A Virtual Power Plant Risk Characterization Method Based on Semantic Entropy and Information Age

By constructing a networked information flow equation for virtual power plants based on semantic entropy and information age, the problem of unified representation of the risks of diverse and heterogeneous information in virtual power plants is solved, and the true reflection and efficient assessment of nonlinear effects are achieved.

CN123117764APending Publication Date: 2026-11-17CHONGQING UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611125449.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-28
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies for quantitatively assessing the risks of diverse and heterogeneous information in virtual power plants suffer from problems such as model stacking, high modeling complexity, inability to characterize nonlinear coupling effects, and difficulty in reflecting the impact of communication delays and data tampering on the frequency regulation performance of virtual power plants.

Method used

By employing a method based on semantic entropy and information age, and by acquiring demand-side resource status information, communication network parameters, and frequency regulation control parameters of a virtual power plant, a networked information flow equation for the virtual power plant is constructed. The equation is then solved in a reduced-order manner by combining the information network topology, thus uniformly representing the risks of multi-dimensional information.

Benefits of technology

It achieves a unified representation of the risks of diverse information in virtual power plants, breaks through the limitations of traditional one-by-one modeling, can truly reflect nonlinear effects, improves evaluation efficiency, and meets the real-time requirements of engineering applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123117764A_ABST
    Figure CN123117764A_ABST
Patent Text Reader

Abstract

This invention provides a method for characterizing virtual power plant risks based on semantic entropy and information age, belonging to the field of power system information and control technology. The method involves extracting information transmission features based on communication network parameters and frequency regulation control parameters to obtain node information age, and combining this with demand-side resource status information to perform channel state deviation analysis, thus obtaining information semantic entropy. Based on the analytical functional relationship between information semantic entropy and node information age, and combined with the information network topology, a networked information flow equation for the virtual power plant is constructed. The information network is equivalent to a multi-port network with demand-side resources as the information source and the management and control platform as the information destination. By ignoring relay nodes, the networked information flow equation for the virtual power plant is solved with reduced order, yielding a unified characterization result for the multivariate information risks of the virtual power plant. This invention solves the problems of model stacking and inability to characterize nonlinear coupling effects in existing technologies when analyzing multivariate heterogeneous risks.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power system information and control technology, and in particular to a method for characterizing the risk of a virtual power plant based on semantic entropy and information age. Background Technology

[0002] In recent years, with the large-scale integration of new energy sources such as wind and solar power, the proportion of adjustable power sources on the power system's source side has continued to decline, necessitating the exploration of adjustability from the demand side. The development of advanced information and communication technologies such as cloud computing and 5G has given rise to virtual power plants. Virtual power plants are mainly composed of demand-side resources such as electric vehicles, distributed energy storage, and flexible loads. Through the perception and decision-making of massive demand-side resources, they provide services such as peak shaving and valley filling, and frequency regulation for the power system. However, virtual power plants mostly use public wireless networks for communication, resulting in low levels of information security protection. They are highly vulnerable to diverse information risks, including spoofed data injection attacks, edge computing denial-of-service attacks, occasional equipment failures, and communication and computing latency, which can constrain frequency regulation performance and even endanger grid security. Therefore, it is urgent to quantitatively assess the impact of diverse information risks on virtual power plants.

[0003] Existing methods for characterizing information risks can be broadly categorized into two perspectives. One is the electrical engineering perspective, which follows a "modeling information risks one by one" paradigm. For example, it establishes observable and controllable models for communication interruptions and time-delay models for latency. While these methods can reflect the individualized characteristics of risk factors, they inevitably face problems of model stacking and extremely high modeling complexity when dealing with concurrent multi-faceted information risks. Furthermore, they easily overlook the highly nonlinear coupling between spurious data injection and latency. The other perspective is the information science perspective, which focuses on the encoding, transmission, and decoding processes of information. Whether it's establishing message queue models based on queuing theory or the nascent semantic information flow models, these approaches are limited to analyzing the impact on the performance of the information network itself, lacking integration with the dynamic processes of continuous physical systems. Therefore, they cannot be used to quantitatively assess the impact of diverse and heterogeneous information risks on the actual frequency regulation performance of virtual power plants.

[0004] In summary, how to overcome the technical bias of modeling one by one and achieve a unified representation and efficient solution of risks from diverse and heterogeneous information is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the aforementioned shortcomings in existing technologies, this invention provides a virtual power plant risk characterization method based on semantic entropy and information age. This method solves the problems of existing methods for characterizing information risks in virtual power plants following a one-to-one modeling paradigm, which leads to model stacking and inability to characterize nonlinear coupling effects when analyzing multivariate heterogeneous risks. Furthermore, the non-additive nature of information results in excessively high order of the global information flow equation, making it difficult to solve.

[0006] To achieve the aforementioned objectives, the technical solution adopted by this invention is: a virtual power plant risk characterization method based on semantic entropy and information age, comprising:

[0007] S1: Obtain demand-side resource status information, communication network parameters, and frequency regulation control parameters of the virtual power plant;

[0008] S2: Based on communication network parameters and frequency modulation control parameters, information transmission features are extracted to obtain node information age that reflects the immediacy, directionality and reachability of transmission. Combined with demand-side resource status information, channel state deviation analysis is performed to obtain information semantic entropy that reflects the semantic transmission relationship between the two ends of the channel.

[0009] S3: Based on the analytical function relationship between information semantic entropy and node information age, and combined with the information network topology, construct a virtual power plant networked information flow equation that covers demand-side resources;

[0010] S4: The information network is equivalent to a multi-port network with demand-side resources as the information source and the management and control platform as the information destination. By ignoring relay nodes, the networked information flow equation of the virtual power plant is solved in a reduced order to obtain a unified representation of the multi-dimensional information risks of the virtual power plant, thus completing the risk representation of the virtual power plant.

[0011] Further, S2 includes:

[0012] Based on the communication network parameters, channel fading characteristics between the communication access point and demand-side resources are extracted. Combined with the management and control platform processing performance parameters extracted from the frequency modulation control parameters, the time difference including information transmission time and information processing time is calculated as the node information age.

[0013] Based on the node information age, the delay time of the demand-side resource status information is determined, and the probability distribution of the status information after being tampered with by information risk is extracted as input data. The mean deviation and variance deviation between the real status distribution and the tampered status distribution are calculated to obtain the relative deviation value.

[0014] The relative deviation value is substituted into the distribution divergence calculation model to calculate the relative deviation, and the information semantic entropy is obtained.

[0015] Furthermore, the expression for the age of the node information is:

[0016] ;

[0017] in, This indicates the age of the node information. This represents the control cycle time extracted based on the frequency modulation control parameters, where the summation term represents the information transmission time, including the transmission time of multiple links. This indicates the size of the data packet extracted based on the demand-side resource status information. This represents the backhaul link bandwidth between resources and edge systems, obtained based on preset parameters. This represents the communication bandwidth extracted based on the communication network parameters. This indicates the signal transmission power obtained based on the base station equipment parameters. This represents the Rayleigh channel coefficients obtained based on channel measurement methods. This represents the Euclidean distance between devices obtained based on location data. This represents the preset path attenuation coefficient. This represents the noise power within the bandwidth obtained through environmental monitoring methods. This indicates the backhaul link bandwidth between the access point and the management platform, obtained based on preset parameters. This indicates the information processing time. This represents the CPU cycle complexity coefficient required to process a unit of data obtained through pre-calibration based on numerical experiments. This indicates the computing power of the management and control platform based on the frequency modulation control parameters extracted.

[0018] Furthermore, the expression for the distribution divergence calculation model is:

[0019] ;

[0020] in, This represents the relative deviation value output by the distribution divergence calculation model. This represents the information parameter used as an integral variable. This represents the probability density function of the extracted true probability distribution before it was altered due to information risk. The probability density function represents the extracted, altered probability distribution. This indicates the altered state information, and The value depends on the tampering model and the input parameters. .

