373 results about "Robustification" patented technology

The invention discloses an intelligent electric equipment monitoring and optimizing method, and relates to the technical field of electric power, and the method comprises the steps: collecting high-dimensional voltage-current time sequence data and transient event marks, calculating topological invariant features, inputting the topological invariant features to a lightweight neural network model, and recognizing the features of all electric equipment; performing abnormal attribution and anti-fact energy efficiency prediction by using causal reasoning and dynamic regularization regression based on the identified characteristics of each electric device, and constructing a multi-objective optimization function through the abnormal attribution and anti-fact energy efficiency prediction; and based on the multi-objective optimization function, generating an equipment operation scheduling strategy through a deep reinforcement learning agent, based on the operation scheduling strategy, sending a control instruction to the electric equipment, and collecting an operation result feedback in real time for optimization and updating. According to the method, through fusion of topological features, causal reasoning and safety reinforcement learning, the precision, robustness and safety of monitoring and optimization of the intelligent electric equipment are improved.

The invention provides a text code generation method based on multi-modal semantic embedding and a dynamic knowledge graph, and belongs to the technical field of artificial intelligence. According to the automatic code generation method based on multi-modal semantic embedding, the dynamic knowledge graph, constraint-driven code generation and context-aware repair, the semantic understanding precision and the code generation quality are remarkably improved by integrating the technologies of multi-head Transform, the graph neural network, reinforcement learning optimization, genetic algorithm sequence adjustment, the generative adversarial network and the like. According to the method, the knowledge graph can be dynamically constructed to enhance structured semantic modeling, high-quality codes conforming to specific industry standards are generated, functionality, efficiency and conciseness are considered, meanwhile, the iteration cost is reduced through an efficient error positioning and repairing mechanism, and the method is suitable for large-scale popularization and application. And the robustness and the flexibility of the system in diversified scenes are improved by utilizing adaptive weight optimization and multi-target balance.

The invention discloses a resume analysis method and system based on multiple large language models, and belongs to the technical field of large language models.The resume analysis method and system based on the multiple large language models.The resume analysis method and system based on the multiple large language models comprise the following specific steps that firstly, model parallel analysis is conducted, simultaneously inputting the data into at least two heterogeneous large language models through an application program interface; and each large language model independently performs information extraction and analysis according to the model structure and the training data of the large language model, and outputs a structured data result containing a plurality of preset fields. Through the multi-model parallel analysis and conflict re-judgment mechanism, the misjudgment risk of a single model is effectively reduced, the robustness of the whole system is improved, the resume analysis accuracy is remarkably improved, the model pool is automatically optimized and updated through the dynamic scoring mechanism, and the problem that the model is difficult to select and update is solved.

The invention relates to the field of computer systems, in particular to an operation stability testing method and system oriented to a heterogeneous computing system, and the method comprises the steps: constructing a system overall operation logic model comprising a resource topological structure, a task scheduling rule and a communication link matrix; identifying a key communication path according to the communication link matrix, analyzing a resource competition hotspot in combination with a task scheduling rule, and establishing an interference model library; calling a specific model instance in the interference model library according to a test target, configuring injection parameters, and dynamically applying interference in a system operation process; key performance indexes in the system operation process are collected in real time; based on the real-time monitoring data, a multi-index weighted scoring algorithm is adopted to calculate an overall stability coefficient, a robustness grade and a self-healing capability index of the system, and a stability evaluation result is generated; and generating a visual report according to the topological structure of the operation logic model, and supporting comparison and analysis of historical versions. Unified centralized management is realized, and test efficiency and evaluation accuracy are improved.

The invention discloses a target multi-attribute identification method based on feature decoupling and cross-task collaboration, and belongs to the technical field of computers of specific calculation models, and the method comprises the following steps: firstly, extracting the initial features of each task through a lightweight backbone network, carrying out feature decoupling in a subspace, and according to the cross-task feature similarity, carrying out feature extraction; according to the target multi-attribute identification method based on feature decoupling and cross-task collaboration, an orthogonal constraint weight is dynamically adjusted, then mutual information confrontation minimization is adopted to further suppress statistical dependence between tasks, task residual errors are injected in a cross-task feature aggregation stage, differentiation enhancement is achieved, and finally unified joint feature representation is formed. According to the method, subspace statistical independence is realized, independence and necessary collaborative information are considered, a stable basis is provided for subsequent fusion, statistical dependence between tasks is further suppressed through mutual information confrontation minimization, complementation information is reserved while independence is ensured, and feature discrimination and robustness are improved.

