81 results about "Madaline" patented technology

The invention discloses a juicy peach yield prediction method based on comprehensive data analysis, and particularly relates to the technical field of agricultural intelligent perception. The method comprises the following steps: collecting multi-source data of an orchard, and constructing a data set containing meteorological, physiological, remote sensing and soil information; fruit tree physiological response parameters are extracted, and a bimodal diagram structure fusing the spatial adjacency relation and the physiological state similarity is established in combination with the dynamic climate anomaly index; inputting the graph structure into a graph neural network model, extracting spatial-temporal characteristics, dynamically adjusting an edge weight and a propagation layer number, and constructing an adaptive model; performing region division and weighted summarization according to a model output result, and finally obtaining a predicted value of the total yield of the orchard; the method improves the prediction accuracy under the conditions of complex climate and unstable data, and is suitable for refined orchard management.

The invention belongs to the technical field of network security, and provides a network threat real-time detection and defense method and system based on artificial intelligence. The method comprises the steps of multi-modal data acquisition and preprocessing, dynamic graph feature engineering and knowledge graph collaborative fusion, dual-adaptive model training and optimization, streaming real-time detection and anomaly scoring, DRL-driven hierarchical defense response and automatic disposal, and feedback-driven model adaptive updating and block chain auditing. According to the method, a mixed model of OS-ELM + dual-adaptive ridge regression + federated learning is designed, the training speed is higher than that of CNN, and over-fitting / under-fitting is avoided by dynamically adjusting a regularization coefficient; the federal learning realizes data local training and parameter uploading, and solves the problem of privacy disclosure; knowledge distillation enables the model volume to be reduced, edge equipment deployment is adapted while the accuracy is maintained, and the generalization ability is obviously superior to that of a traditional static model.

The invention discloses a federal remote sensing large model training method and system based on low-rank self-adaption. The method comprises the following steps: performing personalized heterogeneous data knowledge learning by using respective remote sensing data, a pre-trained large model and a local low-rank self-adaption model; using the trained local low-rank self-adaptive model parameters to initialize global shared low-rank self-adaptive model parameters, performing collaborative federal training based on the pre-trained large model and the global shared low-rank self-adaptive model, and learning global remote sensing task domain knowledge; after alignment of local heterogeneous data knowledge and global remote sensing task knowledge is realized in a different-rank alignment fusion mode, a federated remote sensing large model composed of parameters of a pre-trained large model and an aligned local low-rank adaptive model is output, so that the parameter quantity and communication overhead uploaded by federated training are greatly reduced, efficient training of the model is ensured, and the reliability of the model is improved. And self-adaptive alignment of local heterogeneous data and global remote sensing tasks is also realized, and the universality and robustness of the model are ensured.

The invention relates to the field of motor fault diagnosis, and particularly discloses a permanent magnet synchronous motor fault diagnosis method and system, which dynamically update motor physical model parameters through an online identification algorithm so as to reflect the characteristic change of a motor in real time. Thirdly, predicting the theoretical voltage of the healthy motor under the working condition by using the self-adaptive model and the real-time operation data, and calculating a residual sequence subjected to working condition normalization between the theoretical predicted voltage and the actual voltage; the residual signal is essentially free from the influence of working condition change, and only the abnormal characteristics caused by the fault are highlighted. And finally, inputting the high-robustness residual error sequence into a convolutional recurrent neural network, deeply fusing space-time fault features, and realizing accurate judgment on a motor fault type and confidence thereof, thereby effectively improving variable working condition adaptability and weak fault detection capability of diagnosis.

The invention discloses a knowledge graph-AI fusion fault diagnosis method for operation and maintenance management of a power distribution network, and relates to the technical field of artificial intelligence. The problems of low initial diagnosis accuracy, insufficient diagnosis adaptation capability in different grounding forms, long fault positioning time and the like caused by external interference of zero-sequence current signals are solved, and the conditions that the operation and maintenance efficiency is reduced and the power supply loss is increased are finally avoided. The method comprises the following steps: acquiring multi-source data and constructing a dynamically updated power distribution network knowledge graph; multi-mode fault feature extraction is carried out; using an adaptive AI model to identify a fault type and performing primary judgment; and accurately positioning a fault section and analyzing a spreading path in combination with a knowledge graph. According to the method, the single-phase earth fault preliminary diagnosis accuracy is improved to reduce invalid troubleshooting of operation and maintenance, cross-scene diagnosis errors are reduced to adapt to diversified operation and maintenance scenes, average positioning time consumption is shortened, and the supporting capacity of fault diagnosis on operation and maintenance management of the power distribution network is remarkably improved.

