

(1) The unbalanced power of the ith energy body at time step t is estimated by the following equation:

wherein, Representing the output of power,/>Load representing power,/>Representing schedulable power of the ith energy bin at time stride t,/>Respectively representing the output power of the fuel generator, the renewable generator, the cogeneration and the electricity storage of the ith energy body at the time step t,Representing the power load that must be operated and the power load that is scheduled to be operated, respectively; beta ₂、β₂ is the non-negative coefficient of cogeneration power, electrical storage. /(I)The charge is positive and the discharge is negative.

(2) The unbalanced heat energy of the ith energy body at the time step t is estimated by the following formula:

/>

wherein, Representing the output of heat energy,/>Representing the load of thermal energy,/>Representing schedulable thermal energy of the ith energy volume at time step t,/>Fuel heating plant, renewable heating plant, cogeneration, thermal energy output of heat storage,/>, respectively representing energy at time step t, ith energy bodyRepresenting the thermal energy load that must be operated and the thermal energy load that is scheduled to be operated, respectively; beta ₂、β₂ is the non-negative coefficient of cogeneration heat energy, heat storage.

(3) The unbalanced gas of the ith energy body at the time step t is estimated by the following formula:

wherein, A schedulable gas representing an ith energy volume at a time step t; /(I)A gas output representing the ith energy volume at time step t; /(I)The gas load must be operated for the ith energy body at time step t; /(I)For the planned operation of the gas load at time step t, the ith energy body.

The energy load of each energy body includes an electric load, a thermal load, and a gas load. Each load can be divided into an equivalent must-run energy load and a schedulable energy load.

S1.3, each participant in the energy body is limited, a decision is made according to a group of local constraints, and constraint conditions of an electric heat network in the comprehensive energy management system are as follows:

(1) For electrical energy, the power energy constraints and ramp rate limits of the fuel generator are respectively:

wherein, Slope rate representing two successive time periods of the fuel generator,/> Representing the maximum and minimum values of the fuel generator power of the ith energy body, respectively.

The trade-off constraints for optimality and omnipotency of renewable generators are as follows:

Where b is the positive parameter of the renewable generator limit equation, And/>Representing the maximum and minimum values, respectively, of renewable generator power for the ith energy volume.

Constraints on electrical storage are:

wherein, And/>Respectively represent the maximum charge-discharge rate,/>Representing charge-discharge efficiency,/>R (·) is a function of the SOC range, representing the state of charge of the electrical energy stored in the device.

The limitations of the schedulable power load are:

wherein, Is the maximum electrical power load.

The ratio of schedulable electrical power loads is:

Wherein, psi represents the conversion ratio from SCM/h to MW, SCM/h is the unit for gas delivery, and this value is set to 1/84; gamma _i,g→p、γ_i,g→h、γ_i,h→g represents the ratio of electric power to combined power and gas load, the ratio of heat to combined heat and gas load, the ratio of electric power to combined power and gas load, respectively.

(2) For thermal energy, the constraints of the fuel heating apparatus are as follows:

where a is a positive parameter of the fuel heating apparatus restriction equation, Maximum and minimum values of heat energy of fuel heating device representing ith energy body, respectively,/>Representing the ramp rate of the fuel heating apparatus for two consecutive time periods.

The trade-off constraints between optimality and likelihood of renewable heating devices are as follows:

Where c is the positive parameter of the constraint equation for the renewable heating device, Representing the maximum and minimum values, respectively, of renewable heating device power for the ith energy body.

The constraints on thermal storage are:

wherein, And/>Respectively represent the maximum charge-discharge rate,/>Representing charge-discharge efficiency,/>Representing the state of charge of thermal energy stored in the device, R (·) is related to/>A range function.

The limit of the schedulable thermal load is:

wherein, Is the maximum thermal energy load. /(I)

The ratio of schedulable thermal loads is:

Wherein γ _i,g→h represents the ratio of heat to combined heat and gas load.

(3) In a gas system, the relevant constraints are as follows.

Q (·) is aboutIs a monotonically increasing constraint function of (1), the constraints of the gas supplier are:

wherein, Representing the minimum and maximum values of the gas output of the ith energy body, respectively.

The gas load limits are:

wherein, The maximum gas load for the ith energy volume at time step t.

The ratio of the gas loads is:

wherein, Representing the minimum and maximum values of the ratio of electric power to combined power and gas load respectively,Representing the minimum and maximum values of the ratio of heat to combined heat to gas load, respectively.

In cogeneration, while in the electric and thermal systems, its local operating constraints are:

wherein, And/>Is the coefficient of the linear inequality constraint determined by the feasible operating region of cogeneration,/>Is the ramp rate between two consecutive time steps of the cogeneration generator of the ith energy body;

S2, performing energy control on multiple targets in the energy body in the comprehensive energy management system.