The invention relates to the technical field of artificial intelligence, and provides a multi-agent-based task processing method, system and device and a storage medium, and the method comprises the steps: receiving a user request, carrying out the intention analysis of the user request, and generating an initial task sequence according to an analysis result; according to at least one sub-task to be executed in the initial task sequence, scheduling at least one sub-agent to execute the sub-task, and monitoring an execution result of each sub-task; when the execution result of any subtask is monitored to be abnormal, performing reflection analysis on an abnormal reason, and performing task planning again based on a reflection conclusion and the current context information to generate a new task sequence; and rescheduling the corresponding sub-agent to execute the sub-task based on at least one sub-task to be executed in the new task sequence. According to the method, the replanning process is guided by introducing the reflection analysis when the task execution is abnormal, so that the task execution path is dynamically corrected and adjusted, and the robustness and success rate of task execution are improved.

The invention relates to an agent workflow-oriented dynamic fault sensing and risk propagation prediction method, and belongs to the field of artificial intelligence. Comprising the steps of constructing a dynamic dependency graph according to a target workflow; for each tool node, the input and output credible auditor verifies the workflow through a format compliance layer, an execution state layer and a semantic consistency layer, and execution is continued if all verification is passed; when the verification of any layer is not passed, blocking the return of a fault information structured report, and judging an error severity weight and a data sensitivity weight; and downstream nodes of the tool nodes which do not pass the verification are managed and controlled through a risk propagation prediction model based on the dynamic dependency graph. According to the invention, the initial fault can be intercepted in time and prevented from causing cascade propagation and amplification effects in a complex workflow; the normal form of error processing in the intelligent agent tool flow is changed, so that the intelligent agent tool flow is subjected to passive emergency response conversion in-process accurate control after an event, and the robustness and reliability of the complex workflow are enhanced.

The invention discloses a large language model jailbreak attack method based on alignment mechanism interference, and relates to the technical field of artificial intelligence. The method comprises the following steps: by constructing a consistency loss function, guiding an intermediate layer representation when a model generates a malicious request to approach an intermediate layer representation of a compliance request, thereby interfering a security alignment mechanism of the model and weakening the rejection capability of the model to malicious input. From the perspective of the internal representation space of the model, a jailbreak strategy which can interfere with the security cognitive path of the model is designed, and the model is disguised to the middle representation of malicious input as benign representation by optimizing prompt input, so that the jailbreak attack with higher concealment and stability is realized, the defense capability of the model to vulnerabilities is conveniently trained, and the security of the model is improved. The method can be widely applied to large model security mechanism research, alignment robustness test, red team evaluation and other scenes.

The invention provides a large language model safety protection defense method and device based on dynamic regulation and control, and belongs to the field of artificial intelligence safety protection. The large language model security protection defense method comprises the following steps: constructing a non-security data set and a security data set jail break prompt data set, calculating a gradient average value of parameters of each layer of a large language model during back propagation of various data sets, calculating cosine similarity among gradients, determining a non-security layer, namely a layer most sensitive to non-security content, and performing security protection defense on the non-security layer. Therefore, the subsequent regulation and control are more accurate and effective; in the aspect of dynamic regulation and control of a specified non-security layer, a joint loss function is established to optimize and train a security offset vector, and the security offset vector is applied to a hidden state of the non-security layer in a big language model reasoning process to carry out intervention, so that output of a big language model is dynamically regulated and controlled; therefore, the robustness of the large language model is improved.

The invention provides a pluggable multi-mode large model safety protection method, and belongs to the field of artificial intelligence safety protection, and the method comprises the steps: effectively covering a potential risk scene through a dynamic image-text variant adversarial generation technology, and remarkably enhancing the novel attack defense capability; a double-track collaborative filtering mechanism of a dominant rule filtering engine and a hidden semantic recognition engine is constructed, explicit and implicit violation contents are doubly protected, and malicious questions and normal interaction are accurately distinguished; then, by means of a cross-modal intention disassembling and self-adaptive risk matching technology, multi-modal content is deeply understood, and risk positioning is accurately achieved; finally, an ethical security defense line is constructed by using a value alignment bottom response, the input and output security, robustness and ethical compliance of the multi-modal large model are effectively guaranteed, and a modular pluggable architecture is adapted to various image-text multi-modal large models by constructing a closed-loop security protection link. And the method has model independence, strategy controllability and implementation convenience.

