51 results about "Minimization algorithm" patented technology

The invention is suitable for the technical field of life evaluation, and provides an enhanced high-voltage circuit breaker life evaluation method, which comprises the following steps: constructing a dynamic evolution model of a contact material wear rate; constructing a nozzle degradation dynamic prediction model; constructing residual life probability distribution of the insulating material; according to the dynamic evolution model of the wear rate of the contact material, the dynamic prediction model of nozzle degradation and the residual life probability distribution of the insulating material, constructing a comprehensive life evaluation index; inputting the comprehensive life evaluation index into a preset layered prediction architecture for evaluation; wherein the first layer generates basic life distribution through a Bayesian network, the second layer outputs posterior life distribution correction parameters through a convolutional neural network, and the third layer optimizes posterior distribution through a KL divergence minimization algorithm to obtain a probability density function and a confidence interval of residual electrical life; according to the method, the error of life evaluation can be reduced on the basis of complex working conditions.

The invention discloses a power battery real-time heat flow field regulation and control method based on edge calculation, and the method comprises the following steps: S1, collecting and preprocessing temperature data, and constructing a time sequence data set; s2; organizing a Euclidean heat data structure according to space + time; s3, performing diffusion mapping to obtain non-Euclidean manifold features; s4, constructing a thermal field graph structure model; s5, inputting the graph structure to the graph adaptive convolutional neural network model of the edge node; s6, training the model by using historical data; s7, optimizing model training by using a sharpness perception minimization algorithm; s8, predicting future heat flow distribution in real time; s9, generating a thermal management strategy instruction set; and S10, controlling an execution module to carry out real-time regulation and control. According to the invention, the accuracy of thermal anomaly prediction and the adaptive ability of the cooling strategy are significantly improved, and the method can be widely applied to an intelligent thermal management scene.

The invention discloses a traveling wave fault location improvement method for power distribution network fault location analysis, relates to the technical field of power distribution networks, and solves the problems that a traditional single wave velocity model causes location deviation, and particularly, false fault points are easily generated at the joint of a cable and an overhead line due to refraction and reflection. And calculating an amplitude difference mean value under different translation amounts, automatically identifying an optimal time alignment point, and locking an association time difference t1. According to the process, time errors caused by signal distortion, asynchronous sampling and the like can be effectively corrected, the time difference precision is improved from the microsecond level to the nanosecond level, a foundation is laid for follow-up precise positioning, a plurality of correlation points are randomly selected in a locked range interval, a corrected waveform is constructed based on the attenuation proportion and repeated verification is carried out, and the time difference precision is improved. Screening an optimal solution through a feature mean value minimization algorithm; the mechanism can effectively inhibit the error of a single measuring point, and improves the robustness of a positioning result through the statistical average effect.

The invention provides a method for assisting a movable antenna to realize secure communication by STAR-RIS, and relates to the technical field of wireless communication. According to the method, the safety and the rate of the system are maximized by jointly optimizing the beam forming of the base station, the position of the movable antenna and the STAR-RIS coefficient matrix. The method specifically comprises the following steps: constructing a downlink MIMO system and establishing a joint optimization problem; decomposing the joint optimization problem into three sub-problems by adopting an alternate optimization framework; respectively utilizing a particle swarm optimization algorithm and a weighted minimum mean square error method to be combined with a Majoritization-Minimalization algorithm to solve each sub-problem, and respectively utilizing the particle swarm optimization algorithm and the weighted minimum mean square error method to be combined with the Majoritization-Minimalization algorithm to solve each sub-problem; and alternately executing related steps until system security and rate convergence to obtain a solution of the joint optimization problem. By the adoption of the method, the safety communication performance of the system is remarkably improved, full-space coverage and precise beam regulation are achieved, the algorithm convergence is high, the solving efficiency is high, and the method is suitable for a downlink system with multiple users and multiple eavesdropping nodes.

The invention relates to the technical field of artificial intelligence and energy system operation control, and discloses an artificial intelligence-based energy dynamic aid decision-making method and system, and the method comprises the steps: collecting multi-source heterogeneous data of an energy system, and constructing an observation vector; decoupling the observation vector into an environmental causal factor and an environmental noise factor of a potential space by using a variational structured causal characterization model; pruning a causal factor based on an accumulated anti-fact regret value gradient to obtain effective decision information, generating an anti-fact trajectory by applying intervention to a noise factor to construct an expanded belief state set, and updating a control strategy by using a continuous anti-fact regret minimization algorithm on this basis; and finally, performing physical constraint verification on the sampling action and executing the physical constraint verification. Through causal decoupling and an anti-factual reasoning mechanism, decision robustness, interpretability and physical safety of an energy system in a dynamic non-stationary environment are effectively improved.