The invention belongs to the technical field of hypersonic flight vehicle control, and particularly relates to a hypersonic flight vehicle protection control method considering non-starting of an air inlet channel. The method specifically comprises the steps of establishing a quasi-one-dimensional nominal mechanism model to describe flow field changes of an isolation section and a combustion chamber; adopting a physical information neural network to construct a reduced-order model as a real-time agent of the quasi-one-dimensional model; designing a dual extended Kalman filtering algorithm to realize joint online estimation of model parameters and blind area states, and introducing shock wave position pseudo measurement to form multi-loop feedback; a dual-channel fusion early warning system of a model channel and an intelligent sensor channel is constructed, and robust early warning is realized through adaptive weight; and designing a continuous adaptive model prediction controller, and dynamically adjusting a control strategy according to risk indexes and parameter uncertainty to realize active protection control.

The invention relates to a self-adaptive model updating method and device, electronic equipment and a storage medium. The self-adaptive model updating method comprises the following steps: acquiring distributed external sample data newly generated by a large language model; clustering processing is carried out on the out-of-distribution sample data to obtain an out-of-distribution sample cluster, and the out-of-distribution sample cluster at least comprises out-of-distribution sample data of the same structure label; and performing search training on the out-of-distribution sample cluster, determining a target adapter, and sending the target adapter to a database of the large language model to update the large language model. Static model updating is converted into a time-driven learning process by setting a self-adaptive updating mode, so that a large language model can be subjected to self-adaptive updating operation according to failure result feedback during reasoning in a training process, the model is easier to adapt, error accumulation is reduced, stable updating is ensured, accuracy is improved, and the training efficiency is improved. And the precision and robustness of the model are enhanced.

The invention discloses a three-stage cooperative lane changing method for a vehicle queue in a ramp diversion area in an intelligent network connection environment, and the method comprises the steps: firstly, determining a safety gap of a target lane based on a minimum safety distance and a maximum deceleration, and selecting a vehicle which can change the lane to the safety gap from a CAV queue for preferential lane changing; secondly, establishing a longitudinal trajectory planning model based on self-adaptive model predictive control, performing longitudinal distance adjustment on a lane-changed vehicle by using the model, and reversely calculating an initial position required by the vehicle to smoothly drive into a ramp based on a quintic polynomial trajectory planning method; embedding into adaptive model predictive control as a ramp tail end constraint to form a new safety gap of the target lane; and finally, the vehicles which do not change lanes are controlled to sequentially change lanes to the corresponding safety gaps, and lane changing of the CAV queue is achieved. According to the method, the CAV queue can efficiently and orderly drive into the exit ramp, the vehicle speed fluctuation of a shunting area can be reduced, and the overall passing efficiency of a road section is improved.

An adaptive model partitioning method and system applied to distributed training, which method and system belong to the technical field of deep learning, and aim at solving the technical problem of how to implement, in respect of distributed training, distributed model training by means of combining deep reinforcement learning and a Bayesian optimization algorithm. The method comprises the following steps: constructing a Q network on the basis of a deep neural network, and defining state information, actions and a reward function, wherein the state information comprises feature vectors of partitioned models, and training times, resource utilization rates and inter-node communication overheads of when the partitioned models are subjected to distributed training by means of distributed computing nodes, each action is a partitioning strategy used by an agent under the current state information, the reward function R is used for evaluating the effect of the current partitioning strategy, and the Q network uses the state information as input to predict and output a Q value of each action that the agent may take; and performing multiple iterative training on a deep reinforcement adaptive model, so as to obtain a final partitioning strategy and parameters of the Q network.

The invention discloses a T ferromagnetic annealing performance consistency optimization method based on reinforcement learning, and the method comprises the steps: collecting the annealing temperature, heat preservation time, magnetic field intensity, cooling rate and other key process parameters in real time through an industrial Internet of Things, and improving the data quality through normalization and multi-stage data cleaning processing; intelligent evaluation and label classification of the process state are realized by using a multi-layer neural network; a dynamic reward function weight is introduced in combination with a process state label, and a deep Q network reinforcement learning model is driven to perform strategy optimization; the method supports self-adaptive model updating of periodic data acquisition and strategy rapid switching under significant change of a process state, realizes intelligent cooperative closed-loop control of multiple parameters in the annealing process, and improves the operation stability of the annealing process and the consistency of material performance.