[0021] Furthermore, the expression for the semantic entropy of the information is:

[0022] ;

[0023] in, This represents the semantic entropy of the information. This represents the tamper attack drift coefficient obtained based on historical risk data statistics. This represents the variance of tamper attack propagation obtained based on historical risk data statistics. This represents the average spread of tampering attacks obtained based on historical risk data statistics. This represents the variance deviation of the true state distribution. The mean deviation of the true state distribution is obtained by solving stochastic differential equations based on the unaltered demand-side resource state information. This represents the variance deviation of the tampered probability distribution after introducing the age delay of the node information. This indicates the mean deviation of the tampered probability distribution after introducing the age delay of the node information.

[0024] Further, S3 includes:

[0025] Based on the information network topology, an information transmission map between the information source and the information sink is determined, and an information transmission matrix reflecting the status of the information propagation link is generated.

[0026] Based on the analytical function relationship between the semantic entropy of the information and the age of the node information, the semantic vector of the node information of each network node is determined.

[0027] By multiplying and associating the information transmission matrix with the node information semantic vector, the networked information flow equation of the virtual power plant issued by the demand-side resources is obtained.

[0028] Further, S4 includes:

[0029] Boundary conditions are constructed based on the physical location relationship of the demand-side resources and the management platform in the network, and the network topology relationship, excluding relay nodes other than the information source and the information sink, is equivalent to the multi-port network.

[0030] Substituting the known information probability distribution at the demand-side resource location and the known node information age received at the control platform into the multi-port network, the unknown node semantic parameters in the virtual power plant networked information flow equation are reduced in order and eliminated to obtain the reduced-order information flow equation.

[0031] Based on the reduced-order information flow equation, the semantic distribution data of information about the source at the sink is calculated, and a unified representation result of the multivariate information risk of the virtual power plant is obtained, thus completing the risk representation of the virtual power plant.

[0032] Furthermore, the expression for the reduced-order information flow equation is:

[0033] ;

[0034] ;

[0035] in, This represents the set of comprehensive semantic entropy of multiple information sources to information destinations in the information network before the order reduction. This represents the distribution of multiple real-state variables of the demand-side resources. This indicates the distribution of multiple demand-side resource status variables received by the management platform. This represents the information transfer matrix generated based on the information network topology. This represents a semantic vector containing node information of each network node, including the relay node to be solved. This represents the semantic distribution data of information about a single information source at the destination obtained by the reduced-order solution, i.e., the unified representation result. Indicates the integral information parameter, The probability density function represents the probability distribution of known information at the demand-side resources, which are used as boundary condition inputs. This indicates the age of the known node information received at the control platform. Then, the probability density function of the destination information is obtained after removing the semantic vector of the unknown relay node information.

[0036] The beneficial effects of this invention are as follows: This invention provides a virtual power plant risk characterization method based on semantic entropy and information age, which has the following beneficial effects:

[0037] (1) It breaks away from the traditional paradigm of modeling specific attacks or faults one by one, and innovatively uses the age of node information to uniformly quantify the immediacy, directionality and reachability of information transmission and decision-making, and uses the semantic entropy of information to uniformly quantify the semantic deviation at both ends of the channel. This mechanism can reduce the dimensionality of diverse and heterogeneous risks such as false data injection, delay and equipment failure to a unified representation system, and avoid omissions caused by improper selection of state variables.

[0038] (2) Based on the analytical function relationship between information semantic entropy and node information age, and combined with the information network topology, a networked information flow equation for virtual power plants covering demand-side resources was constructed. This model effectively breaks through the limitations of simple linear superposition in traditional physical models and can truly reflect the nonlinear amplification effect of concurrent risks such as communication delays and data tampering on virtual power plant decision-making.

[0039] (3) To address the problem that networked information flow equations do not satisfy Kirchhoff's laws and that information is not additive, resulting in a solution dimension as high as one million, this invention innovatively equates the information network to a multi-port network with demand-side resources as the information source and the management and control platform as the information destination. By relying only on known boundary conditions at both ends and ignoring the complex relay nodes, the high-dimensional equation is compressed into a single boundary integral formula that can be calculated quickly, greatly improving the evaluation efficiency and meeting the real-time requirements of engineering applications. Attached Figure Description

[0040] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:

[0041] Figure 1 This is an exemplary flowchart illustrating a virtual power plant risk characterization method based on semantic entropy and information age, according to some embodiments of this specification.

[0042] Figure 2 This is a control architecture diagram of a virtual power plant “cloud-pipe-edge-device” as shown in some embodiments of this specification;

[0043] Figure 3 This is a schematic diagram illustrating age changes based on some embodiments of this specification;

[0044] Figure 4 This is a virtual power plant information flow model diagram shown according to some embodiments of this specification;

[0045] Figure 5 This is a map showing the distribution of base stations and demand-side resources in a local area of ​​a city, based on some embodiments of this specification.

[0046] Figure 6 This is a probability density distribution diagram showing the occurrence of a false data injection attack before and after some embodiments of this specification;

[0047] Figure 7 This is a graph showing the change of information age and control signal under basic scenarios according to some embodiments of this specification, where (a) is information age and (b) is control signal;

[0048] Figure 8 This is a comparison diagram of the semantic changes of demand-side resource status information in scenarios 1 to 3 as shown in some embodiments of this specification, where (a) is the semantic entropy of information, (b) is the mathematical expectation of the state variable, and (c) is the variance of the state variable.

[0049] Figure 9 This is a comparison diagram of information age and information semantic entropy in scenario 1 and scenario 4 according to some embodiments of this specification, where (a) is information age and (b) is information semantic entropy;

[0050] Figure 10 This is a comparison diagram of information flow results in scenario 1 and scenario 4 according to some embodiments of this specification, where (a) is the control signal, (b) is the mathematical expectation of the state variable, and (c) is the variance of the state variable;

[0051] Figure 11This is a comparison diagram of information age and information semantic entropy in scenario 1 and scenario 5 according to some embodiments of this specification, where (a) is information age and (b) is information semantic entropy;

[0052] Figure 12 This is a comparison diagram of the information flow results in scenario 1 and scenario 5 according to some embodiments of this specification, where (a) is the control signal, (b) is the mathematical expectation of the state variable, and (c) is the variance of the state variable;

[0053] Figure 13 This is a comparison diagram of information age and information semantic entropy in scenarios 1, 6 and 7 shown in some embodiments of this specification, where (a) is information age and (b) is information semantic entropy;

[0054] Figure 14 This is a comparison diagram of information flow results in scenarios 1, 6 and 7 as shown in some embodiments of this specification, where (a) is the control signal, (b) is the mathematical expectation of the state variable and (c) is the variance of the state variable;

[0055] Figure 15 The graph shows the sensitivity analysis results of a fake data injection attack according to some embodiments of this specification, where (a) is the drift coefficient sensitivity, (b) is the diffusion coefficient sensitivity, and (c) is the variance sensitivity.

[0056] Figure 16 The graph shows the sensitivity analysis results of delay attacks according to some embodiments of this specification, where (a) represents the change in information age and (b) represents the change in information semantic entropy. Detailed Implementation

[0057] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0058] Example

[0059] Figure 1 This is an exemplary flowchart illustrating a virtual power plant risk characterization method based on semantic entropy and information age, according to some embodiments of this specification. Figure 1 As shown, the process includes the following steps. In some embodiments, the process may be executed by a processor.

[0060] S1: Obtain the demand-side resource status information, communication network parameters, and frequency regulation control parameters of the virtual power plant.