The invention provides a state management and anomaly backtracking method and system for a multi-agent collaborative task, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining interaction data, state data and an execution log, building an initial task knowledge graph based on an ontology model, building a bidirectional causal chain through a Bayesian network, and forming a dynamic knowledge graph; and when an exception is detected, analyzing a state transition rule, identifying propagation nodes, constructing an exception propagation sub-graph, generating an influence chain, determining a recovery priority and triggering state rollback. According to the method, an abnormal source can be quickly traced, abnormal diffusion is effectively controlled, and the system robustness and the task recovery efficiency are improved.

The invention relates to the technical field of artificial intelligence and knowledge engineering, and provides a digital human knowledge graph dynamic iteration system and method based on user feedback. According to the method, the timeliness and automation level of knowledge updating are improved; the configuration efficiency of technical resources is optimized; a data-driven iterative verification closed loop is constructed; according to the knowledge service system, the maintainability and robustness of the system are enhanced, the system has the capabilities of tracking, querying and rollback any change, the risk of system service degradation caused by misoperation or invalid updating is greatly reduced, and the stability of the whole knowledge service system is improved.

The invention belongs to the technical field of automatic driving safety, and particularly relates to a multi-task adversarial patch generation method based on adaptive weighted double-layer optimization, electronic equipment and a storage medium. According to the method, a double-layer optimization framework is constructed aiming at the problems of unbalanced attack effect and insufficient intensity of the existing multi-task countermeasure attack. In the inner layer optimization, the weight of each sensing task is adaptively adjusted through a Softmax mechanism, so that optimization resources are dynamically allocated according to task vulnerability; and the outer layer optimization is to update an adversarial patch by using a projection gradient ascending method under the constraint of a physical feasible region so as to maximize the weighted multi-task loss. Through alternate optimization of the inner layer and the outer layer, physical deployable patches which simultaneously generate balanced and remarkable attack effects on multiple tasks such as three-dimensional target detection, semantic segmentation and depth estimation can be generated, and the comprehensiveness and reliability of robustness evaluation of the automatic driving perception system are effectively improved.

The invention belongs to the field of track planning and intelligent control of a reusable vehicle, and relates to an intelligent track planning method for a cruise section of the reusable vehicle. The method comprises the following steps: constructing a three-degree-of-freedom centroid motion dynamics model of a cruise section vehicle, generating an offline optimal trajectory sample library covering wide working conditions based on a Gaussian pseudo-spectral method, designing a mapping relation between deep neural network learning task parameters and trajectory features, and outputting a high-quality trajectory planning initial value online. The distribution of collocation points is dynamically adjusted in combination with a self-adaptive collocation point method, a nonlinear programming problem is iteratively solved with a high-quality initial value as a starting point, and efficient and accurate optimization of the trajectory is achieved. Simulation results show that the method significantly reduces the sensitivity of online optimization to initial guess, still has fast response ability and high robustness under complex constraint and uncertainty working conditions, effectively improves the autonomous trajectory planning ability of the cruise section of the reusable vehicle, and ensures flight reliability and control precision.

The invention relates to the technical field of intelligent control technology and edge computing application, in particular to an actuator real-time control method and system based on edge computing, and the method comprises the steps: carrying out the multi-mode state sensing and real-time preprocessing of an actuator at an edge node, mapping a high-dimensional feature to a low-dimensional evolution space through space-time compression projection, and carrying out the multi-mode state sensing and real-time preprocessing of the actuator; constructing a lightweight prediction model in combination with a linear skeleton and nonlinear residual reasoning, and dynamically generating a confidence interval to quantify uncertainty; further introducing an evolution updating mechanism based on a residual error, correcting a projection matrix, a dynamic operator and an inverse mapping matrix on line, and ensuring that the real states of the model and the actuator are consistent for a long time; and finally, based on the prediction result and the risk perception capability, adaptively generating an optimal control signal within the security constraint. According to the invention, high-precision, low-delay and robust edge real-time control is realized, and the robustness and adaptability of the system under complex working conditions are improved.