The application discloses a double-layer adaptive RVM reliability analysis method for small failure probability, first, a first-layer adaptive RVM model is constructed by combining a Harris Hawks optimization algorithm and an adaptive RVM, and an important sampling sample is generated at a design point according to an iterative updating strategy to approximate the design point; then, the idea of active learning is utilized, a learning function is adopted to continuously update the RVM model, and a failure probability is solved after convergence. The application greatly improves the accuracy of reliability failure probability calculation results, reduces the calculation times, saves the calculation cost, and improves the ability of calculating small failure probability reliability.

The present disclosure proposes an actuator failure compensation and safety control method based on a double-layer evaluation architecture, a storage medium and a system. The bottom layer uses an adaptive model predictive control to output a nominal instruction, a reinforcement learning model is built thereon to output a high-frequency residual compensation instruction to eliminate model mismatch, and a second evaluation network is independently built to evaluate the future global loss-of-control risk of the vehicle. The Lagrange dual optimization is introduced to convert the macro risk evaluation into a mathematical constraint to guide the first strategy network correction; the global risk index is used to feed forward the target cost weight of the predictive control to realize continuous flexible degradation, and the control barrier function is triggered at the limit boundary to realize discrete quadratic programming instruction projection. The present application overcomes the conservative strategy degradation defect caused by the traditional mixed reward, establishes an absolute safety line, and improves the control accuracy and safety robustness of the vehicle in complex nonlinear working conditions.

The invention relates to the technical field of battery management, and discloses a battery thermal management adaptive method and system based on deep learning, and the system comprises a data collection module which is used for obtaining the real-time operation data of a battery; the failure mode diagnosis module is used for identifying specific failure modes such as internal short circuit risk and reversible concentration polarization of the single cell on line by using models such as a convolutional neural network; the joint prediction module is used for synchronously predicting future temperature and health risk indexes by using a bidirectional long-short-term memory network; the adaptive control module selects an optimal control strategy according to the failure mode and reconstructs a model prediction control problem; the instruction generation module is used for solving a model prediction control problem and generating a multi-domain cooperative control instruction of power management and heat management; and the closed-loop feedback module is used for correcting the control deviation. Through deep fusion of a deep learning diagnosis prediction model and a self-adaptive model prediction control framework, conversion from passive response to active pre-judgment is realized.

The invention relates to the technical field of BIM conflict checking, and discloses an NLP-driven district-level BIM data space conflict automatic checking method and system, and the method comprises the steps: receiving a natural language checking instruction of a user, inputting a pre-training multi-modal field adaptive model through an NLP processing module, and carrying out the NLP-driven district-level BIM data space conflict automatic checking. Completing intention recognition and named entity extraction to obtain key semantic elements and generate unified semantic representation; a dynamic task graph is constructed, tasks are simplified, after a target model and a collision rule are called, geometric intersection detection is accelerated by a GPU through a collaborative architecture scheduling engine cluster, and a CPU is responsible for topology analysis; acquiring original collision data, determining collision points by combining engineering knowledge graph clustering and grading, and generating an intelligent interpretation report; conflicts are visualized on a CIM platform, a BIM correction script is generated in combination with user operation and a grading result, and an updating model and a collision rule are synchronously collected and fed back. According to the BIM data processing method and the BIM data processing device, automation and high efficiency of BIM data processing can be considered while natural language instruction driven checking is carried out.

The invention discloses a high-voltage switch cabinet state recognition method based on a gradient boosting tree algorithm, relates to a high-voltage switch cabinet circuit breaker mechanical state recognition method, and aims to solve the problems that an existing high-voltage switch cabinet circuit breaker mechanical fault recognition method is low in recognition precision and high in false alarm rate and cannot be applied in real time across equipment. According to the method, a gradient boosting tree is adopted as a basic learning device, a multi-classification logarithm loss function is adopted as a target, regularization is added, an optimal hyper-parameter combination is determined through grid search, and a training set is utilized to perform greedy training on trees one by one to obtain an initial GBDT classification model; performing fine adjustment on a top tree structure of the initial GBDT classification model by using a small amount of labeled samples in a low-temperature or high-humidity environment of the target area to form an environment adaptive GBDT model; and deploying the environment adaptive GBDT model at an edge computing terminal, performing online reasoning on single-time opening and closing data acquired in real time, and outputting the mechanical state category of the high-voltage switch cabinet circuit breaker. The beneficial effects are that the fault identification accuracy is high and the false alarm rate is low.

