95 results about "Bayes' theorem" patented technology

The invention discloses a multi-source environment monitoring data fusion method based on adaptive Bayesian reasoning. The method comprises the following steps: S1, collecting an environment monitoring data set; s2, obtaining a preprocessed environment monitoring data set; s3, carrying out uncertainty evaluation on the preprocessed environment monitoring data set; s4, forming an initial prior probability; s5, calculating the posterior probability distribution of each environment monitoring data source through the Bayesian theorem, and dynamically updating the weight of each data source in the environment monitoring data fusion process according to the posterior probability distribution; s6, performing weighted fusion on each environment monitoring data source according to the dynamically updated weight to form fused environment monitoring data; and S7, carrying out post-processing on the fused environment monitoring data. According to the invention, the data fusion process can adapt to environmental changes in real time, so that the accuracy of data fusion is improved.

The invention discloses a bridge network anti-seismic performance rapid evaluation method based on capability spectrum analysis. The method comprises the following steps: firstly, carrying out finite element modeling on bridges in a regional bridge network needing to be analyzed by using OpenSees, extracting piers, and carrying out vulnerability analysis on the piers; a bridge vulnerability curve is obtained based on a capacity spectrum method, then the failure probability of the bridge in a certain damage state serves as a post-earthquake damage index, and the bridge failure probability is replaced by the post-earthquake failure probability of the corresponding road section; after the post-earthquake failure probability of each road section exists, all paths between a starting point and an ending point, namely, to-be-researched OD pairs in a network area are analyzed, the passing probability of each path after the earthquake is calculated based on the Bayesian theorem and an event independence calculation method, all the paths under the OD pairs are sorted according to the passing probability, and therefore the optimal passing path is obtained; and the path with small passing probability is avoided. According to the invention, a scientific basis is provided for regional bridge network anti-seismic planning and emergency management.

The invention discloses an online intelligent monitoring and analyzing system for wear of a circuit breaker contact, which relates to the field of contact wear monitoring and comprises a visual module, an acoustic module, a current module, an acquisition module, an analysis module, an alarm module, a processing module and a display module. The method comprises the following steps of: analyzing and establishing a quantum driving feature selection model; constructing a contact motion mechanical control model; setting parameter prior distribution; fusing a likelihood function and prior by utilizing a Bayesian theorem; a vibration chaos non-linear loss value is calculated based on recurrence plot line segment probability distribution analysis, a jump diffusion model is constructed, the number of jumps is counted to calculate Poisson intensity, corresponding display and alarm are carried out, the accuracy of contact abrasion state evaluation is guaranteed, and early abrasion abnormity is found in time.

The invention relates to the technical field of accident responsibility judgment, solves the problem that intelligent driving accident analysis and real-time responsibility judgment cannot be carried out in the prior art, and provides an accident responsibility judgment method, device and equipment in an intelligent driving scene and a storage medium. The method comprises the following steps: analyzing related parameters of an accident in an intelligent driving scene, and determining a plurality of key variables in the related parameters and a causal network corresponding to each key variable; according to a pre-collected traffic database and the causal network, the occurrence probability of each key variable is quantified, and a prior knowledge base related to an intelligent driving accident is determined; and according to the real-time data of the intelligent driving accident and the prior knowledge base, in combination with the Bayesian theorem, accident causes are analyzed, and an accident responsible party is determined. According to the method, man-machine responsibility division of the intelligent driving accident can be realized, and the emergency processing capability and the judgment efficiency of the intelligent driving system after the accident are remarkably improved.

The invention discloses a Bayesian traffic density estimation method based on data driving in a mixed traffic environment, and the method comprises the steps: S1, building a probability relation model between traffic density and vehicle travel time, and between a vehicle and a front vehicle space-time region area based on measurement data; s2, setting prior probability distribution for the vehicle travel time and the model noise variance; s3, calculating joint posteriori distribution of the vehicle travel time and the model noise variance by using the Bayesian theorem, and performing marginalization processing to obtain marginal posteriori distribution of the vehicle travel time; and S4, probability estimation is carried out on the traffic density of the unobserved space-time region through posterior prediction distribution. The technical problem of traffic density estimation in the mixed traffic environment is effectively solved by establishing a complete Bayesian probability framework, and the estimation precision is remarkably improved by adopting the steps of prior distribution setting, posterior distribution calculation, marginalization processing and the like based on CAV dynamic data acquisition.

