63 results about "Constraint satisfaction" patented technology

The invention discloses an intelligent planning method and system for a weak current system in a smart park, and belongs to the technical field of weak current intelligent design. The method comprises the steps of performing feature extraction on the weak current multi-source data of the smart park to form a weak current feature set; a multi-dimensional semantic space is constructed, semantic association features are obtained, and node features, topological relations and constraint rules of the weak current system are determined; generating a weak current knowledge graph based on the information, and performing semantic alignment on the basic information of the park to obtain a final scene demand representation; performing graph reasoning and constraint calculation according to the representation to obtain a feasible region and constraint satisfaction condition, and generating a candidate construction scheme; and screening out an optimal construction scheme from the candidate schemes according to a preset comprehensive optimization strategy and sending the optimal construction scheme to a control center. According to the scheme, the weak current scheme is promoted from demand understanding to scheme optimization, and a coherent and verifiable automatic process is formed; therefore, the manual intervention is less, the design judgment is more accurate, and the finally output construction scheme has higher engineering reliability.

The invention provides a rounded-corner container transportation path dynamic optimization scheduling method and system, and the method comprises the steps: quantifying three constraints of the size, the gravity center and the loading and unloading priority of a rounded-corner container, converting the three constraints into a matching formula, a path constraint threshold value and a weight rule, and constructing a weighted multi-objective optimization function in combination with the transportation cost, the time and the cargo damage risk; a genetic algorithm and ant colony algorithm mixed framework is built, a constraint adaptation layer is embedded to filter invalid solutions, and the iteration efficiency is improved; two types of algorithm operators are improved, and a constraint satisfaction degree, a loading and unloading priority and a dynamic parameter adjustment mechanism are fused; dividing multiple regions into sub-region optimization by adopting a divide-and-conquer strategy, and adapting to a large-scale dynamic scene through cross-region collaboration and local re-optimization; a full-dimension verification scheduling scheme in iteration is carried out, algorithm parameters are automatically adjusted based on constraint violation information, and iteration is terminated or constraint relaxation is started according to preset conditions; an improved algorithm is integrated to a dynamic scheduling system, real-time data are connected, parameters are optimized through a self-learning module, and a manual intervention interface is reserved.

A neural-symbolic hybrid system for generating binary document formats directly from natural language input comprises a binary-aware hierarchical tokenizer operating across four levels (binary bytes, structural elements, semantic content, and concepts), a constraint satisfaction engine with 64 parallel processing cores for enforcing structural integrity and mathematical consistency, format-specific processors for Excel, PowerPoint, PDF, and CAD documents, and a formal verification system generating mathematical proofs of correctness. The system includes custom AI Document Generation Processor (AIDGP) silicon spanning 600 mm2 with specialized cores providing 500 TOPS processing power. Performance characteristics include 99.7% structural accuracy, 100% format compliance, 15.3 second average generation time for complex documents, and distributed capacity of 1,000,000 documents per hour. The system eliminates intermediate conversion steps while maintaining semantic preservation through hardware-accelerated constraint satisfaction and formal verification engines ensuring structural integrity, format compliance, and security through AES-256 encryption and automated regulatory compliance across 25+ international standards.

The invention discloses a competitive evolution multi-task optimization method, system and equipment, and aims to solve the problem that an existing method is difficult to carry out collaborative optimization on fuel cost and gas emission under the condition that power balance and unit capacity constraint conditions are met. The method comprises the following steps: constructing main and auxiliary double-population parallel search; state indexes representing population convergence, diversity and feasibility are calculated in real time; based on the current state, a cooperation mode and evolution operator combined action is adaptively selected through a deep reinforcement learning agent; rewards are calculated according to improvement of the population on cost, emission and constraint satisfaction after action execution; and training the intelligent agent by using the state, the action and the reward, and iteratively optimizing the decision strategy until the Pareto optimal scheduling scheme meeting the constraint is output. According to the method, adaptive intelligent guidance of the evolutionary process is realized, and the optimization quality, the convergence speed and the algorithm robustness of the economic emission scheduling scheme of the power system are remarkably improved.