The embodiment of the invention provides a multi-modal model reasoning method, a framework, equipment and a medium, and the method comprises the steps: carrying out the standardized packaging of multi-modal input data, reducing the adaptation complexity between heterogeneous data and a model through unified data packaging and multi-modal feature standardized mapping, and remarkably improving the expansibility and flexibility of a system. A dynamic resource scheduling strategy based on reinforcement learning is used to realize on-demand allocation and accurate scheduling of computing resources, and the utilization efficiency of the overall cluster resources is significantly improved. The deep fusion hybrid precision calculation and adaptive operator fusion technology realizes significant increase of calculation efficiency, and effectively supports a high-concurrency real-time reasoning scene. A closed-loop feedback and self-adaptive load regulation mechanism is introduced, and the system can dynamically regulate computing resources and task scheduling strategies according to real-time performance indexes, so that the system stability and the service quality under an extreme load are ensured, and the system robustness is enhanced.

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 artificial intelligence and man-machine interaction, and discloses a dynamic evaluation system and method for robustness of a graphical user interface agent, and the system comprises a task execution environment module, an abnormal triggering mechanism module, a state verification module, a data acquisition module and an evaluation index calculation module. The intelligent agent enters a simulation environment and completes a circulation interaction process of perception-decision-execution-feedback, abnormal events are monitored, automatically triggered, abnormal scene generation and recovery are carried out, state verification is carried out, whether tasks are successfully completed or not is judged, multi-source data acquisition of continuously recording the tasks in the whole operation period is carried out, and the task execution time is shortened. The method has the beneficial effects that the evaluation environment is converted from a static state to a dynamic state, the authenticity is obviously improved, a semantic-driven exception injection mechanism is introduced, context-aware interruption is realized, the evaluation is more objective, and the method has good expandability and reusability.

The invention relates to the technical field of dynamic object pose recognition and mechanical arm grabbing control algorithms, and discloses a dynamic object pose recognition and mechanical arm grabbing control algorithm. The framework comprises a multi-source heterogeneous sensing fusion module, a motion state joint estimation engine, a grabbing feasibility online evaluator and a rolling time domain grabbing controller, and low-delay collaboration is achieved through unified state space modeling and event-driven scheduling. Wherein the sensing module fuses RGB-D and inertial data to output a six-degree-of-freedom pose with covariance, the estimation engine adopts self-adaptive unscented Kalman filtering to recursion a complete motion state, the evaluator combines dynamic constraints to screen feasible grabbing postures, and the controller solves an optimal joint trajectory with obstacle avoidance constraints on line based on a self-adaptive target trajectory. And feedforward compensation is introduced to suppress dynamic disturbance. According to the technical scheme, the real-time performance is guaranteed, and meanwhile the grabbing success rate and robustness of high-speed, non-uniform-speed and sudden turning dynamic objects are remarkably improved.

The invention discloses an AI intelligent investment and research decision optimization method based on deep learning, particularly relates to the field of artificial intelligence financial science and technology, and solves the technical problems that an existing intelligent investment and research system is insufficient in multi-source data fusion, static stiffness of feature engineering, insufficient in complex market association modeling, lagged in extreme risk protection and the like. Cross-modal alignment of structured market information and unstructured public opinions is realized through a multi-modal data fusion architecture, a self-evolution feature extraction network is constructed by adopting a deformable convolutional network and a gating circulation unit chain, and a multi-target strategy is generated based on a reinforcement learning and graph neural network coupling engine. In combination with a conditional generative adversarial network simulation extreme scene and a meta-learning framework dynamic hardening risk threshold, a cloud edge collaborative decision closed loop is finally formed, and the decision adaptability, risk control robustness and model interpretability of the system in a complex market environment are improved.

