531 results about "State vector" patented technology

The invention discloses a charging pile cooling control method and system based on AI prediction reinforcement learning, and relates to the technical field of cooling control. The charging pile cooling control method and system based on AI prediction reinforcement learning comprises the following steps: S1, collecting charging heat dissipation data of a charging pile, and preprocessing the charging heat dissipation data; s2, constructing a time sequence characteristic matrix, inputting the time sequence characteristic matrix into a temperature prediction model, outputting a temperature prediction sequence of a controlled target, evaluating a thermal runaway risk in a prediction stage, and constructing a thermal risk identification sequence; s3, constructing a cooling strategy optimization model, inputting the real-time state vector into the cooling strategy optimization model, outputting an adjustment instruction, and issuing and executing the adjustment instruction; and S4, the execution deviation of the adjustment instruction is evaluated, the cooling strategy is adjusted based on the evaluation result, and a cooling strategy feedback sample is generated. The problems of energy consumption waste and cooling imbalance caused by lack of real-time prediction and self-adaptive regulation and control capabilities in the cooling control process of the existing charging pile are solved.

The invention relates to the technical field of industrial automation and intelligent control, in particular to a general control system strategy optimization method based on reinforcement learning, and the method comprises the steps: firstly collecting the operation data of a system, constructing a state vector of reinforcement learning, and enabling a reinforcement learning agent to fully understand the current operation condition of the system; then, the state vector is input into a reinforcement learning strategy network, an action for adjusting a control strategy is generated by the network, and the action can be used for modifying the proportion, integral or differential coefficient of PID and can also be used for adjusting the prediction step length, weight coefficient or constraint strength of model prediction control, so that the adaptive capacity of a controller to external changes is enhanced; then, a reward signal is constructed according to a response result of the reference controller; the reward function comprehensively considers the error size, the steady-state characteristic, the system energy consumption, the control smoothness and the stability requirement, so that the reinforcement learning not only pays attention to the error minimization when optimizing the strategy, but also considers the low energy consumption, the smooth action and the anti-interference performance at the same time.

The invention relates to a multi-driving-style high-risk automatic driving cut-in scene test method and system, and the method comprises the steps: collecting original driving track data, carrying out the clustering of driving styles, and generating a risk cut-in scene track cluster with consistent driving styles through employing a Cutin-TimeGAN network model in combination with physical feasibility constraints; constructing a cost function, calculating a cost function value for each risk cut-in scene trajectory cluster, and selecting a target cut-in reference trajectory as a target reference state vector; constructing a game framework, and combining a target reference state vector to construct a utility function according to the state vectors of the VUT and the test confrontation vehicle in the framework; constructing a risk confrontation utility function based on the predicted collision time; a game optimization problem is constructed and solved, an optimal interaction track of the test confrontation vehicle is obtained, and a closed-loop high-risk automatic driving scene switching test is carried out based on the optimal interaction track; the system is used for implementing the method. Compared with the prior art, the method has the advantages that testing of authenticity and high-risk confrontation of multiple driving styles is considered.

The invention belongs to the technical field of satellite orbit prediction, and provides a low-orbit satellite orbit state prediction method based on artificial intelligence, which comprises the following steps of: dividing step lengths of a predicted orbit according to a time window, calculating a resultant acceleration of each step length, performing stress input on an initial position and speed, and obtaining a nominal orbit state sequence; forming an augmented state vector with aerodynamic synthesis parameters; inputting a time sequence neural network model, and performing rolling updating on the augmented state according to the output; obtaining an original standard deviation through mapping by using uncertainty measurement, calculating a standardized residual error, solving a right side weighted quantile according to a normalized time weight, comparing the right side weighted quantile with a preset threshold value, and carrying out adaptive deviation adjustment; and calculating risk measurement by using the acquired data, comparing the risk measurement with a set threshold value and a priority rule, automatically executing a judgment action on the risk measurement, recording a trigger reason and the judgment action, and updating parameters in the time sequence neural network model.