The invention relates to the technical field of business process modeling, in particular to a meta-model-driven interactive business process intelligent modeling method and system. The method comprises the following steps: constructing a meta-model framework containing core elements such as activities, events and gateways, and establishing a meta-model constraint rule base to define a structure legality rule of a BPMN2.0 specification; based on a domain ontology Schema and the domain exclusive large language model after instruction fine tuning, extracting a business entity and a process relationship thereof from the unstructured text; carrying out information integrity verification and logic consistency detection on the extracted BPMN elements by utilizing a constraint satisfaction algorithm and Petri net reachability analysis; displaying a difference label of the process sketch through an interactive verification interface, and realizing semantic correction and model iteration of a user on the process elements based on a natural language feedback mechanism; an XML code is generated by adopting a template-driven code generation strategy, and a flow chart is generated through a BPMN parser. According to the method, the business process modeling efficiency can be remarkably improved, and the manual intervention proportion in the modeling process is reduced.

The application discloses a kind of multi-agent driven network sentiment regulation method and system based on reinforcement learning, belong to artificial intelligence and network information processing technical field.The method obtains text, speech, image, video, behavior, relationship and propagation path data, constructs space-time heterogeneous interaction graph and extracts graph structure feature;Combined with multi-modal basic feature, establish causal attention sentiment state representation;Multi-agent collaborative decision-making environment, sentiment regulation action set, constraint reward function and counterfactual propagation environment are constructed;Collaborative communication training is carried out using multi-agent constrained reinforcement learning, and the output joint regulation strategy;According to sentiment feedback, behavior feedback and propagation feedback, closed-loop update is carried out, and according to negative sentiment proportion, sentiment entropy, propagation risk index and constraint satisfaction rate, iterative control is carried out.The application can improve the accuracy, collaboration, robustness and constraint controllability of network sentiment regulation.

The application provides a kind of industrial vertical field corpus construction data cleaning method, belong to industrial big data technical field, the application is through collecting multi-modal industrial data and recording original time stamp, unified time reference is generated using dynamic time warping algorithm for time alignment, cross-modal fusion feature vector is generated by multi-modal attention mechanism learning feature interaction weight, key data section is identified based on event-driven trigger rule and is associated as event data package, quality evaluation and abnormal calibration are carried out through step response consistency detection, multi-modal knowledge graph is constructed and cross-modal alignment model based on dual optimization constraint satisfaction framework is used for feature completion, finally, data quality is ensured through adaptive verification process, solve the technical problems that multi-modal industrial data is difficult to ensure alignment accuracy and original information integrity in time dimension alignment and semantic level fusion process.

This invention discloses a hierarchical multi-agent reinforcement learning-based trip recommendation method and system, belonging to the field of trip recommendation technology. The hierarchical multi-agent reinforcement learning-based trip recommendation method includes: acquiring basic POI data, user historical behavior data, and contextual information; constructing a dynamic heterogeneous POI graph and generating POI embedding representations; constructing multi-type user preference representations and calculating the Bayesian user preference posterior distribution, and calculating candidate POI preference scores; constructing a hierarchical multi-agent reinforcement learning model, including, from high to low, a region selection agent, an agent action set generation unit, a POI selection agent, a time scheduling agent, and a path optimization agent; and outputting the final personalized trip route based on the hierarchical multi-agent reinforcement learning model. This invention can efficiently generate executable trips that balance personalization, path quality, and hard constraint satisfaction in large-scale POI scenarios, significantly improving recommendation accuracy and user experience.

The invention discloses an improved reservoir dispatching safety reinforcement learning method based on a Lagrange method, and relates to the technical field of reservoir dispatching. The Lagrange duality method has obvious advantages in the aspect of security constraint processing, and compared with a corrective method and a penalty method, the Lagrange method has the advantages that the constraint weight is adaptively adjusted, so that the target optimization performance is remarkably improved while the constraint is satisfied; experimental results show that the method is superior to other methods in the aspects of constraint satisfaction and generating capacity, and the constraint violation degree is reduced by more than 80%; according to the method, an action mask mechanism is adopted, through the synergistic effect of preposed defense and posterior optimization, the training stability is effectively improved, a complete rule mask can precisely define a safety boundary, the violation rate is reduced to zero from 21%-79%, and meanwhile strategy convergence is accelerated.

The present application relates to steel structure construction and assembly planning, and proposes a method for once-through of steel bar crossing through the through hole of cross steel column based on constraint programming; in order to solve the scratch, deadlock and rollback caused by the constraint of hole structure and fireproof coating, the method is: a scene model is established with the center of the through hole as the reference, the hole, coating, steel bar and reachable parameters are obtained; the contact normal is determined through geometric reasoning, the minimum bending radius is converted into the upper limit of the cumulative turning angle of the path, and the coating contact limit area is calibrated according to the friction and allowable indentation, and the incident angle range and allowed turning angle are given; the tool angle resolution is discretized and a safety margin is set, and the interlocking relationship table is formed in combination with mutual interference; the constraint satisfaction integer programming is carried out with the steel bar penetration sequence, the incident side, the incident angle and the guide as variables, and when it is not feasible, a detachable trumpet guide is installed at the high-risk hole edge for recalculation. The method ensures that the steel bar is once-through, avoids coating scratches and assembly deadlocks, and improves the construction feasibility.

