328 results about "Predictability" patented technology

The invention discloses a multi-agent task arrangement method and system, and relates to the technical field of artificial intelligence. In the method, firstly, a task demand is obtained, key information and intention of the task demand are extracted, and task semantics are obtained; secondly, based on the task semantics and the business standard process template, obtaining a business standard process template matched with the task demand; then, according to the matched business standard process template and task semantics, a task context is constructed, and a domain specific language DSL is generated through reasoning according to the task context; thirdly, the DSL is analyzed, a task dependency graph is constructed according to the DSL analysis result, and the task dependency graph is converted into BPMN process model data; and finally, sub-task scheduling and execution of the sub-agents are carried out according to the BPMN process model data. According to the method provided by the invention, the dependence on professional developers or process engineers can be reduced, the response and planning time of complex tasks can be shortened, the result predictability can be improved, and the reasoning cost can be obviously reduced.

The invention discloses a task scheduling method based on predictable resource state atlas modeling, a platform is oriented to a heterogeneous computing environment, a unified resource state atlas is constructed by collecting multi-dimensional resource state parameters of computing nodes, and performance characteristics and communication topological relations among the nodes are comprehensively described. On the basis, a bidirectional time sequence model and an attention mechanism are fused, and the load trend of each node in a future short time is predicted. The platform constructs a multi-factor scheduling scoring function based on a task feature vector and resource state prediction map, integrates parameters such as resource matching degree, prediction load, communication delay and energy consumption cost, dynamically evaluates the adaptability of tasks and resources, and realizes adaptive scheduling and optimal resource allocation of the tasks. Compared with the prior art, the method has the advantages of being high in resource state predictability, high in task allocation intelligence degree, outstanding in platform evolution capability and the like, and is suitable for intelligent task scheduling application in a large-scale heterogeneous resource environment.

The invention discloses a continuous charging production line intelligent management and control strategy and framework, and the framework further comprises a multi-physics field coupling modeling module which carries out the physical mechanism modeling and state prediction of a production process through a heat-fluid-solid three-field coupling model on the basis of a conventional data collection and control module; the intelligent twin module adopts a mode of dynamically fusing a mechanism model and a data-driven residual model to improve the precision and robustness of system state estimation; the scheduling optimization module is used for realizing dynamic adaptive scheduling of AGV paths and production tasks based on a multi-target deep reinforcement learning strategy; and the multi-layer safety causal reasoning module realizes advanced prediction and active prevention and control of safety risks through a causal map, fuzzy logic and a linkage response mechanism. According to the method, the problem that safety, quality and efficiency are difficult to collaboratively optimize in the prior art is solved, and the predictability, safety, scheduling efficiency and adaptive capacity of the production line are remarkably improved.

The invention relates to the technical field of data processing and analysis, in particular to an intelligent large health system and a data processing method thereof, and the method comprises the steps: collecting the data of a multi-source heterogeneous data source in real time, and comprehensively capturing the multi-dimensional health data of a user; a multi-modal health knowledge graph is dynamically constructed and optimized through an efficient data preprocessing and feature alignment technology, and the integration value and the utilization efficiency of data are improved; a graph neural network and a time sequence analysis model are utilized to realize accurate evaluation of the health state of the user and timely prediction of future risks, and predictability and pertinence of health management are effectively improved; based on a reinforcement learning algorithm, a highly personalized intervention sequence can be generated according to the individual condition of a user, the accuracy of health intervention is enhanced, and the positive change of the health behavior of the user is promoted; an intervention instruction is executed through intelligent equipment, user feedback is monitored in real time, an intervention strategy is dynamically adjusted, and the flexibility and adaptability of intervention measures are ensured.

The invention relates to the technical field of mineral processing visual detection, and discloses a visual analysis system for detecting the grade of a phosphorite flotation froth layer. The system comprises an image acquisition and decomposition module, a parallel feature extraction module, a dynamic feature fusion module, a foam evolution analysis module and a grade decision output module. The system performs multi-scale decomposition on a foam image, extracts physical and semantic features in parallel, constructs a dynamic fusion network based on bidirectional mapping to perform iterative interaction, and generates a multi-modal feature descriptor. Therefore, self-organizing growth of a foam evolution graph is driven, an evolution track of a key foam primitive is positioned and tracked, a foam grade state vector is formed, and finally, a regulation and control decision is output in combination with external control parameters. According to the system, deep fusion of physical and semantic features and deep analysis of the foam dynamic evolution process are achieved, the accuracy and predictability of foam grade state sensing are improved, and an effective means is provided for accurate control over the flotation process.

