106 results about "Adaptive integration" patented technology

The invention discloses a road traffic flow prediction method based on a space-time mixed attention network, and the method breaks through the limitation of a conventional time sequence model and a single deep learning architecture based on the systematic analysis of urban road traffic flow space-time heterogeneity, periodic non-stationarity and road network topological relevance, constructs the space-time mixed attention network, and achieves the prediction of road traffic flow. Spatial heterogeneous correlation of road network nodes is captured through a graph convolution network, dynamic time sequence evolution characteristics of traffic flow are modeled by adopting a hybrid architecture, a residual attention mechanism is introduced to realize layer-by-layer refining of multi-scale spatio-temporal characteristics, and the overall architecture of the method has remarkable advantages in the aspects of spatial topology modeling and time dynamic capture compared with a traditional model. Feature decoupling learning is carried out on multi-source heterogeneous data such as weather and events, adaptive integration of environment sensitive features is realized through a parameterized gating fusion strategy, and the prediction error fluctuation amplitude in an extreme weather scene is reduced by 34.8%.

The invention relates to the technical field of intelligent fire fighting, in particular to a building electrical fire identification method and system based on multi-source information fusion, and the method comprises the steps: firstly collecting the multi-source monitoring data of an electrical system in real time, then carrying out the preprocessing of the multi-source data, including noise filtering, time alignment and feature extraction, and constructing a dynamic weight fusion model; the weight of each data source is adaptively allocated based on the feature entropy, then fusion features are input to the fire risk grading model, a fire probability index and an early warning level are output, and when the early warning level exceeds a threshold value, a linkage control instruction is triggered and an alarm is given; according to the method, through dynamic fusion of multi-source information, the electrical fire identification precision is remarkably improved, multi-dimensional parameters such as current, temperature, smoke and arc are adaptively integrated by adopting an entropy weight assignment mechanism, the problems of false alarm and missing alarm of traditional single-threshold detection are solved, the missing alarm rate and the false alarm rate are remarkably reduced, and the reliability of fire early warning is ensured.

The invention provides a six-degree-of-freedom flexible adaptive integrated large-scale connector automatic docking method based on vision and force feedback. According to the method, a parallel structure is used as a design basis, and the difficulty of automatic docking is solved by combining an active compliance thought of taking a force sensor as feedback and a thought of taking a passive compliance docking mechanism as mechanism passive compliance docking. A set of active and passive compliant control integrated large-scale connector automatic docking system based on combination of vision and force feedback is designed. As a research focus in the robotics field, the active compliance control technology can be deployed in numerous accurate docking application scenes depending on force feedback, the active compliance docking strategy is formulated for the design difficulty of the active compliance docking control, and the strategy is simple, fast, low in cost and high in reliability. The anti-interference performance and the reliability of the control system are greatly improved, the system is suitable for automatic docking occasions in the industrial automation field in some complex environments, and contributions are made to further research and expansion of the technology. The high-precision flexible butt joint technology is applied more and more widely nowadays with the continuous development of science and technology, compared with a traditional butt joint device, the automatic butt joint device designed by the invention is higher in assembly efficiency and precision and higher in intelligent degree, damage to all paths of plug-in devices is smaller, meanwhile, economic losses can be further reduced, and the automatic butt joint device is suitable for popularization and application. And human resource cost is saved.

The invention discloses a deep geothermal target area intelligent positioning method based on a machine learning algorithm, and relates to the technical field of geothermal resource exploration. Comprising the following steps: S1, multi-source heterogeneous data fusion acquisition and preprocessing; s2, geological feature entropy quantification and spatial autocorrelation analysis are carried out; s3, deep feature extraction and multi-modal feature fusion are carried out; s4, performing multi-model adaptive integration and dynamic weight optimization; s5, reinforcement learning dynamic adjustment participation abnormal threshold determination; and S6, performing intelligent positioning and risk assessment on the three-dimensional geothermal target area. According to the positioning technology, geological structure, geophysical field, geochemistry and remote sensing data are integrated through a machine learning algorithm, high-dimensional feature vectors are constructed, the limitation that the prior art depends on physical detection singly is solved, reinforcement learning is introduced to dynamically optimize model parameters, real-time geothermal well data updating is combined, and the positioning accuracy is improved. The model can adapt to geological condition changes, and the prediction precision is greatly improved compared with a traditional method.