The invention discloses a generalized Nash equilibrium search method and system based on a neural network, and relates to the technical field of distributed control, and the method comprises the specific steps: aggregation game problem definition and dynamic model construction, disturbance approximation and constraint transformation design, aggregation estimation and optimization method design, and parameter range determination and monitoring and verification. According to the method, the Lyapunov function is constructed through the Lyapunov stability theory, and the value range of the adjustment parameters is scientifically determined in combination with key parameters such as the Lipschitz constant, so that the innovation not only ensures the convergence of the distributed time-varying generalized Nash equilibrium optimization method, but also improves the optimization efficiency of the distributed time-varying generalized Nash equilibrium optimization method. In addition, mathematical requirements such as a global Lipxiss condition, non-empty closed convexity and continuous differentiable convexity are met, a design-verification closed loop is formed by monitoring an intelligent agent strategy evolution curve and a constraint satisfaction condition, and it is ensured that the optimal state can be continuously kept in the actual operation process.

The application discloses a dynamic optimization method for unmanned aerial vehicle online trajectory planning, and belongs to the technical field of unmanned aerial vehicle autonomous control, and comprises the following steps: constructing a finite time domain optimal control problem containing a target function and multiple constraints at each sampling time; converting continuous control input into a segmented function by control parameterization technology, converting an infinite dimension problem into a finite dimension nonlinear programming problem; converting time-varying speed and obstacle avoidance constraints into integral penalty items for processing by using a constraint transcription method; introducing a terminal controller and a terminal area constraint to ensure recursive feasibility and asymptotic stability of the system; solving the nonlinear programming problem on-line, and applying the first control instruction to the unmanned aerial vehicle. The application realizes real-time, safe and efficient trajectory generation of the unmanned aerial vehicle in a complex environment by optimizing problem reconstruction and calculation simplification, and provides reliable technical support for unmanned aerial vehicle autonomous tasks under the premise of strictly ensuring constraint satisfaction and closed loop stability.

The invention relates to steel structure construction and assembly planning, and provides a one-time penetrating method for a reinforcing steel bar to penetrate through a cross-shaped steel column through hole based on constraint planning. In order to solve the problems of scratching, deadlock and rollback caused by orifice construction and fireproof coating constraint, the method comprises the steps that a scene model with the center of a through hole as the benchmark is established, and orifices, coatings, reinforcing steel bars and reachable parameters are obtained; determining a contact normal direction through geometric reasoning, converting the minimum bending radius into a path accumulated corner upper limit, calibrating a coating contact limiting area according to friction and allowable indentation, and giving an incident angle range and an allowable corner; discretizing according to the tool angle resolution, setting a safety margin, and forming an interlocking relation table by combining mutual interference; the rib penetrating sequence, the incident side, the incident angle and the guide piece are used as variables for constraint, integer programming is met, and when not feasible, a detachable horn mouth guide piece is additionally arranged on the high-risk hole edge for recalculation. According to the method, single-time penetration of the steel bars is ensured, coating abrasion and assembly deadlock are avoided, and the construction feasibility is improved.

The invention discloses a program control instruction generation method based on large model federation prompt fine tuning, which realizes the collaborative utilization of multi-party data on the premise of protecting the data privacy of each party by constructing a multi-level conditional diffusion model federation learning architecture. The problem of rapid adaptation of new equipment is solved through a heterogeneous document-oriented hierarchical retrieval enhancement generation method, and the problems of information fragmentation and text discontinuity in the traditional RAG technology are solved by constructing an operation process knowledge graph and a multi-granularity semantic representation learning framework and adopting a continuous perception retrieval mechanism. And finally, solving a fundamental contradiction between large model probability generation and test instruction preciseness through a constraint satisfaction deterministic generation mechanism, and converting a probability generation process into a strict logical reasoning process by adopting a beam search algorithm guided by a complete deterministic constraint. By constructing a three-dimensional constraint system of grammar constraint, semantic constraint and security constraint and a multi-stage verification mechanism, each generated instruction is ensured to have uniqueness, correctness and security.