The invention discloses an efficient single-stage domain adaptive target detection method for an open scene. The method comprises the following steps: acquiring a to-be-detected open scene target domain image; inputting the target domain image into a single-stage domain adaptive model, wherein the single-stage domain adaptive model is a model obtained by circularly executing cognitive closed-loop iterative training including a learning step, a practice step and a reflection step on the label-free target domain data; and performing target detection processing on the target domain image based on the single-stage domain adaptive model to obtain bounding box position and category information of each target in the target domain image. By constructing a single-stage cognitive closed-loop training architecture fusing the steps of learning, practice and reflection, self-iterative evolution is realized in the same detection model, the training efficiency is improved by one order of magnitude, error accumulation is effectively inhibited, and the detection accuracy is improved. The problems of low training efficiency, bloated architecture and performance limitation caused by lack of a feedback correction mechanism due to adoption of a complex multi-stage teacher-student framework in an existing domain self-adaptive method are solved.

The application discloses a kind of based on multi-modal perception and adaptive model's cold-rolled emulsion concentration intelligent closed-loop control method and system.Method includes: through heterogeneous sensor array multi-modal perception, data is fused using neural network based on attention mechanism, obtain comprehensive state information;Adaptive model predictive controller and the meta-learning optimizer based on deep reinforcement learning are used to generate control instructions by two-layer intelligent decision engine constituted;Combining digital twin model carries out forward-looking simulation and optimization;Finally drive high-precision actuator to complete closed-loop control.System includes multi-modal perception module, intelligent decision and control module, digital twin and optimization module, high-precision execution module and data communication network.The application solves the problems of low sensor reliability, rigid control model and lack of prediction ability in the prior art, and realizes high-precision, adaptive and intelligent closed-loop control of emulsion concentration.

The application discloses a kind of based on low-rank self-adaption's federal remote sensing big model training method and system, comprising: using respective remote sensing data, pre-training big model and local low-rank self-adaption model is carried out individualized heterogeneous data knowledge learning;Using the local low-rank self-adaption model parameter initialized global shared low-rank self-adaption model parameter after training, based on pre-training big model and global shared low-rank self-adaption model is carried out collaborative federal training, learns global remote sensing task field knowledge;By heterogeneous rank alignment fusion mode, the alignment of local heterogeneous data knowledge and global remote sensing task knowledge is realized after, output is composed of pre-training big model and the low-rank self-adaption model parameter of alignment after local by federal remote sensing big model, such as greatly reduce the parameter quantity and communication overhead of federal training upload, guarantee the efficient training of model, also realize local heterogeneous data and global remote sensing task self-adaption alignment, guarantee the versatility and robustness of model.

The invention discloses a wind power prediction method and device based on adaptive segmentation and a medium, and relates to the field of wind power prediction. The method comprises the following steps: for each candidate segment number in a plurality of candidate segment numbers, independently constructing a training set training model, and obtaining a corresponding adaptive Transform model based on the training set training model; based on the verification set, determining a target adaptive Transform model with the minimum verification error and the number of target candidate segments from all the adaptive Transform models; based on a greedy search strategy, in the real-time time sequence, searching for a target segmentation point enabling the data distribution difference value to be maximum for the target segmentation number, and dividing the real-time time sequence according to the target segmentation point to obtain a plurality of real-time sub-periods; and inputting the plurality of real-time sub-periods into the target adaptive Transform model to obtain a real-time wind power prediction value, thereby improving the prediction precision of the model through segmented prediction.

The invention relates to a sequential test design method based on BDNN and key area density weighting, and relates to the technical field of radars. The method comprises the steps of performing general feature extraction on an initial data set through a constructed dynamic adaptive BDNN model, and outputting a continuous prediction result and a classification prediction result; a multi-target chaotic particle swarm optimization algorithm is adopted to identify key areas in the multi-dimensional factor space, after an optimal solution set is obtained through iterative optimization, comprehensive scores are calculated and screened, and high-priority key areas are output; on the basis of the high-priority key area, density weighted sequential sampling is implemented to generate multi-source candidate points, the comprehensive weight of each candidate point is calculated, and final sampling points are screened out by maximizing the sum of the comprehensive weight and the spatial distance; and verifying the final sampling point, and fusing the verified sample with the initial data set for retraining the dynamic adaptive BDNN model until a preset convergence condition is met. The performance of the detection system can be improved.