[0061] A virtual power plant is a system architecture that aggregates massive, dispersed demand-side resources such as electric vehicles, distributed energy storage, and flexible loads to provide peak shaving, valley filling, frequency regulation, and other control services for the power system. For example, a virtual power plant may include a cluster of control platforms deployed in the cloud, an energy management system at the edge, and adjustable electrical equipment and energy storage devices distributed in residential and industrial areas.

[0062] In some embodiments, the processor can utilize a high-performance computer cluster as a cloud management platform, establish communication connections via a public wireless network or fiber optic communication network, and perceive and integrate the operating status and frequency regulation capabilities of demand-side resources distributed in various locations in real time, thereby forming a virtual power plant that is uniformly subject to grid dispatch.

[0063] Demand-side resource status information is a set of numerical data reflecting the real-time operating conditions and adjustability of distributed resources within a virtual power plant at a specific moment. For example, demand-side resource status information may include the real-time charging and discharging power, maximum energy capacity, and initial state of energy of distributed energy storage, as well as specific operating parameters such as the current charging rate and remaining power of electric vehicles.

[0064] In some embodiments, the processor can collect real-time operating voltage, current and power data of various load devices through sensing modules and edge-side energy management systems deployed on the resource side, and package and convert these collected underlying physical quantities according to a preset control cycle and communication protocol to obtain demand-side resource status information.

[0065] In some embodiments, to obtain the mean and variance of the true probability distribution of the demand-side resource status information, a dynamic model of the demand-side resources needs to be established first. The expression of the dynamic model of the demand-side resources is:

[0066] ;

[0067] in, This represents the change in active power of demand-side resources, i.e., the state variable of demand-side resources. Indicates the time parameter. This represents the inherent time constant of demand-side resources obtained based on the inherent physical properties of the equipment. This refers to control inputs issued by the virtual power plant cloud management platform, and is limited by the physical upper and lower bounds of the equipment. This represents the Gaussian white noise measurement error obtained based on the device's sensing accuracy. This represents the error coefficient obtained based on the equipment error calibration.

[0068] Based on the above model, the mean deviation of the true state distribution after introducing the age delay of the node information is obtained. With variance bias The expression is:

[0069] ;

[0070] ;

[0071] in, This indicates the initial value measured at the initial moment of the demand-side resource state quantity. This indicates the age of the node information. This represents the time variable for integration.

[0072] Communication network parameters are a set of physical characteristic indicators that describe the information transmission capabilities and environmental interference conditions between communication nodes and links in a virtual power plant information network. For example, communication network parameters may include the communication bandwidth of the backbone network or access network, the data transmission power of the base station signal, the channel attenuation coefficient, the path data transmission coefficient, and the environmental noise power within the bandwidth range.

[0073] In some embodiments, the processor can query device configuration information by calling the base station network management system, or use channel measurement equipment to detect the fading characteristics of the wireless channel in complex environments such as urban high-rise buildings in real time, and statistically analyze the power loss and background noise level of the signal during transmission to obtain communication network parameters.

[0074] Frequency regulation control parameters are the time and computing power allocation rules used by a virtual power plant to constrain and guide computational decision-making and command issuance processes when responding to the frequency regulation needs of the power system. For example, frequency regulation control parameters may include the control cycle length of the frequency regulation command issued by the power system dispatch center, as well as the computing power of the central processing unit of the virtual power plant cloud management platform for processing massive amounts of data and the unit data processing complexity coefficient.

[0075] In some embodiments, the processor can obtain frequency modulation control parameters by receiving time control constraints set by the upper-level power grid control center and combining them with benchmark test results of the cloud server hardware to evaluate the number of processor cycles that the cluster can execute per second and the cycle overhead required to process a single bit of data.

[0076] S2: Based on communication network parameters and frequency modulation control parameters, information transmission features are extracted to obtain node information age that reflects the immediacy, directionality, and reachability of transmission. Combined with demand-side resource status information, channel state deviation analysis is performed to obtain information semantic entropy that reflects the semantic transmission relationship between the two ends of the channel.

[0077] Node information age is an indicator used to characterize the freshness of information transmission and decision-making processes in information networks, reflecting the immediacy, directionality, and accessibility of information. For example, node information age can include the time difference between the absolute time of a node in the information network and the time when the information source sends the information used for the current decision update. It usually presents as a sawtooth distribution of data that increases linearly with time and decreases due to decision updates.

[0078] In some embodiments, the processor can extract the timestamps from the information data packets, calculate the transmission time of each segment of the information in the communication access network and backbone network, and add the computation time required by the management and control platform to process this status information based on its computing power, and add the transmission time and computation time to obtain the node information age.

[0079] In some embodiments, to support the calculation of node information age, it is necessary to calculate the channel transmission rate and signal-to-interference-plus-noise ratio between the communication access point and the demand-side resources, and their expressions are as follows:

[0080] ;

[0081] ;

[0082] in, This represents the signal-to-interference-plus-noise ratio (SINR) calculated based on channel fading characteristics. This represents the channel transmission rate calculated based on the signal-to-interference-plus-noise ratio and bandwidth. Indicates signal transmission power. Represents the Rayleigh channel coefficient. This indicates the Euclidean distance between the devices. Represents the path attenuation coefficient. Indicates the noise power over the bandwidth. The above parameters, which represent communication bandwidth, are all included in the communication network parameters.

[0083] In some embodiments, the calculation of node information age depends on the precise breakdown of transmission time for each segment.

[0084] The expressions for the information transmission time and information processing time of each level of the link are as follows:

[0085] ;

[0086] ;

[0087] ;

[0088] ;

[0089] ;

[0090] in, This represents the first transmission time from the demand-side resources to the edge management system, calculated based on data volume and bandwidth. This represents the second transmission time from the edge system to the communication access point, calculated based on the data volume and channel rate. This represents the third transmission time from the access point to the cloud management platform, calculated based on data volume and bandwidth. This represents the total communication transmission time, including the time consumed by each segment. Indicates the total information processing time. This indicates the size of the data packet as perceived by the sensing module. , These represent the preset backhaul link bandwidth. This represents the channel transmission rate derived from the signal-to-interference-plus-noise ratio (SINR). This represents the complexity coefficient obtained during calibration. This indicates the computing power of the cloud management platform.

[0091] Meanwhile, the expression for the Rayleigh channel gain required to derive the channel transmission rate is:

[0092] ;

[0093] in, This represents the channel gain obtained based on physical environment characteristics. This represents the Rayleigh channel coefficients obtained from actual measurements. This represents the Euclidean distance between devices calculated based on location. This represents the preset path attenuation coefficient.

[0094] Information semantic entropy is a metric based on continuous numerical information characteristics, used to quantify the degree to which the received information deviates from the true information of the source due to various information risks such as tampering attacks after information has undergone network transmission. For example, information semantic entropy can include a relative deviation value reflecting the semantic transmission relationship between the two ends of the channel, which is reflected as the combined divergence data of the mean and variance between the tampered demand-side resource state variables and the true variables.

[0095] In some embodiments, the processor can construct the Gaussian distribution probability density function of the real state variables of the demand-side resources, derive the probability distribution of the state variables after being affected by the risk by combining the channel attack model, and calculate the average logarithmic deviation integral when the received distribution is approximated by the real distribution to obtain the information semantic entropy.

[0096] In some embodiments, when the information risk faced is a linear spoofing attack, i.e., the diffusion variance of the tampering attack is constant, and considering the time difference of information delay, the expression for the information semantic entropy is specifically simplified as follows:

[0097] ;

[0098] in, This represents the constant diffusion coefficient of a linear tampering attack, obtained based on historical risk data statistics. The definitions of the remaining parameters are the same as those for the semantic entropy parameter in the Gaussian attack mode.