S2.1: consider an integrated energy system with n energy bodies. At the same time, there are m _i participants per energy volume (i.e. renewable power generator, fuel power generator, renewable heating device, cogeneration, electricity storage, gas supply, and dispatchable energy load, etc.). Finding an optimal operation with maximum social benefit, optimal energy distribution and minimum delivery cost among energy bodies, wherein the definition of the optimization targets is as follows:

wherein, Is the maximum benefit value,/>For minimum transport costs, obj is the optimization objective;

in the case of electrical energy which is to be used, The following formula is shown: /(I)

As to the thermal energy of the heat energy,The following formula is shown:

s2.2: definition x _ij∈R³ is a three-dimensional vector of power, thermal energy, and gas of energy-body participants, where i represents an energy-body node and the i-th energy-body is the i-th energy-body node. Some of the elements may be zero, depending on the characteristics of the participants. Definition of the definition Is a three-dimensional vector of power, heat and gas composition of the energy load that must be run j times for the ith energy volume. Converting the variable of each participant into a form of x _ij as shown in the following formula;

Wherein, if x _ij represents a controllable energy load, then B _ij＝-I₃; otherwise, B _ij＝I₃; wherein I ₃ is a three-dimensional identity matrix; omega _ij is a partially closed convex set constrained by a partial inequality.

The cost function corresponding to the zero variable is set to any type of strong convex function. In addition, the endoplasmic reticulum functions to exchange energy and information with other endoplasmic reticulum. To estimate the value of the exchanged energy in a distributed energy network, i.eAndA virtual variable x _ij is assigned to the energy router in the energy bank, letting j=0. A strong convex cost function is assigned to x _i0 and is bounded to zero. Let W (x _ij) then represent the corresponding cost function or negative utility function, as shown by:

S2.3: the objective function and its constraints are converted into the following form:

wherein, And/>Is a locally closed convex set determined by local inequality constraints.

S3, defining three energy distribution diagrams corresponding to different physical links of power, heat and gas among energy bodies.

The distributed communication network is represented as a graphWherein/>Representing a set of nodes. /(I)Representing edge sets, each edge set referring to a communication link. /(I)Representing the adjacency matrix. If node i can receive information from node j, node j is referred to as node i's neighbor. The neighbor set of node i is defined asIf/>Then a _ij = 1; otherwise, a _ij =0. Furthermore, we define/>For the diagram/>Laplace matrix of (2), wherein/>Hypothesis/>Is connected when no attack occurs. In this case, all eigenvalues of L are non-negative; at the same time, there is only one zero eigenvalue. All eigenvalues of L are in the order 0=λ ₁＜λ₂≤,…,≤λ_n

S4, establishing models of DOS attack and FDI attack

S4.1: DOS attacks refer to an attacker occupying network resources or system resources, blocking sensor or actuator communication channels, preventing measured data or control information from normally reaching their respective target locations, causing loss or delay of data, which is a behavior that breaks the availability of data. Such attacks are made by a plurality of computers or devices distributed in different geographical locations that cooperate to simultaneously send a large amount of requests or traffic to the target system, so that the target system is overwhelmed and cannot provide normal service.

For a distributed algorithm, each agent needs to share information with its neighbors to perform local computation. They operate under a distributed communications network. If the communication network is disrupted, the convergence and optimality of the distributed algorithm will correspondingly be disrupted. In this context, we consider that multiple opponents cooperatively launch a denial of service attack. Each adversary's task is to interrupt the transmission of information on its target communication channel (edge). This means that during an attack no information sharing can be done on the attacked channel. In this sense, the impact of a denial of service attack on the communication network, and thus the performance (convergence and optimality) of the distributed algorithm, is affected. The total time series consists of two parts. One is normally without any attack. The other is a denial of service attack.

For simplicity, some symbols are defined. The time interval for the kth attack is defined as [ t _k ^Φ,t_k ^Φ+Δ_k ^Φ ], where t _k ^Φ is the moment of attack initiation and Δ _k ^Φ is the duration of the corresponding attack. Next, we can set let ：Ξ_a(t^Φ)＝∪[t_k ^Φ,t_k ^Φ+Δ_k ^Φ)∩[t₀ ^Φ,t^Φ] denote the union of communication moments that are subject to DoS attacks, and let Ξ_s(t^Φ)＝∪[t_k ^Φ+Δ_k ^Φ,t_k+1 ^Φ)∩[t₀ ^Φ,t^Φ] denote the union of communication moments that are not subject to attacks. Where t ^Φ ₀ is the initial time. We defineTotal duration of attack for [ t0, t ].

Furthermore, note that not all agents are attacked. It is defined that agent i is an attacked node if any of its communication channels is attacked. The set of attacked nodes is denoted An. A successful attack means that it makes graph G non-connected during t e t _k ^Φ,t_k ^Φ+Δ_k ^Φ). In the worst case, each attack launched is considered a successful attack.