Systems and methods iteratively train, using training data, a natural language processing (NLP) model to interpret conversational input during a conversation between an agent and a user by predicting key statements to be used in the prediction of a user intent, the training comparing outputs to a target variable during each iteration and adjusting parameters of the NLP model during each iteration to improve predictability of the user intent from the conversational input. Real-time conversational data is transmitted to the NLP model and the trained NLP algorithm derives key statements predicted to indicate intents and predicts one or more user intents based on the data from the conversation. One or more pre-filled forms predicted to effectuate the one or more user intents is generated, the pre-filled forms including generated text derived from information from the data of the conversation, and the form is transmitted to an agent device.

The invention discloses a three-dimensional dynamic safety management and control method and system for power transmission and transformation hoisting operation, and relates to the technical field of construction safety, and the method comprises the steps: obtaining point cloud data of a target operation area through ground three-dimensional laser scanning; according to the point cloud data, a three-dimensional grid model is constructed through an oblique photography technology and a Beidou real-time dynamic carrier phase difference technology, and a semantic electronic fence is generated; according to a pure vision technology, through target detection, three-dimensional reconstruction and time sequence tracking, a hanging object 4D digital branch is constructed, and a hanging object prediction track is obtained; the shortest three-dimensional distance is calculated through an efficient collision detection algorithm according to the hanging object prediction track and the semantic electronic fence, multi-dimensional risk assessment is carried out, and graded early warning signals are generated; and carrying out comprehensive alarm prompting according to the graded early warning signal. According to the invention, a traditional passive protection mode depending on single threshold alarm is changed, and quantitative evaluation and predictive management and control of dynamic risks are realized.

The invention discloses a cold-chain logistics temperature control supervision method and system based on GPS linkage, and relates to the related technical field of cold-chain logistics, and the method comprises the steps: installing a vehicle-mounted terminal on a cold-chain logistics transport vehicle, and collecting the vehicle position information and carriage temperature data in real time; encrypting and uploading to a monitoring processing center, and calling cold-chain logistics temperature control analysis dual channels of the target cold-chain goods; obtaining a temperature comparison result with a preset safe temperature threshold value, and matching and determining a target temperature control analysis channel; and performing temperature control analysis, determining a temperature control adjustment strategy parameter, and performing cold-chain logistics temperature control supervision. The technical problems that in the prior art, temperature control supervision lags behind, deep fusion analysis with the vehicle position and the transportation state is not carried out, specific reasons of temperature abnormity cannot be distinguished, and the regulation and control strategy is lack of accuracy and predictability are solved, and real-time, accurate and intelligent cold chain transportation temperature control abnormity recognition and supervision are achieved. And the reliability, the safety and the efficiency of cold-chain logistics are improved.

The invention relates to the technical field of geotechnical material local modeling, in particular to a geotechnical material data driving local modeling method and device based on a knowledge migration strategy, and the method comprises the steps: building a first mapping data pair based on the stress-strain physical quantity of a test curve sampling point, and building a high-fidelity data set of a geotechnical material; constructing a second mapping data pair based on the stress-strain state of the simulation curve sampling point and the intermediate physical quantity label, and establishing a low-fidelity data set of the geotechnical material; pre-training the target data-driven constitutive model by using the low-fidelity data set to generate an initial geotechnical material data-driven constitutive model; and driving the model to be fine-tuned by using the high-fidelity data set for the initial geotechnical material data, and generating final geotechnical material data to drive the model. Therefore, the problem of insufficient generalization ability and prediction precision of a data-driven model due to limited training data and unpredictability of a stress path in the prior art is solved.

The invention discloses a network topic hot spot extraction method based on bullet screen semantic recognition, and aims to solve the problems of bullet screen data semantic sparsity, semantic offset, noise interference and the like. The method is characterized by comprising the following steps: carrying out semantic coding on a bullet screen by utilizing a Transform structure; a dynamic semantic evolution perception model is constructed, semantic offset is measured through KL divergence, and topological correlation analysis is carried out through GCN; constructing a space-time density field in combination with a video time axis to realize space-time coupling feature fusion; automatically extracting a hot spot cluster by adopting an improved density peak clustering algorithm; and predicting a hotspot evolution trend by using an LSTM model. By means of the technical scheme, topic hotspots can be accurately captured, semantic evolution logic can be recognized, and the purity and predictability of hotspot extraction are improved.