The invention provides a communication perception fusion waveform design method based on space-time constructive interference, which belongs to the field of wireless communication and comprises the following steps: establishing a downlink multi-user communication perception fusion system model; establishing a communication model and a sensing model; establishing an optimization problem for minimizing the transmitting power, wherein the optimization problem has a space-time constructive interference constraint condition and a CRB constraint; after the CRB constraint of the optimization problem is relaxed into linear constraint by using a continuous convex approximation technology, a second optimization problem is obtained through iterative solution, the second optimization problem is simplified through a Lagrangian multiplier method and a Carush-Kuhn-Tuck condition, and the simplified second optimization problem is solved through iterative solution of a main optimal minimization algorithm; and obtaining a designed transmitted waveform vector. According to the invention, the symbol error rate of transmission and the transmitting power of the transmitting end can be effectively reduced under the condition that the communication and sensing performance requirements are met.

The present invention provides a method for selecting cache nodes in an Internet of Vehicles (IoV) under complex scenarios, and belongs to the technical field of IoV. It solves the technical problems of communication interruption and cache node load imbalance caused by obstacles blocking the link and multiple requests from a single vehicle. Its technical solution is: including the following steps: S1: constructing three models; S2: proposing a link quality assessment algorithm; S3: based on the file load upper limit of the cache vehicle, dynamically allocating file requests, and proposing a cache node load balancing algorithm; S4: preferentially selecting cache nodes and task nodes, and proposing a cache node minimization algorithm; S5: giving a cache node selection algorithm process under complex scenarios. The beneficial effects of the present invention are: the present invention considers the maximum number of requests that can be served by the cache vehicle as a load constraint, dynamically allocates file requests, and achieves load balancing.

The present application provides a method and system for calculating three-dimensional strain of deformed objects, comprising the following steps: for a set of input three-dimensional images of the object after strain deformation, using a zero-normalized cross-correlation function to calculate the pixel displacement and deformation gradient of all pixels in the three-dimensional image after deformation; using an alternating minimization algorithm to decompose the problem into two sub-problems and define an augmented Lagrangian function; solving the first sub-problem to obtain a solution to the first sub-problem; solving the second sub-problem to obtain a solution to the second sub-problem; updating the dual variables in the Lagrangian function based on the solutions to the two sub-problems; iteratively calculating the Lagrangian function after the updated variables using alternating minimization, stopping the iteration when a preset iterative stopping condition is met, and obtaining the three-dimensional strain random field of the object's strain deformation as the global displacement field of the last iteration. The high-efficiency three-dimensional strain calculation method for deformed objects provided by the present application has the advantages of higher computational efficiency, strong noise resistance, and fast convergence speed.

The application discloses a 5G communication network data dynamic privacy publishing method and system based on reinforcement learning, and belongs to the technical field of information security. The steps of the application are as follows: 1, determining the interval of the data to be published histogram; 2, using a data stream sampling method based on a sliding window to sample the data in the sliding window at the current time to obtain a sampling set M; 3, obtaining the approximate statistical result and the noisy statistical result of all intervals in the sliding window at the current time; then, according to the obtained result, using an error minimization algorithm based on the comparison between the approximation and the noise to obtain a better result; 4, using a grouping noise adding method based on reinforcement learning to obtain a locally optimal grouping result for noise adding, and obtaining the finally publishable noisy histogram data. When the application is used for the differential privacy histogram publishing of the sliding window data stream, the application can quickly and accurately publish the data meeting the user demand, improves the data availability and reduces the privacy budget.

A modelling method for making a virtual model of a user's head includes operations of: acquiring a plurality of depth images of a user's head in different head poses; processing the acquired depth images; localising on the two-dimensional images first marker points corresponding to characteristics of the head; estimating the head by estimating the yaw, pitch and roll angles of the head and the positioning of the head along three axes of a reference system of Cartesian axes for each of the two-dimensional images of the depth images, performing a transformation of the head pose; projecting the first marker points located in the respective cloud of points; associating iteratively in pairs the clouds of points until obtaining a cloud of synthesis points; aligning the cloud of synthesis points; and deforming the statistical head model by means of a minimization algorithm of a predefined cost function.