The invention discloses an elastomer layer carbon fiber composite material damage positioning method based on elliptic probability fusion, which comprises the following steps: taking different sensors as excitation sources, and collecting Lamb wave signals of a plurality of sensing paths; performing continuous wavelet transform by using a complex Morlet wavelet, and extracting actually measured flight time; introducing a layered wave velocity correction model to optimize an elliptical orbit method, and calculating theoretical flight time and damage probability; introducing a dynamic short-axis optimization model to optimize a probability weighting method, and calculating a damage probability; generating prior distribution by using results calculated by an elliptical orbit method and a probability weighting method; constructing a likelihood function based on the difference between the actually measured flight time and the theoretical flight time; the posterior distribution of the damage position is obtained through the Bayesian theorem in combination with the prior distribution and the likelihood function; sampling the posterior distribution by using an adaptive MCMC algorithm, and generating a probability cloud picture of the damage position; and outputting a final imaging result of the damage position, and positioning the damage position.

The invention provides a biomedical causal relationship inference method and system based on a Bayesian knowledge graph, and relates to the technical field of biomedical data mining and artificial intelligence, and the method comprises the steps: carrying out the multi-source evidence fusion of biomedical data, and obtaining a structured triple containing Bayesian confidence; analyzing the triple by using priori knowledge and obtaining a conditional probability table through parameterized filling; carrying out posteriori updating by using the Bayesian theorem; and analyzing the updated knowledge graph state by using a graph neural network model to obtain an inference result. Wherein the Bayesian inference module is combined with the graph neural network model, the former provides priori knowledge with confidence, the latter provides a fine path dependency relationship, and the accuracy and robustness of inference are remarkably improved. According to the method, the problems of evidence isomerism fragmentation, causal inference subjectization and knowledge discovery inefficiency are solved, and intelligent and automatic inference of the causal relationship is realized.

The invention discloses an intelligent segmental arc correlation method based on a minimum optical admissible domain, belongs to the field of intelligent calculation optimization of space target monitoring, solves the problem of large calculation amount of segmental arc correlation, and comprises the following steps: constructing the minimum optical admissible domain by a first observation segmental arc; uniformly sampling in the minimum optical tolerance domain, performing orbit propagation on each sampling point under a J2 perturbation model to obtain theoretical optical observation data of a second observation arc section, and calculating an angle root-mean-square error of the second observation arc section to obtain an angle loss function; using a DBSCAN clustering algorithm to identify wave troughs; selecting an initial value of a simplex method in each wave trough to obtain an angle optimal solution; a probability loss function is obtained through Bayesian theorem calculation; substituting the angle optimal solution into a probability loss function, and obtaining a probability optimal solution by using an LM algorithm; a threshold value is set according to the probability loss function obeying chi-square distribution, and when the probability optimal solution is smaller than the threshold value, the two arc segments are associated successfully; according to the invention, the calculation cost is saved, and the space target observation accuracy is improved.

The invention discloses an occupation grid map construction method based on distance weighting, and belongs to the field of intelligent vehicle digital maps and high-precision positioning. The method comprises the following steps: acquiring pose data of an intelligent vehicle and sensing data of a surrounding environment; constructing a distance weight factor; generating an occupation grid map according to the obtained pose data of the intelligent vehicle and the perception data of the surrounding environment, dividing the occupation grid map based on the Bayesian theorem and a distance weight factor, and endowing each grid with state information; and dynamically updating the occupied grid map by using a double-end linked list structure. According to the method, the distance weight factor is constructed, and the grid map is divided through the Bayesian theorem and the distance weight factor, so that the obstacle state can be accurately expressed in a dynamic non-structural environment, and the problem of track noise is avoided.

A rock burst risk dynamic assessment method based on a Bayesian method comprises the steps that a stope working face is divided into M * M regular grids, and each grid serves as an independent assessment unit; collecting geological and mining data, selecting a plurality of influence factors based on historical data and expert knowledge, and quantifying the plurality of influence factors by adopting a static evaluation model; determining weights of a plurality of influence factors by using an analytic hierarchy process, and calculating an initial risk score of each evaluation unit by using a comprehensive index method; normalizing the initial dangerousness score into prior distribution by adopting an S-type function, and constructing an initial dangerousness prior probability field; a power law attenuation model is introduced to calculate influence weights among the evaluation units, and an influence weight matrix is constructed; real-time monitoring data are obtained through a micro-seismic monitoring system, the risk probability of each evaluation unit is updated in real time based on the Bayesian theorem, and dynamic evaluation and evaluation result output are achieved. According to the method, the accuracy and the real-time performance of rock burst risk assessment can be effectively improved.

