36results about How to "Improve fitting ability" patented technology

The invention discloses a drilling pump pressure prediction method based on an artificial neural network, and the method comprises the following steps: S1, data collection; s2, data preprocessing; s3, constructing a neural network model; s4, model training and optimization; and S5, model verification and deployment. According to the method, multi-channel data fusion and time synchronization optimization are innovatively adopted, so that the data quality and consistency are improved; in combination with a deep neural network and a self-adaptive optimization strategy, the precision and generalization ability of pump pressure prediction of the drilling pump are improved; and an online updating mechanism is introduced, so that the model can be dynamically optimized according to real-time data, the defects of low prediction precision, poor adaptability and difficulty in real-time updating of a traditional method are overcome, and an efficient and reliable prediction means is provided for intelligent drilling control.

The invention provides a sewage plant carbon source accurate adding method based on online water quality monitoring and automatic machine learning. The method comprises the following steps: using 21 online water quality monitoring indexes as explanatory variables, including water inlet and outlet indexes: CODcr, NH3-N, SS, TN, TP, C / N, pH and FLOW; the five removal rate indexes comprise the CODcr removal rate, the NH3-N removal rate, the SS removal rate, the TN removal rate and the TP removal rate; taking the carbon source dosage as a target variable, and establishing a model by adopting an automatic machine learning method; the prediction model is used for predicting the adding amount of the required carbon source, and the method is a novel method for adding the water treatment carbon source of the sewage plant.

The invention relates to the field of carbon emission prediction, in particular to a subway station construction stage carbon emission prediction method and system. Measuring and calculating carbon emission of the subway integrated station at different stages according to a carbon emission coefficient method, and performing data preprocessing to obtain carbon emission monitoring data; performing feature and target variable separation on the carbon emission monitoring data to obtain a feature matrix and a target variable vector; constructing a carbon emission prediction model to perform feature processing on the carbon emission monitoring data to obtain carbon emission feature data; the method comprises the following steps: constructing a carbon emission prediction integrated model based on LightGBM and XGBoost, identifying carbon emission characteristic data through the carbon emission prediction integrated model, and optimizing parameters in the LightGBM model and the XGBoost model by using grid search to obtain a carbon emission prediction amount in a subway station construction stage.

The application relates to the field of battery life prediction, in particular to a multi-feature fuel cell life prediction method based on a long short-term memory network; the method comprises the following steps: collecting voltage attenuation data and collecting running state data to construct an original input data set; analyzing the running state data, calculating a correlation coefficient according to the voltage attenuation data and each running state data, analyzing the correlation coefficient, and obtaining an input feature set; processing the input feature set and the voltage attenuation data to construct a neural network training data structure; dividing the training data structure into a training set and a test set; constructing a network architecture solution space to be searched, analyzing the network architecture solution space, constructing a fitness function, and determining an optimal network architecture parameter combination based on the fitness function; and inputting the training set into the optimal network architecture parameter combination for training to generate a target detection prediction result. The application can improve the prediction accuracy of the voltage attenuation trend.

The present disclosure provides a data reconstruction method based on Radon domain sparse representation and related equipment, relating to the technical field of communication. The method comprises: obtaining low-resolution Radon coefficients; inputting the low-resolution Radon coefficients into a neural network of an encoding-decoding structure to obtain first high-resolution Radon coefficients; and increasing the sparsity of the first high-resolution Radon coefficients through an adaptive soft threshold function cascaded at the back end of the neural network to obtain second high-resolution Radon coefficients. The method can reduce the data size, operation time, and operation cost of time-domain inversion of reduced-time variable Radon transform, and improve the stability and resolution of reconstructed data.

The application is suitable for the medical technology field, and particularly relates to a speech recognition method and device based on artificial intelligence, computer equipment and medium. The application obtains a first speech enhancement matrix and a second speech enhancement matrix missing different semantic information by randomly shielding rows and columns of a mel spectrum matrix; calculates a metric sub-loss to perform self-supervised training on an encoder according to the first speech frame features and the second speech frame features extracted by the encoder; obtains speech fusion features and inputs the speech fusion features into a decoder to obtain mapped characters, combines preset characters to calculate a prediction loss to perform supervised training on a speech recognition model, weights and adds the prediction loss and the metric loss according to the number of zero characters and non-zero characters to obtain a target loss to train the encoder and the decoder, and combines the self-supervised training mode and the supervised training mode to improve the recognition accuracy of the speech recognition model, greatly improving the instantaneity, convenience and accuracy of information input in the medical technology field.