S4.2: false data Injection attack (FALSE DATA Injection) refers to that an attacker obtains the direct access right of a communication system, and after grasping information such as certain system parameters, topological structures, detection mechanisms and the like, randomly modifies data measurement values in the communication network to realize hidden attack. For FDI attacks that exist in a power system, an attacker interferes with the normal operation of the system by injecting a false data set into the critical information of the various power components. The system comprises a monitoring information system for attacking the power system, a SCADA measurement system and the like, wherein the attacked node externally presents abnormality, and for the attacked node, iteration information of the attacked node is injected with false data, and the type of attack only needs to acquire data currently transmitted by the comprehensive energy management system, so that an attacker can maliciously superimpose the data on the basis of original data without acquiring model knowledge of the system. Sensor and actuator data subject to FDI attacks can be described as:

/>

Representing malicious data with superimposed integrated energy management system control inputs (i.e. actuator data), Representing malicious data with superimposed output values (i.e., sensor data) corresponding to the integrated energy management system,/>Respectively, whether the executor channel and the sensor channel are subjected to FDI attack, wherein the attack is 1, and the attack is not 0.

Assume that: attack vectorAre bounded in a collection and haveHere/>Is a known positive scalar

S5, using the parallel CNN-BILSTM model to detect network attack.

In order to judge whether the network is attacked or not and judge the attack type when the network is attacked, and realize network attack detection, we propose a parallel CNN-BiLSTM model. The parallel CNN-BiLSTM network can process different parts of the input data simultaneously and make full use of the parallel computing capabilities. This significantly improves computational efficiency and speeds up the model training and reasoning process. Furthermore, the parallel CNN and BiLSTM layers facilitate the simultaneous extraction and integration of data features. This enables us to capture information about various aspects of the data and provide an efficient model representation, thereby improving prediction accuracy.

In particular, the CNN component is operative to extract intrinsic features between different data types within a defined time step. At the same time BiLSTM is used to capture temporal features that are deeper by taking into account information from both the "forward" and "backward" directions. The parallel architecture of CNNs and BiLSTM allows for independent extraction of intrinsic features from various data types and extraction of deeper temporal features from the input data. These features are then connected into a final feature vector that is used to determine the attack type.

S5.1: tensor modeling is respectively carried out on the characteristic data and the power data of the power grid, the air grid and the heat supply network in the comprehensive energy management system, and the tensor modeling is used as X _Λ and P _Λ. Meanwhile, the attack type judged by the neural network model is defined as

X_Λ＝[X_ΛV,X_ΛH,X_ΛP]^T

P_Λ＝[P_ΛD,P_ΛQ,P_ΛR]^T

X_Λi＝[x_Λ(i-L),x_Λ(i-L+1),…,x_Λ(i-2),x_Λ(i-1)]^T

Wherein the method comprises the steps ofInput characteristic data required for attack detection representing the ith sampling time point, wherein the characteristic data comprise relevant quantities such as voltage of a power grid, heat of a heat supply network, pressure of an air network and the like. X _ΛVi represents the input grid voltage required for attack detection at the i-th sampling time point, X _ΛHi represents the input heat supply grid heat required for attack detection at the i-th sampling time point, and X _ΛPi represents the input air supply grid pressure required for attack detection at the i-th sampling time point. L represents the time step of the input data of the parallel CNN-BiLSTM model, note that X _Λi contains time-series data.

S5.2: after normalization, X _Λi is considered as a grayscale image and serves as input to Conv 2D; in addition, X _Λi is planarized and used as an input to BiLSTM.

X _Λi represents the characteristic data acquired at the ith sampling time point; p _Λi represents the comprehensive energy management system power of the ith sampling time point, P _ΛDi represents the grid system power of the ith sampling time point, P _ΛQi represents the air grid system power of the ith sampling time point, and P _ΛRi represents the heat grid system power of the ith sampling time point; Representing the attack type detected by the parallel CNN-BiLSTM model.

S5.3: the judgment of the attack type is carried out,

If X _Λ＝[X_ΛV,X_ΛH,X_ΛP]^T is zero, the communication is interrupted, the attack type is DOS attack, and then S5.2 is returned;

If X _Λ＝[X_ΛV,X_ΛH,X_ΛP]^T is changed drastically and fluctuates greatly, the FDI attack is temporarily judged and then S5.2 is returned;

If X _Λ＝[X_ΛV,X_ΛH,X_ΛP]^T has no change or small amplitude fluctuation, judging that the attack is not generated, and returning to S5.2;

S5.4: outputting the network attack detection result and outputting For DOS attack, output/>For FDI attack, output/>Is attack-free. And outputting local information, wherein the characteristic data comprise relevant quantities such as voltage of a power grid, heat of a heat supply network, pressure of an air network and the like, and power data of the power grid, the air network and the heat supply network as input S6 of a free Jacobian descent algorithm, and setting an elastic initial free Jacobian descent algorithm. By using real-time information. To resist DOS attacks and FDI attacks in an untrusted communication network.