The invention discloses an SRAF placement method based on a Bayesian model. The method comprises the steps that historical SRAF configuration parameters and photoetching simulation data are collected and preprocessed; based on the preprocessed data, using kernel density estimation to construct prior probability distribution; constructing a likelihood function of multiple photoetching result indexes in combination with a Hopins photoetching model; calculating approximate distribution of a posterior probability through variation inference by using the Bayesian theorem; and finally selecting an optimal SRAF configuration parameter according to the posterior distribution, and verifying the manufacturing feasibility of the process window and the mask. According to the method, historical prior knowledge and real-time photoetching data are organically integrated through the Bayesian model, and high-precision, high-efficiency and high-robustness optimization of the SRAF layout is realized by combining specific technical characteristics such as preprocessing, kernel density estimation, a Hopkinson physical model and variation approximation; the problems of complex rule base, time-consuming calculation and insufficient adaptability in the prior art are effectively solved.

This invention relates to a method and system for optimizing the quenching of large-diameter bearings based on multi-model set membership estimation. The method includes: determining multiple state parameters of the target quenching measurement system; constructing a system model based on the fusion of multiple set membership estimation models using the state parameters and a fully symmetric multi-cell method, employing a Kalman filter; determining the predicted state equation and probability density function of each set membership estimation model; updating the probability density function of the predicted set of each set membership estimation model according to the dimension of the output matrix of the real-time quenching measurement system; calculating the predicted set of the system model using Bayes' theorem based on the probability density function of the predicted set of each set membership estimation model; and solving for the optimal gain matrix of the target quenching measurement system by minimizing the set radius based on the predicted set and error set of the system model. This invention improves the stability and robustness of the optimized control for quenching large-diameter bearings through multi-model set membership estimation and Bayes' theorem.

The present invention discloses an intelligent arc segment correlation method based on the minimum optical tolerance domain, belonging to the field of intelligent computing optimization for space target surveillance, which solves the problem of large computational volume of arc segment correlation, and includes: constructing a minimum optical tolerance domain from the first observation arc segment; uniformly sampling in the minimum optical tolerance domain, and propagating the orbit of each sampling point under the J2 perturbation model to obtain the theoretical optical observation data of the second observation arc segment and calculating the root mean square error of the angle of the second observation arc segment to obtain an angle loss function; using the DBSCAN clustering algorithm to identify the wave valleys; selecting the initial value of the simplex method in each wave valley to obtain the optimal angle solution; calculating the probability loss function by the Bayesian theorem; substituting the optimal angle solution into the probability loss function and using the LM algorithm to obtain the optimal probability solution; setting a threshold according to the chi-square distribution of the probability loss function, and when the optimal probability solution is less than the threshold, the two arc segments are successfully correlated; the present invention saves computational costs and improves the accuracy of space target observation.

The invention provides a user satisfaction prediction method and device and a medium, and relates to the technical field of computers. The method comprises the following steps: based on a first index, using a user satisfaction model obtained by training according to user complaint data in advance to obtain a first probability that a first user is not satisfied; based on a second index, obtaining a second user who has a poor quality event in the first users, and correcting the first probability of the second user by using the Bayesian theorem to obtain a second probability that the second user is not satisfied; and predicting an unsatisfied fourth user according to the first probability of the third user who does not generate the poor-quality event in the first user and the second probability of the second user. According to the method, the probability that the user is not satisfied according to the user complaint prediction is corrected through the Bayesian theorem according to the poor-quality event, so that user satisfaction evaluation errors caused by the fact that some users are not satisfied with commodities or services actually but do not complaint are avoided.