The invention belongs to the technical field of offshore wind power operation and maintenance and reliability engineering, and discloses an offshore wind turbine generator system-level health degree evaluation method based on dynamic weight distribution, and the method comprises the following steps: S1, collecting and preprocessing the multi-source data of an offshore wind turbine generator; s2, constructing a layered health degree evaluation index system; s3, fusing the mechanism model, the data driving model and the statistical reliability model, and constructing a subsystem degradation model to obtain a subsystem health degree; s4, performing weight fusion in combination with the static weight and the dynamic correction factor to obtain a dynamic weight; s5, fusing the subsystem health degree and the dynamic weight to obtain a system-level health degree; and S6, performing confidence interval sorting and risk grade division according to the system-level health degree. According to the evaluation method provided by the invention, through fusion of a mechanism model, a data driving model and a statistics model, the physical interpretability is reserved, the fitting capability of nonlinearity and coupling degradation is improved, and the misjudgment and missed judgment risks are reduced.

The invention discloses a small sample data-oriented fruit external quality image classification method, and relates to the technical field of image classification. According to the method, a fruit external quality classification model is constructed and is mainly divided into three layers: a fruit classifier, a fruit defect recognizer and a fruit quality classifier. Firstly, the fruit classifier improves a Faster R-CNN model, a multi-scale subnet is added in a feature extraction part to adapt to fruit feature extraction of different scales, and the accuracy of fruit classification is improved; secondly, the main function of the fruit recognizer is to detect the classification layer to output the defect category of the fruit, and in order to better fit the marginalization or centralization of the feature of the defect, the pooling strategy of the CNN model is changed, the pooling mode is selected and changed in a targeted manner, and the defect classification effect is better realized; and finally, a fruit classifier is established through fruit feature dependence maximization, and accurate classification of fruits under small sample data is realized.

This invention discloses a heterogeneous two-order wind power prediction method and system considering physical constraints, belonging to the field of wind power prediction and microgrid energy management technology. The method includes: acquiring and preprocessing raw data; constructing input features including wind speed, wind direction encoding, temperature, humidity, and historical power; dividing the data into training, validation, and test sets; training an XGBoost multi-step direct master prediction model using Bayesian optimization; training an LSTM model based on the prediction residuals and outputting residual prediction values; superimposing the master prediction values ​​and residual prediction values ​​to obtain a two-order prediction result; applying power boundary and ramp rate constraints, and outputting the final prediction result. This invention, by constructing a heterogeneous two-order prediction architecture of XGBoost and LSTM combined with physical constraints, improves prediction accuracy while adapting to small sample scenarios, thus enhancing the engineering usability of the prediction results.

The application discloses a cyanobacterial community succession prediction method and device fusing ecological mechanism and causal enhancement. The method comprises the following steps: collecting multi-source high-frequency monitoring data of a target water area, and constructing an environmental driving factor characteristic matrix; constructing a causal weight matrix through causal relationship aggregation analysis, and outputting a causal enhancement characteristic matrix; constructing a species ecological strategy embedding matrix; performing multi-scale decomposition on algal density time series data; constructing a multi-species interaction prediction model based on a multi-head self-attention mechanism, fusing environmental characteristics, causal enhancement characteristics, ecological strategy embedding and multi-scale time series characteristics, and predicting the biomass and relative abundance of each dominant cyanobacterial species; and outputting a cyanobacterial community succession process and a water bloom risk grade. Through deep fusion of causal enhancement, ecological strategy embedding and multi-head attention mechanism, the precision, stability and ecological interpretability of cyanobacterial community succession prediction are significantly improved, and the application is suitable for water bloom early warning and ecological management of lakes, reservoirs and rivers.