The invention provides an intelligent road management and maintenance system and method based on multi-model fusion, and relates to the technical field of road maintenance and intelligent traffic. The method comprises the following steps: acquiring road condition data and positioning data through a multi-source data acquisition module; the disease recognition module is combined with a neural network model to analyze the multi-modal data, so that the disease detection and recognition precision is improved; the big data processing early warning module mines a disease rule through a big data model, and realizes disease early warning and predictive maintenance; the digital twin modeling module is used for constructing a road digital twin model and realizing real-time monitoring and simulation analysis; and the decision scheduling module realizes automatic maintenance scheduling based on the output of the modules. Through multi-model deep fusion, the problems of high manual dependence, low precision, lack of predictability and scheduling lag in traditional road management and maintenance are solved, the road maintenance efficiency, quality and intelligent level are remarkably improved, and the maintenance cost is reduced.

The invention relates to the technical field of tunnel engineering informatization and intelligent construction safety monitoring, in particular to a tunnel construction-oriented blasting data acquisition management platform system, which comprises a data standardization acquisition module for obtaining a standardized management input data set; the compliance conflict sensing module is used for generating a space-time mismatching degree index; the dynamic trust decision module is used for generating a security adjustment threshold window allowing the execution parameter to deviate from the reference value; the execution supervision module compares and verifies the to-be-executed construction parameters with the safety adjustment threshold window, judges compliance attributes of the to-be-executed construction parameters, and generates a corresponding release instruction or a blocking instruction; the process tracing module is used for recording the authorization basis and the environment background of each parameter adjustment to form a non-tampering audit log; according to the method, the problem that the geological uncertainty cannot be accurately represented by the traditional means is effectively solved, and the predictability and the accuracy of construction risk identification are remarkably enhanced.

The invention provides an AI-based distribution network unmanned aerial vehicle inspection defect identification and work order self-generation method and system. According to the method, inspection images, historical images, environment context and actual working condition constraint data are obtained, and an equipment state calculation rule base is constructed to describe a rule of defect characteristics evolving along with time and environment; carrying out state analysis and deduction in combination with current data to generate a defect evolution prediction set; risk assessment and work order decision making are carried out based on the prediction result and resource configuration, work order recommendation is formed, work order feasibility verification and optimization are carried out in combination with a time window and geographical constraints, and an execution work order set is output. According to the method, the problems that the existing inspection data analysis dimension is single, the defect cognition is lack of predictability, and the operation and maintenance decision and resource allocation depend on artificial experience and are disjointed with actual execution conditions are solved.

The invention relates to the technical field of thunder and lightning monitoring and protection, in particular to a thunderbolt positioning and energy analysis method and system based on a lightning rod monitoring network. The method comprises the following steps: collecting and verifying a lightning current waveform and a microsecond timestamp in real time, and measuring grounding resistance regularly; the terminal carries out filtering analog-to-digital conversion on the signal and extracts characteristic parameters, after packaging, the priority is set according to the current amplitude, and the priority is uploaded to the cloud platform; the cloud platform is associated with multi-terminal synchronous data, when the number of effective terminals is not less than three, coordinates of a thunderbolt point are calculated, integral operation is performed on a lightning current waveform in combination with actually measured grounding resistance to quantify single-time lightning stroke energy, and historical data are analyzed by using a long-short-term memory network model to realize equipment health diagnosis and life prediction. And displaying the thunderbolt point distribution thermodynamic diagram and the state evaluation report through a visual interface. According to the invention, the whole process management from data acquisition to intelligent decision making is realized, and the accuracy and predictability of lightning monitoring and protection are improved.

The invention discloses an intelligent state monitoring method and system based on petrochemical equipment, and relates to the technical field of intelligent state monitoring, and the method comprises the steps: constructing a petrochemical equipment digital twin based on the parameters and physical characteristics of target petrochemical equipment, and collecting multi-source time sequence data; inputting the multi-source time sequence data into the digital twin of the petrochemical equipment, and performing real-time simulation to obtain theoretical operation state data; and based on the multi-source time sequence data and the theoretical operation state data, obtaining deviation data between the measured value and the theoretical value, and updating the parameters of the digital twin of the petrochemical equipment by using the deviation data to obtain equivalent real operation state data. Through deep fusion of a mechanism model, data driving and an intelligent algorithm, accurate prediction of equipment life, effective distinguishing of fault types and automatic generation of a diagnosis report are realized, and the industrial problems of high false alarm rate, poor interpretability and lack of predictability in the prior art are fundamentally solved.

