CN-122027031-A - Collaborative control method and system based on space-time holography and neural reasoning

CN122027031ACN 122027031 ACN122027031 ACN 122027031ACN-122027031-A

Abstract

The invention discloses a cooperative control method and a system based on space-time holography and neural reasoning, which are applied to an optical cable embedded optical fiber wireless heterogeneous communication scene. The method comprises the steps of constructing a system comprising a central server/gateway, a weakly-coupled multi-core optical fiber backbone network and a distributed micro-nano node group, constructing and updating a space-time hologram by the nodes, deducing average field statistics, realizing group collaborative optimization based on average field game and a neural network, and completing individual decision through neural initiative reasoning to form closed-loop collaborative control. The invention solves the contradiction of energy uncertainty, power dimension disaster and perception-decision splitting in the prior art, improves network survival time, cooperative efficiency, data quality and environmental robustness, and is suitable for high-precision perception scenes such as intelligent power grids, long-distance pipelines and the like.

Inventors

XIAO ZIYANG
WANG HUA
LI LUMING
PENG CHAO
ZHANG ZHIGUO
GU XUELIANG

Assignees

国网江西省电力有限公司信息通信分公司

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. The cooperative control method based on space-time holography and neuro reasoning is characterized by comprising the following steps of: S1, constructing a distributed micro-nano sensing network system, wherein the distributed micro-nano sensing network system comprises a central server/gateway, an optical fiber energy transmission/communication backbone network and a distributed micro-nano sensing node group, the optical fiber energy transmission/communication backbone network adopts a weak coupling multi-core optical fiber, and the distributed micro-nano sensing node group is embedded in or attached to an optical cable; s2, each micro-nano sensing node builds and updates a space-time hologram, wherein the space-time hologram is a lightweight distributed world model of the joint probability distribution of key physical and information states related to the decision of a coding network; s3, carrying out group collaborative modeling and solving based on average field game; s4, each micro-nano sensing node makes individual decisions based on nerve initiative reasoning; and S5, repeating the step S2 to the step S4 to form closed loop cooperative control of 'node updating STH-receiving average field-executing MFG-AIF joint decision'.
2. The collaborative control method based on space-time holography and neuro-reasoning according to claim 1, wherein step S3 is specifically: S31 defining individual status Individual control actions Including perceptual mode/frequency selection, communication decisions and computational load distribution, The energy is left over for the node and, T is a matrix/vector transpose symbol, which is a low-dimensional latent variable of the space-time hologram; S32 defining an average field For all node states Empirical probability distribution over a state space, each node acquiring the average field Low order statistics of (2); S33 definition node From the slave Expected cost function to terminal time T: ; in the formula: A desired cost function from the current time T to the terminal time T is used as an ith node; for mathematical expectation calculation, T is the current time, T is the termination time, The instant expected free energy of the ith node at time tau, For the control action vector of the node, R is the control effort weight matrix, In order to control the quadratic cost term of the action, The terminal cost function of the ith node at the terminal moment T is used as the terminal cost function; s34, approximately solving a coupling partial differential equation of the MFG through an average field neural network, wherein the average field neural network comprises a strategy network And average field evolution network The gateway trains off-line and fine-tunes the average field neural network on-line, each node obtains the approximate optimal control action through the strategy network based on the self state and the average field low order statistic, wherein Is the control action vector of the i-th node, Is the state vector of the node.
3. The collaborative control method based on space-time holography and neuro-reasoning according to claim 1, wherein step S4 is specifically: S41, constructing a generating model In the following For the observational quantity, s is the hidden state, Is a priori probability distribution of the environmental state, Is a priori probability distribution of the internal states, Observing a conditional probability distribution of o for a given hidden state s; s42, performing state estimation based on Variable Free Energy (VFE) minimization, wherein the variable free energy calculation formula is as follows: in the following Is the variation free energy at the time t, For the relative divergence of KL, Hidden state for time t Is used for the distribution of the variational posterior probability, Hidden state for time t Is used to determine the prior probability distribution of (c), Is the hidden state at the moment t, Is based on variational posterior distribution Is used for the mathematical expectation operation of (a), The observed quantity at the time t; s43, performing action selection based on the expected free energy minimization, wherein the expected free energy calculation formula is as follows: ; in the formula, beta, gamma and lambda are weight coefficients, The energy costs required to implement policy pi, To measure the divergence of the difference between the two probability distributions, To be environmental state under policy pi The variance posterior prediction distribution of (c), For a priori preference distribution of the environmental states, Information gain for implementing policy pi; And S44, generating a candidate strategy set, calculating expected free energy of each candidate strategy through a lightweight strategy evaluation neural network, and selecting the strategy with the minimum expected free energy for execution.