This disclosure provides a method and system for enhancing low-exposure images based on embedded McLaurin and convolution, relating to the field of image processing technology. The method involves acquiring a low-exposure image to be enhanced, processing and decomposing the low-exposure image, inputting the low-exposure image into a backbone network to extract multi-scale features, introducing a dynamic adjustment factor into the backbone network, using the dynamic adjustment factor to learn a convolution weight calibration map, calibrating the convolution weights of the backbone network with the learned weight calibration map, performing convolution operations on the features using the calibrated convolution kernels to obtain the luminance component of the low-exposure image, and outputting a long-exposure enhanced image that varies with the dynamic adjustment factor using the luminance component.

The application discloses a kind of based on digital twinborn model's diversion tunnel structure health assessment method, it is related to water conservancy engineering health monitoring field, including: obtaining the geometric parameter and material parameter of diversion tunnel, constructs initial digital twinborn model;Carrying out numerical simulation including fluid-structure coupling effect, obtain the simulation physical response data of tunnel structure under preset load condition;Simulation physical response data are compared with actual monitoring data, and model error index is calculated;When model error index exceeds preset threshold, construct fitness function with minimizing model error index as target, iteratively optimize model key parameters using particle swarm optimization algorithm, obtain optimal parameter combination;Based on updated digital twinborn model, diversion tunnel structure health assessment is carried out.It realizes accurate simulation and health assessment to diversion tunnel structure response by high-precision digital twinborn model after parameter calibration.

This invention belongs to the field of tunnel engineering stability analysis, specifically disclosing an intelligent inversion analysis method for tunnel surrounding rock parameters based on the XGBoost optimization algorithm, including the following steps: S1: Establish a numerical simulation calculation model to obtain a sample library for surrounding rock parameter inversion; S2: Perform correlation and sensitivity analysis on the displacement and parameter data in the sample library, evaluate the feasibility of each parameter as the parameter to be inverted, and quantitatively evaluate the rationality of displacement feature combinations; S3: Use the CART algorithm to determine the parameters to be inverted and perform displacement feature combination screening; S4: Use the CART algorithm as the base learner to establish an XGBoost ensemble algorithm model for intelligent inversion analysis of tunnel surrounding rock parameters; S5: Apply the single variable control method and Bayesian optimization method to optimize the XGBoost algorithm; S6: Input the displacement features into the XGBoost ensemble algorithm model to obtain the predicted surrounding rock parameter values. This invention achieves optimization of the XGBoost ensemble algorithm model through hyperparameter optimization, resulting in high model stability and prediction accuracy.

This invention discloses an uncertainty analysis method for a spacecraft thermoacoustic-vibration coupling system, relating to the field of aerospace computing. The method includes: discretizing the geometric model of the spacecraft's structure and internal sound field to obtain a thermoacoustic-vibration coupling finite element model; selecting sample points in the uncertainty parameter space to establish a data-driven multidimensional parallelepiped model; selecting observation points in the spacecraft's structure and internal sound field; selecting experimental points in the uncertainty parameter space and inputting them into the finite element model to perform thermoacoustic-vibration coupling finite element analysis, obtaining the corresponding structural velocity response and sound pressure level response values ​​at the observation points; establishing a radial basis function neural network model based on the experimental points and their corresponding response values; and calculating the response intervals of the structural velocity response and sound pressure level response of the multidimensional parallelepiped model based on the radial basis function neural network model. This invention improves the accuracy of uncertainty calculation for spacecraft thermoacoustic-vibration coupling systems while maintaining computational efficiency.

The invention relates to the technical field of electric field grid-connected control, in particular to a wind power plant unit grid-connected synchronous detection and debugging system which comprises a data acquisition module, a synchronous detection algorithm module, a historical data training module, a self-adaptive compensation debugging module and a cooperative control module. The data acquisition module synchronously acquires original electrical parameters of a power grid side and a unit side; the synchronous detection algorithm module adopts an improved second-order generalized integrator phase-locked loop algorithm, filters out harmonic waves, extracts fundamental wave parameters and calculates four types of detection parameters; the historical data training module establishes a mapping model of power grid fluctuation and compensation amount through machine learning, and outputs an optimal compensation coefficient combination; the self-adaptive compensation debugging module is combined with a dynamic compensation and feed-forward pre-judgment algorithm to output the excitation current and the rotating speed adjustment amount of the unit; and the cooperative control module establishes closed-loop linkage until the detection parameter meets the grid-connected threshold value, and outputs a grid-connected ready signal.