The invention discloses a detection method, equipment and medium for a new energy high-power direct-current charging pile, and relates to the technical field of charging pile stability detection, and the method comprises the steps: controlling a direct-current electronic load to generate a load disturbance signal at N characteristic working points, collecting instantaneous value change data through a broadband data collection center, and calculating the stability of the charging pile according to the instantaneous value change data; and based on the dynamic load disturbance signal and the instantaneous value change data, calculating the pulse response characteristics of the to-be-tested DC charging pile at the N characteristic working points through a correlation function analysis method, constructing a dynamic impedance characteristic model, and based on the safe working area three-dimensional boundary map, the electric energy quality indexes of the N characteristic working points and the dynamic impedance characteristic model, calculating the dynamic impedance characteristic of the to-be-tested DC charging pile. And evaluating and diagnosing the comprehensive performance of the direct current charging pile to be tested. According to the direct current charging pile comprehensive performance detection method based on dynamic load disturbance, accurate physical examination of the overall performance of the charging pile is achieved, and the detection depth, precision and predictability are improved.

The invention relates to the technical field of intelligent exhibition hall control, and discloses an intelligent exhibition hall control system based on the Internet of Things. The system synchronizes the operation state of a physical exhibition hall through a digital twin model, and performs operation situation slicing and feature extraction on a virtual mirror image to obtain a situation feature vector. And constructing an exhibition hall operation situation evolution graph based on historical data, mapping the real-time feature vectors to graph nodes, and performing multi-hop state deduction along the edges, thereby generating a future situation prediction path and identifying potential abnormal nodes. And finally, simulation regulation and control are performed in the digital twinborn model according to a deduction result, and a preventive regulation and control instruction is generated. According to the scheme, a closed loop from real-time state sensing to future evolution prediction and active regulation and control is realized, and the predictability and management and control perspectiveness of the operation risk of the complex exhibition hall are improved.

The invention provides an interface generation method, which comprises the following steps: acquiring a first interface generation intention, and determining first scene information corresponding to the first interface generation intention; generating an interface module corresponding to the first interface generation intention according to the first scene information, a preset component combination layout rule set and a preset front-end component library; generating a first interface according to the interface module and a preset module combination layout rule set; according to the embodiment of the invention, the front-end component library and the component combination layout rule set are preset, the generated first interface has the characteristics of high controllability and high certainty, the predictability of the interface is improved, and an interface interaction scheme with both flexibility and stability is provided. The disclosure also provides an electronic device, a computer readable medium, and a computer program product.

The invention discloses a remote rail fault intelligent detection system based on big data driving, and relates to the technical field of rail traffic intelligent monitoring and state maintenance. The problems that a traditional detection method is high in manual dependence, slow in response and lack of prediction and self-repairing capacity are solved. According to the method, multi-source data are acquired through a fixed sensing network and a mobile inspection vehicle, a multi-scale digital twin is constructed and updated, and a probabilistic fault evolution graph is generated based on a random phase field model; collaborative judgment is carried out based on distributed edge computing nodes, and a maintenance scheme is generated by adopting a hybrid geometric distance retrieval immune memory bank; and finally, collaborative repair is executed through mixed reality and a micro-robot cluster, and continuous optimization of the system is realized based on federated learning and case-based reasoning. According to the system, integrated closed loop of track state sensing, fault prediction, intelligent decision making and self-repairing is achieved, and the detection foreseeability, the maintenance accuracy and the system self-adaptive capacity are improved.

The invention relates to the field of landfill leachate treatment, and discloses a landfill leachate treatment equipment management method and system based on big data, and the method comprises the steps: collecting the operation parameters of landfill leachate treatment equipment and the water quality parameters in the treatment process in real time; extracting multi-dimensional features based on the operation parameters and the water quality parameters; inputting the multi-dimensional features into a trend prediction model; the trend prediction model outputs a short-term prediction result of the water quality parameters; on the basis of the short-term prediction result, the current operation parameters and the current operation type of the equipment, technological parameter adjustment suggestions are output through a control strategy model trained by a reinforcement learning algorithm, and a control strategy is generated; and the control strategy is sent to an equipment controller, and the equipment controller regulates and controls the corresponding processing equipment based on the control strategy. By means of the technical scheme, short-term prediction of the water quality parameters of the landfill leachate is achieved, and the predictability of the treatment process is improved.