The application discloses an evolution modeling and prediction method and system for extreme risk coupling of an aviation hub network, and the method comprises the following steps: acquiring static section data of an aviation transportation system; constructing a space coupling dynamics model according to the static section data, wherein each airport node in the space coupling dynamics model corresponds to an attraction weight, and the attraction weight represents the co-mobility characteristics of each airport node under an extreme load state; using a maximization minimization algorithm to inverse the attraction weight; based on the inverted attraction weight, acquiring a space core parameter group through a Markov chain Monte Carlo algorithm, and obtaining regional congestion co-mobility early warning information based on the space core parameter group. The application can accurately simulate the dynamic growth logic of a multi-center airport group, and can more quantitatively identify the cascade load pressure distribution within the space dependence under the extreme scenario of the failure of a specific core hub.

The invention provides a method and a system for eliminating a pulsating pressure interference peak in a pump source pipeline. A sparse interference peak filtering model for a pulsating pressure amplitude logarithmic spectrum is established by using a zero-phase filtering principle and a total variation filtering method, and then a pulsating pressure amplitude spectrum effective signal iterative calculation method is formed through a main minimization algorithm. The effective signal of the pulsating pressure can be extracted from the amplitude spectrum containing the sparse interference peak through iterative calculation, and the accuracy of a pulsating pressure signal measurement result is improved. In order to solve the problem of signal distortion of a result obtained by a total variation filtering model, the invention proposes to carry out secondary processing on the pulsating pressure obtained by the filtering model by using a moving least square method, so that the divergence problem of the extraction result of the filtering model is solved, and the reliability of a pulsating pressure filtering algorithm is enhanced.

This invention discloses a method for 3D modeling of small celestial bodies based on SFS (Self-Focused Spectroscopic Method), belonging to the field of intelligent 3D modeling and image reconstruction technology for small celestial bodies. To improve the accuracy of 3D modeling of small celestial bodies, this invention performs modeling under the condition that the camera's extrinsic parameters are determined. This includes setting the positions of the camera and the small celestial body in a simulation environment and setting the camera's motion trajectory to obtain multiple frames of small celestial body observation images; extracting the contours from the small celestial body observation images; transforming the 3D modeling problem of small celestial bodies into a pseudo-Boolean function minimization problem; applying local minimization and hierarchical minimization algorithms to solve for the minimum value of the pseudo-Boolean function to obtain a 3D voxel model of the small celestial body. For modeling under the condition that the camera's extrinsic parameters are uncertain, this includes converting the 3D voxel model of the small celestial body into a polygon model; selecting 3D model points; matching the 3D model points with 2D image points; and applying the LM algorithm to optimize the matching error between the 3D model points and 2D image points. This patented method is applied to the process of small celestial body detection.

The invention discloses a reconfigurable antenna pattern design method and system based on zero-forcing precoding, and belongs to the technical field of wireless communication, and the method comprises the steps: building a non-convex optimization model for maximizing a user signal-to-noise ratio in a single mode through employing the user receiving signal-to-noise ratio under zero-forcing precoding; solving the non-convex optimization model for maximizing the signal-to-noise ratio of the user in the single mode through a continuous convex approximation SCA algorithm, and designing a related convex problem by using first-order Taylor expansion to realize iterative solution of the non-convex optimization model; and designing a new proxy function by using a dominating-minimizing MM algorithm to simplify the solving process of the non-convex optimization model, and calculating to obtain a reconfigurable antenna pattern design scheme based on zero-forcing precoding. According to the method, the convergence speed of the model is high, the time complexity of the solving process is low, the efficiency of solving the single-mode design problem of the directional diagram reconfigurable system under zero-forcing precoding is improved, and application of the directional diagram reconfigurable antenna in a single-mode scene can be promoted.

The present invention proposes a novel image segmentation method based on mixed offset field correction. The steps are as follows: constructing a mixed offset field image decomposition model based on the input image to obtain the image's multiplicative offset field and additive offset field, coupling the image decomposition theory to a piecewise smooth image segmentation model; adding regularization terms of the multiplicative offset field and additive offset field, as well as constraints of the multiplicative offset field, to construct a novel image segmentation model, and using the Lagrange multiplier method to transform the image segmentation model into an unconstrained optimization problem; and utilizing an alternating minimization algorithm to iteratively solve each subproblem of the unconstrained optimization problem to obtain the final segmentation result. By considering the image's multiplicative offset field and additive offset field, the present invention can not only effectively remove the offset field present in the image, but also accurately segment natural and medical images with uneven intensity, demonstrating robustness to various image segmentation tasks.