This invention discloses a method for predicting the dynamic evolution of pollution plumes based on a physical perception spatiotemporal large model, comprising the following steps: acquiring multi-source heterogeneous monitoring data and constructing a static spatial topology matrix; constructing a PF-STGNN model, which adopts an encoder-decoder architecture; the encoder captures the long-term dependencies of pollution plume evolution through a context-aware temporal self-attention mechanism, and couples dynamic graph convolution with the static spatial topology matrix through a dynamic flow field interactive spatial solution module, outputting a feature matrix containing historical diffusion memories; the decoder ensures temporal unidirectionality through a causal masking temporal self-attention module, injects historical evolution laws into future predictions through a cross-spatiotemporal evolutionary context cross-attention module, and performs spatial morphology verification again through a dynamic flow field interactive spatial solution module; finally, a concentration prediction matrix for future periods is output through a parallel generator. The advantages of this invention are: it can be applied to groundwater pollution monitoring and environmental risk early warning systems.

The invention discloses a battery aging trajectory adaptive extrapolation method fusing physical time sequence prior, which comprises the following steps: collecting full life cycle constant current charging data of a battery, extracting an incremental capacity curve, and constructing characteristic peak evolution to form multi-dimensional physical time sequence prior; a state space backbone network is introduced, and trend perception is achieved through a priori initialization implicit state; constructing an attention anomaly detection module to discriminate real relaxation and noise; monotonic constraint is released through a micro-relaxation layer for a relaxation event, and data is corrected through Kalman filtering for noise; and constructing a dynamic trade-off loss function to adaptively adjust the physical constraint strength, and iteratively training the model based on the corrected data to realize stable extrapolation prediction. The method effectively solves the problems that in the prior art, real physical relaxation and measurement noise cannot be distinguished, and prediction deviation is caused by physical constraint rigidity.

The embodiment of the invention provides a KAN-LSTM-based target trajectory prediction method. The KAN-LSTM-based target trajectory prediction method comprises the following steps: S1, obtaining target observation data containing D-dimensional features; s2, data preprocessing: removing abnormal values in the target observation data, and complementing missing values; s3, data normalization: scaling data of each dimension of the target observation data to a range of [0, 1] in an equal proportion; s4, a target trajectory prediction network model based on KAN-LSTM is constructed, and model training is completed; and S5, inputting the target observation data into the target trajectory prediction network model, and obtaining a target motion state prediction result at the next moment. According to the embodiment of the invention, the fitting capability of the model to a nonlinear relationship is effectively improved, the learning capability of the model to a complex data mode is enhanced, a new solution is provided for processing a marine target trajectory prediction task, and higher accuracy and reliability are shown in the dynamic behavior prediction of a marine target.

This invention discloses a series-parallel type switched linear composite envelope tracking power supply, comprising a multilevel converter (VSC), a linear amplifier (VLA), and a switching converter (CSC). The multilevel converter (VSC) and the linear amplifier (VLA) are connected in series, and the switching converter (CSC) and the linear amplifier (VLA) are connected in parallel. The multilevel converter (VSC) tracks the envelope signal and outputs a stepped-wave voltage to fit the load voltage v. o Simultaneously, the stepped waveform voltage output by VSC serves as the input voltage of CSC; the linear amplifier VLA realizes the load voltage v o For the reference signal v ref The switching converter (CSC) outputs current to match the load current. This invention significantly reduces the switching frequency of the CSC and the voltage stress on the switching transistor without increasing the complexity of the series-parallel ET power supply, thereby improving the overall efficiency of the ET power supply.

The application discloses a lithium battery residual life prediction method based on an adaptive Wiener process, obtains the degradation amount of a sample lithium battery at different moments by accelerating the life experiment of multiple sample lithium batteries; then acquires the model prior parameters of a target lithium battery by using a maximum likelihood estimation algorithm; and then the model parameters are updated by using a Kalman filtering algorithm and a transfer learning algorithm, and the updated parameters are used to calculate the residual life probability density function of the target lithium battery at the current moment, so that the lithium battery residual life prediction method has the characteristics of high prediction accuracy, good real-time performance and strong self-adaptive capability.