[0099] In some embodiments, the processor can extract channel fading characteristics between the communication access point and demand-side resources based on the communication network parameters, and combine the management and control platform processing performance parameters extracted by the frequency modulation control parameters to calculate the time difference between information transmission time and information processing time, which is used as the node information age; determine the delay time of demand-side resource status information based on the node information age, and extract the probability distribution of status information after information risk tampering as input data, calculate the mean deviation and variance deviation between the true status distribution and the tampered status distribution to obtain the relative deviation value; substitute the relative deviation value into the distribution divergence calculation model to calculate the relative deviation and obtain the information semantic entropy.

[0100] Channel fading characteristics describe the physical properties that cause random fluctuations in signal strength when wireless communication signals propagate in complex urban environments due to multipath effects such as building obstruction, refraction, and reflection. For example, channel fading characteristics may include data such as Rayleigh channel coefficients conforming to an independent and identically distributed quasi-static Rayleigh distribution, Euclidean distances between devices, and path attenuation coefficients characterizing the signal attenuation rate.

[0101] In some embodiments, the processor can calculate the Euclidean spatial distance by obtaining the geographic location coordinates between the edge-side energy management system and the communication access point, and combine this with on-site electromagnetic environment measurements or statistical channel modeling analysis to analyze the multipath power envelope distribution of the wireless signal after scattering, thereby obtaining the channel fading characteristics.

[0102] The processing performance parameters of the control platform are computing power indicators that reflect the ability of the cloud-based computer cluster of the virtual power plant to process massive amounts of demand-side resource status information and generate frequency regulation combination strategies. For example, the processing performance parameters of the control platform may include the total number of CPU cycles that the cloud control platform can execute per second, as well as the complexity coefficient required for the system to process one bit of status data, and other computing power-related data.

[0103] In some embodiments, the processor can obtain the number of available CPU cores and clock frequency of the cluster by calling the performance monitoring interface of the underlying operating system of the cloud management platform, and obtain the processing performance parameters of the management platform by statistically analyzing the computing power overhead of a specific frequency modulation control algorithm for processing unit bit information through numerical experiments.

[0104] In some embodiments, the expression for the node information age is:

[0105] ;

[0106] in, This indicates the age of the node information. This represents the control cycle time extracted based on the frequency modulation control parameters, where the summation term represents the information transmission time, including the transmission time of multiple links. This indicates the size of the data packet extracted based on the demand-side resource status information. This represents the backhaul link bandwidth between resources and edge systems, obtained based on preset parameters. This represents the communication bandwidth extracted based on the communication network parameters. This indicates the signal transmission power obtained based on the base station equipment parameters. This represents the Rayleigh channel coefficients obtained based on channel measurement methods. This represents the Euclidean distance between devices obtained based on location data. This represents the preset path attenuation coefficient. This represents the noise power within the bandwidth obtained through environmental monitoring methods. This indicates the backhaul link bandwidth between the access point and the management platform, obtained based on preset parameters. This indicates the information processing time. This represents the CPU cycle complexity coefficient required to process a unit of data obtained through pre-calibration based on numerical experiments. This indicates the computing power of the management and control platform based on the frequency modulation control parameters extracted.

[0107] The latency of demand-side resource status information is the time span from when the information is generated and sent from the underlying resource device until it is received and processed by the cloud management platform. For example, the latency of demand-side resource status information can include the communication transmission time of the information on the backhaul link, the communication time from the edge access point to the cloud, and the computing time required by the cloud management platform to consume computing power.

[0108] In some embodiments, the processor can extract the generation timestamp of the state information data packet on which the current decision depends at the information source, and read the current system time when the cloud management platform has completed the decision processing and is ready to issue the instruction. By comparing the difference between these two timestamps, the processor can obtain the delay time of the demand-side resource state information.

[0109] State information probability distribution is a mathematical characteristic data used to describe the probabilistic laws governing the values ​​of state variables such as active power of demand-side resources at a given moment, which have stochastic characteristics. For example, state information probability distribution can include the Gaussian probability density function of the real state variables derived from a first-order linear stochastic differential equation, as well as the tampered state variable probability density distribution formed by superimposing tampering drift coefficients and diffusion coefficients introduced by attackers.

[0110] In some embodiments, the processor can construct a demand-side resource dynamic model that includes control inputs and Gaussian white noise measurement errors, calculate the mathematical expectation and Ito equidistant variance of the equation, determine its mean and variance characteristics, and then construct the distribution function of the continuous random variable to obtain the probability distribution of state information.

[0111] The relative deviation value is a measure used to quantitatively characterize the overall difference between the probability distribution of tampered or delayed information received by the receiver and the probability distribution of the true information emitted by the source. For example, the relative deviation value can include integrated nonlinear deviation data calculated using KL divergence, reflecting the difference between the tampered mean and variance and the true mean and variance after introducing the information age delay effect.

[0112] In some embodiments, the processor can extract the probability density functions of the true state distribution and the tampered state distribution respectively, divide the density function of the tampered distribution by the density function of the true distribution and take the logarithm, and then multiply it by the density function of the true distribution and perform an anomalous integral operation over the entire real number domain to obtain the relative deviation value.

[0113] A distribution divergence calculation model is a mathematical analysis structure used to measure the similarity or difference between the probability distributions of two continuous random variables in order to assess the semantic fidelity of information transmission. For example, a distribution divergence calculation model may include an integral equation based on Kullback-Leibler divergence, which uses the average logarithmic deviation of the probability density of the information at the sink to approximate the probability density of the true information at the source.

[0114] In some embodiments, the processor can calculate the logarithmic expectation of the ratio of the two density functions by setting the probability density functions of two contrasting distributions as inputs and using one of the true distributions as a reference benchmark. This mathematical analysis logic is then transformed into executable computer code logic and integral solution steps to obtain the distribution divergence calculation model.

[0115] In some embodiments, the expression for the distribution divergence calculation model is:

[0116] ;

[0117] in, This represents the relative deviation value output by the distribution divergence calculation model. This represents the information parameter used as an integral variable. This represents the probability density function of the extracted true probability distribution before it was altered due to information risk. The probability density function represents the extracted, altered probability distribution. This indicates the altered state information, and The value depends on the tampering model and the input parameters. .

[0118] In some embodiments, it is necessary to construct an information tampering model and calculate the state distribution characteristics before the introduction of delay. The expressions for the tampered demand-side resource state information in the state without delay and the tampering expression after the introduction of delay are as follows:

[0119] ;

[0120] ;

[0121] in, This represents the state variable after it has been altered due to risk. This represents the drift coefficient from tampering attacks extracted based on historical data. Represents the real state variable that has not been tampered with. This represents the diffusion coefficient obtained based on historical attack characteristics. This represents the actual state variable after introducing known node information, age, and time delay.

[0122] For a true state distribution that has not been tampered with and has not introduced delays, its mean and variance are expressed as follows:

[0123] ;

[0124] ;

[0125] in, This represents the mean of the true state distribution obtained by solving the differential equation. This represents the variance of the true state distribution obtained by integrating the isometry property of Itoh. Indicates the current moment. Represents the inherent time constant of the resource. This represents the initial state measurement value. Indicates control input, This represents the error coefficient.

[0126] To ensure the control security of demand-side resources, the expression of the control input must satisfy:

[0127] ;

[0128] in, and These represent the lower and upper bounds of the control inputs obtained based on the device's hardware parameters, respectively.

[0129] In some embodiments, the expression for information semantic entropy is:

[0130] ;

[0131] in, This represents the semantic entropy of the information. This represents the tamper attack drift coefficient obtained based on historical risk data statistics. This represents the variance of tamper attack propagation obtained based on historical risk data statistics. This represents the average spread of tampering attacks obtained based on historical risk data statistics. This represents the variance deviation of the true state distribution. The mean deviation of the true state distribution is obtained by solving stochastic differential equations based on the unaltered demand-side resource state information. This represents the variance deviation of the tampered probability distribution after introducing the age delay of the node information. This indicates the mean deviation of the tampered probability distribution after introducing the age delay of the node information.