The invention discloses an iterative task execution method and system based on dynamic feedback and causal fault tolerance, belongs to the technical field of artificial intelligence and data analysis, and remarkably improves the execution success rate and system robustness of a complex data analysis task by fusing multi-modal perception, dynamic feedback and causal reasoning. According to the method, the disassembling deviation caused by information splitting is avoided, the task complexity is quantified through the graph neural network, self-adaptive task disassembling is achieved, a dynamic feedback mechanism is combined, the task structure is corrected in real time in the execution process, and insertion, combination and sequence adjustment of subtasks are supported. A causal fault-tolerant strategy based on a historical failure trajectory is introduced, a high-risk path can be pre-judged, defensive operation can be automatically embedded, and error propagation is blocked. The whole process execution track is stored in a memory bank and is used for continuously optimizing the model and the strategy to form the closed-loop learning ability.

The invention discloses a reaction steam turbine through-flow method and system based on variable reaction degree optimization, and belongs to the technical field of steam turbine through-flow design. The method comprises the steps that an initial geometric model of a steam turbine through-flow part and thermal boundary conditions of all stages are obtained; inputting the initial geometric model and the thermal boundary conditions into a flow field solver, and solving a flow control equation to obtain an original flow field solution; constructing a functional taking through-flow efficiency as a target, and deducing an adjoint equation corresponding to the flow control equation; solving the adjoint equation to obtain gradient information of the target functional to each level of reaction degree; inputting the gradient information into a sequential quadratic programming optimizer, and iteratively updating the reactance of each stage by combining the maximum equivalent stress upper limit of the moving blade and the lower limit of the through-flow throat area to generate a variable reactance distribution scheme; and adjusting molded line geometric parameters of each stage of stationary blade and moving blade, and outputting optimized three-dimensional molded line data and a digital model file for numerical control machining. According to the method, the in-stage flow loss is reduced, and the design robustness is enhanced.

The invention discloses a power distribution optimization scheduling method and system suitable for a commercial complex microgrid. According to the invention, a modular layered structure is introduced in modeling, and the whole system is divided into a main power supply scheduling layer, a renewable output prediction response layer and an energy storage flexible adjustment layer; and a mathematical optimization + reinforcement learning cooperation mechanism is adopted in solving, a deterministic part is accurately optimized by using an MILP model, an uncertain part is subjected to online generation of a strategy by using a reinforcement learning algorithm, and rapid iterative updating and feedback control of a scheduling result are realized. And meanwhile, through a boundary coupling coordination algorithm, synchronous updating of state variables among the sub-models and global feasibility of a scheduling solution are guaranteed, so that the response speed, the solution quality and the system robustness of the whole scheduling system are improved.

Multi-stage hybrid network integrates relationship regularization links and explainable elements to improve alignment with human values, explainability, robustness, and efficiency. The network comprises neural components, event prediction elements, and probability models across multiple stages, with relationship constraints enforcing structured knowledge representation. Explainable elements provide interpretable rationales for decisions, enhancing transparency. Training incorporates supervised learning, human-guided refinement, semi-automated knowledge engineering, and adversarial robustness techniques. A Socratic reasoning module detects contradictions and refines outputs for logical consistency. Indexed model elements enable dynamic memory optimization for improved efficiency. Candidate outputs may be scored, verified, or selected using neural and symbolic criteria. The invention supports retry loops and configurable subsystem pipelines to improve output quality. Applications include text generation, speech recognition, translation, and decision support. By combining structured constraints, human oversight, and modular architectures, the system improves the trustworthiness, safety, and adaptability of AI systems across diverse modalities and tasks.

The invention belongs to the technical field of energy system optimization, and particularly provides a multi-objective collaborative optimization method and system for an integrated energy system. Comprising the steps of generating a multi-dimensional uncertainty time sequence scene of the integrated energy system; establishing a high-fidelity dynamic behavior model library; defining decision variables, and constructing a multi-target two-stage stochastic programming optimization model by taking the economy, environmental protection and toughness of the comprehensive energy system as a three-dimensional optimization target; solving the multi-objective two-stage stochastic programming optimization model by adopting a self-adaptive multi-objective evolution algorithm assisted by an agent model; and according to a solving result, realizing a mixed multi-attribute decision based on a fuzzy analytic hierarchy process and a multi-criterion compromise solution sorting method. According to the method, the park-level IES global optimal, high-robustness and high-efficiency planning design is realized from uncertainty modeling, equipment characteristic description and large-scale solution to multi-attribute decision full-chain innovation.