In this section, a distributed resilient uninitialized jacobian descent (DRIFJD) algorithm is presented to resist attacks in an untrusted communication network. We define three control variables, denoted as And/>Each control variable needs to be adjusted during the attack. The detailed expression of DRIFJD algorithm is as follows:

Using combined measurements

And design control variables

/>

Wherein the method comprises the steps ofAnd/>Is an auxiliary variable; /(I)And/>Is within attack interval/>And/>Is determined by the estimation of (a); /(I)And/>Is the latest received or measured state of success of the kth attack. Thereafter, network communications will be blocked for the attacking node during [ t _k ^Φ,t_k ^Φ+Δ_k ^Φ ].

S6.1, firstly, inputting local information, includingAnd/>AndAny value of (2). The characteristic data comprise the voltage of the power grid, the heat of the heat supply network, the pressure of the air network and other related quantities and the power data of the power grid, the air network and the heat supply network

S6.2: each energy network attempts to exchange with its neighborsAnd/>Information of (2);

s6.3: if the communication between the energy sources i and j is successful, the energy sources i and j are not attacked, and the control variable is selected as And/> For updatingEquation/>And/>For auxiliary variables/>And/>Is updated according to the update of the update program.

Is of physical significance with respect to/>Is a dual variable of (c). /(I)AndThe main function of (a) is to collectively converge all z _i to the same value by distributed computing.

S6.4: if the communication between the energy source i and the energy source j is unsuccessful, the attack is suffered, whether the i is the node under attack is judged, and if the i is not the node under attack, the control variable is still selected asAnd/> For updating/>Equation(s)And/>For auxiliary variables/>And/>Is updated according to the update of the update program.

S6.5: if the communication between the energy source i and the j is unsuccessful, the node is attacked, whether the node i is a node under attack is judged, if so, the node i under attack cannot receive neighbor information during the attack period, namely/>And/>These missing data will result in agreed protocol/>And/>Is a failure of (a). To solve this problem, the latest received data/>, is usedAndCalculating/>, replacing those lost dataAnd/>These two data are further used as estimates to calculate the evolution of the system. Selecting control variable as/>And For updating/>Equation(s)And/>For auxiliary variables/>And/>Is updated according to the update of the update program.

Note that: 6.3-6.5 consider two cases. In one aspect, the system operates during normal periods. In this case, each agent can control updates using real-time information. On the other hand, the system is subject to attack. If node i is not the node under attack, it will still be updated dynamically using real-time information. If node i belongs to the attacked node, switching to dynamic updating with the estimation. Meanwhile, a switching strategy is adopted to defend attacks. Specifically, in order to achieve this, three control variables are performed as the communication environment changes, namelyAnd/>

S6.6: judging variableIf the iteration is ended, the iteration is continued by returning to S6.1 until the convergence of the result is ended

And S7, outputting the optimal energy exchange quantity which is carried out when each energy network is completely converged to the optimal state and the optimal energy generation and consumption is achieved among the energy bodies, and controlling the energy generation and transmission of the energy bodies according to the output data by each energy body.

Compared with the prior art, the invention designs the corresponding decision variables and objective functions of each participant in the comprehensive energy system based on global supply and demand constraint, and builds an energy distribution model for exchanging information for the energy router. A detection method and a recovery strategy for resisting DoS attacks and FDI attacks are provided. An analysis model of DoS attacks and FDI attacks is established.

Compared with the existing network attack detection method, the invention provides a new parallel CNN-BILSTM model for network attack detection. Including digital physical models and neural network models (i.e., parallel networks of convolutions and two-way long-short-term memory models) for reflecting the physical operating mechanisms, for capturing hidden spatiotemporal features. The parallel CNN-BiLSTM network can process different parts of the input data simultaneously and make full use of the parallel computing capabilities. This significantly improves computational efficiency and speeds up the model training and reasoning process. Furthermore, the parallel CNN and BiLSTM layers facilitate the simultaneous extraction and integration of data features. This enables us to capture information about various aspects of the data and provide an efficient model representation, thereby improving prediction accuracy.

Compared with the existing method, the invention provides a distributed elastic initialization-free Jacobian descent algorithm. The algorithm designs three switching control protocols, so that the algorithm can reasonably use the estimated value to replace lost information when an attack occurs. The method embeds second-order information, thereby accelerating convergence speed. The method has the advantages of high convergence speed, no special initialization condition, asynchronous communication and the like. Meanwhile, the convergence and the optimality can be maintained in the duration of denial of service attack, and the algorithm can exponentially converge to the global optimal solution of the studied problem.

The present disclosure may be a system, method, and/or computer program product. The computer program product may include a computer readable storage medium having computer readable program instructions embodied thereon for causing a processor to implement aspects of the present disclosure.

The computer readable storage medium may be a tangible device that can hold and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: portable computer disks, hard disks, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static Random Access Memory (SRAM), portable compact disk read-only memory (CD-ROM), digital Versatile Disks (DVD), memory sticks, floppy disks, mechanical coding devices, punch cards or in-groove structures such as punch cards or grooves having instructions stored thereon, and any suitable combination of the foregoing. Computer-readable storage media, as used herein, are not to be construed as transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., optical pulses through fiber optic cables), or electrical signals transmitted through wires.