The invention discloses a power grid operation fault prediction and abnormal trend early warning method, equipment and a medium, and aims to construct a tetrad fault sequence based on power grid multi-source data and realize second-level fault evolution dynamic tracking. By updating the power grid topology base map in real time, the problem of model lag is solved, and the propagation path precision is improved. The electrical coupling strength is introduced as the edge weight of the graph convolutional network, the electrical topography of the power grid is duplicated in the vector space, the calculation is simplified, and the accuracy is improved. And similarity diffusion is carried out by using the topological embedded vector, so that a high-precision influence range sub-graph can be generated in milliseconds, and prediction distortion is avoided. And converting the subgraph into a local monitoring area, quickly constructing a low-dimensional fault state vector, integrating the low-dimensional fault state vector into a dynamic fault propagation map, and clearly displaying fault traceability, path and termination logic. And finally, the node risk probability and the multi-dimensional influence index are output through one key by means of the graph attention network, an early warning instruction is automatically generated, and the second-level control and protection capability of the power grid is remarkably enhanced.

The invention relates to the technical field of thermal field modeling and control, and particularly provides a molecular beam epitaxy substrate temperature field control method and system based on artificial intelligence, and the method comprises the following steps: obtaining the all-region temperature data of the surface of a substrate, and the heating power and environment parameters of each region of a heating cavity; inputting the heating power and the environmental parameters into the temperature field model, and outputting a temperature distribution matrix and temperature gradient data of the surface of the substrate; predicting temperature prediction data according to the temperature distribution matrix, the heating power and the sequence data of the environmental parameters in the historical time period; and inputting a state vector formed by the temperature distribution matrix, the temperature gradient data, the heating power data and the temperature prediction data into a deep reinforcement learning model, and then outputting the heating power adjusting quantity of each area to control a temperature field. According to the invention, the problems of insufficient prediction precision and real-time performance of the substrate temperature in the prior art are solved, the influence of environmental factors and heating power on the substrate temperature is considered, and the temperature field control capability is improved.

The invention discloses a full-closed-loop automatic spraying method, system and device based on geometrical characteristic driving and multi-physical quantity feedback and a storage medium. The method comprises the following steps: performing three-dimensional scanning on a workpiece and extracting geometric features; generating a spraying track and a spraying posture for keeping a topological structure based on the geometrical characteristics; motion, process and coating quality parameters are collected in real time through a sensor, and are processed and fused to form a state vector; inputting the state vector into an intelligent control module, performing optimization by combining model predictive control, fuzzy logic and a neural network, and outputting trajectory and process parameters; a control strategy is updated according to the quality deviation of online detection, and closed-loop correction is carried out; spraying is automatically executed in the closed spraying cabin, and unmanned operation of the whole process is achieved. According to the method, complex curved surface spraying can be completed without manual programming, and the method has the advantages of high flexibility, multi-parameter closed loop, self-adaptive optimization and the like.

The invention relates to the technical field of rotary steerable drilling system dynamics analysis and control, in particular to a rotary steerable drilling system dynamics analysis system which comprises a data acquisition module used for acquiring ground engineering parameters and underground dynamic parameters in real time and intercepting drilling parameter adjusting instructions; the digital twin modeling module is used for outputting a complete state vector of a digital twin model; the noise prediction module is used for generating a control source noise signal; the signal reconstruction module is used for executing self-adaptive hedging processing on the original mixed signal so as to reconstruct a pure geological signal; the cooperative control module is used for predicting the vibration risk caused by the change of the front stratum and generating an optimal control instruction for actively avoiding the vibration risk; the occurrence probability of malignant vibration is remarkably reduced, the drilling efficiency is improved, and the service life of a drilling tool is prolonged.