The invention relates to the technical field of protein prediction, and discloses an efficient protein stability prediction method for selective state space modeling. According to the method, the BiMama core module and the CoGNN core module are adopted, and limitation of an existing method on calculation efficiency and multi-scale information processing is broken through in a mode of combining selective state space modeling and the collaborative graph neural network. According to the invention, a local sampling strategy based on a k-hop sub-graph is provided, and efficient calculation is realized by focusing a key environment around a mutation site; a bidirectional selective state space model is adopted to model a long-range dependency relationship with linear complexity, and the calculation bottleneck of a traditional Transform architecture is effectively overcome. The gating fusion module developed by the invention can adaptively integrate long and short range features, so that the model can flexibly adjust a feature combination strategy for different types of mutations, and the accuracy and efficiency of protein delta delta G prediction are improved.

The invention belongs to the technical field of industrial system integration, and provides a design method for rapid adaptive integration of a cross-domain system. Through the method, developers extract business logic rules of the industry, standardize characteristic mapping and select, develop and integrate functional modules, logic processes and scene components of the cross-domain system, so that the cross-domain system which conforms to industry characteristics and meets business circulation is formed, and meanwhile, the development efficiency of the cross-domain system is improved. The newly added industry logic component can be classified into each industry characteristic model based on the knowledge base, and a multi-dimensional characteristic vector is generated in the platform, so that cross-industry integrated component multiplexing is realized, and the cross-field applicability and flexibility of the platform are further improved. Modularized construction and configuration type development are supported, mapping and matching functions are carried out on various industrial manufacturing industry characteristics such as machining, electronic and electrical appliances and equipment sets, the requirements for subdivision production in the industrial field and rapid adaptation of business processes are met, the system integration cost is reduced, and the application development efficiency is greatly improved.

The invention discloses a power transformer periodic time sequence prediction method and a multivariable periodic time sequence prediction method based on a hybrid model, and the method comprises the steps: collecting the operation data of a transformer, including oil temperature and load variables, and carrying out the standardization, missing value processing and periodic time feature embedding; constructing a multi-branch hybrid model, fusing a bidirectional state space encoder, a time convolution network, a local window attention mechanism and a periodical enhancement network, and respectively capturing long-term periodical dependence, multi-scale local periodical features, a short-term periodical mode and periodical feature enhancement representation; the multi-branch features are adaptively integrated through a global context weighted fusion mechanism, efficient nonlinear transformation is performed on the fusion features by adopting a multi-order Chebyshev polynomial projection module, and a prediction sequence of oil temperature and load variables is generated.

The invention discloses a drug-target interaction prediction method based on double-flow collaborative attention and sparse feature fusion, and belongs to the technical field of computational biology. The method solves the problem that the existing method cannot capture the sub-structure discrimination features and the key binding region features of the drug-target interaction pair. According to the method, a double-flow collaborative attention strategy combining a multi-scale space attention mechanism and a channel enhanced attention mechanism is adopted to cooperatively capture discriminative features of substructures, the multi-scale space attention mechanism utilizes a multi-branch convolutional layer to adaptively integrate space substructure features, and molecular representation of each substructure is enhanced; the channel-enhanced attention mechanism mitigates the inconsistency of substructure features. The sparse attention mechanism can highlight the key features while suppressing the noise, and the cross attention mechanism improves the extraction capability of the features of the key combination region through feature interaction between the modeling drug and the target. The method can be applied to drug-target interaction prediction.