4. The collaborative control method based on space-time holography and neuro-reasoning according to claim 1, wherein in step S2, the statistical summary of neighboring node states includes neighbor average energy and specific event occurrence frequency, network average field The low order statistics include the mean Sum of variances 。
5. The collaborative control method based on spatiotemporal holography and neuro-reasoning according to claim 2, wherein in step S3, the coupled partial differential equations of the average field game include HJB equation and FPK equation: HJB equation: ; In the middle of As a cost function at time t and state x, Is that Is used for the gradient of (a), Is in state of Is used as a dynamic equation of (a), As an average field distribution function of the network at time t, Is the dot product of the gradient vector and the kinetic equation; FPK equation: in the following As a function of the average field distribution, For an optimal control action the control device is provided with, In order to operate on the degree of divergence, For optimal control actions at time t and state x, Is the product term of the average field distribution and the state evolution.
6. The collaborative control method based on spatiotemporal holography and neuro-reasoning according to claim 3, wherein in step S4, the neural implementation of neuro-active reasoning comprises an encoder neural network Prior network Cost prediction network, information gain prediction network and policy evaluation network The encoder neural network is used for approximating the posterior probability of state estimation, and the prior network is used for estimating the posterior probability of state estimation according to STH latent variable And dynamically generating prior probabilities by the average field information, wherein the cost prediction network is used for predicting the energy consumption of a given strategy, the information gain prediction network is used for predicting the information gain of the given strategy, and the strategy evaluation network is used for outputting an approximation value of expected free energy.
7. The collaborative control method based on space-time holography and neuro-reasoning according to claim 1, wherein in step S1, each node in the distributed micro-nano sensing node group comprises an energy acquisition unit, a sensing unit, a communication unit and a calculation and storage unit, wherein the energy acquisition unit is a photocell and is used for receiving light energy from a PoF special optical fiber core and converting the light energy into electric energy, the sensing unit comprises a vibration sensor, a temperature sensor, a strain sensor and a local discharge sensor, the communication unit supports a Fi-Wi mode, data is received/sent through an optical fiber and short-distance wireless communication is carried out through a micro radio frequency antenna, and the calculation and storage unit comprises an ultra-low power consumption microcontroller unit and a memory and is used for running a neuro-initiative reasoning decision algorithm.
8. The collaborative control method based on space-time holography and neuro-reasoning according to claim 3, wherein in step S44, the candidate strategy set is generated by locally sampling a basic action output by an upper average field game module, wherein the basic action includes medium frequency vibration monitoring, dormancy and a low power consumption mode.
9. The collaborative control method based on space-time holography and neuro-reasoning according to claim 1, wherein the micro-nano sensing network system further comprises a collaborative control software module, wherein the collaborative control software module is deployed at each node and gateway and comprises a space-time hologram construction and update module, an average field information processing module and a neuro-initiative reasoning engine.
10. The distributed micro-nano sensing network cooperative control system based on space-time hologram and average field nerve active reasoning is characterized by comprising a central server/gateway, an optical fiber energy transmission/communication backbone network and a distributed micro-nano sensing node group, wherein the central server/gateway is deployed at an optical cable starting end or a data center and is used for initializing and maintaining global average field information, operating a lightweight average field neural network to rapidly solve an average field equation, broadcasting a solving result to the network and simultaneously receiving abstract information uploaded by nodes to update a global state part in the space-time hologram; The optical fiber energy transmission/communication backbone network adopts a weak coupling multi-core optical fiber, wherein at least one core is used for transmitting high-power energy light, and the other cores are used for transmitting data signals; The distributed micro-nano sensing node group is embedded or attached to the optical cable at a preset interval and is used for constructing and updating the space-time hologram, receiving average field information, executing MFG-AIF joint decision and executing an optimal strategy.