This invention belongs to the field of high-precision positioning technology. It discloses an inertial navigation positioning calibration method and system based on multi-source error compensation. The method includes acquiring acceleration and angular velocity data output by an inertial measurement unit (IMU), obtaining position, velocity, and attitude information through integration, and constructing navigation state variables under a unified time series. Errors in the navigation state variables are characterized and decomposed into zero-bias error, scaling factor error, dynamic coupling error, and environmental disturbance error, forming a multi-source error superposition relationship and generating an error sequence that evolves over time. Error evolution prediction modeling is performed based on the error sequence, dividing the error sequence into steady-state intervals and transition intervals. An evolution prediction function constrained by motion is constructed within the steady-state interval, and an error-driven enhancement prediction function is constructed within the transition interval. Continuity constraints are applied to adjacent intervals to obtain continuous error prediction results, thereby improving the accuracy and stability of inertial navigation positioning.

The application discloses a frequency-driven point cloud sequence human behavior recognition method, relates to the technical field of behavior recognition, and aims at the problem that existing methods lack explicit modeling of structural and rhythmic frequency characteristics in human behavior.The application constructs a frequency-driven space-time representation network, which comprises a space geometry-frequency collaborative perception module, a time transient-frequency collaborative perception module and a frequency response module.The space geometry-frequency collaborative perception module extracts point cloud frequency information in the spatial dimension to enhance the understanding of posture structure rules and capture limb texture distribution in combination with geometric description.The time transient-frequency collaborative perception module describes periodic rhythmic motion through frequency decomposition in the time dimension and models non-periodic mutation motion in combination with statistical moment dynamic feature modeling.The frequency response module performs nonlinear transformation on frequency characteristics through complex residual mapping and a learnable spline.The application significantly improves the representation ability of potential change rules of human behavior and has the advantages of high recognition accuracy and strong generalization ability.

The application belongs to the technical field of traffic flow prediction, and particularly relates to a spatio-temporal traffic flow prediction method fusing multi-graph structure and knowledge enhancement, comprising the following steps: obtaining historical observation data of traffic monitoring nodes at continuous multiple time steps; constructing enhanced features based on the historical observation data; splicing the enhanced features and the historical observation data to obtain enhanced input features; inputting the enhanced input features into a pre-trained traffic flow prediction model; and the traffic flow prediction model predicting traffic flow at one or more future time steps. The application realizes joint modeling of multi-source spatial correlation, complex time dynamics and historical period knowledge in traffic flow data, and improves the accuracy and robustness of future multiple time step traffic flow prediction.

The invention provides a multi-class and multi-scale modeling method for a fracture-vuggy carbonate reservoir, and relates to the technical field of petroleum geology modeling. The multi-class multi-scale modeling method for the fractured-vuggy carbonate reservoir comprises the steps of S1, unified standardization and space alignment of multi-source multi-scale data: acquiring seismic data, logging data, core digital data and production dynamic data of a target reservoir, and performing abnormal value elimination, missing value compensation and dimension unified processing on the multi-source data, and realizing alignment of different data in a unified space coordinate system through space coordinate conversion and scale mapping. A unified and reliable data basis is provided for subsequent modeling through generation of a standardized multi-source data set, elimination of data dimension differences and elimination of abnormal values, a macroscopic fracture-cavity boundary is extracted by adopting an edge detection operator and an adaptive threshold segmentation method, the recognition precision of geometric contours of fractures and karst caves is remarkably improved, and the recognition efficiency is improved. And identifying a micro pore-crack structure based on an image segmentation model.