The invention discloses a predictive cooperative control method for low-temperature heating of a battery of a pure electric vehicle. Comprising the steps of obtaining current vehicle state data and future disturbance data, obtaining the current vehicle state data and the future disturbance data, solving to obtain an optimal control sequence used for cooperatively controlling PTC power and the rotating speed of an air conditioning system compressor, and selecting a control strategy capable of bringing higher predicted endurance mileage. A final control instruction is generated, and the PTC heater and the air-conditioning system compressor execute the control instruction; through multi-source data fusion and a long-short-term memory neural network model, accurate prediction of available waste heat of a motor system, the energy efficiency ratio of an air conditioning system and the net heat load demand of a battery in a future time domain is realized, a smooth and reasonable control instruction is generated in advance, the precision and stability of battery temperature control are improved, and the battery temperature control efficiency is improved. Frequent start and stop and power impact of key actuators such as a PTC heater and an air conditioning system compressor are reduced, and the service life of the key actuators is prolonged.

The invention relates to the technical field of geological disaster early warning, in particular to a risk grading method for tunnel seismic wave data and geological information fusion, which comprises the following steps: S1, collecting tunnel multi-temporal seismic wave amplitude, frequency response and propagation delay information, identifying an abnormal response event, extracting a spatial propagation path of the abnormal response event in combination with a tunnel structure, generating a seismic wave evoked response event chain set; s2, geological information such as a drilling profile, a geological structure line and lithologic distribution is fused, and a structure chain type conduction diagram is output; and S3, counting the activeness, the derivation direction and the time sequence trend of the structural unit in the conduction diagram, calculating a trend activation factor, and generating a risk level map with predictability and evolution ability. According to the method, the earthquake response event chain and the structure conduction network are constructed, the multi-source geological information is fused, and the trend activation factor is introduced, so that the dynamic, accurate and predictable grading evaluation of the tunnel structure risk is realized.

The invention discloses an intelligent question-answering system of a water affair service system, which adopts a four-layer architecture of a sensing layer, a control layer, a computing layer and an application layer, realizes closed-loop intelligent management from data sensing to equipment execution, and ensures data quality from the source by the sensing layer through a self-checking water affair data acquisition device; and the calculation layer processes multi-source data by using a time sequence mechanism double-drive fusion algorithm, analyzes user questions by using an intelligent question and answer engine, and generates answers containing business decisions and equipment operation instructions in combination with a water affair knowledge base. The system has the prominent advantages that after a question and answer equipment linkage control unit of a deep linkage control layer receives an instruction, the real-time mechanical performance of field equipment is verified in advance, it is ensured that the instruction can be executed and then issued, the execution progress is monitored in the whole process, synchronous feedback is conducted, the data reliability and decision scientificity are fundamentally improved, and the system is suitable for popularization and application. Question and answer output is established on credible data, and decision predictability and accuracy are improved in emergency scheduling and other scenes.

The invention discloses a multi-dimensional data fusion analysis method and system for fine chemical production, and belongs to the technical field of industrial process control, and the method comprises the steps: obtaining multi-source production data, and generating a unified time sequence data set through timestamp alignment and redundant data filtering; extracting multi-scale features and fusing the multi-scale features and the process knowledge parameters into a joint feature vector; calculating a production state score and grade through weighted fuzzy reasoning; predicting a deviation trend of the key parameters based on the score and historical optimal production data; actual sampling data are obtained, deviation comparison is carried out on the actual sampling data and the deviation trend data, and optimization calculation is triggered to generate optimization control parameters; the reaction condition is adjusted through the distributed actuator, and the new data is used for updating the unified time sequence data set. According to the method, state evaluation is carried out by fusing process knowledge and multi-scale features, and deviation trend prediction and closed-loop optimization control are carried out in combination with the state transition probability, so that predictability, stability and adaptive optimization capability of the production process can be improved.

The invention provides an offshore sand wave coupling pile foundation scouring monitoring system based on a multi-beam echo sounding technology, which belongs to the technical field of offshore pile foundation structure scouring monitoring and comprises a data acquisition module, a data processing and analysis module, a sand wave recognition and feature extraction module, a pile foundation scouring evaluation module and a result output and early warning module. The system further comprises a calibration module and a solidified soil protection material database. According to the system, dynamic evaluation and early warning of scouring pits around piles are achieved through multi-source data collaborative acquisition (multi-beam sounding, layered flow velocity and pile foundation parameters) and precision calibration, coupling hydrodynamic force-scouring simulation (considering pile foundation turbulent flow) and sand wave migration feature extraction; and when the scouring exceeds the threshold value, a protection scheme can be output based on the stabilized soil database. The system solves the problems that traditional monitoring data are isolated, analysis is split and no protection suggestions exist, and monitoring comprehensiveness, predictability and engineering practical value are improved.