The invention discloses a tourism investment project risk assessment method and early warning system, and the method comprises the steps: S1, determining tourism investment project risk assessment indexes which comprise external risks, investment subject internal risks and project engineering risks; s2, based on the Bayesian theorem, constructing a risk assessment model used for calculating the occurrence probability of each risk index and the influence on the overall risk of the project, defining a prior probability for each risk index according to historical data or expert experience, continuously collecting latest data related to the risk assessment indexes, and according to the newly collected data, establishing a risk assessment model; using a Bayesian algorithm to update the posterior probability of each risk index; s3, comprehensively evaluating the posterior probability of each risk index, and calculating the overall risk level of the project; according to the method, the external risk, the internal risk of the investment subject and the project engineering risk are covered by defining the risk assessment indexes, and the comprehensive coverage of the risk points is ensured, so that the problem of incomplete risk assessment caused by missing important risk points can be reduced.

The invention provides a method for establishing a high-dimensional feature posterior probability estimator based on a sparse Bayesian method. The method comprises the following steps: (1) constructing a sparse Bayesian posterior probability estimation model; (2) model parameter representation based on the Bayesian theorem; (3) performing model parameter solving by using the sample data; and (4) realizing model sparsification based on automatic correlation confirmation. The invention designs a high-dimensional feature posterior probability estimator by using a sparse Bayesian method in order to solve the problem that the feature category affiliation probability is difficult to solve due to the fact that the high-dimensional feature probability distribution is difficult to estimate.

A multi-source heterogeneous data driven large rotating machinery fault diagnosis method, which first collects one-dimensional data composed of low-frequency vibration signals, medium-frequency vibration signals and temperature signals, and two-dimensional data composed of operation images at the same time, then fuses the one-dimensional data and two-dimensional data by Bayes theorem to obtain a diagnosis result, at the same time, obtains a diagnosis result B according to the low-frequency vibration signals and medium-frequency vibration signals, and obtains a diagnosis result C according to the temperature signals, then sums up the diagnosis results A, B and C according to the weighted summation method to obtain the final diagnosis result, if the final diagnosis result is greater than or equal to 0.5, it is judged that there is a fault, if it is less than 0.5, it is judged that there is no fault. The design not only monitors multiple signals, but also has good fault diagnosis effect.

The present invention provides a method for disambiguating personal names based on multi-relationship deep retrieval text matching, which relates to the technical field of personal name disambiguation. The present invention obtains cross-modal data related to enterprises and people, performs data alignment and data fusion on the cross-modal data through an entity alignment algorithm to form a person-enterprise structured data set, generates semantic vectors by establishing a multi-relationship deep retrieval model based on a pre-trained language model, calculates semantic similarity according to the semantic vectors, and generates a personal embedding vector according to the co-occurrence frequency and spatio-temporal correlation features calculated from the person-enterprise structured data set. An anti-disambiguation recognition model is established based on an adversarial neural network, personal name disambiguation is performed according to the personal embedding vector, a structure update model is established through a graph attention network to update the graph structure in real time, the confidence levels of semantic similarity and co-occurrence frequency are calculated according to Bayes' theorem, and the weights of semantic similarity and co-occurrence frequency are updated according to the confidence levels.

This invention relates to the field of seismic data processing technology, specifically disclosing a method for removing strong reflection shielding in hidden river channels based on adaptive hybrid L0-L1 norm. The method includes: first, establishing an optimization objective function based on Bayes' theorem using hybrid L0-L1 norm; second, dynamically adjusting the L0-L1 norm weights using an adaptive weight function driven by the seismic signal, combined with reflection coefficient amplitude and residual information, enhancing sparsity constraints in strong reflection zones and reducing constraint strength in weak reflection zones; then, constructing a convex upper bound for the objective function using a minimization framework, and solving it iteratively in stages using an accelerated rapid iterative threshold shrinkage algorithm, while incorporating prior knowledge of seismic wave propagation laws and river channel deposition patterns to ensure the geological rationality of the solution. This invention solves the technical problems of traditional sparse processing methods, such as fixed parameters, lack of geological constraints, and inability to simultaneously address strong reflection suppression and weak signal protection, significantly improving the separation accuracy of strong reflections and the recovery rate of weak signals, effectively overcoming the strong reflection shielding effect.

The invention discloses a heading machine hob abnormal state early warning method and device. The method comprises the steps that real-time heading parameter data, hob vibration data, hob infrared video data, hob temperature data and rock slag image data are acquired; acquiring historical tunneling parameter data according to a specified mileage, inputting the historical tunneling parameter data and the real-time tunneling parameter data into a first hob damage early warning model, and outputting a first early warning result; inputting the hob vibration data into a second hob damage early warning model, and outputting a second early warning result; inputting the hob infrared video data and the hob temperature data into a third hob damage early warning model, and outputting a third early warning result; inputting the rock slag image data into a fourth hob damage early warning model, and outputting a fourth early warning result; and based on the Bayesian theorem, fusing the first early warning result, the second early warning result, the third early warning result and the fourth early warning result to obtain a hob state early warning result. The accuracy of early warning of the abnormal state of the hob can be improved.