The application discloses a large model multi-round dialogue contradiction recognition and correction method, system, device and medium, and belongs to the technical field of artificial intelligence. The method comprises the following steps: S1, extracting semantic information, and adopting an incremental updating algorithm to construct and update a dialogue state graph; S2, performing multi-level contradiction detection including a semantic layer, a constraint layer and an intention layer based on the dialogue state graph; S3, judging whether the user actively changes according to a preset change mode word, and if not, performing a first locking, clarification generation or annotation and recording strategy according to a contradiction level to obtain a corrected reply; and S4, inputting the corrected reply into a quality evaluation model, calculating four scores of fact consistency, logical coherence, fluency and constraint satisfaction degree, and determining output or triggering a secondary correction process according to a weighted comprehensive result of the four scores. The application solves the problems of single contradiction detection and fragmented correction and detection in the prior art, realizes high-quality consistency maintenance of multi-round dialogue, and does not need model fine tuning.

The invention discloses a micro-grid multi-target rolling optimization operation method considering voltage stability and energy storage cooperation, and belongs to the technical field of power system operation and control. The invention aims to solve the problems of low precision, poor robustness and high operation cost when an existing micro-grid energy management strategy is used for processing the uncertainty of renewable energy sources. The method comprises the following steps: preprocessing original photovoltaic output and meteorological data, carrying out smoothing processing by adopting an improved Akima interpolation method, screening related input characteristics through a Pearson's correlation coefficient, and decomposing the related input characteristics into different frequency band sub-signals by adopting a wavelet packet; training the LSTM prediction sub-model, and carrying out linear weighted fusion output; establishing a micro-grid system mathematical model, and taking power generation cost and environment cost as optimization targets; solving is carried out based on an improved NSGA2 algorithm, a Pareto optimal solution set is obtained by introducing an adaptive learning mechanism and a constraint satisfaction strategy, and a TOPSIS method is adopted to decide and select an optimal scheduling scheme; and executing rolling horizon optimization. The method is used for micro-grid sustainable energy management.

This invention relates to the field of compliance verification technology, specifically to a method, system, and electronic device for compliance verification of generative model output. The method includes: acquiring the output content, evidence set, and compliance policy of the generative model, and converting the compliance policy into a set of constraint rules; performing constraint satisfaction processing based on the output content and the set of constraint rules to obtain a set of constraint judgment results; generating verifiable proof data based on the set of constraint judgment results, and generating a compliance proof package containing the verifiable proof data, a policy fingerprint of the compliance policy, and a hash value of the output content; digitally signing the compliance proof package and outputting it; and having the verifier directly verify the compliance of the output content based on the verifiable proof data. This invention effectively solves the problems of existing technologies where compliance verification relies on repeated semantic parsing and constraint matching processing, resulting in high computational overhead, low efficiency, and a lack of credibility in the verification process.

The present application relates to the technical field of artificial intelligence and industrial process modeling, in particular to a time series data generation method and device based on Gamma variational autoencoder, comprising: modeling the latent variables by using Gamma distribution in the variational autoencoder, and when training the variational autoencoder, first, according to the constraint condition, normalizing each variable value of the original time series data in the training set as the training input, and then introducing a differentiable reparameterization sampling mechanism to sample the Gamma distribution constructed by the encoder output parameter, so as to solve the training instability problem existing in the traditional sampling method under the condition of small shape parameter, so that the trained variational autoencoder can significantly improve the constraint satisfaction rate and physical consistency of the generated time series data. The present application can stably generate high-quality time series data with good engineering usability.

The invention discloses a large-scale academic conference intelligent scheduling system and method based on multi-algorithm fusion, belongs to the technical field of conference management, and aims to solve the problems of low resource allocation efficiency, scattered themes, author time conflict and inconvenient user query in existing large-scale academic conference scheduling. The system comprises a data acquisition module, a theme clustering module, a scheduling optimization module and a front-end display module. According to the method, conference paper data are obtained through a web crawler, text features are extracted through a TF-IDF algorithm, paper theme classification is achieved in combination with K-means clustering, then an optimized scheduling scheme is generated through a greedy algorithm or an integer linear programming (ILP) algorithm, and finally a date / conference room screening query function is provided through a front-end interface. According to the method, the conference room use cost can be reduced, the greedy algorithm is suitable for real-time large-scale scheduling, the ILP ensures the constraint satisfaction degree, and the conference scheduling efficiency and the conference participation experience are remarkably improved.

