52results about How to "Accurate and Efficient Prediction" patented technology

The invention relates to the technical field of flash memories, and discloses a NAND FLASH service life prediction and management method and related equipment. The method comprises the following steps: performing multi-dimensional acquisition and feature extraction on physical and electrical parameters and environmental error parameters of a target flash memory, and performing time sequence alignment and preprocessing on obtained basic feature data to obtain multi-modal parameter data; and through multi-level feature extraction and correlation analysis, constructing a multi-dimensional feature space, generating an initial fusion feature matrix based on feature pairing and fusion analysis, and through error coefficient and performance degradation calculation and performance prediction of a multi-P / E cycle scale, generating a life prediction result. Based on the performance state parameters, threshold division and deviation calculation are carried out, weighting calculation and state evaluation of margins are carried out, and safety operation margin and health state evaluation results are obtained; and finally, generating an operation management result based on state grade division and parameter mapping. And efficient and accurate prediction and management of the service life of the flash memory are realized.

The invention provides an underwater image enhancement quality evaluation method based on quality perception domain adaptation. The method comprises the following steps: S1, acquiring a plurality of original underwater images to generate a training set and a test set; s2, constructing an underwater image enhancement quality evaluation network by using a deep learning framework; s3, training the underwater image enhancement quality evaluation network by using the training set, calculating a root-mean-square error between a prediction quality score and a label corresponding to each first enhancement image, and obtaining network loss in combination with binary cross entropy loss; s4, repeating the step S3 for multiple rounds to obtain an underwater image enhancement quality evaluation network model; and S5, testing each second enhanced image in the test set by using the underwater image enhancement quality evaluation network model to obtain a corresponding predicted quality score as an underwater image enhancement quality evaluation result. The method has the beneficial effects that the quality of the underwater enhanced image can be efficiently and accurately predicted.

The invention provides a photovoltaic power generation power prediction method, system and device and a storage medium, and relates to the technical field of photovoltaic power generation power prediction, and the method comprises the steps: collecting historical data and real-time data of a photovoltaic array region; performing synchronous preprocessing to generate a historical data set and a real-time data set; establishing an initial generation power model, and drawing a curve graph of generation power changing along with time; analyzing climate data and cloud cluster image data in the real-time data, judging the illumination condition of the photovoltaic array area, and calculating the moving direction and the moving speed of the photovoltaic array area under the illumination condition; calculating the actual illuminated area of the photovoltaic array area under each time node, introducing the information into the initial power generation power model for optimization, and generating new prediction data; and carrying out comparison verification on the optimized prediction data and subsequent actual generation power. According to the method, the prediction accuracy is remarkably improved, and the influence of environmental factors on the prediction result is effectively reduced.

The invention discloses an unsupervised anomaly detection method for time series data, and relates to the technical field of anomaly detection of the time series data, which comprises the following steps of: filling a current null value by using an average value of a forward value and a backward value of the current null value data, and performing data normalization processing on numeric data in a data set. The method comprises the following steps: normalizing data to standard normal distribution with a mean value of 0 and a standard deviation of 1, and separating a training set and a test set from a given time sequence data set; according to the unsupervised anomaly detection method for the time series data, potential representation is performed on the data through a beta-VAE encoder, a long-term and short-term memory network module is adopted to estimate the long-term correlation of the potential representation, the forward and backward dependency relationship of the data is captured, and time series prediction is performed on the data. And finally, through average fusion of prediction results and reconstruction of a beta-VAE decoder, a reconstruction error is calculated to detect time sequence abnormity. The method has excellent anomaly detection performance on a plurality of public time series data sets.

The invention discloses a concrete durability prediction method based on data expansion and machine learning, and the method comprises the steps: 1, collecting concrete mix proportion data and environmental parameters in an engineering environment, and building a standard concrete durability database; 2, performing data expansion on the standard concrete durability database based on a Markov Monte Carlo algorithm; 3, constructing an artificial neural network model, taking the expanded database as training sample input, and optimizing and selecting model hyper-parameters; and 4, selecting the artificial neural network model of the hyper-parameter combination with the optimal performance for training, and realizing accurate prediction of the concrete durability under different mix proportions and environmental influences. According to the method, the Markov Monte Carlo algorithm is adopted for data expansion, the loss of the parameter relationship is avoided by adopting an internal preset model method, the defect that huge data is needed to support the establishment of an accurate machine learning model is overcome, and accurate and efficient prediction of the concrete durability chloride ion diffusion coefficient in an engineering environment is realized.