The invention discloses a Bayesian model averaging-based power prediction model evaluation method and a related device, and belongs to the technical field of power system prediction. The method comprises the following steps: constructing a candidate model set comprising a plurality of power prediction models; secondly, historical data are used for training all the models, the posterior model probability of each model is calculated based on the Bayesian theorem, the posterior model probability serves as the scientific weight of the model, and the goodness of fit and complexity of the model are considered in the weight at the same time; and finally, for a new prediction input, performing weighted average on the prediction distribution of each candidate model by taking the posterior probability as the weight to generate comprehensive probability prediction distribution. According to the method, the advantages and disadvantages of each candidate model are scientifically evaluated through the posterior probability, and a comprehensive and probabilistic prediction result is finally generated, so that the robustness and reliability of prediction are improved, and richer decision information is provided for power grid dispatching.

The invention relates to the technical field of refueling machine calibration, in particular to a batch right protection case intelligent filing system and method based on multi-source data analysis, and the method comprises the steps: collecting right protection case data from a multi-source channel, marking key information to construct a data set, training a model through a convolutional neural network, and extracting related information in a text; distributing a unique identifier for an entity, associating data to form a case portrait, constructing a knowledge graph, and calculating a probability through the Bayesian theorem to determine a case type; various right protection case filing rules and legal regulations are collected, classified and sorted to be used for automatically auditing case information, integrating the case information, rule judgment and prediction results, and providing a basis for generation of a filing strategy; and integrating case information conforming to case filing conditions, automatically filling a document template to generate a case filing application document, and submitting an online case filing application to a court case filing system through an AP I interface. The method covers all types of evidences such as contracts, audios and videos, chat records and the like, and avoids case filing blocking caused by evidence missing.

A meta continual-learning and inferring method uses an implementation of Bayes' theorem, and includes: calculating a likelihood of learning data for a given latent variable by a data distribution learner; performing a sequential Bayesian update and calculating a final posterior distribution of the latent variable by using prior distribution of the latent variable and the calculated likelihood by a Bayes' calculator; sampling the latent variable from the final posterior distribution; and inferring test output data based on the sampled latent variable and test input data by an inference engine, wherein respective meta parameters of a neural network of the data distribution learner, the prior distribution of the latent variable of the Bayes' calculator, and a neural network of the inference engine are trained by a meta learning.

The invention discloses a core component performance prediction method, system and device based on industrial Internet of Things, and a medium, and relates to the technical field. The method is implemented through a five-platform architecture including a user platform, a service platform, a management platform, a sensing network platform and a production object platform, multi-dimensional feature vectors of core components in an operation cycle are obtained in real time, feature alignment processing is executed, a clustering algorithm is adopted to cluster performance degradation modes of the core components, and the performance degradation modes of the core components are obtained. Prototype vectors reflecting different decline stages are generated, a knowledge graph of a performance decline evolution path is established in combination with equipment working condition labels, prediction results are calculated through an inference method of the Bayesian theorem in combination with the knowledge graph, meanwhile, relation edge weights and confidence coefficients are updated according to prediction and actual results, and performance decline of core components can be accurately predicted. The prediction accuracy and reliability are improved, the service life of equipment is prolonged, the maintenance strategy is optimized, and the non-planned downtime and the maintenance cost are reduced.

A method of diagnosing medical conditions using artificial technology, which combines elements of Ockham's Razor and modified utilization of likelihood Ratio / Bayesian Theorem. The system employs standardized smart weight to assess presenting symptoms, assigning scores to each diagnosis related sign and symptom, and setting a generalized cut-off point to confirm diagnoses and link them with evidence-based treatments based on severity of illness scores calculated by the system, as well as simplified smart weight algorithms to achieve accurate results without relying on complex sensitivity and specificity data.