The computer readable program instructions described herein may be downloaded from a computer readable storage medium to a respective computing/processing device or to an external computer or external storage device over a network, such as the internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmissions, wireless transmissions, routers, firewalls, switches, gateway computers and/or edge servers. The network interface card or network interface in each computing/processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium in the respective computing/processing device.

The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction Set Architecture (ISA) instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including an object oriented programming language such as SMALLTALK, C ++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions may be executed entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider). In some embodiments, aspects of the present disclosure are implemented by personalizing electronic circuitry, such as programmable logic circuitry, field Programmable Gate Arrays (FPGAs), or Programmable Logic Arrays (PLAs), with state information of computer readable program instructions, which can execute the computer readable program instructions.

Finally, it should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that: modifications and equivalents may be made to the specific embodiments of the invention without departing from the spirit and scope of the invention, which is intended to be covered by the claims.

Claims (10)

1. A network attack elasticity detection recovery method based on deep Jacobian comprises the following steps:

s1, performing energy control on multiple targets in each energy body network in a comprehensive energy management system, establishing an objective function with the minimum cost function of the comprehensive energy management system, and establishing constraint conditions of an electric heating network;

s2, under the established constraint condition, the signal acquisition module acquires related information of an Energy Body (EB) and an Energy Router (ER), and the objective function and the constraint condition are simplified;

S3, building an energy distribution model among energy bodies, defining three energy distribution diagrams, and corresponding to different physical links of power, heat and gas quantity among the energy bodies;

s4, establishing DOS and FDI network attack models;

S5, performing network attack detection by using a parallel CNN-BILSTM model, respectively performing tensor modeling on characteristic data and power data of a power grid, an air network and a heat supply network in the comprehensive energy management system, and then connecting the characteristics into a final characteristic vector, wherein the characteristic vector is used for judging attack types;

S6, setting an elastic initial free Jacobian descent algorithm, and dynamically updating by utilizing real-time information to resist DOS attacks and FDI attacks in an untrusted communication network;

And S7, when each energy network meets constraint conditions and meets an objective function, judging that the energy network converges to an optimal state, and controlling energy generation and transmission of the energy body according to the data by each energy body when the energy bodies reach optimal energy exchange quantity.

2. The method for detecting and recovering the elasticity of the network attack based on the deep jacobian as claimed in claim 1, wherein the method comprises the following steps:

The step S1 includes:

S1, establishing a cost function for each device in a comprehensive energy management system, and establishing an objective function with the minimum cost function of the comprehensive energy management system, wherein the constraints of supply and demand balance and the constraints of output of each device are considered, and respective constraint conditions are set for unbalanced power, heat and gas of an electric heating network;

s1.1: the cost function required in the calculation is proposed,

The cost function of the fuel generator and the fuel heating device and the cost function of the cogeneration are respectively obtained by the fuel cost:

wherein, Is a non-negative cost coefficient for a cost function of the fuel generator; /(I)Is a non-negative cost coefficient for the cost function of the fuel heating apparatus,/>Is a non-negative cost coefficient of a cost function for cogeneration,/>The time step distance t and the power and heat energy of cogeneration of the ith energy body are respectively;

the cost functions of the renewable power generator and the renewable heating device are respectively:

wherein, And/>A non-negative cost coefficient that is a cost function for the renewable power generator; /(I)Penalty coefficients for cost functions for renewable power generators; /(I)And/>A non-negative cost coefficient that is a cost function for the renewable heating device; /(I)Penalty coefficients for cost functions for renewable heating devices;

the cost functions of electrical and thermal storage are:

wherein, And/>Is a cost coefficient for the cost function of the electrical storage; /(I)And/>Is a cost coefficient for a cost function of electrical and thermal storage;

The cost function of the gas supply is:

wherein, And/>A non-negative cost coefficient for the gas supply cost function; Is convex within the constraint;

The energy load utility function considering the demand response is:

wherein, And/>Is the non-negative utility coefficient of the energy load effect function of the demand response,/>And/>Respectively represent controllable power, thermal energy and gas load;

revenue/cost of selling/purchasing energy to other energy bodies is:

wherein, And/>Is a market settlement price; /(I)A non-negative coefficient of value for selling or purchasing energy; /(I)And/>Respectively representing power, heat energy and gas exchanged in the energy network;

The total cost of the comprehensive energy management system is as follows:

s1.2, in the comprehensive energy management system, the power is subjected to global supply and demand balance constraint, and unbalanced power, heat and gas constraint conditions of an ith energy body are as follows:

(1) The unbalanced power of the ith energy body at time step t is estimated by the following equation:

wherein, Representing the output of power,/>Load representing power,/>Representing the schedulable power of the ith energy volume at time stride t,/>Fuel generator, renewable generator, cogeneration, electricity storage output power respectively representing the ith energy body at time step t,/>Representing the power load that must be operated and the power load that is scheduled to be operated, respectively; beta ₂、β₂ is the non-negative coefficient of cogeneration power, electricity storage,/>Setting the charge as positive and the discharge as negative;