[0132] In some embodiments, to verify the accuracy of the adopted semantic entropy, a conventional geometric measure model is introduced for parallel computation verification. The expressions for mutual information, Wasserstein distance, and total variation distance are as follows:

[0133] ;

[0134] ;

[0135] ;

[0136] in, This represents the mutual information result used to characterize the degree of dependency. This represents the joint probability density function of two variables. and These represent the marginal density functions of the corresponding variables. This indicates the Wasserstein distance verification result. This represents the coupling mapping function between distributions. This represents the validation result of the total variation distance.

[0137] To conduct extreme testing, the verification expression for the bimodal distribution attack is constructed as follows:

[0138] ;

[0139] in, This represents the result of a bimodal distribution probability model generated by artificially distorting the original distribution. This represents the established Gaussian distribution benchmark function.

[0140] S3: Based on the analytical function relationship between information semantic entropy and node information age, and combined with the information network topology, construct a virtual power plant networked information flow equation that covers demand-side resources.

[0141] An information network topology is a network skeleton model that describes the physical connections and logical information flow between information sources, communication relay nodes at all levels, and information sinks in a virtual power plant. For example, an information network topology may include a list of connections between demand-side resources, edge-side energy management systems, 4G / 5G communication access points, and cloud management platforms, hierarchical architecture information, and a data graph set containing nodes and connections.

[0142] In some embodiments, the processor can read the registration access information and routing table entries of devices at all levels in the communication network management system, sort out the end-to-end communication link connection relationship and direction attributes from the underlying sensing module to the edge gateway, and then to the cloud server through the public network base station, and obtain the information network topology.

[0143] The virtual power plant networked information flow equation is a mathematical model system that comprehensively considers the transmission characteristics of information elements and the characteristics of network topology, and is used to globally describe the semantic changes of information from multiple sources to destinations. For example, the virtual power plant networked information flow equation can include a system of equations that equate the information network to matrix multiplication, where the left side is the set of comprehensive semantic entropies of each source to destination, and the right side is the product of the information transmission matrix and the semantic vector of the node information.

[0144] In some embodiments, the processor can use the analytic function relationship between the semantic entropy of each information branch and the age of the node information as a component model, combine it with the information transmission graph that reflects the ordered serial and parallel topology of the network to generate a system-level Laplace matrix, and then combine the two matrices to obtain the networked information flow equation of the virtual power plant.

[0145] In some embodiments, the construction of the networked information flow equation for a virtual power plant is based on the analysis of a single information source path. The semantic entropy derivation expression for a single information source flow is:

[0146] ;

[0147] in, This represents the semantic entropy of information from a single source to a destination. This represents a relay node in the network. This indicates the set of nodes covered by the information flow. This indicates that the probability density function at the information source is known. and Let represent the probability density functions of adjacent nodes, respectively. Represents the integral parameter. This represents a row vector consisting entirely of 1s, generated based on the number of nodes. This represents the Laplacian matrix generated by extracting the topology of this information stream. This represents the semantic vector of the node information corresponding to the information flow.

[0148] In some embodiments, the processor can determine the information transmission map between the information source and the information sink based on the information network topology, and generate an information transmission matrix that reflects the state of the information propagation link; determine the node information semantic vector of each network node based on the analytical function relationship between the information semantic entropy and the node information age; and multiply and correlate the information transmission matrix with the node information semantic vector to obtain the virtual power plant networked information flow equation issued by the demand-side resources.

[0149] An information transmission graph is a directed graph data structure that reflects the propagation path and node coverage of information in a virtual power plant communication network, embodying the non-ergodicity and sparsity of information flow. For example, an information transmission graph can include a visualized or matrix-based graph theory model consisting of a set of demand-side resource nodes, a set of relay nodes for communication equipment at all levels, sink nodes, and directional information flow connections between them.

[0150] In some embodiments, the processor can extract the historical routing trajectory of all routers, base stations, and gateways that a single piece of information passes through after being sent from a specific information source, and perform graph theory set aggregation and edge weight assignment on all parallel transmission trajectories generated by tens of thousands of massive demand-side resources in the system to obtain an information transmission graph.

[0151] The information transmission matrix is ​​derived from the information transmission graph and is a highly sparse matrix data used in mathematical equations to quantify the connectivity of information propagation links and the relationships between network nodes. For example, the information transmission matrix may include matrix blocks composed of Laplace matrices corresponding to each single-source information flow arranged in a specific diagonal structure; if an information flow does not pass through a specific node, the corresponding row of the matrix is ​​zero.

[0152] In some embodiments, the processor can construct a local connectivity matrix for each source-to-destination path in the information transfer graph by parsing the directed connections in the graph, and then use the Kronecker product or diagonal concatenation operation to integrate the local matrices of all sources in a global dimension to obtain the information transfer matrix.

[0153] An analytic functional relationship is a mathematical formulaic expression describing the quantitative mapping relationship between the information age parameter and the information semantic entropy parameter at the channel input and output ports. For example, an analytic functional relationship can include an analytic expression reflecting how, when the information age increases due to risks such as time delay, the decisions made by the receiver based on historical data deviate from the actual situation, resulting in a non-linear monotonically increasing characteristic of the information semantic entropy.

[0154] In some embodiments, the processor can recalculate the mean and variance of the state distribution after introducing the time lag variable due to communication and computation delays into the tampering attack model, and derive a new divergence calculation formula containing the delay variable by substituting it into the distribution divergence integral formula, thereby obtaining the analytical function relationship.

[0155] A node information semantic vector is a one-dimensional array of data containing quantified semantic values ​​of information corresponding to each specific node in a particular information flow path within a network. For example, a node information semantic vector may include a column vector consisting of the logarithmic integral of the original information emitted by the source at its own node, and the logarithmic integral of the corresponding probability of the information received at each subsequent relay node and the destination node, arranged in order.

[0156] In some embodiments, the processor can calculate the local probability density function of an information flow when it passes through a specific relay node, combine the analytical function relationship to calculate the logarithmic integral value of the information semantics at that node, and then combine the calculations of all network nodes by applying the rule that the semantic value is set to zero if the information does not pass through that node, and then combine the results into an array to obtain the node information semantic vector.

[0157] In some embodiments, such as Figure 2As shown, this invention provides a "cloud-pipe-edge-device" control architecture for a virtual power plant. In this architecture, "cloud" refers to the virtual power plant cloud management platform, supported by a high-performance computer cluster, responsible for frequency regulation capability modeling and scheduling decisions; "pipe" represents communication networks such as 4G, 5G, and WiFi, enabling bidirectional data and command transmission; "edge" includes communication access points and edge-side energy management systems, ensuring secure access to massive resources; and "device" refers to demand-side resources such as electric vehicles and energy storage, along with their sensing and communication modules, responsible for state perception and flexible regulation. This architecture provides physical and informational carriers for the occurrence and propagation of diverse information risks.

[0158] In some embodiments, such as Figure 3 As shown, the information age exhibits a sawtooth distribution, increasing linearly over time and then plummeting at the moment of decision update. The information age decreases when the source sends initial state information and it is used to make decisions at the destination. The peak value of the information age is jointly determined by the control cycle, communication transmission time, and the computation time of the management platform, intuitively reflecting the immediacy of information transmission and decision-making. The relative size of the information age at different nodes reflects the directionality of transmission, while the boundedness of the information age at the destination characterizes the accessibility of the information.