The invention belongs to the field of power distribution network parameter identification, and discloses a power distribution network line parameter identification method and system based on a graph neural network, and the method comprises the steps: employing a unified projection and attention weighting mechanism, obtaining the point features of a power distribution network line, and carrying out the self-adaption integration of multi-source heterogeneous observation data; initial edge features are explicitly constructed in neural network propagation of each layer of graph, node and edge representation is alternately updated, and the line parameter prediction precision is improved; power grid core laws such as Ohm's law and node power injection constraint are converted into a loss function to constrain a training process, consistency constraint of a parameter prediction result and a power system physical law is realized, and interpretability and credibility of the prediction result are improved; and inputting the power grid data into the graph neural network optimization model, and outputting a power distribution network line parameter identification result, thereby solving the problems that in the prior art, dynamic information interaction between nodes and edges is neglected, so that parameter identification is sensitive to network structure change, and physically incredible parameter output is easy to generate.

The invention discloses a broadband scattering RCS acquisition method and system based on an adaptive integral method and a pseudo-spectral derivative method, and mainly solves the problems of overlarge memory occupation, low calculation efficiency and unstable precision in the prior art. According to the implementation scheme, boundary conditions are applied to a composite electromagnetic target to obtain an EFIE-PMCHWT integral equation of an electromagnetic field, and discretization, Galerkin testing and fast Fourier transform are carried out on the EFIE-PMCHWT integral equation to obtain a near-field impedance matrix; selecting GLC interpolation sampling points at the center frequency based on a pseudo-spectral method to generate a pseudo-spectral derivative matrix, and performing linear combination on the pseudo-spectral derivative matrix by a near-field impedance matrix to obtain a high-order impedance matrix; setting a rectangular grid used for surrounding the composite target, projecting current and charges of the composite target to grid points, and expanding the current into Taylor series and expanding the frequency band of the current by using an asymptotic waveform technology; and the broadband electromagnetic scattering characteristic is calculated according to the frequency band and the moment coefficient. According to the method, memory occupation is reduced, calculation efficiency and precision are improved, and the method can be used for analyzing the broadband characteristic of the dielectric / metal composite object.

The invention discloses a cloud data center virtual machine adaptive integration method based on Autoformer and an enhanced double-Q network, and relates to the technical field of cloud computing resource management and optimization, the resource utilization rate of each physical host is monitored in real time, historical load time sequence data is collected, a pre-trained Autoformer model is loaded, and double-Q network parameters and environment state variables are initialized; and predicting a host resource utilization rate based on an Autoformer model, and dividing an overload host, a low-load host and a normal host in combination with the current host resource utilization rate. According to the method, the Autoformer model is adopted for load prediction, trend terms and period terms are decomposed through an autocorrelation mechanism, the multi-scale periodicity in the cloud load is effectively captured, compared with a traditional LSTM model, the Autoformer is higher in modeling capacity for a complex time sequence mode, the misjudgment rate of an overload / low-load host can be remarkably reduced, a more accurate host state detection basis is provided for virtual machine integration, and the method has the advantages of being high in practicability and easy to popularize. Therefore, a resource allocation strategy is optimized.

The invention discloses a dual-path multi-scale feature extraction method with differential information compensation for an LSTF, and relates to the technical field of long-term time series prediction. According to the method, a new network structure DMWCNet is designed, and lightweight and efficient long-term time sequence prediction can be realized by adopting an integral design of reversible normalization, dual-path multi-scale feature extraction and differential perception compensation. Dynamic normalization and anti-normalization of data distribution are realized through the RevIN layer, and the stability and generalization ability of the model under cross-scene and cross-distribution conditions are improved; a core backbone VATMLP module adopts a main and auxiliary double-path structure, a main path models a long-term trend through dimension expansion and progressive compression, an auxiliary path retains local details through a compression and reconstruction mechanism, and two paths of features are adaptively integrated through cascade and fusion projection, so that the expression ability of the model under different time scales is remarkably enhanced.