The invention discloses a human-post personalized matching recommendation method fused with a dynamic knowledge graph, which relates to the technical field of human resource management, and comprises the following steps: carrying out relation extraction on an original acquisition item set, generating a candidate triple set, loading the candidate triple set, generating a post map data set, and carrying out post map data processing on the post map data set; establishing a version index record for the position map data set, and generating a position map packet; performing position skill closure processing, shortest repair path search and reachability threshold calculation on the basis of the position map packet, and generating a position reachability table; carrying out difference set comparison according to the position reachable capability table to generate a reachable candidate list, carrying out time stamp screening and region screening on the reachable candidate list, and carrying out duplicate removal to generate a candidate fusion set; and performing relation sorting on the candidate fusion set, generating a disturbance scheme, performing constraint satisfaction verification on the disturbance scheme, and generating a causal enhancement sorting result. The reasonability and adaptability of the recommendation result are enhanced, and the credibility of the recommendation result is enhanced.

This invention provides a heuristic optimization method for vehicle workshop production scheduling, relating to the field of vehicle manufacturing technology, and includes the following steps: S1, Input Validation Step: Validating the completeness and validity of the vehicle order information, production constraints, and production goals input by the user; S2, Scheme Initialization Step; S3, Iterative Optimization Step: Before meeting the user-defined algorithm termination condition, the following operations are performed iteratively: S31, Scheme Adjustment; S32, Scheme Evaluation and Comparison: Calculating the production goal achievement degree and constraint violation degree of the new scheme; if the new scheme is not inferior to the current optimal scheme in all goal achievement degrees and constraint violation degrees, then the new scheme is updated to the current optimal scheme; S4, Scheme Output Step. This invention solves the problems of strategy conflict and evaluation ambiguity in multi-objective production scheduling through one-time overall optimization and quantitative evaluation, overcomes the limitations of empirical strategies, and significantly improves scheduling efficiency and quality.

This invention relates to the interdisciplinary field of intelligent control and constraint programming, and provides a stage intelligent control method and system based on multi-device collaboration. To address the inherent spatiotemporal lag between equipment execution and actor behavior caused by the "post-event response" mode of traditional stage control, the method establishes a unified spatiotemporal reference, captures the continuous motion trajectory flow of actors in real time, and directly encodes it into dynamic spatiotemporal constraints. These constraints are then fused with the physical response characteristics of the stage execution equipment to construct a multi-dimensional fusion constraint space. Feasible state sequences are solved in real time within this multi-dimensional fusion constraint space, generating instruction sets with execution timestamps, which are then distributed to each stage execution device for synchronous execution. Through a constraint satisfaction mechanism of "direct behavior encoding - constraint fusion - real-time solution," real-time and precise collaboration between multiple stage devices and actor behavior is achieved, significantly enhancing the artistic expression of stage performances and the creative freedom of actors.

The invention provides an end side image quality inspection method and system for constraint satisfaction degree coding, and belongs to the field of computer vision and edge computing, and the method comprises the steps: obtaining a structured constraint set; extracting feature data of the target image; calculating a constraint satisfaction degree corresponding to each constraint; aggregating the plurality of constraint satisfaction degrees to generate a constraint satisfaction degree vector; and processing the vector according to a preset decoding rule to generate an end-side quality inspection result. According to the method, the structured constraint satisfaction degree vector is generated on the end side, the information density of the quality inspection result is improved, the communication overhead and the privacy risk are reduced, the vector can be reused by subsequent tasks, and the robustness and expandability of the system are enhanced.

This invention relates to the field of multi-source data analysis and processing technology, specifically to a slope stability analysis method and system based on multi-source sensing. The method includes: collecting slope monitoring data, generating a multi-source time-mapping sample set, constructing a multi-source time reference function, and using the multi-source time reference function to reconstruct the time series of the slope monitoring data; constructing event nodes according to preset event scale rules, constructing cross-scale causal constraint edges, generating multi-layer event subgraphs according to event scales, and constructing a cross-scale event sequence reconstruction map; generating an event chain sequence, selecting a set of target paths according to constraint satisfaction criteria and encoding them as a real event chain sequence; performing structural analysis and parameter statistics on the real event chain sequence, and generating slope time series analysis results. This invention achieves slope stability analysis by reconstructing multi-source monitoring data using a unified time reference and restoring the real event chain order based on a cross-scale event map.