The invention discloses a shadow carbon price-based power transformation path analysis method and system, and relates to the technical field of power transformation path analysis, and the method comprises the following steps: dividing a path in a power transformation process into a plurality of node regions, obtaining the power supply state data of each region, extracting an abnormal power supply coefficient, an equipment overload coefficient and an abnormal communication coefficient from the power supply state data; constructing a power supply fault model based on a neural network by using the power supply abnormal coefficient, the equipment overload coefficient and the communication abnormal coefficient, and generating a power supply fault coefficient; and acquiring a power supply fault coefficient of each node region, drawing a power supply-fault curve graph, judging a region with a fault in the power transformation process based on the power supply-fault curve graph, and performing priority processing. Through power supply fault analysis, the problem that various abnormal faults of power supply cannot be found and eliminated in time in the power transformation process is solved.

The invention relates to a power grid resource allocation method and device, computer equipment, a computer readable storage medium and a computer program product, and can be used in the technical field of electric power. The method comprises the following steps: acquiring multi-source data of a power grid; the multi-source data at least comprises power grid operation data, environment data and behavior data; inputting the multi-source data into a resource prediction model to obtain a resource prediction result of the power grid; generating a resource allocation scheme of the power grid according to the resource prediction result; and allocating the resources of the power grid according to the resource allocation scheme. By adopting the method, the resource allocation efficiency of the power grid can be improved.

The invention discloses a gold mine rapid prospecting method suitable for a plateau desert region, and belongs to the field of gold mine prospecting, and the method comprises the steps: obtaining geological and chemical prospecting data, and selecting a prospecting region based on the geological and chemical prospecting data; acquiring a high-resolution satellite image of the prospecting area, and acquiring geological, construction and mineralization alteration characteristics of an abnormal area of the prospecting area based on the high-resolution satellite image to construct a geological data set; a gold ore metallogenic area prediction model is constructed based on the long and short term memory network and the convolutional neural network; and inputting the geological data set into the gold ore metallogenic area prediction model for calculation, and generating a gold ore metallogenic area delineation result. Compared with a traditional geological prospecting method, the method has the advantage that the prospecting cost can be remarkably reduced by utilizing the high-resolution satellite image and the deep learning model to perform gold ore metallogenic prediction. Meanwhile, due to the fact that the accuracy of the prediction result is higher, unnecessary drilling and sampling work can be reduced, and resources and time are further saved.

The invention belongs to the technical field of hydrology and ocean, and particularly discloses an ocean three-dimensional temperature and salt flow prediction method and system based on a Fourier neural operator, and the method comprises the following steps: obtaining ocean temperature data, ocean salinity data and ocean flow velocity data of different depth layers; constructing an ocean three-dimensional temperature and salt flow prediction model based on a Fourier neural operator architecture; and inputting the ocean temperature data, the ocean salinity data and the ocean flow velocity data into the ocean three-dimensional temperature-salt flow prediction model, and outputting to obtain an ocean three-dimensional temperature-salt flow prediction result. According to the method, the problems of high computing resource consumption and long time in the research of the traditional ocean three-dimensional temperature and salt flow numerical simulation are solved, the prediction precision and efficiency are improved, and powerful support is provided for ocean scientific research and application.

The invention discloses a wind power plant wake flow prediction method based on an improved physical information neural network. The wind power plant wake flow prediction method comprises the following steps: constructing a basic physical information neural network fusing a laser wind finding radar module, a two-dimensional Navier-Stokes equation and an actuating disc model; a cosine annealing algorithm and a dynamic loss function weight are adopted for the basic physical information neural network, and a step-by-step time saving strategy is introduced for training to construct an improved physical information neural network; constructing a training data set based on the radial wind speed data of the sparse measurement points of the laser wind finding radar, and training the improved physical information neural network to obtain a wind power plant wake flow prediction model; and inputting spatial position data and time data to obtain the predicted wake wind speed and air pressure of the wind power plant. According to the invention, the improved physical information neural network combines the measurement data with the physical law, and achieves the accurate and efficient prediction of the wake flow of the whole wind power plant through employing a precision improvement strategy and a step-by-step time saving strategy.