[0159] In some embodiments, such as Figure 4 As shown, the virtual power plant information flow model consists of demand-side resources representing information sources, a communication network containing numerous relay nodes, and a management and control platform serving as the information destination. This model constructs a networked structure based on information semantic entropy and node information age through the ordered serial and parallel connection of information elements, thereby replacing the traditional power flow network. This allows for macroscopic tracking of the semantic decay and latency accumulation of multi-source heterogeneous information throughout the entire virtual power plant network.

[0160] S4: The information network is equivalent to a multi-port network with demand-side resources as the information source and the management and control platform as the information destination. By ignoring relay nodes, the networked information flow equation of the virtual power plant is solved in a reduced order to obtain a unified representation of the multi-dimensional information risks of the virtual power plant, thus completing the risk representation of the virtual power plant.

[0161] An information network is composed of various information transmission and processing elements connected in series and parallel. It serves as the communication carrier for sensing state data and issuing control commands between the virtual power plant cloud platform and massive edge resources. For example, an information network may include communication access networks using wireless operator networks such as 4G / 5G and WiFi, as well as infrastructure such as communication backbone networks using fiber optic communication to connect the cloud platform and communication access points.

[0162] In some embodiments, the processor can logically connect and configure interconnection routes for these physical devices by integrating smart gateway devices deployed in residential, commercial, and industrial areas, regional wireless base station towers, and high-performance switches in data centers, using network communication protocols such as TCP / IP, to obtain an information network.

[0163] Multiport networks are a macroscopic abstract network model that focuses on the characteristics of input and output ends, addressing the problem of high-order solution difficulties caused by the non-additivity of information flow. It simplifies and equates complex information networks. For example, a multiport network can include an equivalent structural model with N demand-side resources as N input source ports, a single virtual power plant cloud management platform as a single receiving destination port, and shielding all internal relay transmission nodes.

[0164] In some embodiments, the processor can obtain a multi-port network by stripping the detailed state variables of various intermediate transmission links such as base stations and routers in the original information network that are not involved in the final decision-making reception, and retaining only the node characteristics of the source of massive information generation and the reception characteristics of the final aggregation point, and logically constructing an end-to-end direct mapping relationship.

[0165] The unified characterization result of multi-source information risks in virtual power plants is the final output data that comprehensively quantifies the impact of heterogeneous risks such as spoofed data injection, edge computing denial-of-service attacks, equipment failures, and latency on the frequency regulation performance of virtual power plants after solving reduced-order equations. For example, the unified characterization result of multi-source information risks in virtual power plants can include a set of comprehensive information semantic entropy distribution data for each information source calculated at the information sink.

[0166] In some embodiments, the processor can substitute the obtained communication delay, attack tampering parameters, and computing power impairment parameters into the constructed reduced-order information flow equation, and perform a divergence integral calculation process under multi-port boundary conditions. This reduces the dimensionality of the coupled effects of multiple risks and outputs them as intuitive numerical evaluation indicators, thus obtaining a unified characterization result of the multiple information risks of the virtual power plant.

[0167] In some embodiments, the processor can construct boundary conditions based on the physical location relationship between the demand-side resources and the management platform in the network, and convert the network topology relationship, excluding relay nodes other than the information source and the information sink, into the multi-port network; substitute the known information probability distribution at the demand-side resources and the known node information age received at the management platform into the multi-port network, and perform order reduction and elimination processing on the unknown node semantic parameters in the networked information flow equation of the virtual power plant to obtain a reduced-order information flow equation; calculate the information semantic distribution data of the information source at the information sink based on the reduced-order information flow equation, and obtain the unified characterization result of the multivariate information risk of the virtual power plant, thus completing the risk characterization of the virtual power plant.

[0168] Boundary conditions are known state parameters and constraint baseline data set at both ends of a multi-port network when simplifying and solving complex information flow equations. For example, boundary conditions may include the probability density function of the original state information of the demand-side resources input on the source side, which has not been tampered with, and the known node information age data reflecting the global transmission delay, which has been accurately measured on the sink side.

[0169] In some embodiments, the processor can determine the source probability distribution by extracting the first and second moments of the state variables at the source under normal system operation, and calculate the end-to-end total reception delay by extracting the source timestamps of the data packets received by the cloud management platform. These determined peripheral parameters are set as known quantities required to solve the equations to obtain the boundary conditions.

[0170] Network topology, stripped of its physical entity attributes, is a logical abstraction describing the connectivity and organizational structure of nodes in an information network from a purely graph theory perspective. For example, network topology can include directed graph edge sets and node degree distribution data describing how multiple edge-side energy management systems converge to a single communication access point, and how these access points are connected in a star topology to a central cloud platform.

[0171] In some embodiments, the processor can use a web crawler or device management interface to traverse and obtain the adjacency routing table of all communication nodes in the network, filter out transmission physical quantities such as specific bandwidth and power, retain only the Boolean values ​​of connectivity between nodes and convert them into the adjacency matrix representation in graph theory to obtain the network topology.

[0172] The known information probability distribution is the inherent mathematical and statistical data of the original information at the source end, used as the boundary condition for solving reduced-order equations, when it is not subject to interference from various information risks in the intermediate transmission network. For example, the known information probability distribution can include the probability density function expression of a Gaussian random variable whose mean and variance have not been introduced with additional drift or diffusion terms by hackers, based on physical parameters such as the time constant of the demand-side resources themselves and measurement white noise.

[0173] In some embodiments, the processor can read the dynamic model parameters of the demand-side resources independently tested and calibrated before grid connection or in a secure isolation environment, solve the mathematical expectation and variance of the first-order differential equation under the influence of no external control input and tampering, and thus construct an initial distribution function to obtain the probability distribution of known information.

[0174] The known node information age is the deterministic total delay time data included in the information received at the sink end, obtained directly from observation or calculation as a macroscopic boundary condition when solving a multi-port network. For example, the known node information age may include the specific peak information age value, including transmission and calculation time, directly recorded by the system at the moment when the cloud management platform receives a status information packet and is preparing to execute a frequency modulation decision.

[0175] In some embodiments, the processor can read the current system clock at the moment the control decision algorithm is triggered on the virtual power plant cloud platform, parse the generation timestamp field in the header of the latest arriving data packet being processed, subtract the two and filter out jitter errors to obtain the known node information age.

[0176] Unknown node semantic parameters are unknowns in the complex, high-dimensional virtual power plant network information flow equations, corresponding to the unsolved semantic changes in information caused by invisible attacks or delays experienced by numerous intermediate routers or base stations within the network. For example, unknown node semantic parameters can be included in the node information semantic vector, which, in addition to the information source itself and the final destination, corresponds to the specific unsolved values ​​of the logarithmic integrals of the local probability distributions of hundreds or thousands of access network nodes and backbone network nodes.

[0177] In some embodiments, the processor can set the input-output semantic relationships of each node in the intermediate transmission link as a variable symbol system in the equations when constructing the system global information flow equations, and identify and classify these symbol systems in a matrix before preparing to use a multi-port order reduction algorithm that ignores relay nodes, to obtain the semantic parameters of unknown nodes.

[0178] The reduced-order information flow equation is derived to address the difficulty in solving the original high-order networked information flow model due to the non-additivity of information. It is a simplified mathematical model that removes a large number of redundant internal node parameters. For example, the reduced-order information flow equation can include a single boundary integral formula that retains only the integral form of the known probability density function of the information source and directly substitutes the macroscopic information age parameter into the density function of the information sink as a time delay correction, thereby directly solving the end-to-end semantic entropy.

[0179] In some embodiments, the processor can use the equivalence principle of multiport networks to perform elimination and removal of unknown node variables in the higher-order equation in the form of the product of the original information transfer matrix and the node semantic vector, and replace the corresponding terms in the remaining formula with the known boundary conditions at both ends to obtain the reduced-order information flow equation.