Description

Collaborative control method and system based on space-time holography and neural reasoning Technical Field The invention relates to the technical field of industrial Internet of things and optical fiber communication perception, in particular to a collaborative control method and a collaborative control system based on space-time holography and neuro reasoning. Background With the development of the energy internet, the global, real-time and high-reliability sensing requirements on infrastructure such as a power grid, a pipeline and the like are increasingly urgent. The optical cable embedded micro-nano sensing network utilizes the optical fiber as a composite medium for communication, sensing and energy supply, and forms the prior art architectures such as an optical fiber wireless heterogeneous communication architecture, a weak coupling multi-core optical fiber transmission medium, a miniaturized low-power node technology and the like. However, the existing scheme faces three core contradictions in the construction of an actual network management and control system, namely the contradiction between the high uncertainty of energy under the constraint of a physical law and a static management strategy, which leads to unreasonable node decision, poor network lifetime and efficiency, the contradiction between a large-scale heterogeneous coordinated 'dimension disaster' and a limited computational power of the node, the too fast increase of the computational complexity of the traditional algorithm along with the number of the nodes, incapability of adapting to the computational power of the node, and the contradiction between perceived ambiguity in a complex physical field and a 'perceived-communication-decision' splitting architecture, and the lack of the local data value judging capability of the node, so that bandwidth and energy waste are caused. Disclosure of Invention The invention aims to provide a collaborative control method and a collaborative control system based on space-time holography and neuro-reasoning, which are used for solving the problems of three core contradictions of limited energy, complex calculation and perception blurring in the prior art. The technical scheme for solving the technical problems is as follows: The cooperative control method based on space-time holography and neural reasoning is applied to an optical cable embedded optical fiber wireless heterogeneous communication (Fi-Wi) scene, and comprises the following steps: S1, constructing a distributed micro-nano sensing network system, wherein the micro-nano sensing network system comprises a central server/gateway, an optical fiber energy transmission/communication backbone network and a distributed micro-nano sensing node group, the optical fiber energy transmission/communication backbone network adopts a weak coupling multi-core optical fiber (WC-MCF), and the distributed micro-nano sensing node group is embedded in or attached to an optical cable; S2, constructing and updating a space-time hologram (STH) of each micro-nano sensing node, wherein the space-time hologram is a lightweight distributed world model of the joint probability distribution of key physical and information states related to coding network decision, and the core variables comprise the current residual energy of the node Fixed physical location of nodes along fiber optic cableOptical fiber energy transfer (PoF) physical model prediction-based t-moment received optical powerEnvironmental field vector composed of local environmental physical quantity perceived by nodeAnd an estimated value of the importance of the node to the local data to be transmittedWherein the received light powerThe calculation formula of (2) is as follows: ; wherein: the optical power injected at the time t; As a temperature dependent attenuation coefficient of the optical fiber, Is thatTime nodeLocation of the positionAmbient temperature of (2); Is a node Distance from the light injection end; Each node obtains the statistical abstract of the states of adjacent nodes through local wireless communication, and obtains low-dimensional latent variables through the lightweight variable self-encoder (VAE) coding by combining the states of the nodes And based on the latent variableInferring network average fieldsLow order statistics of (2); S3, carrying out group collaborative modeling and solving based on average field gaming (MFG): S31 defining individual status Individual control actionsIncluding perceptual mode/frequency selection, communication decisions, and computational load distribution; S32 defining an average field For all node statesEmpirical probability distribution over a state space, each node acquiring the average fieldLow order statistics of (2); S33 definition node From the slaveExpected cost function to terminal time T: ; wherein: R is a control effort weight matrix; e represents the expected operation; s34, approximately solving a coupled partial differential eq