The invention discloses a fuselage turbulence micro-rib resistance reduction prediction method based on a Bayesian probability model, and belongs to the technical field of micro-rib resistance reduction. According to the method, the problems of high calculation cost and high difficulty caused by large calculation grid quantity and lack of calculation methods when the resistance reduction characteristics of the fuselage micro-ribs are solved by a traditional numerical simulation method are solved. According to the method, parameterization extraction is carried out on related features of fuselage micro-ribs, shape parameters, environment parameters, flight state parameters and the like of the micro-ribs are obtained, parameter dimension reduction is carried out, then modeling is carried out on historical data through a Bayesian neural network method, and finally accurate and efficient prediction of resistance increment of unknown micro-rib structures is achieved. And uncertainty evaluation is carried out on a prediction result. According to the method, the problems that a traditional CFD calculation method is high in calculation cost are effectively solved, the test efficiency is further improved, and accurate and efficient resistance reduction income evaluation of the fuselage micro-rib can be completed.

The present invention belongs to the technical field of big data analysis and management of photovoltaic power operation and maintenance, and specifically relates to a distributed photovoltaic intelligent monitoring method, system, equipment and medium. Meteorological data and pre-processed photovoltaic operation data are obtained, and the pre-processed photovoltaic operation data include photovoltaic operation power generation and component temperature; the meteorological data, photovoltaic operation power generation and component temperature are input into a pre-trained prediction model to obtain a photovoltaic power generation prediction value; wherein the pre-trained prediction model is obtained based on the FA algorithm and the GAT algorithm. The method of the present invention takes into account photovoltaic operation data including photovoltaic operation power generation and component temperature, and predicts photovoltaic power generation in combination with real-time meteorological data, thereby improving the accuracy of photovoltaic power generation prediction; by combining the efficient FA algorithm and the highly accurate GAT algorithm, the efficiency and accuracy of power generation prediction are further improved.

The invention relates to a heterogeneous reservoir pressure prediction method based on multi-modal data and knowledge hybrid driving. The method comprises the following steps that a heterogeneous reservoir permeability field and pressure field multi-modal data set is established; establishing a deep neural network model for knowledge guidance, and training the deep neural network model based on the multi-modal data set; and inputting the acquired permeability field and spatio-temporal information into the trained deep neural network model to obtain the predicted oil reservoir pressure. According to the method, the requirement of a traditional deep learning method for the high-cost training data volume is remarkably reduced, the algorithm performance and prediction precision are improved, and the method has important significance for intelligent development of an oil field and improvement of the recovery ratio.

The invention discloses an interpretable painless endoscopy hypoxia risk assessment method and a related device, and belongs to the technical field of machine learning, and the method comprises the steps: obtaining clinical features of a patient to be subjected to painless digestive endoscopy; inputting the clinical features into a pre-trained hypoxia risk assessment model to obtain a hypoxia risk assessment result; the acquisition method of the hypoxia risk assessment model comprises the following steps: acquiring an importance sorting result of each clinical feature in a historical sample data set; obtaining an interpretable model according to an importance sorting result; the negative sample data set is constructed based on the historical sample data set, the improved interpretable model is trained through the historical sample data set and the negative sample data set, the hypoxia risk assessment model is obtained, the problems that corresponding rules are difficult to learn in the model training process and overfitting is caused due to few sample data are solved, and the hypoxia risk assessment accuracy is improved. And accurate and efficient interpretable prediction of the low-oxygen event risk is realized.

The invention discloses a wind power plant layout optimization method based on a Gaussian wake flow model and an improved genetic algorithm, and the method employs a wake flow analysis model with higher precision, namely an Ishihara-Qian wake flow model, to evaluate the power generation amount of a wind power plant, and the model can efficiently and accurately predict the wind speed loss and additional turbulence intensity of a fan wake flow region. The variation rate and the crossover rate of the genetic algorithm are improved according to the population fitness, so that the two parameters can be dynamically adjusted according to the population fitness during each iteration, and when the fitness difference in one population is small (the population is relatively uniform), the corresponding crossover rate or variation rate is increased, so that the diversity of the population is increased, and the robustness of the genetic algorithm is improved. And falling into a local optimal solution is avoided.