The application discloses a kind of competitive evolution multitask optimization method, system and equipment, to solve the problem that existing method is difficult to carry out collaborative optimization to fuel cost and gas emission under meeting power balance and unit capacity constraint condition.The method comprises the following steps: constructing main, auxiliary double population parallel search;Real-time calculation state index representing population convergence, diversity and feasibility;Based on the current state, through deep reinforcement learning agent self-adapting selection cooperation mode and evolution operator combination action;According to the improvement of population after action execution in cost, emission and constraint satisfaction, calculate reward;State, action and reward are used to train agent, iteratively optimize decision strategy, until the pareto optimal scheduling scheme that meets the constraint is output.The application realizes the adaptive intelligent guidance of evolution process, significantly improves the optimization quality, convergence speed and algorithm robustness of power system economic emission scheduling scheme.

The invention provides an intelligent course arrangement system architecture based on multi-dimensional constraints and data snapshots and an implementation method, and is applied to the technical field of data processing. According to the method, basic business data and multi-dimensional constraint parameters are obtained, entity time sequence association is completed through modular decoupling design, structured modeling and resource overall planning are constrained, and a standardized input data set is generated. A data state is locked by creating a course arrangement task exclusive data snapshot, an initial course arrangement scheme is generated in combination with a core algorithm and a conflict detection mechanism, and manual fine adjustment and real-time verification optimization are supported. Historical data are associated based on a snapshot tracing mechanism, a closed-loop optimization system is constructed, and finally, an evaluation report containing indexes such as course arrangement efficiency and a constraint satisfaction rate and a visual course schedule are output through a multi-dimensional course schedule display model and a snapshot filing retrieval function, so that efficient, compliant and traceable intelligent course arrangement is realized.

A problem-solving collaboration system and method, including: a semantic structured annotation module, which labels each information segment with a type and assigns a role; a dynamic deduction whiteboard module, which provides an interactive visual canvas for displaying the initial form of a standardized mathematical model, receiving user-set tentative solutions, and providing real-time feedback on objective function values and constraint satisfaction states; a conflict diagnosis and heuristic guidance module, which identifies key variables and their impact coefficients that cause resource overruns when user-set tentative solutions violate any constraint conditions, and provides visual prompts for preferred adjustment directions and approximate correction magnitudes; and a conclusion extraction interaction unit, which generates visual charts based on solution distribution datasets and pushes guided questions. This system not only improves students' ability to solve complex application problems, but also effectively promotes the development of their high-order cognitive abilities such as logical reasoning, critical thinking, and model-based thinking, making it highly practical and technically advanced for teaching.

The disclosed concepts relate to interactive constraint solving with generative language models. For example, a generative language model can be prompted to generate a constraint data structure expressed in natural language that represents user preferences. The constraint data structure can be parsed to extract constraint parameters that can be programmatically solved by a constraint solver. The generative language model can also be prompted to generate constraint checking code that can be invoked by the constraint solver.

This invention discloses a distributed flexible job shop intelligent scheduling method oriented towards process-bound constraints, belonging to the field of computer vision technology. The method constructs a distributed scheduling environment simulator and constraint manager to output a normalized three-dimensional state feature vector; it filters the legal action set through constraint verification; it then constructs an improved duel depth Q-network agent to separate state value and action advantage flow to improve estimation accuracy; it introduces a priority experience replay mechanism to accelerate convergence, employs an ε-greedy strategy combined with action masking to ensure action validity, and triggers a heuristic backoff strategy when no valid action is found; it iteratively trains the agent guided by a composite reward function to achieve multi-objective optimization of completion time, load balancing, and compliance with process-bound constraints; finally, it extracts the optimal scheduling scheme from non-dominated solutions. This invention overcomes the shortcomings of traditional algorithms in constraint handling and susceptibility to local optima, achieving a synergistic improvement in constraint satisfaction and multi-objective optimization.

A kind of nuclear power unit power regulation control scheme automatic generation method and system based on deep reinforcement learning, by constructing the hybrid surrogate model of fusion CNN, GRU and self-attention mechanism, the dynamic response of reactor is represented, PPO-Lagrangian training strategy network is used, the accurate constraint control of axial power deviation in the regulation process of conventional AO method is realized;With current AO and target AO as input, the preset power and control rod position full-process operation sequence that meets the safety constraint is automatically generated;On the basis of existing operation, the remaining sequence is continued to be written, and the constraint violation caused by existing operation is repaired, the strategy network based on MLP introduces normalized space sampling to maintain training exploratory, hard constraint is used in test stage to ensure operation safety;Lagrangian multiplier self-adaptingly balances control performance and constraint satisfaction.The AO deviation can be controlled within the target value ±1.0%, and the generation time is less than 1 second after training is completed, which can provide real-time decision support for operators.