The invention discloses a modular sensor adaptive integration and data processing method and system in a complex terrain environment, and the method comprises the steps: constructing an extensible sensor sensing layer through a standardized interface module, and achieving the plug and play of various environment monitoring sensors; performing an inspection task in a complex terrain area based on a mobile device, and performing fusion processing on real-time data acquired by multiple sensors through an edge calculation layer; according to the environment data characteristics, the terrain complexity and the running state of the mobile device, the inspection behavior strategy of the mobile device is dynamically adjusted, and follow-up data monitoring and processing are carried out; the processed real-time data is uploaded to a cloud collaboration layer for collaborative analysis in a self-adaptive selection communication mode, and fundamental conversion from a static and fixed monitoring mode to a dynamic and adaptive mode is realized through modular sensor integration, mobile inspection, edge intelligent processing and self-adaptive communication; and the coverage rate, the efficiency and the intelligent level of environment monitoring are remarkably improved.

The invention discloses a virtual-real fusion traffic twinborn simulation test system and method oriented to air-ground coordination, belongs to the field of intelligent traffic design optimization, verification or simulation, and aims to provide a highly flexible multi-scene experiment platform for teaching and scientific research through an entity twinborn-dynamic mapping concept. The system is based on a dynamic display technology, real traffic scenes such as urban roads, emergency dispersion and three-dimensional parking are twinned to a small-scale experiment space, rapid switching of scenes and space layout is supported, and seamless mapping of virtual and real environments is achieved. The central control system synchronizes motion trails and virtual traffic flows of physical equipment such as an unmanned aerial vehicle and an intelligent vehicle in real time through a virtual-real fusion technology to form a small space entity-central control virtual two-way interactive closed loop. A multi-scene adaptive integration idea is taken as a core, the space-ground cooperation capability is combined, dynamic reconstruction and nonlinear migration of tasks such as logistics distribution and emergency response are supported, and a cross-scene experiment can be completed without hardware transformation.

The invention belongs to the technical field of a strength analysis method of a lattice structure, discloses an efficient finite element modeling and analysis integrated method for static analysis of the lattice structure, and establishes an efficient and high-precision finite element modeling and analysis integrated method for the lattice structure. Aiming at the limitations that in traditional finite element analysis, high-quality hexahedron mesh generation is difficult and units need to ensure welt processing, the method adopts structured meshes to quickly divide the meshes, improves a self-adaptive integral strategy in a traditional finite cell element method, and further improves simulation efficiency. Meanwhile, a penalty function method is adopted to weakly apply displacement boundary conditions. In addition, in post-processing, a result calculated by a finite cell method is mapped to a smoother visual model. Through example comparison, it is verified that the method can perform strength check domain analysis on the complex lattice structure in a stable and efficient mode, and a feasible scheme is provided for efficient analysis of a large-scale complex engineering structure.

The invention discloses a multi-modal visual target tracking method based on dynamic adaptation. The method comprises the following steps: constructing a dynamically adaptive multi-modal feature extraction and fusion network; constructing a dynamic bridging fusion module; constructing a parameter efficient adaptation mechanism based on low-rank adaptation; constructing a complete multi-modal visual target tracking model; training the model on computing equipment such as a server, and optimizing network parameters by reducing an overall loss value of a network loss function until the network converges; tracking a specified single target in a to-be-tracked video sequence by using the trained visual target tracking model; and performing performance evaluation on the trained model. According to the method, adaptive integration and unified target tracking of different visual modal information are realized through dynamic modal fusion and an efficient parameter adaptation mechanism.

A flood similarity intelligent analysis method based on multi-mode Transform and comparative learning comprises the steps that firstly, static features are processed through a grouping full-connection network, and the characterization capacity is enhanced through feature specific transformation; a CNN-Transform-Attention hybrid network is constructed to extract dynamic process line features, a convolutional layer captures local morphological features, a Transform encoder captures a global time sequence dependency relationship through a self-attention mechanism, key hydrological stages are adaptively focused through an attention weight, multi-modal features are adaptively integrated by adopting a gating fusion mechanism to generate unified embedded representation, and the dynamic process line features are extracted by adopting a convolutional neural network (CNN)-Transform-Attention hybrid network. A comparison learning and difficult sample mining strategy is introduced, discriminative characterization is learned in a low-dimensional embedding space through a comparison loss function training model based on the Euclidean distance, and the similarity is mapped into an interpretable probability of a [0, 1] interval; the technical problem that multi-source heterogeneous features are insufficient in utilization is effectively solved, accurate quantification and interpretable evaluation of flood similarity are achieved, and technical support is provided for flood forecasting, historical flood matching and flood control dispatching decision making.