The invention relates to the technical field of radar observation, and discloses a radar observation intelligent collaborative assimilation method, which comprises the steps of constructing a Python-Fortran hybrid programming platform, commanding a weather forecast mode simulation environment, generating multi-scale background field data, and enabling the programming platform and a weather forecast mode to share data mutually. A variational auto-encoder and a decoder are used to realize rapid mutual conversion between a physical space and a low-dimensional hidden space, the encoder realizes data dimension compression through a neural network, and the decoder realizes data physical reconstruction through the neural network; in the hidden space, radar observation is fused into a mode background by using a minimization algorithm to realize multivariable nonlinear assimilation. The physical environment state after assimilation is output to form a continuous time sequence which can be used for three-dimensional environment monitoring service, and an assimilation result at the latest moment is used as an initial state of a weather forecasting mode, so that the weather subjected to numerical simulation naturally expands to the future in a time domain, and forecasting service can be formed. According to the method, the efficiency and precision of three-dimensional environment monitoring and forecasting are improved.

The invention provides a self-anti-S box construction method and system based on m-sequence generation and logic optimization, and belongs to the technical field of cryptography and hardware security. The method comprises the following steps: firstly, constructing an n-level linear feedback shift register according to a set S box number n and a primitive polynomial, generating an m sequence in a non-all-zero initial state, and adding an all-zero state at the tail; a self-reverse S box lookup table is constructed through a symmetric mapping strategy, so that the ith element in the sequence and the element at the symmetric position form mapping, and the same S box is used for encryption and decryption. Thirdly, the lookup table serves as a Boolean function truth table, is simplified into a simplest expression through a logic minimization algorithm and is decomposed into a layered logic circuit; during actual processing, N data blocks are converted into a bit plane format from block type storage, then parallel operation is carried out through a hierarchical circuit, and finally a result is transposed back to block type storage. According to the method, the same S box hardware module can be reused in the encryption and decryption process, and hardware resources are remarkably saved.

The invention belongs to the technical field of circuit fault diagnosis, and particularly relates to a solving method and device based on multiple observation aggregation fault diagnosis and a medium. In order to solve the problems of one-sided single observation information, redundant diagnosis results and the like, multiple times of fault observation are performed on a circuit system, and multiple independent diagnosis results are generated and integrated into a diagnosis set; establishing a mapping relation between a diagnosis set and a unique keyword by using a map function of the STL, and converting the component combination into an integer identifier to simplify operation; calling a BWSS algorithm to perform collision calculation on the mapped keywords to obtain a candidate solution set covering all observation scenes; after the keyword solution set is converted to an original diagnosis form, deleting a superset according to a redundancy rule to form a simplified intermediate diagnosis solution set; and finally, through a branch and bound minimization algorithm, according to potential classification and ascending order arrangement of the diagnosis sets, high-order supersets are deleted layer by layer by adopting a heuristic strategy, and finally, a minimum diagnosis solution set of all faults is output.

A method is proposed of using low-resolution images of at least one product produced by one or more imaging processes, and imaging models characterizing the imaging processes, to determine values for plurality of numerical parameters which collectively define a product model of the at least one product. The determination of the values is performed by forming a loss function based on the acquired images, the imaging models, and the numerical parameters of the model, and performing a minimization algorithm to minimize the loss function with respect to the numerical parameters. Due to prior knowledge of the product encoded in the loss function, the product model may comprise reconstructed images which have a higher resolution than the low-resolution images.

The present invention discloses a single-bit direction-of-arrival estimation method, system, terminal and storage medium with signal amplitude recovery. The method includes: obtaining signals of each snapshot moment of a signal source from a uniform linear array receiver, introducing a non-zero threshold to perform single-bit quantization on the signals of each snapshot moment to obtain a signal model; reconstructing the signal model, using a smooth hyperbolic tangent function to replace the sign function of the reconstructed signal model and performing reconstruction to obtain a target signal model; based on the target signal model, using the truncated Frobenius norm and the L 2,0 norm to construct a single-bit DOA estimation optimization problem, and using the semi-quadratic technique to convert the optimization problem into a target optimization problem; using the proximal alternating minimization algorithm and the forward-backward splitting algorithm to solve the target optimization problem to obtain the estimated direction of the signal source, the estimated number of signal sources and the estimated amplitude of the signal. The present invention eliminates the prior dependence on the number of signal sources and accurately estimates the amplitude of the signal source.