The invention discloses a drug-drug interaction prediction method based on multi-channel feature fusion, which effectively combines a drug molecular substructure, a drug similarity matrix and a molecular fingerprint by utilizing various features of drug data and combining a multi-channel feature fusion module and a cross attention mechanism, thereby predicting drug-drug interaction. And then realizing efficient and accurate DDI prediction by using a KAN neural network. According to the method, various characteristic information data, such as drug substructures, drug similarities and drug molecular fingerprint data, of drugs are efficiently integrated, and efficient and accurate drug-drug reaction prediction based on multi-channel characteristic fusion is realized; the MCF-DDI model provided by the invention shows excellent performance in interpretable experiments and contrast experiments, not only are evaluation indexes and stability improved to different degrees, but also key substructures in drug molecules can be identified, and the importance of the substructures in drug interaction prediction can be verified; powerful theoretical support is provided for medicine research and development and clinical application.

The invention discloses a method for predicting the grain morphology of a titanium alloy cast ingot. The method comprises the steps that S1, longitudinal cutting structures of the titanium alloy cast ingot under different processes are collected, and an equiaxed crystal zone and a columnar crystal zone are distinguished; s2, establishing a titanium alloy ingot numerical model through software; s3, actual smelting parameters are input into the titanium alloy ingot numerical model, and solidification dynamic information is output; s4, a CET marking program is used for making the longitudinally-cut tissue correspond to the solidification dynamic information in the spatial position; s5, drawing a G-R macroscopic structure map; s6, the critical temperature gradient G * and the critical solidification rate R * are obtained according to the macroscopic structure atlas, and titanium alloy as-cast structure prediction is achieved. According to the prediction method for the grain morphology of the titanium alloy cast ingot, temperature field and flow field information corresponding to the cast ingot under different smelting parameters is obtained through numerical simulation, a certain relation is established between solidification dynamic information in the cells and an actual solidification structure, and the macrostructure morphology of the titanium alloy cast ingot is predicted more rapidly and accurately.

The invention provides a plant cell type prediction method and device based on interpretable machine learning, and the method comprises the steps: obtaining single cell data of a target organ of a target plant; preprocessing the single cell data to obtain standardized single cell data; performing feature screening on the standardized single cell data to obtain a hypervariable gene of the single cell data, the expression level of the hypervariable gene in a target cell being higher than a first expression threshold, and the expression level of the hypervariable gene in other cells being lower than the first expression threshold; performing normalization based on single cell data and hypervariant genes to obtain a cell gene matrix; determining a target multi-classification model corresponding to the target organ of the target plant; inputting the cell gene matrix into a target multi-classification model to obtain a predicted cell type output by the target multi-classification model; according to the invention, efficient and accurate prediction of cell types of various organs of various plants can be realized.

The invention relates to the field of material science and engineering, and discloses a big data-driven intelligent gold bonding wire system performance prediction method, which comprises the following steps: constructing a gold wire physical model of a gold bonding wire, defining a data-driven network architecture of the gold bonding wire, and determining a data-driven model of the gold bonding wire; defining a GANs network of the intelligent gold bonding wire system, analyzing the gold wire feature tuple of the gold bonding wire, and training the gold wire physical model and the data driving model to obtain a trained gold wire physical model and a trained data driving model; analyzing physical model performance and driving model performance of the training gold wire physical model and the training data driving model, and coupling the training gold wire physical model and the training data driving model to obtain a mixed gold wire analysis model; and capturing a performance factor dependency relationship of the gold bonding wire, analyzing a potential performance coefficient of the gold bonding wire by using the mixed gold wire analysis model, and constructing a performance analysis report of the gold bonding wire. The performance analysis effect of the gold bonding wire can be improved.

The invention discloses a light field orbital angular momentum coding quantum key distribution performance improvement method based on a neural network. The method comprises the following steps: when a plurality of photons are subjected to communication transmission in atmospheric turbulence, the transmission efficiency of the plurality of photons is obtained, and the plurality of photons are coded through orbital angular momentum in the transmission process; determining an initial security key rate of the plurality of photons under high-dimensional quantum key distribution based on the transmission efficiency of the plurality of photons; and based on the initial security key rate of the plurality of photons under the high-dimensional quantum key distribution, obtaining the security key length of the high-dimensional quantum key distribution under the finite length effect. According to the method, the technical problems of wide calculation range, low speed and long calculation time when quantum communication is transmitted in a free space and a security key is calculated by a polarization coding technology are solved.