(2) The unbalanced thermal energy of the ith energy at time step t is estimated by:

wherein, Representing the output of heat energy,/>Representing the load of thermal energy,/>Representing the schedulable thermal energy of the ith energy volume at time step t,/>Fuel heating plant, renewable heating plant, cogeneration, thermal energy output of heat storage,/>, respectively representing the ith energy body at time step tRepresenting the thermal energy load that must be operated and the thermal energy load that is scheduled to be operated, respectively; beta ₂、β₂ is the non-negative coefficient of cogeneration heat energy, heat storage;

(3) The imbalance of the ith energy cell at time step t is estimated by:

wherein, Representing the amount of dispatchable gas of the ith energy volume at time step t; /(I)Representing the gas output of the ith energy body at time step t; /(I)The necessary operating gas load for the ith energy body at time step t; /(I)For a planned operating gas load of the ith energy body at time step t;

the energy load of each energy body comprises an electric load, a thermal load and a gas load, and each load can be divided into an equivalent energy load which is required to be operated and a schedulable energy load;

S1.3, each participant in the energy body is limited, a decision is made according to a group of local constraints, and constraint conditions of an electric heat network in the comprehensive energy management system are as follows:

(1) For electrical energy, the power energy constraints and ramp rate limits of the fuel generator are respectively:

，

wherein, Slope rate representing two successive time periods of the fuel generator,/> Maximum and minimum values of fuel generator power representing the ith energy body, respectively;

the trade-off constraints for optimality and omnipotency of renewable generators are as follows:

，

Where b is the positive parameter of the renewable generator limit equation, And/>Representing the maximum and minimum values of renewable generator power of the ith energy volume respectively,

Constraints on electrical energy storage are:

wherein, And/>Respectively represent the maximum charge-discharge rate,/>Representing charge-discharge efficiency,/>Representing the state of charge of the electrical energy stored in the device, R (·) being a function of the SOC range;

the limitations of the schedulable power load are:

wherein, In order to be a maximum electrical power load,

The ratio of schedulable electrical power loads is:

Wherein, psi represents the conversion ratio from SCM/h to MW, SCM/h is the unit for gas delivery, and this value is set to 1/84; gamma _i,g→p、γ_i,g→h、γ_i,h→g represents the power ratio of electric power to cogeneration and gas load, the power ratio of heat to cogeneration and gas load, the power ratio of gas load and heat to cogeneration, respectively;

(2) For thermal energy, the constraints of the fuel heating apparatus are as follows:

，

where a is a positive parameter of the fuel heating apparatus restriction equation, Maximum and minimum values of heat energy of fuel heating device representing ith energy body, respectively,/>A ramp rate representing two successive time periods of the fuel heating apparatus;

The trade-off constraints between optimality and likelihood of renewable heating devices are as follows:

，

Where c is the positive parameter of the constraint equation for the renewable heating device, Maximum and minimum values of renewable heating device power representing the ith energy body, respectively;

The constraints of thermal energy storage are:

，

wherein, And/>Respectively represent the maximum charge-discharge rate,/>Representing charge-discharge efficiency,/>Representing the state of charge of thermal energy stored in the device, R (·) is related to/>A range function;

The limit of the schedulable thermal load is:

，

wherein, Is the maximum thermal energy load;

The ratio of schedulable thermal loads is:

Wherein γ _i,g→h represents the ratio of heat to combined heat and gas load;

(3) In a gas system, the relevant constraints are as follows;

Q (·) is about Is a monotonically increasing constraint function of (1), the constraints of the gas supplier are:

wherein, A minimum value and a maximum value of gas output representing the ith energy body, respectively;

The gas load limits are:

，

wherein, For maximum gas load of the ith energy body at time step t,

The ratio of the gas loads is:

，

wherein, Representing the minimum and maximum values of the ratio of electric power to combined power and gas load respectively,Representing the minimum and maximum values of the ratio of heat to combined heat to gas load, respectively;

in cogeneration, while in the electric and thermal systems, its local operating constraints are:

wherein, And/>Is the coefficient of the linear inequality constraint determined by the feasible operating region of cogeneration,/>Is the ramp rate between two consecutive time steps of the cogeneration generator of the ith energy.