[0180] In some embodiments, the expression for the reduced-order information flow equation is:

[0181] ;

[0182] ;

[0183] in, This represents the set of comprehensive semantic entropy of multiple information sources to information destinations in the information network before the order reduction. This represents the distribution of multiple real-state variables of the demand-side resources. This indicates the distribution of multiple demand-side resource status variables received by the management platform. This represents the information transfer matrix generated based on the information network topology. This represents a semantic vector containing node information of each network node, including the relay node to be solved. This represents the semantic distribution data of information about a single information source at the destination obtained by the reduced-order solution, i.e., the unified representation result. Indicates the integral information parameter, The probability density function represents the probability distribution of known information at the demand-side resources, which are used as boundary condition inputs. This indicates the age of the known node information received at the control platform. Then, the probability density function of the destination information is obtained after removing the semantic vector of the unknown relay node information.

[0184] Information semantic distribution data is obtained by solving reduced-order information flow equations. It is a set of results that quantitatively presents the distortion and misrepresentation of meaning that occurs when information from various demand-side resources arrives at the cloud management platform. For example, information semantic distribution data may include specific numerical matrices or charts of the final semantic entropy of thousands of electric vehicles and distributed energy storage devices in the system, calculated under specific communication network conditions and the superposition of various heterogeneous risk attacks.

[0185] In some embodiments, the processor can substitute the initial boundary conditions of each resource into the reduced-order integral formula, and use the powerful numerical integration computing capabilities of the computer cluster or the Monte Carlo simulation algorithm to quickly solve the KL divergence results of each different source information at the sink, and organize and archive these results into structured data reports to obtain information semantic distribution data.

[0186] In some embodiments, to ensure that the risk characterization results of the virtual power plant accurately reflect the actual operating conditions of the urban power grid, the communication network parameters, the frequency regulation control parameters, and related physical constants are all given specific preferred values ​​or value ranges, as follows:

[0187] Regarding the communication network parameters: the communication bandwidth Preferably 10MHz; the signal transmission power Preferably 27 dBm; the Rayleigh channel coefficient Preferably, it is 1; the path attenuation coefficient Preferably 4; noise power within the bandwidth range Preferably -104dBm; the bandwidth of each level of backhaul link (e.g. , The preferred configuration is 20Mbps.

[0188] Regarding the equipment and attack constant parameters involved in the virtual power plant control process: the inherent time constant of the demand-side resources. The value range is configured as [0.01, 0.1] seconds; the mean value of the Gaussian white noise measurement error is... Set to 0 or 0.0001, error variance The value range is configured as [0.2, 0.9]; the error coefficient The preferred setting for equipment accuracy is typically between 0.011 and 1. For different types of demand-side resources, the adjustable active power constraint range is set as follows: [2,4]kW for residential loads, [50,100]kW for commercial loads, [100,1000]kW for industrial loads, and [2,10]kW for distributed energy storage. The complexity coefficient of the number of CPU cycles required to process a unit of data is also specified. Based on the server architecture and the complexity of the control algorithm, the fixed scalar value was determined through offline numerical simulation experiments.

[0189] In some embodiments, such as Figure 5 As shown, to verify the effectiveness of this method, a virtual power plant test system covering a local area of ​​a city was constructed. The blue crosses in the figure indicate the locations of 4G LTE communication access points, and the surrounding Voronoi polygons delineate the communication coverage areas of each access point. Within this area, a large number of WiFi access points are deployed in densely populated commercial and residential areas, while industrial areas are relatively sparse. Thousands of electric vehicles, distributed energy storage systems, and various types of adjustable load equipment are widely distributed throughout the area, constituting a highly representative city-level virtual power plant verification environment.

[0190] In some embodiments, such as Figure 6 The figure illustrates the changes in probability density of demand-side resources before and after a data injection attack. The true state variables, when unaffected, follow a baseline Gaussian distribution; when attacked, one pattern causes a constant shift in the mean of the state information, while another pattern distorts the probability density function, forming a bimodal distribution with the same mean and variance. The figure visually reflects the heterogeneous disruption of the underlying data's statistical characteristics caused by different tampering patterns.

[0191] In some embodiments, such as Figure 7As shown, in a normal basic scenario (Scenario 1) where no information attacks or equipment failures occur, Figure 7 (a) shows that due to the small communication delay, the information age of resources such as distributed energy storage remains within a low bounded range, exhibiting regular sawtooth fluctuations. Figure 7 (b) indicates that the control signals issued by the control platform can be dynamically adjusted in a stable and continuous manner based on the timely updated status data, and the frequency adjustment decision-making process of the entire system is in a healthy state.

[0192] In some embodiments, such as Figure 8 As shown, this illustrates a comparison of the semantic changes in information across scenarios 1 (no attack), 2 (linear fake data injection), and 3 (Gaussian fake data injection). Figure 8 As shown in (a), the semantic entropy of information in scenario 1 is close to 0, while the semantic entropy of information in scenarios 2 and 3 surges, indicating that the information received by the cloud platform deviates significantly from the true value. Figure 8 (b) Show how the mathematical expectation of the state variables changes as the control signal is updated; Figure 8 (c) This shows that the variance tends to a constant value over time, proving that the information semantic entropy index used in this invention can accurately capture and quantify the degree of data deviation caused by attacks.

[0193] In some embodiments, such as Figure 9 As shown, scenario 1 is compared with scenario 4, where an edge computing denial-of-service (DoS) attack occurs. Due to the DoS attack occurring in the first second, the edge server's computing power is exhausted, and state information is forced to be offloaded to the cloud for processing. Figure 9 (a) The peak age displayed in scenario 4 is significantly higher than normal; this is due to the influence of age delay coupling. Figure 9 (b) The peak value of semantic entropy in scenario 4 also increases significantly, quantifying the dual damage to system immediacy and final semantic fidelity caused by the loss of computing power.

[0194] In some embodiments, such as Figure 10 As shown, this reflects the actual control results of the information flow in scenarios 1 and 4. Figure 10 (a) The DoS attack shows that the control signal issuance in scenario 4 is significantly delayed; Figure 10 (b) indicates that the mathematical expectation of the demand-side resource state variables has therefore been delayed. This further confirms that impaired computing power not only slows down communication but also directly distorts the control system's effective scheduling of physical resources.

[0195] In some embodiments, such as Figure 11 As shown, scenario 1 is compared with scenario 5, where an edge-side energy management system failure occurs. Figure 11(a) shows that the failure occurred after the 6th second, and the inability to upload information caused the information age to continuously spike towards infinity, disrupting accessibility; however, in Figure 11 In (b), the corresponding semantic entropy of information remains at 0 after the 6th second, which reveals that when information is completely interrupted, the cloud no longer makes decisions based on erroneous information, reflecting the special logic of risk masking in the physical control process.

[0196] In some embodiments, such as Figure 12 As shown, the impact of scenarios 1 and 5 on the distribution of underlying resource states is illustrated. Figure 12 (a) After a fault is displayed, the control command is interrupted and the signal returns to zero; Figure 12 (b) and Figure 12 (c) indicates that the expectation of the resource state variables on the demand side gradually decays, while the variance continues to evolve according to its own physical laws and is not affected by the control signal. This shows that when the directionality of information is blocked, resources can only rely on their inherent inertia to operate.

[0197] In some embodiments, such as Figure 13 As shown, scenarios 1, 6 (where only a latency attack occurs), and 7 (where latency and Gaussian tampering occur concurrently) are compared. Figure 13 (a) Scenes 6 and 7 show that the peak information age increased significantly due to latency attacks; Figure 13 (b) It reveals that under concurrent attacks, the semantic entropy of information in scenario 7 is much higher than that in scenario 6 under a single attack, and this value is not a simple linear superposition of the consequences of latency and tampering alone, verifying that there is a highly nonlinear coupling amplification effect between multiple risks.