The invention discloses a reference crop evapotranspiration prediction method suitable for limited meteorological conditions. The reference crop evapotranspiration prediction method comprises the following steps: acquiring historical meteorological data through a meteorological station and preprocessing the historical meteorological data; calculating a reference crop evapotranspiration standard value by adopting a P-M formula; extracting important factors based on an integrated feature selection method, training a LightGBM model by using data after feature selection, automatically determining an optimal hyper-parameter combination of the LightGBM by using a northern eagle optimization algorithm, and further obtaining a trained NGO-LightGBM prediction model; and preprocessing the real-time meteorological data, and inputting the determined important factors into the model to obtain a predicted value. According to the method, the meteorological factors having significant influence on prediction of the evapotranspiration of the reference crop can be effectively extracted, the hyper-parameters in the model can be accurately adjusted, the complex nonlinear relationship between the key meteorological influence factors and the ET0 can be accurately fitted, and the ET0 can be efficiently and accurately predicted under the limited meteorological condition.

This invention provides a training and prediction method, system, device, and storage medium for a pressure prediction model, belonging to the field of urban intelligent water supply network technology. The training method for a multi-monitoring-point network pressure prediction model includes: acquiring historical pressure data of the monitored urban water supply network; constructing a pressure prediction model, which includes a temporal convolutional network model and a long short-term memory network model; jointly training the convolutional network model and the long short-term memory network model based on historical pressure data to obtain a trained pressure prediction model; and using this trained pressure model to predict the pressure data of each monitoring point on the monitored urban water supply network based on the historical pressure data. This invention provides a training method for a pressure prediction model that, based on the correlation and periodicity of multiple monitoring points on an urban water supply network, efficiently and accurately predicts the pressure data of each monitoring point on the network.

The invention discloses an optimal selection method and device for a favorable area of a multi-thin-layer tight gas reservoir and electronic equipment. The optimal selection method comprises the following steps: calculating an interlayer difference coefficient of each gas-bearing layer according to the pressure and permeability of each gas-bearing layer; calculating the single-layer open-flow capacity of each gas-bearing layer according to the porosity, the gas saturation, the permeability and the gas-bearing layer thickness of each gas-bearing layer; multiplying the interlayer difference coefficient of each gas-bearing layer by the single-layer open-flow capacity, and accumulating results to obtain single-well combined-layer open-flow capacity; classifying the single wells according to the single well combined layer open flow capacity, and predicting the favorable area of the multi-thin-layer tight gas reservoir according to the classification result in combination with the sand body distribution characteristics. According to the optimal selection method for the favorable area of the multi-thin-layer tight gas reservoir, interlayer difference and interference characteristics are fully considered, the pressure and the permeability are adopted as parameters, the quantitative optimal selection method for the favorable area of the multi-thin-layer laminated tight gas reservoir is provided, development benefits can be improved, and the yield can be kept stable.

The invention relates to the field of water resource management, in particular to a method for establishing an electricity-to-water coefficient prediction model, which comprises the following steps: step S100, acquiring an electricity-to-water coefficient, water consumption and an influence factor data set of the electricity-to-water coefficient obtained by field test; step S200, performing correlation analysis on the electricity-to-water coefficient and the influence factor thereof by using a Pearson correlation analysis method to obtain a high-correlation influence factor data set; and step S300, preprocessing the high-correlation impact factor data set, and then inputting the high-correlation impact factor data set into a convolutional neural network prediction model for training to obtain an electricity-to-water coefficient prediction model with relatively high prediction precision. According to the method, the Pearson correlation analysis and the convolutional neural network are combined, the prediction precision of the electricity-to-water coefficient can be remarkably improved, the method has high generalization ability, high prediction accuracy can be kept in different regions and under different conditions, the prediction efficiency is greatly improved, and the method is suitable for popularization and application. The prediction model provided by the invention can efficiently and accurately predict the coefficient of water conversion by electricity.

This invention discloses an intelligent analysis method and system for the bottom support status of caissons based on correlation features, relating to the field of caisson construction monitoring technology. The method includes: acquiring strain time-series data of the target monitoring location at the bottom of the caisson during the drainage sinking stage; preprocessing the strain time-series data and classifying and labeling the support status, constructing labeled sample data, where the support status includes both voided and non-voided states; performing various correlation analyses on the labeled sample data to mine multiple correlation information between the sample data and the support status; constructing a classification prediction model for each correlation information, and training the classification prediction model using each correlation information; deploying the trained classification prediction models to the non-drainage sinking stage, and comprehensively judging the bottom support status of the caisson during the non-drainage sinking stage based on the prediction results of each classification prediction model. This invention can achieve accurate and efficient identification of the bottom support status of caissons during the non-drainage sinking stage.