The invention discloses a staged adaptive model pruning method, which belongs to the technical field of federal learning and comprises an initial pruning stage and a further pruning stage. In the initial pruning stage, a lightweight initial model is generated through parameter importance evaluation, and the initial communication and calculation overhead is reduced; in the further pruning stage, a model structure is optimized based on multilateral end cooperation and periodic reconfiguration, and dynamic adjustment is carried out to cope with heterogeneous data and resource limited conditions. According to the method, the communication and computing resource consumption of each round of training can be remarkably reduced, meanwhile, the model precision is ensured, and an efficient solution is provided for resource-limited federated learning.

The invention discloses a remote sensing crop identification domain adaptive model based on feature alignment, and the model comprises the following steps: 1, making a sample set; step 2, constructing a whole network architecture of the model; step 3, a detailed construction scheme of the model framework; step 4, a loss function fusion strategy; according to the feature alignment-based remote sensing crop recognition domain adaptive model, a cross-domain feature alignment strategy is introduced, so that effective fusion of a source domain and a target domain in a feature space is realized, and the adaptability and generalization ability of the model to different regions and time phase data are improved; the method not only improves the robustness and precision of crop identification, but also provides a technical path with higher generalization for remote sensing agricultural intelligentization.

The invention provides an abnormal working condition diagnosis method for a rainwater collection system, and the method comprises the steps: constructing a basic data layer through multi-modal sensor deployment, protocol standardization and time sequence synchronization, improving the data quality through the combination of sliding window filtering and normalization, building a self-adaptive multi-scale causal atlas through expert knowledge and historical fault data, and carrying out the diagnosis of the abnormal working condition of the rainwater collection system. The causal intensity is dynamically updated through Bayesian reasoning according to real-time observation; after the graph attention network is embedded, accurate recognition and confidence quantification of an abnormal propagation path and a root node are achieved, if the diagnosis confidence is insufficient, the self-adaptive model is retrained, the interpretability, accuracy and operation and maintenance efficiency of large system abnormality diagnosis are improved, and fault early warning and maintenance decision optimization are effectively supported.

The invention introduces an AI / ML-based CSI compression and reconstruction system for next-generation wireless networks, enabling efficient and scalable Channel State Information (CSI) reporting. It extends traditional spatial and frequency domain compression to include the temporal domain, significantly reducing feedback overhead while maintaining high downlink throughput and beamforming accuracy. The system leverages adaptive AI / ML models such as autoencoders, LSTMs, and transformers to exploit multi-dimensional CSI correlations. A server-driven architecture selects optimal models and hyperparameters based on real-time CSI statistics, enabling deployment across base stations and user equipment (UE) with minimal signaling. Compressed CSI is transmitted using standard uplink / downlink channels and reconstructed at the base station to recover full channel or precoding matrices. Quality metrics and feedback indicators ensure model consistency and performance monitoring. This approach achieves superior CSI fidelity, reduces quantization loss, lowers processing complexity, and supports dynamic adaptation to channel conditions making it ideal for 5G-Advanced and future 6G systems with massive MIMO configurations.

This invention discloses an adaptive model reuse method based on adapter fine-tuning. Taking object detection in computer vision as an example, firstly, a model pre-trained on a large-scale image dataset is obtained as the original model, and candidate adapters are set for it according to the structure of the pre-trained model. Then, the Taylor expansion method is used to calculate the global gain of adding adapters at different positions in the network for images with localized bounding boxes and class labels in the target domain, thereby achieving adaptive adapter configuration. In the fine-tuning stage, the global gain of adapters at different positions for labeled image samples is recalculated. The original model parameters are frozen, and the adapter parameters are updated using gradient backpropagation with a learning rate weighted by the global gain softmax. Finally, the fine-tuned model is reused in downstream tasks to locate objects in images and provide class predictions. This invention achieves adaptive adapter configuration based on global gain, realizing more accurate object localization and class prediction in object detection tasks.