The present invention discloses a method for intelligent positioning of deep geothermal target areas based on a machine learning algorithm, and the present invention relates to the field of geothermal resource exploration technology. The method comprises the following steps: S1, multi-source heterogeneous data fusion acquisition and preprocessing; S2, geological feature entropy quantification and spatial autocorrelation analysis; S3, deep feature extraction and multimodal feature fusion; S4, multi-model adaptive integration and dynamic weight optimization; S5, reinforcement learning dynamic adjustment to participate in abnormal threshold determination; S6, three-dimensional geothermal target area intelligent positioning and risk assessment. This positioning technology integrates geological structure, geophysical field, geochemistry and remote sensing data through a machine learning algorithm to construct a high-dimensional feature vector, thus solving the limitation of the existing technology that relies solely on physical detection. At the same time, it introduces reinforcement learning to dynamically optimize model parameters, combined with real-time geothermal well data updates, so that the model can adapt to changes in geological conditions, and the prediction accuracy is greatly improved compared with traditional methods.

The invention provides an underwater image enhancement method based on structure-color collaborative modeling, and relates to the technical field of image enhancement, and adopts a parallel double-branch architecture: on one hand, efficient long-range dependence modeling is performed on horizontal and vertical directions through ''directional frequency Mamba'' so as to maintain edge and structure topology; on the other hand, a differentiable histogram is introduced to guide self-attention, and global color distribution is explicitly coded to correct color cast and dynamic range compression. On each scale, context and fine-grained details are adaptively integrated by using lightweight'gated multi-scale fusion ', and a result with natural color, improved contrast and clear structure is obtained through decoding and reconstruction. The method is low in calculation overhead, high in interpretability and suitable for real-time deployment of the low-power-consumption terminal.

The invention discloses a smart city big data acquisition and summarization method, and particularly relates to the technical field of big data processing, and the method comprises the steps: S1, data perception and pre-acquisition, predicting future data demands through a business demand prediction engine, and dynamically adjusting an acquisition strategy, S2, dynamic routing and storage, and according to a demand prediction result and a real-time portrait of a storage node, carrying out data acquisition and summarization. The method comprises the steps of S1, intelligently distributing appropriate storage nodes for data, S2, carrying out self-adaptive integration, predicting and generating a minimum conversion set according to requirements, and only converting and fusing necessary data, and S4, carrying out data service, and pushing fused data in a subscription-release mode through a uniform interface. According to the method, demand prediction is preposed and runs through the whole data processing flow, so that the conversion from passive full-quantity acquisition to active precise service is realized, the technical problems of data redundancy, high transmission load, contradiction between storage and calling and the like in a traditional method are effectively solved, and the efficiency and the intelligent level of data acquisition and summarization are remarkably improved.

The present invention discloses a method for recognizing large-angle license plate images, comprising: obtaining a training data set and a test data set, wherein the training data set comprises a first number of simulated large-angle motion-blurred license plate images with license plate numbers annotated, and the test data set comprises a second number of real large-angle motion-blurred license plate images; inputting the training data set into a pre-built adaptive license plate recognition network for end-to-end training to obtain a trained adaptive license plate recognition network, and inputting the test data set into the trained adaptive license plate recognition network for model evaluation and optimization to obtain an optimized adaptive license plate recognition network, wherein the adaptive license plate recognition network comprises a feature extraction module, a spatial transformation parameter prediction module, a scale-aware feature adaptation module, a feature adaptive integration module, an upsampling module, and a character recognition module; and inputting a license plate image to be recognized into the optimized adaptive license plate recognition network for license plate recognition to obtain a license plate recognition result.