The invention relates to the technical field of data processing and mode recognition, in particular to a bearing fault diagnosis method based on multi-scale feature fusion, which comprises the following steps: fusing multi-source data such as vibration, acoustic emission and rotating speed, performing angle domain resampling by using rotating speed data, generating a two-dimensional order spectrogram, and stacking to construct a three-dimensional working condition information tensor; a master-slave modulation heterogeneous neural network is adopted, high-dimensional spatial-temporal features are extracted through a main branch three-dimensional convolutional network, time sequence details are extracted from an original sequence through an auxiliary branch one-dimensional convolutional network, affine transformation parameters are generated, and dynamic modulation is achieved on the high-dimensional features; and the output state vector is mapped to a fault evolution knowledge graph, probability prediction is carried out through a graph attention network and by introducing a Monte Carlo discarding mechanism, a probability mean value is calculated as a fault classification result, and the diagnosis confidence is quantified by a probability variance. According to the invention, through multi-scale feature fusion and dynamic modulation, the problem of insufficient feature discrimination caused by scale mismatch under variable working conditions is solved.

The invention discloses a generator state estimation method and system considering noise and parameter uncertainty constraints. The method comprises the following steps: acquiring model parameters and dynamic state vectors of a generator, and establishing augmented state vectors; performing unscented transformation on the augmented state vector to obtain a particle set; improving to obtain robust mixed Kalman particle filtering, and in the process of performing unscented Kalman filtering on an augmented state vector, taking correlation entropy maximization of a measurement information sequence as a target function, and solving by adopting a fixed point iteration method to obtain a filtering gain; determining a filtering gain according to the updated state of the measurement information; robust mixed Kalman particle filtering is executed, physical constraints of model parameters serve as a feasible region, after resampling, projections of the model parameters in new-generation particles exceed the feasible region, the model parameters are set to be closest boundary points, and then resampling is conducted again; a weighted average value of the particle set is an optimal joint estimation value, a dynamic state estimation value and a model parameter identification result are separated, and reliable uncertainty quantization is provided for state and parameter estimation.

The invention discloses a new engineering course teaching evaluation method based on a knowledge-ability-quality triple atlas, and belongs to the technical field of intelligent education and intelligent control. Collecting data of three dimensions of knowledge, ability and quality of students to construct a unified state vector, and describing dynamic evolution of a learning process by combining a state updating and prediction model of a residual network; a multi-cell filtering method is adopted to carry out state set estimation, linear propagation and residual local linearization are combined in the prediction step, and a Lipschitz upper bound is utilized to carry out external connection on a nonlinear residual to ensure the safety and credibility of state interval estimation; in the updating step, set tightening is achieved through prediction strip intersection and generator contraction, coverage rate calibration based on quantiles is introduced, and the confidence level of set estimation is ensured. According to the method, the learning state of the student can be dynamically, stably and interpretably estimated, accurate and efficient course adjustment optimization is realized, and the method has a good application prospect and popularization value.

The invention provides a vehicle working condition virtual debugging simulation method based on multi-modal perception fusion, and relates to the technical field of digital twinning, and the method comprises the steps: achieving the precise fusion of multi-modal perception data through the synchronous collection of visual, radar and acoustic signals and the unification of a time reference and a coordinate system; constructing a vehicle dynamics state model based on mutual information weighting and recursive filtering, and obtaining dynamically consistent state vectors; combining space-time modeling and robust coding to generate driving characteristics capable of being physically explained, and inputting the driving characteristics into a simulation environment; in the digital twinborn model, parameter adaptive tracking is realized through online correction and joint recursive updating; and closed-loop feedback debugging is formed by using the health index and the fault probability, so that virtual and real consistent and dynamically adjustable vehicle working condition simulation and intelligent debugging are realized. According to the method, fusion of multi-mode sensing data can be realized, so that the simulation model dynamically tracks the working condition of the vehicle.