This invention discloses a semi-blind OFDM receiving method that combines channel estimation, phase noise compensation, and signal detection. The method first constructs a constrained objective function for joint channel estimation, phase noise compensation, and signal detection based on phase noise modulo-1 constraints and the mapping constraint of the signal to be detected. Then, given the phase noise and the signal to be detected, the optimal channel estimate is obtained, and this estimate is brought back to the original optimization objective to obtain an updated objective function. Subsequently, by introducing a penalty term, a penalty factor, and auxiliary variables, an augmented Lagrangian function of the optimization objective is constructed, and the corresponding optimization problem is solved using the cross-direction multiplier method (ADMM) and the minimization-maximization (MM) algorithm. When the iteration termination condition is met, the joint estimate of the channel, phase noise, and transmitted signal is obtained. This invention ensures optimal performance from the optimization objective and can effectively save pilot overhead, thereby improving spectral efficiency.

The invention discloses a small celestial body three-dimensional modeling method based on SFS, and belongs to the technical field of small celestial body intelligent three-dimensional modeling and image reconstruction. In order to improve the three-dimensional modeling precision of the small celestial body, the modeling under the condition that the external parameters of the camera are determined comprises the following steps: in a simulation environment, setting the positions of the camera and the small celestial body and setting the motion trail of the camera so as to obtain a plurality of frames of small celestial body observation images, and carrying out contour extraction on the small celestial body observation images; converting a three-dimensional modeling problem of the small celestial body into a minimization problem of a pseudo Boolean function; and solving the minimum value of the pseudo Boolean function by applying a local minimization algorithm and a hierarchical minimization algorithm to obtain a three-dimensional voxel model of the small celestial body. Modeling is carried out under the condition that the external parameters of the camera are uncertain, wherein a three-dimensional voxel model of the small celestial body is converted into a polygon model; selecting three-dimensional model points; matching the three-dimensional model points with the two-dimensional image points; and optimizing a matching error between the three-dimensional model point and the two-dimensional image point by using an LM algorithm. The method is applied to a small celestial body detection process.

The invention belongs to the field of laser welding numerical simulation, and discloses a laser welding numerical model high-precision efficient calculation method and system based on a physical information neural network. Generating model basic training data based on the laser welding multiphase flow numerical model; a 3D ConvLSTM-based laser welding physical field prediction model is constructed, energy input information is introduced into a hidden layer, and the randomness of the laser welding process is restored; designing a loss function containing various error sources; adjusting the network weight by using a gradient minimization algorithm, and minimizing a loss function; and inputting the three-dimensional physical field data of T moments to predict the change of subsequent T moments, extracting the data of (T-T) moments after the input data and the output data to reform a model for input, and predicting the physical field change of the subsequent moments. The invention provides a high-precision and high-efficiency novel solving method for laser welding numerical model calculation.

The application discloses an industrial surface defect detection method and system based on sampling disturbance optimization and history reuse, and relates to the technical field of industrial visual defect detection. In view of the problems that the background of an industrial defect image is complex, the loss topography is highly non-convex, which leads to model overfitting, and the traditional sharpness perception minimization algorithm relies on double back propagation, which leads to excessive algorithm cost, a lightweight optimization method is provided. The method introduces a sampling sharpness estimation mechanism in the network parameter update stage, generates a disturbance tensor by sampling within the preset norm boundary to explore the flat minimum value, and introduces a history reuse mechanism to dynamically evaluate and reuse the historical state, so as to bypass the back propagation calculation of the current calculation graph in part of the training period. With less memory cost, the application greatly reduces the disturbance calculation cost, effectively crosses the non-convex loss valley, significantly improves the generalization accuracy of industrial surface defect detection, and has a wide application prospect.

The present application relates to the technical field of vehicle control, and discloses a demonstration-assisted reinforcement learning suspension control method, system and storage medium, the method comprising: initializing a control strategy using a clonal imitation learning algorithm according to a demonstration sample; generating a set of state transition quadruplets according to a preset deep deterministic policy gradient algorithm; learning an evaluator network using a Bellman error minimization algorithm; learning an actor network according to a strategy parameter that maximizes an estimated value of an action-value function; storing the state transition quadruplets in a first replay cache; sampling multiple state transition quadruplets from the first replay cache, and updating the evaluator network and the actor network; obtaining a value function loss according to the mean square error between the value function estimate of the current state and the target; using the minimum value of the difference between the current control strategy and the demonstration sample generator as the strategy loss; and performing stochastic gradient descent to optimize the parameters of the control strategy according to the total loss to improve the efficiency of training samples.