3. The method for detecting and recovering the elasticity of the network attack based on the deep jacobian as claimed in claim 1, wherein the method comprises the following steps:

The step S2 includes:

S2.1: setting an integrated energy management system containing n energy bodies, wherein each energy body has m _i participants, finding out the optimal operation with the maximum social benefit, the optimal energy distribution and the minimum delivery cost among the energy bodies, and defining the optimal targets as follows:

wherein, Is the maximum benefit value,/>For minimum transport costs, obj is the optimization objective;

in the case of electrical energy which is to be used, The following formula is shown:

as to the thermal energy of the heat energy, The following formula is shown:

s2.2: definition x _ij∈R³ is a three-dimensional vector of energy body participants' power, heat and gas quantity, where i represents the energy body node, i is the i energy body node, definition Is a three-dimensional vector consisting of power, heat and gas quantity of the energy load which must be operated j times of the ith energy body, and converts the variable of each participant into a form of x _ij as shown in the following formula;

Wherein, if x _ij represents a controllable energy load, then B _ij＝-I₃; otherwise, B _ij＝I₃; wherein I ₃ is a three-dimensional identity matrix; omega _ij is a local closed convex set constrained by a local inequality;

Setting the cost function corresponding to the zero variable as any type of strong convex function to estimate the value of the exchange energy in the distributed energy network And/>A virtual variable x _ij is allocated to the energy router in the energy body, let j=0, a strong convex cost function is assigned to x _i0 and set to zero, and then let W (x _ij) represent the corresponding cost function or negative utility function as shown in the following formula:

S2.3: the objective function and its constraints are converted into the following form:

wherein, And/>Is a locally closed convex set determined by local inequality constraints.

4. The method for detecting and recovering the elasticity of the network attack based on the deep jacobian as claimed in claim 3, wherein the method comprises the following steps:

the participants in the energy body include renewable power generators, fuel power generators, renewable heating devices, cogeneration, electricity storage, gas supply, and dispatchable energy loads.

5. The method for detecting and recovering the elasticity of the network attack based on the deep jacobian as claimed in claim 3, wherein the method comprises the following steps:

The step S2 further includes:

The related information collected by the signal collection module comprises: the power, heat energy and gas output, load of each energy body participant and the power, heat energy and gas exchange quantity between energy body networks, and the unbalanced power, heat energy and gas exchange quantity of the ith energy body at the time step t.

6. The method for detecting and recovering the elasticity of the network attack based on the deep jacobian according to claim 5, wherein the method comprises the following steps:

The step S3 includes:

the distributed communication network is represented as a graph Wherein/>A set of nodes is represented and,Representing edge sets, each edge set referring to a communication link,/>Representing an adjacency matrix, when node i can receive information from node j, node j is referred to as node i's neighbor, and node i's neighbor set is defined asIf/>Then a _ij = 1; otherwise, a _ij = 0;

Definition of the definition For the diagram/>Laplace matrix of (2), wherein/>Setting/>Is connected when no attack occurs, in which case all eigenvalues of L are non-negative; meanwhile, only one zero eigenvalue exists, and the sequence of all eigenvalues of L is 0=λ ₁＜λ₂≤,…,≤λ_n;

If an attack is detected, the lost data is replaced by the latest received data, and for a computing unit which is not attacked by the network, each energy-body participant dynamically exchanges information with its neighbor, then the energy router dynamically exchanges information with its neighbor energy router, and the power, heat and gas output of each energy-body participant and the power, heat and gas exchange of the energy router are updated.

7. The method for detecting and recovering the elasticity of the network attack based on the deep jacobian as set forth in claim 6, wherein the method comprises the following steps:

The step S4 includes:

S4.1: the time interval for the kth attack is defined as t _k ^Φ,t_k ^Φ+Δ_k ^Φ, where t _k ^Φ is the moment of attack initiation, a _k ^Φ is the duration ,Ξ_a(t^Φ)＝∪[t_k ^Φ,t_k ^Φ+Δ_k ^Φ)∩[t₀ ^Φ,t^Φ] of the respective attack, the union ,Ξ_s(t^Φ)＝∪[t_k ^Φ+Δ_k ^Φ,t_k+1 ^Φ)∩[t₀ ^Φ,t^Φ] of communication moments subject to DoS attack represents the union of communication moments not subject to attack, where t ^Φ ₀ is the initial time, Total duration of attack for [ t0, t ];

It is defined that if any communication channel of agent i is attacked, agent i is the attacked node, the set of attacked nodes being denoted An;

S4.2: sensor and actuator data subject to FDI attacks are:

Representing superimposed malicious data of integrated energy management system control input,/> Malicious data representing superimposed output values corresponding to comprehensive energy management system,/>Respectively representing whether the executor channel and the sensor channel are subjected to FDI attack, wherein the attack is 1, and the attack is not 0;

setting attack vector The collections are bounded and there are/>Here/>Is a known positive scalar.