The invention relates to the technical field of industrial vehicle safety control, in particular to a forklift safety online industrial monitoring management system which comprises a data fusion unit used for collecting kinematics data, position and attitude information and load distribution data of a vehicle and conducting space-time alignment processing on the collected data to generate a real-time state vector; the margin calculation unit is used for calculating a plurality of independent safety margin components and carrying out fusion processing based on the safety margin components to generate a comprehensive safety margin; the risk prediction unit is used for predicting a future safety margin value based on the time sequence of the comprehensive safety margin, and comparing and analyzing the future safety margin value with a preset safety threshold and a preset warning threshold to generate a risk level; the intervention control unit is used for generating and executing a corresponding hierarchical intervention strategy in response to the risk level; according to the invention, a complete technical link from multi-source data sensing to closed-loop control is constructed, and the conversion from passive safety to active safety is realized.

The invention provides a satellite weak and small target on-orbit observation system and method based on intelligent closed-loop feedback, and the system comprises a detection module which carries out the image enhancement and time sequence consistency enhancement of an infrared image, obtains an enhancement feature, and obtains a candidate target set based on the enhancement feature; the tracking module is used for carrying out target matching in the search area; when the matching succeeds, the candidate position is used as an observation value, and the observation value is input into a Kalman filter to obtain a target state vector; when the matching fails, taking a prediction state of the Kalman filter as a target state vector, expanding a search window by taking a prediction position as a center, and executing re-identification; the control module is used for mapping the target state vector into an attitude error and generating a control instruction by adopting a single-neuron self-adaptive PID (Proportion Integration Differentiation) controller; and the closed-loop scheduling module is used for dynamically adjusting operation parameters of at least one module according to the execution error and the detection confidence coefficient. According to the invention, the continuous tracking precision and attitude control stability of the weak and small target are improved.

The invention provides an autonomous navigation and intelligent obstacle avoidance method of a biomimetic robot, and relates to the technical field of biomimetic robots, which realizes real-time estimation of six-degree-of-freedom poses by collecting environmental data; visual features are extracted by using a convolutional network, an enhanced memory unit is constructed in combination with pose information, a triple index path memory library is maintained through a dynamic memory updating mechanism, and feature similarity retrieval and navigation instruction generation in a backtracking mode are supported; defining a state vector based on a resultant force of a robot pose, a target direction and an artificial potential field, training a strategy network through deep reinforcement learning, and outputting a motion instruction according to Q value maximization; fusing an artificial potential field and a dynamic window method, and optimizing an objective function to obtain optimal linear velocity and angular velocity; and solving a joint angle through inverse kinematics, and combining attitude error feedback to realize body attitude maintenance and foot end trajectory tracking.

The invention discloses an underground water pollution diffusion model construction method based on multi-modal data fusion, and the method comprises the steps: fusing the dynamic time sequence characteristics of a monitoring well and the static space attributes of a geological field, and generating an initial state vector with rich information for a graph node; based on the instantaneous water level difference between the nodes and the equivalent permeability coefficient, a dynamic graph topological structure which evolves along with time and represents the hydraulic connection relation is constructed; deducing a future state through numerical integration by using a graph neural network model with a node state derivative as output; and constructing a composite loss function containing a data fitting item and a physical law residual item, and training the model through back propagation. According to the method, the dynamic graph of physical driving is constructed and the physical equation constraint is introduced, so that the model prediction has the data driving precision and the reliability of the physical mechanism, and the generalization ability and the prediction precision of the model under the complex hydrogeological condition are improved.

The invention relates to the technical field of virtual reality experiment teaching, and discloses an experiment teaching simulation method and system based on virtual reality. The method comprises the steps that a continuous behavior sequence generated by user operation is collected, and a teaching behavior chain is generated through discretization; constructing a state vector by identifying the space-time state of the behavior unit, and comparing the state vector with a standard template to mark an abnormal behavior unit; tracing the context to generate an abnormal behavior description block, and querying an error knowledge base to obtain error information; meanwhile, analyzing a state change event of the virtual experiment equipment; and finally, fusing the error information and the state event to generate a simulation teaching evaluation result. According to the method, refined multi-dimensional analysis of experimental operation behaviors and intelligent error influence quantitative evaluation are realized, and the accuracy and guidance of teaching feedback are improved.