The invention discloses an intelligent water affair management system based on deep learning, and the system comprises a data collection and preprocessing module which is used for collecting the data of a water supply network and constructing a standardized water affair input tensor; the topology modeling module is used for generating a water supply pipe network diagram structure and diagram structure representation thereof; the state initialization module is used for fusing the input tensor and the graph structure to generate an initial state vector; the state evolution calculation module is used for establishing an augmented state evolution equation and carrying out adaptive integration to generate a prediction state sequence and source sink estimation; the multi-objective decision module is used for constructing a comprehensive cost function and generating a pump station and valve control instruction; the physical constraint projection module is used for executing water pressure, flow velocity and water age constraint projection and outputting an executable scheduling instruction; and the dynamic simulation feedback module is used for performing digital twin simulation and updating the model and sparse estimation weight. According to the invention, fine control and dynamic feedback optimization of the water affair system are realized, and the operation efficiency and the scheduling intelligence level are improved.

The present application relates to the technical field of Xinchuang storage and database integration, and discloses a database adaptive integration method, system, device and medium based on Xinchuang storage, wherein the database adaptive integration method based on Xinchuang storage comprises the following steps: extracting a protocol element definition set and a medium performance characteristic parameter matrix of a source system and a target system; generating a protocol difference vector by using an edit distance algorithm; generating a three-dimensional affinity tensor of a data block-medium-encoding format by comprehensively storing overhead, access efficiency score and medium performance parameters; generating a version-aware data encoding conversion rule set and a medium allocation mapping table by solving and generating through a genetic algorithm optimization; performing streaming data reading and online format conversion; and directly writing the converted data stream into a corresponding target storage medium. The method realizes integrated processing of protocol adaptation and medium optimization, effectively shortens the migration window, and avoids temporary storage of an intermediate format.

The application provides a radar scattering cross section (RCS) acquisition method based on CBFM-AIM-EDM, and the implementation steps are as follows: initializing parameters; calculating main characteristic basis functions of each sub-region based on CBFM-AIM; calculating secondary characteristic basis functions of each sub-region based on CBFM-AIM-EDM; constructing a reduction matrix and acquiring the radar scattering cross section (RCS). The main characteristic basis functions and the secondary characteristic basis functions of each sub-region are calculated by using the characteristic basis function method (CBFM), then the RWG basis functions and their divergence on each sub-region can be projected on the matrix grid respectively by using the adaptive integration method (AIM) to obtain the Topelitz matrix, so that the calculation speed is accelerated, the RCS acquisition efficiency is improved under the condition of ensuring the accuracy, and the coupling between two sub-regions of the target is equivalent to the coupling between equivalent dipoles by using the equivalent dipole moment method (EDM), so that the RCS acquisition efficiency is further improved.

The invention relates to a text classification method fusing semantic information and structural information, and aims to solve the problems of text graph noise interference, incomplete semantic capture, rigid information fusion and the like in existing graph neural network text classification. The method comprises the steps that firstly, a target text is preprocessed, and phrase blocks with complete semantics are extracted through BERT sequence labeling and B-I-O labeling; constructing an enhanced text graph by taking the phrase blocks as nodes and combining various relationships such as self-loop edges and syntactic dependency edges and cross-sentence anaphora connection; afterwards, redundant edges in the picture are cut through attribute-enhanced personalized PPR, and noise interference is weakened; and finally, inputting the optimized text graph into gradient gating fusion GNN, adaptively integrating BERT context features and global dependency information, outputting node features and completing classification. The method effectively breaks through the limitation of a traditional method, realizes deep fusion of semantic and structural information, has higher classification accuracy, stronger generalization ability and good stability, and is suitable for various short text, long text and professional field text classification scenes.