[0198] In some embodiments, such as Figure 14 As shown, Figure 14 In (a), the control signals for scenarios 6 and 7 show a significant lag compared to scenario 1; corresponding to Figure 14 The expected value of the state variables in (b) also deviates significantly from the baseline trajectory, while Figure 14 The variance evolution trend in (c) remains consistent. This figure clearly demonstrates the most severe disruptive effects of concurrency risks on the flexible control process of the virtual power plant.

[0199] In some embodiments, such as Figure 15 As shown, this demonstrates the system's sensitivity to parameters in response to linear spoofing attacks. Figure 15 Figures (a), (b), and (c) plot the upward trend of information semantic entropy as the drift coefficient, diffusion coefficient, and Gaussian attack variance increase. The curves show that the greater the tampering intensity, the more severe the divergence between the information received in the cloud and the true value. The mean and peak values ​​of information semantic entropy both exhibit non-linear growth, effectively measuring the consequences of different attack intensities.

[0200] In some embodiments, such as Figure 16 As shown, a sensitivity analysis of pure delay attacks is presented. Figure 16 (a) and (b) show that as the set delay increases stepwise, the peak values ​​of information age and information semantic entropy not only rise significantly, but the timing of the peaks also exhibits a clear lag effect. Furthermore, the growth rate of semantic entropy shows a non-linear convergence characteristic relative to the increase in delay, accurately depicting the non-linear transmission mechanism of network latency on decision fidelity.

Claims

1. A method for characterizing the risk of a virtual power plant based on semantic entropy and information age, characterized in that, Includes the following steps: S1: Obtain demand-side resource status information, communication network parameters, and frequency regulation control parameters of the virtual power plant; S2: Based on communication network parameters and frequency modulation control parameters, information transmission features are extracted to obtain node information age that reflects the immediacy, directionality and reachability of transmission. Combined with demand-side resource status information, channel state deviation analysis is performed to obtain information semantic entropy that reflects the semantic transmission relationship between the two ends of the channel. S3: Based on the analytical function relationship between information semantic entropy and node information age, and combined with the information network topology, construct a virtual power plant networked information flow equation that covers demand-side resources; S4: The information network is equivalent to a multi-port network with demand-side resources as the information source and the management and control platform as the information destination. By ignoring relay nodes, the networked information flow equation of the virtual power plant is solved in a reduced order to obtain a unified representation of the multi-dimensional information risks of the virtual power plant, thus completing the risk representation of the virtual power plant.

2. The virtual power plant risk characterization method based on semantic entropy and information age according to claim 1, characterized in that, S2 includes: Based on the communication network parameters, channel fading characteristics between the communication access point and demand-side resources are extracted. Combined with the management and control platform processing performance parameters extracted from the frequency modulation control parameters, the time difference including information transmission time and information processing time is calculated as the node information age. Based on the node information age, the delay time of the demand-side resource status information is determined, and the probability distribution of the status information after being tampered with by information risk is extracted as input data. The mean deviation and variance deviation between the real status distribution and the tampered status distribution are calculated to obtain the relative deviation value. The relative deviation value is substituted into the distribution divergence calculation model to calculate the relative deviation, and the information semantic entropy is obtained.

3. The virtual power plant risk characterization method based on semantic entropy and information age according to claim 2, characterized in that, The expression for the age of the node information is: ; in, This indicates the age of the node information. This represents the control cycle time extracted based on the frequency modulation control parameters, where the summation term represents the information transmission time, including the transmission time of multiple links. This indicates the size of the data packet extracted based on the demand-side resource status information. This represents the backhaul link bandwidth between resources and edge systems, obtained based on preset parameters. This represents the communication bandwidth extracted based on the communication network parameters. This indicates the signal transmission power obtained based on the base station equipment parameters. This represents the Rayleigh channel coefficients obtained based on channel measurement methods. This represents the Euclidean distance between devices obtained based on location data. This represents the preset path attenuation coefficient. This represents the noise power within the bandwidth obtained through environmental monitoring methods. This indicates the backhaul link bandwidth between the access point and the management platform, obtained based on preset parameters. This indicates the information processing time. This represents the CPU cycle complexity coefficient required to process a unit of data obtained through pre-calibration based on numerical experiments. This indicates the computing power of the management and control platform based on the frequency modulation control parameters extracted.

4. The virtual power plant risk characterization method based on semantic entropy and information age according to claim 2, characterized in that, The expression for the distribution divergence calculation model is as follows: ; in, This represents the relative deviation value output by the distribution divergence calculation model. This represents the information parameter used as an integral variable. This represents the probability density function of the extracted true probability distribution before it was altered due to information risk. The probability density function represents the extracted, altered probability distribution. This indicates the altered state information, and The value depends on the tampering model and the input parameters. .

5. The virtual power plant risk characterization method based on semantic entropy and information age according to claim 2, characterized in that, The expression for the semantic entropy of the information is: ; in, This represents the semantic entropy of the information. This represents the tamper attack drift coefficient obtained based on historical risk data statistics. This represents the variance of tamper attack propagation obtained based on historical risk data statistics. This represents the average spread of tampering attacks obtained based on historical risk data statistics. This represents the variance deviation of the true state distribution. The mean deviation of the true state distribution is obtained by solving stochastic differential equations based on the unaltered demand-side resource state information. This represents the variance deviation of the tampered probability distribution after introducing the age delay of the node information. This indicates the mean deviation of the tampered probability distribution after introducing the age delay of the node information.

6. The virtual power plant risk characterization method based on semantic entropy and information age according to claim 1, characterized in that, S3 includes: Based on the information network topology, an information transmission map between the information source and the information sink is determined, and an information transmission matrix reflecting the status of the information propagation link is generated. Based on the analytical function relationship between the semantic entropy of the information and the age of the node information, the semantic vector of the node information of each network node is determined. By multiplying and associating the information transmission matrix with the node information semantic vector, the networked information flow equation of the virtual power plant issued by the demand-side resources is obtained.

7. The virtual power plant risk characterization method based on semantic entropy and information age according to claim 1, characterized in that, S4 includes: Boundary conditions are constructed based on the physical location relationship of the demand-side resources and the management platform in the network, and the network topology relationship, excluding relay nodes other than the information source and the information sink, is equivalent to the multi-port network. Substituting the known information probability distribution at the demand-side resource location and the known node information age received at the control platform into the multi-port network, the unknown node semantic parameters in the virtual power plant networked information flow equation are reduced in order and eliminated to obtain the reduced-order information flow equation. Based on the reduced-order information flow equation, the semantic distribution data of information about the source at the sink is calculated, and a unified representation result of the multivariate information risk of the virtual power plant is obtained, thus completing the risk representation of the virtual power plant.

8. The virtual power plant risk characterization method based on semantic entropy and information age according to claim 7, characterized in that, The expression for the reduced-order information flow equation is: ; ; in, This represents the set of comprehensive semantic entropy of multiple information sources to information destinations in the information network before the order reduction. This represents the distribution of multiple demand-side resource real-state variables. This indicates the distribution of multiple demand-side resource status variables received by the management platform. This represents the information transfer matrix generated based on the information network topology. This represents a semantic vector containing node information of each network node, including the relay node to be solved. This represents the semantic distribution data of information about a single information source at the destination obtained by the reduced-order solution, i.e., the unified representation result. Indicates the integral information parameter, The probability density function represents the probability distribution of known information at the demand-side resources, which are used as boundary condition inputs. This indicates the age of the known node information received at the control platform. Then, the probability density function of the destination information is obtained after removing the semantic vector of the unknown relay node information.