The invention relates to a large model fine tuning method, device and equipment of a power system, a storage medium and a program product, and relates to the field of power systems. Comprising the following steps: acquiring multi-modal time sequence data of a power system, inputting the multi-modal time sequence data into a pre-trained multi-modal large model, and acquiring a prediction sequence of each signal output by the multi-modal large model in a time dimension; scoring the importance of each signal at each time point based on the volatility index of the prediction sequence; screening the signal time sequence correlation points with the importance scores exceeding a preset threshold value; for each signal time sequence correlation point, calculating a deviation between a predicted value and a historical steady-state mean value, fusing deviation correlation characteristics of different signals at the same time point, and constructing a state vector; inputting the state vector into a reinforcement learning strategy network to generate a fine tuning action; and performing parameter updating on the corresponding model substructure in the multi-modal large model according to the fine tuning action. According to the method, the fine tuning efficiency can be improved, and the response efficiency and reliability of the multi-modal large model in a complex power scene are improved.

The invention discloses a thermal power plant DCS fault prediction and diagnosis method and system, and relates to the technical field of fault prediction and diagnosis, and the method comprises the steps: obtaining equipment information and operation parameter reference values, and generating a to-be-detected equipment list; target equipment is determined based on the importance degree, a communication channel is established to collect real-time data, a parameter fluctuation characteristic matrix is constructed, a reference value is dynamically calibrated, and a calibration parameter set is generated; historical fault feature data are collected, and a fault feature library is established; generating a state vector matrix based on the calibration parameter set and the real-time data, and calculating an early warning index in combination with the fault feature library; calculating a risk coefficient during monitoring, dynamically optimizing an early warning threshold value, and completing early warning configuration; on the basis, diagnosis is executed, a standardized diagnosis process is generated, and a diagnosis knowledge base is formed; and when an exception occurs, triggering hierarchical rollback and updating the knowledge base to realize continuous optimization. The system comprises a master control module, a configuration module, a prediction diagnosis module, a transmission module and a display module.

The invention relates to the technical field of industrial equipment state monitoring and fault diagnosis, and discloses a dynamic generation method for a multi-stage early warning threshold of a dynamic equipment fault, which comprises the following steps of: dividing a multi-dimensional process parameter space into a plurality of discrete working condition subspaces by utilizing a clustering algorithm, constructing a phase space manifold main hyperplane and a standard tangential velocity field atlas, and constructing a multi-stage early warning threshold of the dynamic equipment fault. Calculating an orthogonal projection residual error and a tangential velocity field consistency deviation of the real-time phase space state vector, synthesizing a space-time coupling deviation index based on the orthogonal projection residual error and the tangential velocity field consistency deviation, and quantifying the ergodic damage degree of each state, and dynamically generating a multi-stage early warning system based on the ergodic damage degree of each state and the space-time coupling deviation index. According to the method, the manifold geometry and dynamics evolution double constraint model is constructed, and the probability distribution evolution analysis is combined, so that multi-dimensional and graded early warning of the dynamic equipment fault under the variable working condition is realized.

The invention relates to the technical field of cold-chain logistics transportation and storage temperature control, in particular to an air compressor vehicle-mounted refrigeration and refrigeration house collaborative temperature control system based on AI scheduling, which comprises a multi-dimensional state sensing module used for collecting real-time temperature field data, environmental meteorological parameters, compressor operation states and goods category information in a refrigerator car and a refrigeration house, and sending the real-time temperature field data to a server; a structured multi-dimensional state vector is output; and the thermodynamic evolution prediction module is used for setting a prediction time domain with a finite length based on the multi-dimensional state vector, and calling and constructing a mathematical model of a heat balance relationship between the refrigerating capacity generated by the operation frequency of the compressor and the environment and the cargo heat load. According to the method, the comprehensive cost function fusing the energy consumption, the cargo quality loss and the equipment wear is constructed, and the hidden quality attenuation is converted into the dominant economic cost by using the Arrhenius model, so that the global operation benefit maximization of the system on the premise of ensuring the cargo safety is realized.