The invention discloses a hypergraph propagation source positioning method based on interaction enhancement. A traditional traceability method mainly depends on a pairwise interaction hypothesis of a simple graph structure, and the positioning precision under a complex topology is limited. The method comprises the following steps: firstly, constructing a hypergraph propagation model and obtaining a complete observation snapshot of a network node; then, extracting the dynamic state and global spectrum features of the nodes, and constructing a basic feature vector; an improved double-flow interaction hypergraph neural network is utilized, hyperedge internal key information is aggregated through local attention flow, and multi-scale topology diffusion flow is utilized in parallel to capture overall long-range dependence; and finally, introducing a gating residual fusion mechanism to adaptively integrate the double-flow features, and accurately outputting the propagation source probability based on a weighted cross entropy loss function. According to the method, the path reconstruction deviation of a traditional method can be effectively overcome and the robustness can be improved in the aspect of processing high-order interaction and long-distance dependence. According to the method, the performance bottleneck based on simple graph traceability is broken through, and powerful technical support is provided for propagation source positioning.

The invention discloses a drivable area detection method based on a progressive gating decoder, and mainly solves the problems of low drivable area detection precision, complex network structure and low calculation efficiency in a weak feature scene in the existing drivable area detection technology. The implementation scheme is as follows: 1) acquiring a data set and a detection label; 2) constructing a drivable area detection model; 3) constructing a loss function; 4) training a drivable area detection model; and 5) obtaining a drivable area detection result. According to the driving area detection model constructed by the invention, multi-dimensional modeling of a target is realized by combining a visual field heterogeneity extractor and extracting heterogeneity features, self-adaptive integration of feature information of different scales is realized through a cross-scale feature fusion module, and self-adaptive integration of feature information of different scales is realized through a progressive gating decoder. And by utilizing the efficient multi-scale convolution module design, the calculation amount is reduced, and meanwhile, an excellent segmentation effect is achieved.

The application discloses a runoff prediction method based on STL-SVMD decomposition and adaptive integration, and comprises the following steps: collecting runoff and multi-source meteorological data; constructing a trend item and a seasonal item responding to long-term climate driving by using a generalized STL method; constructing a multivariate SVMD target function containing a driving alignment item and a physical penalty item, and decomposing runoff residuals into mode subsequences with clear physical correlation on the premise of meeting causality constraints; performing closed-loop screening on the modes based on information retention rate and physical consistency double indexes; constructing a component-level integrated prediction model containing monotonicity, annual integral and extreme value sensitivity constraints; and finally, reconstructing runoff prediction results by using a multi-objective loss function and a time-varying adaptive Bayesian optimization strategy to dynamically adjust model parameters. The application improves the physical consistency and precision of runoff prediction.

The invention belongs to the technical field of lattice structure modal solution, and discloses an adaptive finite cell modal analysis method for a complex lattice structure. The automatic voxelization modeling method is invented for solving the problems that traditional body-fitted grid pretreatment is difficult, grid division consumes long time and the like. In order to solve the problems of large storage capacity, unnecessary consumption of computing resources and the like caused by using a consistency unit, an adaptive boundary unit subdivision technology based on an octree is provided. Meanwhile, in order to solve the problem that the integral domains of the mass array are different due to the fact that the sizes of the units are inconsistent, the self-adaptive integration technology of the mass array is invented, and the technology can effectively capture the discontinuous property of an integrated function on the boundary and achieve efficient integration of the mass array. According to the method, the problem of complex lattice structure modal solving can be solved in a stable and efficient mode, and a new technical path is provided for research and application in related fields.

The invention discloses a weighted trapezoidal integral white light interference signal peak searching method, which comprises the following steps of: acquiring a group of three-dimensional light intensity data through a white light interference measuring instrument, determining a self-adaptive integral parameter according to the central wavelength and the spectral width of a light source, selecting an integral weighting function, creating an integral window, and then carrying out weighted trapezoidal integral to obtain a white light interference signal peak searching result. And calculating a zero optical path position of the position, traversing each pixel position one by one, and calculating positions of all interference peak values to realize surface topography reconstruction. According to the method, the interference signal peak searching problem in white light interference measurement is solved, and the provided method has high anti-noise capability, high peak searching accuracy and high calculation speed, can be applied to different white light interference measurement instruments, and is wide in application range.