8. The method for detecting and recovering the elasticity of the network attack based on the deep jacobian as set forth in claim 6, wherein the method comprises the following steps:

the step S5 includes:

S5.1: tensor modeling is respectively carried out on characteristic data and power data of a power grid, a gas network and a heat supply network in the comprehensive energy management system, and as X _Λ and P _Λ, attack types judged by a neural network model are defined as

X_Λ＝[X_ΛV,X_ΛH,X_ΛP]^T，

P_Λ＝[P_ΛD,P_ΛQ,P_ΛR]^T，

X_Λi＝[x_Λ(i-L),x_Λ(i-L+1),…,x_Λ(i-2),x_Λ(i-1)]^T，

Wherein the method comprises the steps ofInput characteristic data required for attack detection at the ith sampling time point are represented, wherein the characteristic data comprise voltage of a power grid, heat of a heat supply network and pressure of an air network, X _ΛVi represents input power grid voltage required for attack detection at the ith sampling time point, X _ΛHi represents input heat of the heat supply network required for attack detection at the ith sampling time point, X _ΛPi represents input air network pressure required for attack detection at the ith sampling time point, L represents time steps of input data of a parallel CNN-BiLSTM model, and X _Λi comprises time sequence data;

S5.2: after normalization, X _Λi is considered a grayscale image and as input to Conv2D, X _Λi is flattened and used as input to BiLSTM;

X _Λi represents the characteristic data acquired at the ith sampling time point; p _Λi represents the comprehensive energy management system power of the ith sampling time point, P _ΛDi represents the grid system power of the ith sampling time point, P _ΛQi represents the air grid system power of the ith sampling time point, and P _ΛRi represents the heat grid system power of the ith sampling time point; representing attack types detected by a parallel CNN-BiLSTM model;

S5.3: the judgment of the attack type is carried out,

If X _Λ＝[X_ΛV,X_ΛH,X_ΛP]^T is zero, the communication is interrupted, the attack type is DOS attack, and then S5.2 is returned;

If X _Λ＝[X_ΛV,X_ΛH,X_ΛP]^T is changed drastically and fluctuates greatly, the FDI attack is temporarily judged and then S5.2 is returned;

If X _Λ＝[X_ΛV,X_ΛH,X_ΛP]^T has no change or small amplitude fluctuation, judging that the attack is not generated, and returning to S5.2;

S5.4: outputting the network attack detection result and outputting For DOS attack, output/>For FDI attack, output/>In order to avoid attack, local information is output, and the characteristic data comprise relevant quantities such as voltage of a power grid, heat of a heat supply network, pressure of an air network and the like and power data of the power grid, the air network and the heat supply network as input of a free Jacobian descent algorithm.

9. The method for detecting and recovering the elasticity of the network attack based on the deep jacobian as set forth in claim 7, wherein the method comprises the following steps:

The step S6 includes:

A distributed elastic non-initialized Jacobian descent (DRIFJD) algorithm is proposed to resist attacks in an untrusted communication network, defining three control variables, expressed respectively as And/>Each control variable needs to be adjusted during the attack, and the DRIFJD algorithm is expressed as follows:

Using combined measurements

And design control variables

Wherein the method comprises the steps ofAnd/>Is an auxiliary variable; /(I)And/>Is within attack interval/>And/>Is determined by the estimation of (a); and/> The latest received or measured state that is successful for the kth attack, blocking network communications for the attacking node during [ t _k ^Φ,t_k ^Φ+Δ_k ^Φ);

S6.1, inputting local information, including And/>And/>The characteristic data comprises related quantities such as voltage of a power grid, heat of a heat supply network, pressure of an air network and the like, and power data of the power grid, the air network and the heat supply network;

S6.2: each energy network attempts to exchange with its neighbors And/>Information of (2);

S6.3: if the communication between the energy sources i and j is successful, the energy sources i and j are judged not to be attacked, and the control variable is selected as

And/> For updatingEquation/>And/>For auxiliary variables/>And/>Is updated according to the update of (a); /(I)For/>Is/are of the dual variable of (1)And/>The main function of (a) is to collectively converge all z _i to the same value by distributed computing;

s6.4: if the communication between the energy sources i and j is unsuccessful, judging that the node is attacked, judging whether the node can communicate with two adjacent energy sources, and if the node cannot communicate, judging that the node is attacked; if communication is possible, it is determined that i is not a node under attack, and the control variable is selected as And/> For updating/>Equation(s)And/>For auxiliary variables/>And/>Is updated according to the update of (a);

S6.5: if the communication between the energy source i and the energy source j is unsuccessful, judging that the energy source i is attacked, judging whether the energy source i is a attacked node, and if the energy source i cannot communicate, judging that the energy source i is a attacked node; if communication is possible, determining that i is not the node under attack; if yes, the attacked node i can not receive the neighbor information during the attack period And/> With most recent received data/>AndCalculating/>, instead of lost dataAnd/>These two data are further used as estimates to calculate the evolution of the system, selecting the control variable as/>And For updating/>Equation(s)And/>For auxiliary variables/>And/>Is updated according to the update of (a);

s6.6: judging variable If the iteration is converged, ending the iteration if the iteration is not converged, returning to the step S6.1 to continue the iteration until the convergence of the result is ended.

10. A deep jacobian-based network attack resilience detection recovery system, the system being used by the deep jacobian-based network attack resilience detection recovery method according to any one of claims 1-9, comprising:

The signal acquisition module is used for acquiring the related information of the energy body and the energy router;

the attack detection module is used for detecting network attacks and judging attack types;

and the recovery module is used for resisting the network attack by utilizing the dynamic update of the real-time information.