The invention discloses a dialogue modeling and multi-stage fine tuning method oriented to professional field document differentiation, belongs to the technical field of artificial intelligence and natural language processing, and aims at data processing and model fine tuning of professional field documents. The method comprises the four steps of data analysis and preprocessing, differential data construction, multi-stage model fine tuning and model evaluation and optimization: firstly collecting and classifying professional documents, analyzing and cleaning so as to extract question and answer pairs, then constructing multi-type question and answer pairs, performing de-resampling, filtering invalid data to construct an evaluation set, and performing multi-stage model fine tuning and model evaluation and optimization. And performing LoRA fine tuning by using different data sets and learning rates in four stages, and finally evaluating and iteratively optimizing the model based on multiple indexes. According to the method, the question and answer accuracy and the context understanding ability of the large model in the professional field are remarkably improved, the badcase alignment identity cognition is gradually corrected, and the reliability and robustness of the model in the professional scene are enhanced.

The invention relates to a collaborative learning method and system based on edge sample intelligent grading and cloud intelligent decision, and belongs to the technical field of edge intelligence and big data processing, and the method comprises the steps: (1) AI model real-time reasoning and difficult sample preliminary screening; (2) intelligently grading edge end samples; (3) uploading data with priority labels; (4) cloud server state aggregation; (5) running a cloud intelligent training decision algorithm; (6) carrying out on-demand training and version management on the model; (7) online updating and closed-loop feedback of the model; according to the method, scene self-adaption and autonomous evolution of the AI model are realized, the efficiency and the economical efficiency of model iteration are remarkably improved, the robustness and the quick response capability of the system are enhanced, and the automation and intelligence levels of the system are improved.

The invention provides a self-adaptive weighted hybrid modeling method and system for data drift sensing. The method comprises the following steps: acquiring multi-dimensional time series data and performing feature interaction analysis and screening to generate a fusion factor set; inputting the fusion factor set into a mixed structure comprising at least one ensemble learning model and at least one sequence model for training; based on the verification set and the dynamic evaluation indexes, determining fusion weights of all models in the mixed structure, and constructing a dynamic weighted mixed model; and setting a data distribution drift detection mechanism, comparing the distribution difference between a current data window and a historical reference data set, and automatically starting a retraining process of the hybrid structure to update the model when the difference reaches a trigger condition. According to the method, through cooperation of data drift perception and an adaptive weighting mechanism, the problems that an existing hybrid model is rigid in fusion strategy and lags behind a retraining mechanism are solved, and the robustness and long-term effectiveness of the model in a non-stationary data stream are remarkably improved.

The invention discloses a model robustness automatic evaluation method based on large model driving and related equipment. The method comprises the following steps: inputting a task template of a tested model into a dynamic Prompt engine module to obtain a sample generation instruction; generating an adversarial sample according to the sample generation instruction through a sample generation and verification module, and performing model evaluation on the adversarial sample; if the confrontation sample passes the model verification, extracting a high-dimensional multi-modal feedback vector of the tested model through a multi-modal feedback analysis module; a reusable semantic instruction is generated through a strategy control module according to the high-dimensional multi-modal feedback vector so as to optimize a generation strategy; and if the confrontation sample does not pass the model verification, performing effect evaluation on the confrontation sample through an evaluation and interpretation generation module and generating an interpretability evaluation report. According to the method, progressive generation of the adversarial sample and exponential-level improvement of the model robustness evaluation efficiency can be realized, a clear model improvement direction is provided for developers, and the method can be widely applied to the technical field of computers.

The invention relates to the technical field of sales prediction, in particular to an electronic product sales data prediction method and system based on artificial intelligence, and the method comprises the steps: obtaining historical sales time sequence data of a to-be-predicted product; constructing an initial sales prediction model according to the historical sales time sequence data; based on one or more predefined market impact morphological functions, generating an adversarial training sample, the market impact morphological functions being used for simulating external events causing abnormal fluctuation of sales volume; performing adversarial training on the initial sales prediction model by using a training data set containing adversarial training samples to obtain a prediction model with enhanced robustness; and predicting the future sales volume of the to-be-predicted product based on the prediction model with enhanced robustness. Therefore, the problems of strong external feature dependence, weak generalization ability to unknown impact, single model structure and the like in the prior art are solved.