The application discloses a converted wave multiple wave prediction method and related devices, and the method comprises the following steps: establishing multiple wave prediction models corresponding to various types of converted wave multiple wave propagation paths according to the various types of converted wave multiple wave propagation paths; transforming P-wave seismic time domain data and converted wave seismic time domain data to obtain P-wave seismic frequency domain data and converted wave seismic frequency domain data; predicting converted wave multiple wave frequency domain data according to the P-wave seismic frequency domain data, the converted wave seismic frequency domain data and the multiple wave prediction models corresponding to the various types of converted wave multiple wave propagation paths; and inversely transforming the converted wave multiple wave frequency domain data to obtain converted wave multiple wave time domain data. According to the P-wave seismic frequency domain data, the converted wave seismic frequency domain data and the multiple wave prediction models corresponding to the various types of converted wave multiple wave propagation paths, the converted wave multiple wave of the complex multiple propagation paths is efficiently and accurately predicted.

The invention discloses a construction method, a prediction method and equipment of a full-flow-channel flow field prediction model of turbomachinery, and belongs to the technical field of turbomachinery. According to the method, a fluid domain of the turbomachinery is decomposed into different types of physical domains, the different types of physical domains are respectively projected into corresponding rule calculation domains, corresponding agent models are set for training, agent models of the different types of physical domains are obtained, and then a turbomachinery full-flow-channel flow field prediction model is formed; according to the method, the difference between different types of physical domains of the turbomachinery is considered, the corresponding proxy models are trained for the different types of physical domains respectively, and the continuous condition of the physical quantity of each region interface is utilized in the prediction process, so that the prediction of the full-flow-channel flow field is realized by adopting iterative calculation; and the full-flow-channel flow field of the turbomachinery can be efficiently and accurately predicted.

The invention relates to the technical field of engineering computational fluid mechanics, and provides a rapid calculation method for water-sediment two-phase flow of suspended load fine particles in a water pump, and the rapid calculation method comprises the following steps: firstly, under the nominal sediment concentration condition, defining an interaction retardation factor to represent the interaction retardation effect between water-sediment two-phase movement; determining a sand phase velocity calculation formula; then determining an interactive retardation source item, and adding the interactive retardation source item into a water phase momentum equation to reflect the momentum change of a water phase under the influence of suspended load fine particles; and finally, sequentially solving the sand phase velocity and the water phase velocity under a transient numerical calculation framework of the water pump. According to the rapid calculation method, separated solution of the concentration field and the velocity field of the water-sand two-phase flow of the suspended load fine particles in the water pump is fully achieved, the characteristics of the water-sand two-phase flow of the suspended load fine particles in the water pump can be accurately and efficiently predicted, and the aim of considering calculation precision and calculation efficiency is achieved; therefore, a new thought is provided for large-scale complex engineering calculation.

A method and device for predicting the properties of an electrolyte for a lithium battery based on deep learning, relating to the technical field of molecular property prediction, wherein the method collects a number of electrolyte molecules to form an electrolyte molecule database; performs high-throughput calculations through a first-principle calculation strategy to obtain set property data of electrolyte molecules; encodes and classifies the set property data of electrolyte molecules; divides the set property data of the coded and classified electrolyte molecules into Atom point level features and Edge key level features; constructs an electrolyte molecule property prediction model based on a GCN model and a feedforward neural network FFN, and trains it; inputs the Atom point level features and Edge key level features of the electrolyte molecules into the trained electrolyte molecule property prediction model to predict the properties of the electrolyte molecules. The present invention can efficiently and accurately predict the key physical properties of the electrolyte, and provide guidance for the optimal design of the electrolyte.

The application discloses a 6G wireless communication prediction channel modeling method based on large language model fine tuning, relates to the technical field of channel prediction, and comprises the following steps: processing channel measurement data, matching corresponding text data, dividing a training set and a test set, and constructing a channel prediction data set; designing a channel encoder and a dual-domain fusion module to extract channel features, and designing a text-driven encoder to extract text features; fine tuning a large language model by using the extracted channel and text features, enhancing multi-modal perception and transfer learning capability; designing a fine tuning module, fine tuning the output of the large language model, and projecting to predicted future channel state information; designing an angle consistency loss function, training a prediction algorithm based on large language model fine tuning in combination with prediction loss, and obtaining a trained network architecture; and iteratively predicting the space-time domain channel state by using the trained network, and outputting a channel prediction result. The system has high-precision prediction performance, and has outstanding practical value and popularization prospect.