The invention relates to the technical field of vehicle control, and discloses a reinforcement learning suspension control method and system based on demonstration assistance and a storage medium, and the method comprises the steps: carrying out the initialization of a control strategy through employing a clone imitation learning algorithm according to a demonstration sample; generating a state transition tetrad set according to a preset depth deterministic strategy gradient algorithm; learning an evaluator network by using a Bellman error minimization algorithm; learning an actor network according to an estimated value of the policy parameter maximization action-value function; the state transition tetrad is stored in a first playback cache; sampling a plurality of state transition tetrads from the first playback buffer, and updating the evaluator network and the actor network; obtaining value function loss according to a mean square error between the value function estimated value of the current state and the target; taking the minimum value of the difference value between the current control strategy and the demonstration sample generator as strategy loss; and executing parameters of a stochastic gradient descent optimization control strategy according to the total loss, and improving the training sample efficiency.

A method for constitutive response model calibration employs a neural network-based proxy model in place of the output from a corresponding finite element simulation. When the application program receives user selection, the constitutive response model is calibrated, user selection comprises test data output from an experiment, a finite element model for simulating the experiment, constitutive response model selection, historical output quantity, a numerical value minimization algorithm, error measurement, constitutive response model parameters, error measurement of test data relative to finite element simulation output, constitutive response model parameters and a parameter space sampling technology. Architecture setting of an architecture of the proxy model, and a training algorithm for a proxy model training algorithm.

This invention discloses a semi-blind direct localization method for thermal noise communication systems, comprising: based on the frequency domain received signal model of the thermal noise communication system, utilizing the statistical characteristics of the received signal to construct a maximum likelihood estimation problem concerning the mobile station position and channel fading coefficient, obtaining an objective function containing logarithmic determinant terms and trace function terms of a matrix; the objective function simultaneously includes a contribution from pilot power and an expected contribution from the power of the received data signal; using an optimization minimization algorithm framework, alternately solving for the mobile station position and channel fading coefficient until a preset convergence condition is met, obtaining the localization result; wherein, by constructing a substitution function containing the statistical characteristics of blind pilots and data, a closed-form solution for the channel parameters is derived in each iteration, and the mobile station position is updated through a two-dimensional grid search. This invention mines the statistical information of the data portion, effectively improving the utilization rate of the received signal and increasing the localization accuracy under low signal-to-noise ratio conditions.

The invention provides an automatic driving key scene generation method and system based on a nested game algorithm, relates to the technical field of automatic driving, and aims to solve the problems that a single game model is adopted in the prior art, strategy updating is not clear, calculation complexity is high, the convergence speed is low, and interpretability is poor. The method comprises the following steps: constructing an automatic driving antagonism test environment, respectively constructing game models of two stages for vehicle driving, determining a game player set, an action space, a revenue function and a constraint condition, and performing optimization test on the automatic driving antagonism test environment by using a nested anti-fact regret minimization algorithm. And obtaining an optimized adversarial test environment, performing a simulation test on the optimized adversarial test environment in a simulation environment, and generating a key safety scene for testing the safety of automatic driving. The problems in the prior art are solved, the representativeness and diversity of the generated scene are improved, and the efficiency and accuracy of automatic driving safety verification are improved.

The invention discloses a high mobility scene channel estimation and prediction method based on a channel map, and relates to the field of wireless communication system research, and the method comprises the steps: building channel key parameter prior distribution based on a power spectrum density function provided by the channel map; constructing and converting a pilot frequency training model considering inter-carrier interference into a parameterized linear estimation problem; and iteratively solving the problem by using a maximization-minimization algorithm, and obtaining a current channel estimation value and a subsequent channel prediction value based on a channel multipath structure. Compared with an existing method, the method has the advantages that the channel estimation precision can be improved by 10 times, high-precision time domain channel prediction can be realized, and a high-precision channel estimation and prediction solution is provided for high-mobility scenes such as high-speed railways and unmanned aerial vehicles. Instantaneous channel state information estimation of a double-selective channel can be realized, and time domain channel prediction in a quasi-static time interval can be realized.