The invention discloses a game script data processing method, and relates to the field of game script data processing, and the method comprises the steps: wrong data collection: employing Flume to carry out multi-source centralized collection and send the collected data to a unified destination by utilizing the feature difference of a normal log and an error log, and carrying out the preprocessing of a data source of the unified destination, useless information is removed, so that storage space occupation is reduced, and processing efficiency is improved; classification of the collected data: carrying out data classification on the collected data sources subjected to duplicate removal based on the Bayesian theorem; processing various types of data: performing targeted processing on the classified data sources by adopting a corresponding mode; and optimization of daily problems: evaluating and optimizing methods corresponding to processing of various data at regular intervals to improve the efficiency and accuracy of the methods. According to the method, the error logs in the game script are automatically collected, preprocessed and classified on the basis of a simple and easy-to-construct mode, so that the rear end can better process potential errors in the game.

The present invention relates to a method for servo control of axial force of foundation pit steel supports based on digital twin, and the steps are as follows: S1. A foundation pit engineering simulation module constructs a foundation pit deformation prediction surrogate model with key soil parameters and servo steel support axial force as inputs; S2. An Internet of Things monitoring module collects foundation pit deformation data and servo steel support axial force data through an Internet of Things monitoring system; S3. A soil parameter update module uses the foundation pit deformation prediction surrogate model in S1 as a forward calculation model in probabilistic back analysis, combines with Bayes' theorem, and updates the posterior samples of soil parameters through data assimilation technology; S4. A deformation prediction and axial force control module predicts the foundation pit deformation in the next stage and proposes an axial force servo control scheme. This method is based on digital twin technology, deep learning algorithms and data assimilation technology, analyzes the influence of the axial force set value in the servo system on the foundation pit deformation, and provides a scientific and reasonable servo control scheme for the axial force of foundation pit steel supports.

The invention provides a rapid calculation method for limit clearing time based on an improved Bayesian optimization algorithm, and relates to the field of power systems, and the method comprises the steps: obtaining original parameters of different faults of a power system, and discretizing the original parameters into an integer optimization model; the method comprises the following steps: initializing an exploration domain, sampling to generate an initial data set, constructing a Gaussian process model based on the initial data set, calculating a corresponding acquisition function to obtain a next most potential sampling point, and updating the exploration domain; adding observation data corresponding to the new sampling points into the initial data set, updating a Gaussian process model according to the Bayesian theorem, and repeatedly calculating a corresponding acquisition function until a preset convergence criterion is met or the maximum number of iterations is reached, so as to obtain an optimal model; and calculating the lower limit clearing time of each fault of the power system based on the optimal model. According to the method, the continuous optimization problem is discretized, the exploration interval is dynamically cut in the iteration process, the evaluation number is reduced, and a solution close to the global optimum is quickly found.

The invention discloses a concrete strength springback error probability calibration method based on computer vision, and the method comprises the steps: obtaining original pictures of local concrete at different positions of the surface of a component in an existing concrete building; carrying out a rebound test on the concrete at the image shooting position; core drilling sampling is conducted on the component, and the actual strength value of internal concrete is measured; calculating the error between the concrete strength rebound value and the actual strength of the drill core; training a concrete strength springback error prediction model, predicting prediction errors of concrete springback values and actual values at different positions of the surface of the to-be-detected member, and then calculating strength prediction estimation values of concrete at different local positions; and taking the strength prediction estimation values of the concrete at different positions as noise observation data of the real strength of the component, constructing a normal likelihood function, and deducing the probability distribution of the real strength of the concrete in the component through the Bayesian theorem in combination with the prior strength distribution.

The application discloses a runoff reconstruction method for a glacier-covered area without actual measurement data, relates to the technical field of geographic information and hydrology, and comprises the following steps: acquiring multi-source heterogeneous data of a target basin; extracting an area time sequence of an ice lake based on remote sensing images, constructing an area-water level-storage capacity relationship curve based on a digital elevation model, and determining an ice lake storage capacity change time sequence; constructing an ice hydrology physical model and integrating an ice lake reservoir module based on ice distribution data and meteorological reanalysis forcing data; determining a prior distribution through a cross-basin parameter migration technology, constructing a likelihood function based on the ice lake storage capacity change time sequence, determining an optimal parameter set based on Bayes theorem, and running the ice hydrology physical model to obtain a primary simulated runoff sequence; and performing deviation correction through a correction model based on the primary simulated runoff sequence, the ice lake storage capacity simulation time sequence and an environmental characteristic sequence to generate a reconstructed runoff sequence. The application can improve the accuracy and reliability of runoff reconstruction.