The application relates to the technical field of intelligent control of refrigeration equipment, and discloses an intelligent energy-saving control method and system for a refrigeration display cabinet. The method comprises the following steps: collecting temperature distribution and passenger flow data in the refrigeration display cabinet through a multi-parameter sensor network, inputting the data into an ARIMA model after sliding average filtering processing, and predicting a refrigeration load fluctuation trend; a dynamic energy efficiency ratio indicator is established based on the ratio of compressor power to the predicted value of the load, and the energy-saving effect is quantified; the compressor frequency and the fan rotating speed are adjusted in combination with a nonlinear prediction control algorithm, and a cooperative control instruction is generated; the influence of the instruction on the system load is analyzed through a refrigerant flow tendency function, the defrosting cycle is optimized synchronously, an energy-saving control strategy is formed, and the strategy is executed. The application improves the predictability and cooperativeness of energy efficiency management of the refrigeration display cabinet.

The invention discloses a breathing equipment data processing method which comprises the following steps: synchronously collecting multichannel physiological signals through a built-in sensor of breathing equipment, carrying out noise processing and data standardization, and segmenting according to a time window to form data fragments; extracting breathing amplitude, frequency and morphological variability features of each data fragment, and combining to generate a multi-dimensional dynamic feature vector sequence; constructing a breathing state modeling system based on a hidden Markov model, training model parameters through a Baum-Welch algorithm, and calculating a prospective risk transition score by using a forward algorithm; and establishing an adaptive threshold strategy based on the risk transition score, carrying out structured labeling on the original respiration data, and generating processing data containing a prospective risk early warning label. Technical support can be provided for intelligent upgrading and personalized treatment of respiratory treatment equipment, and the accuracy, predictability and clinical practicability of respiratory anomaly detection are improved.

The invention discloses a tire electromagnetic induction heating system control method based on multi-physics field coupling simulation. The method comprises the following steps: carrying out multi-physics field coupling simulation modeling; performing system optimization analysis; deploying a real-time control system; and performing performance verification and iterative optimization. According to the method, control over the complex physical process is achieved through multi-physics field coupling simulation, a closed-loop coupling system of physical entity-digital twinborn body-intelligent optimization-real-time control is constructed, data two-way circulation and state real-time synchronization are achieved, and the accuracy and timeliness of system regulation and control are improved; the global optimization of the system performance is realized by adopting a multi-objective optimization method, and the problem of local optimization is avoided; and through deep integration of a digital twinning technology and real-time control, predictability and controllability of the production process are realized, and the tire quality is remarkably improved.

The invention relates to the technical field of internal combustion engine testing and industrial automation control, in particular to a data acquisition and cooperative control method and system of a piston oscillation cooling test system. The system comprises a multi-source heterogeneous data acquisition and synchronization module, a cooperative control logic engine, an intelligent algorithm execution module, a safety interlocking management module and a man-machine interaction and data management module. According to the method, through a predictive pre-cooling strategy, the industrial problem of temperature overshoot of the inertial system is fundamentally solved, temperature control is achieved, and the control precision reaches + / -0.5 DEG C; besides, by constructing a parameter self-tuning PID cluster based on multi-source information, predictive suppression of coupling disturbance in the system is achieved, the hysteresis problem of head pain and foot pain of a multivariable control system is solved, interactive interference in the dynamic adjustment process is reduced by 60% or above, and the overall stability of the system is remarkably improved.

The invention relates to the field of reinforcement learning, in particular to a reinforcement learning-based large language model antagonism fine-tuning enhancement system. The state prediction module is used for acquiring system state data and robot action data; based on the robot action data, generating a prediction state vector through a neural network prediction model, and comparing the prediction state vector with the system state data to obtain a residual vector; the action decoding module inputs the residual vector into a feedforward neural network, codes the residual vector to generate a natural language state report, and performs autoregression decoding through a large language model to obtain a loss value; the strategy optimization module converts a penalty function value into a reinforcement learning reward signal through a nonlinear function, and the PPO algorithm calculates the total loss according to the reinforcement learning reward signal; and calculating, optimizing and updating the gradient of the low-rank adaptive parameter in the large language model based on the total loss. According to the invention, through the feedforward neural network and the PPO algorithm, the tiny dynamic deviation is captured, and the predictability and safety of the system are improved.