The invention discloses a self-adaptive integrated navigation method based on geometric accuracy factors, which belongs to the technical field of underwater navigation and positioning, is used for underwater navigation and positioning, and comprises the following steps: initializing an inertial navigation system and an integrated navigation filter, and setting a state vector and a noise covariance matrix; predicted navigation parameters of the recursive carrier are mechanically arranged through inertial navigation, and the state transition matrix is used for state prediction; and dynamically adjusting an observation noise covariance matrix according to the geometric precision factor value, further calculating a Kalman gain, performing optimal estimation and correction on a prediction state by fusing acoustic observation information, and finally outputting a high-precision carrier position, speed and attitude. According to the method, the statistical characteristics of observation noise are dynamically remodeled by calculating and feeding back geometric precision factor values in real time, so that the Kalman filter has the capabilities of knowing the own geometric situation and adjusting the trust degree of each information source, and the global optimal navigation precision and reliability are realized under any motion track.

The invention relates to the technical field of laser control, and discloses a multi-axis laser flight compensation control method and device, and the method comprises the steps: collecting the position, speed and acceleration data of multi-axis motion, carrying out the fusion and denoising, and obtaining a smooth multi-axis state vector; constructing an inter-axis coupling matrix to determine the dynamic interaction strength, and extracting a dominant axis pair and a preliminary deviation trend vector when the dynamic interaction strength exceeds a threshold value; simulating trajectory evolution to obtain prediction deviation distribution, and constructing a compensation control law to determine a multi-axis cooperative correction increment; and when the track stability does not reach the standard, iteratively updating the matrix to obtain a refined model, generating a real-time adjustment instruction and feeding back the instruction to an actuator, and updating a state vector after an effect is verified. According to the method, the multi-axis laser flight path can be accurately compensated, the control precision and the dynamic adaptability are improved, and technical support is provided for precision machining.

The invention relates to the technical field of computer risk control, in particular to a big-data-based collaborative self-optimization method and system for a self-growth risk control system, and the method comprises the steps: obtaining the operation data of the risk control system, so as to construct a state vector; inputting the state vector into a reinforcement learning agent, and selecting one coordination action from a preset coordination action space for output; the coordination action is executed in the risk control system, and a scalar reward value used for evaluating the effect of the coordination action is obtained through calculation; and forming an empirical data tuple by using the state vector, the coordination action, the scalar reward value and a new state of the system after the action is executed, and training the reinforcement learning agent to update a decision strategy of the reinforcement learning agent. According to the risk control system, the rule engine and the model engine in the risk control system can be considered as a unified whole, and respective parameters are dynamically and cooperatively adjusted, so that the overall comprehensive efficiency of the system is maximized, and the self-adaptive capability and the iteration efficiency of the system are improved.

The invention discloses a geometric error identification and compensation system and method for a numerical control machine tool, and belongs to the technical field of numerical control machine tools, and the method comprises the steps: collecting the position, temperature and cutting force data of the machine tool in real time, and calculating the components of a position error, a thermally induced equivalent error and a force induced equivalent error to construct a multi-dimensional state vector; according to the method, a compensation value is calculated through a dynamic compensation model, the core of the method is that a dynamic modulation factor is introduced, the factor integrates the heat effect, the force effect and the multi-axis coupling effect to dynamically adjust the compensation weight, in addition, a grading trigger compensation strategy based on comprehensive error amplitude evaluation is designed, and the compensation accuracy is improved. The reconstruction compensation can update model parameters on line to cope with long-term changes of the system, intelligent distribution of compensation resources and self-adaptive adjustment of system changes are achieved, time-varying errors are effectively restrained, and precision, efficiency and robustness are taken into consideration.