The invention belongs to the technical field of semantic communication, and particularly relates to an intelligent metasurface-assisted semantic communication transmission method, which innovatively combines a programmable wireless environment with deep learning to realize joint optimization of a channel environment and semantic feature extraction. Two training algorithms, namely a dynamic random environment self-adaptive multi-signal-to-noise-ratio algorithm, are adopted, a signal-to-noise ratio value is dynamically adjusted in the training process, a single model is established to adapt to different signal-to-noise-ratio conditions, deployment complexity and resource requirements are greatly reduced, and the method is suitable for resource-limited equipment in Internet of Things application with frequently-changed channel conditions; according to the two-stage training algorithm, firstly, pre-training is carried out under a fixed signal-to-noise ratio, and then fine tuning is carried out for a specific signal-to-noise ratio, so that a special optimization model is provided for each signal-to-noise ratio environment, and the balance between a professional model and an adaptive model is deeply understood. According to the invention, a programmable wireless environment is combined with semantic communication, so that the reliability and efficiency of semantic information transmission are effectively improved.

The application discloses a kind of based on adaptive model predictive control curve lane keeping method and system, first according to the curve working condition information of vehicle travel determines the safe vehicle speed of vehicle over curve;Then the optimal time domain parameters of MPC controller under different road information are matched using genetic algorithm;Again, neural network MPC controller is constructed, so that the safe vehicle speed of vehicle over curve and the time domain parameters of MPC controller are adaptively adjusted according to the change of curve working condition;Finally, according to the trained neural network, the vehicle longitudinal controller and vehicle lateral controller are integrated, to realize the control of curve lane keeping.The application effectively solves the problem of vehicle sideslip or rollover caused by high speed over curve, and the poor lane keeping accuracy caused by fixed MPC time domain parameters, improves the accuracy of unmanned vehicle curve lane keeping, ensures the stability of vehicle driving, and has practical significance for promoting the rapid development of unmanned technology.

The invention discloses a hydrogen-doped natural gas pipeline brittleness failure intelligent prediction method and host equipment, and relates to the technical field of hydrogen brittleness evaluation of hydrogen-doped natural gas pipelines. The invention discloses an intelligent prediction method for brittleness failure of a hydrogen-doped natural gas pipeline. The method comprises the following steps: data preparation; calculating hydrogen embrittlement performance parameters; establishing a feature data set; intelligent model establishment: establishing a basic machine learning model and a domain self-adaption model, and performing domain self-adaption in combination with an adversarial neural network technology; model training and evaluation: dividing a feature data set into a training set and a test set through the feature data set, and training a basic machine learning model and a field adaptive model; and model fusion: integrating the trained basic machine learning model and the domain adaptive model into a fusion model yfinal, and outputting a final prediction result. The problem that the hydrogen embrittlement performance can only be predicted through an empirical formula or a single mechanical model in the prior art is solved, and natural gas pipeline brittleness failure prediction is intelligent.

This invention discloses a predictive control method and system for energy conservation in urban rail transit electromechanical systems based on an adaptive model, belonging to the field of energy conservation control technology for urban rail transit. Existing control methods for ventilation and air conditioning systems in subway stations struggle to balance environmental comfort and operational energy consumption. Online simulation is time-consuming, online strategy trial-and-error carries significant risks, and real-time adaptive adjustment is difficult. This method deploys an offline-trained neural network proxy model and a control strategy model on edge computing devices in subway stations. The neural network proxy model outputs predicted temperature field distribution data and equipment energy consumption data for multiple areas of the station based on outdoor environmental parameters, passenger flow parameters, and equipment operating parameters. The control strategy model uses this proxy model as an interactive environment to construct a reward function based on thermal comfort and equipment energy consumption indicators for reinforcement learning pre-training. This invention is primarily used for energy-saving optimization control of ventilation and air conditioning systems in subway stations.

The invention discloses a power grid equipment fault case detection method and system based on multi-granularity retrieval, and belongs to the technical field of intelligent operation and maintenance of power systems. The method comprises the following steps: cleaning and structuring a power grid equipment fault case text through an OCR (Optical Character Recognition) and a domain adaptive NLP (Network Length Polymorphism) model, and constructing an offline case knowledge base; the method comprises the following steps: decomposing a case into'equipment-defect-reason-processing-result 'quintuple knowledge units, and establishing a multi-granularity semantic index; a two-way retrieval mechanism (dense vectors and sparse keywords) is adopted to realize high-precision recall, and multi-factor fine arrangement is performed in combination with features such as time and regions; and finally, a question and answer result with a reference case is generated based on the RAG architecture, so that the interpretability and the user credibility are improved. The system supports off-line construction and on-line incremental updating, and is suitable for fault diagnosis and operation and maintenance decision support of power grid equipment such as transformers and circuit breakers. The method has remarkable advantages in case retrieval accuracy, question and answer interpretability and knowledge updating capability.