The invention discloses an intelligent control method and system for dynamic balance of three floating bodies for crude oil transfer in deep and far sea. The method comprises the following steps: (1) data acquisition and state vector construction; (2) constructing a dynamic response prediction model: constructing a floating body historical motion database based on the initial motion data set, fusing hydrodynamic interference coefficient matrix calculation and multi-body motion mixed constraint analysis, and designing a space-time prediction algorithm based on a graph neural network and a recurrent nerve process to obtain a dynamic response prediction model; establishing a dynamic response prediction model under close-range coupling interference of the three floating bodies; (3) evaluating and correcting data confidence; (4) self-adaptive thrust control and early warning are carried out; and (5) visual monitoring and log management. The problems that in the prior art, positioning precision is low, response lags behind, the abnormal rate of sensor data is high, hydrodynamic coupling interference is not fully considered, and visual monitoring and intelligent early warning are lacked are solved, and the safety, stability and operation efficiency of the deep and far sea crude oil transfer process are effectively improved.

The invention relates to the technical field of ground source heat pump control, in particular to an intelligent cooperative control system for a ground source heat pump system and building energy management. The method comprises the steps that a data collecting and preprocessing unit collects multiple types of data in real time and preprocesses the data to form system state vectors; the load and power generation prediction unit predicts future cooling and heating load and photovoltaic power generation power of the building through a machine learning regression model and a time sequence prediction model; the operation optimization unit generates an optimal operation strategy of the ground source heat pump based on a depth deterministic strategy gradient algorithm; and the multi-source energy coordinated control unit realizes multi-device coordinated control through multi-target mixed integer linear programming. Through combination of prediction data and an optimization algorithm, a dynamic matching relation between load fluctuation and energy supply is fully considered, and the problems of soil heat imbalance and unstable comprehensive energy efficiency ratio of the ground source heat pump system in a traditional fixed parameter operation mode are solved.

The invention relates to the technical field of refrigeration house energy-saving control, and provides a refrigeration house energy-saving control method based on deep reinforcement learning, which comprises the following steps: dividing a refrigeration house according to a multi-scale grid, collecting data in real time, weighting and aggregating according to spatial characteristics, calculating spatial semantic codes and confidence indexes of each grid, and obtaining a standardized and traceable state vector; real-time energy consumption, grid temperature space gradient, long-term exposure economic loss, event condition triggering and weighted summation form a reward function, and a deep strategy network is adopted to train parameters in an accumulated reward mode; constructing a grey box and residual dynamic prediction model, shrinking the original temperature and humidity constraint, solving the prediction model through scenario under the shrinkage constraint, and carrying out weighted fusion on the depth strategy obtained by training and the solution of the prediction model according to confidence; and calculating a statistical value in the sliding window, constructing confidence distribution of strategy return, triggering a hierarchical response according to a detection result, and inputting all triggers into an audit log for threshold self-calibration.

The invention discloses a dust explosion dynamic blocking method based on multi-modal data fusion. The method comprises the following steps: S1, constructing five core accident trees; s2, collecting multi-source data of dust explosion in real time and preprocessing the multi-source data, and extracting a multi-dimensional feature vector containing temporal and spatial variation features; s3, converting the multi-dimensional feature vector into a real-time state vector of five core accident trees through an event activation rule, and completing multi-source data fusion; s4, according to the real-time state vector, calculating a top event probability of each accident tree by using a Monte Carlo simulation method, and then obtaining an accident tree dynamic coupling risk value through multi-tree coupling analysis; s5, according to the dynamic coupling risk value, using a spatial interpolation method and voxelization processing to obtain a normalized risk value, and generating a dynamic risk thermodynamic diagram; and S6, according to the risk value of each area in the dynamic risk thermodynamic diagram, determining the risk level of dust explosion, then carrying out multi-level interlocking control, and triggering a corresponding blocking strategy to realize dynamic blocking.