The invention relates to the technical field of thermal environment assessment, and discloses an artificial intelligence-based indoor thermal environment dynamic assessment method and system, and the method comprises the steps: carrying out the temperature compensation of an infrared thermal imaging sequence in preprocessed multi-source thermal sensing data through a temperature compensation algorithm based on time-frequency domain conjoint analysis, obtaining a pure infrared thermal imaging sequence, and extracting environment temperature time sequence characteristics of an environment thermal sensing data sequence in the preprocessed multi-source thermal sensing data; performing multi-scale skin temperature decomposition and enhancement processing on the pure infrared thermal imaging sequence, and extracting sequential change characteristics of the skin temperature; and dynamically evaluating the indoor thermal environment by using the thermal environment evaluation model fused with the skin temperature perception. According to the method, temperature compensation, multi-scale skin temperature enhancement and skin temperature feature extraction are performed on infrared thermal imaging, and dynamic evaluation of the indoor thermal environment is realized by adopting a thermal environment evaluation model fusing skin temperature perception and environment temperature time sequence features.

This invention relates to the fields of clinical laboratory medicine and artificial intelligence, and provides a risk prediction method and device based on machine learning for clinical mass spectrometry. The risk prediction method includes: defining N+M dimensions of features based on a liquid chromatography-tandem mass spectrometry system to obtain an N+M dimension feature vector structure; obtaining a virtual training dataset based on the N+M dimension feature vector structure, acquiring parameter values ​​of each dimension of the current batch through a data acquisition interface, and assembling them into N+M dimension feature vector data; performing format verification and invalid value filtering on the N+M dimension feature vector data using the N+M dimension feature vector structure to obtain a feature matrix; and obtaining a standardized real-time risk score based on the feature matrix and a trained fusion-integrated risk prediction model. This invention utilizes a machine learning model to uncover the complex nonlinear relationship between configuration parameters and dynamic parameters, thereby achieving more accurate and forward-looking risk warnings than traditional single-threshold methods.

The invention provides a hydrofoil proxy model optimization method and a hydrofoil ship, and the hydrofoil proxy model optimization method comprises the steps that a two-stage adding point Kriging proxy model is used for accurately predicting the hydrofoil performance, then a NSGA II algorithm is used for hydrofoil optimization, the calculation efficiency is ensured, the prediction precision of the hydrofoil strong nonlinear working condition performance is remarkably improved, and the hydrofoil performance is optimized. And dynamic balance between global optimization and local refinement is realized.

The application discloses a network security threat event report generation method and system, and belongs to the network security field.The method comprises the following steps: acquiring a network security threat event log, pre-processing the log, extracting key information by using a pre-trained ERNIE-Gram model, and generating structured knowledge points; integrating the knowledge points into a feature vector, inputting the feature vector into a KAN model for nonlinear feature extraction and multi-level feature fusion, and generating fusion feature data; inputting the fusion feature data into an OverLoCK model, performing multi-stage feature extraction and aggregation through a dynamic context guiding mechanism of the model, and generating summary semantic analysis information; standardizing the semantic analysis information, inputting the semantic analysis information into a pre-trained federal decision tree model, matching a decision path of the federal decision tree model according to a global split rule, and outputting an interpretable security event report containing a confidence distribution.The application can improve the accuracy and interpretability of threat event report generation.

The application provides a weakly supervised fine-grained image classification method and system guided by local diversity, constructs a weakly supervised fine-grained image classification network guided by local diversity, and the classification network comprises a basic backbone network, a cross-layer attention interaction module and a bilinear pooling module connected after Layer 3 and Layer 4 of the basic backbone network, and a random selection strategy is connected between the cross-layer attention interaction module and the bilinear pooling module; the weakly supervised fine-grained image classification network guided by local diversity is trained to obtain a weakly supervised fine-grained image classification network model guided by local diversity; and preprocessed training data sets are sent into the weakly supervised fine-grained image classification network model guided by local diversity to obtain an image classification result. The application solves the problem of inaccurate classification results caused by the problems that distinguishable features are too fine to be captured and local information is not effectively utilized in the existing fine-grained image classification task.

The application discloses a kind of based on deep neural network's voiceprint recognition method, belong to speech processing technical field.The application includes: training voiceprint feature extraction network and voiceprint recognition scoring network, based on the voiceprint feature extraction network of trained prediction registration voice voiceprint feature vector, based on registration voiceprint feature vector constructs voiceprint feature database;Acoustic feature to be identified is obtained, based on the scoring result of voiceprint recognition scoring network with voiceprint feature database is obtained to determine recognition result.The voiceprint feature vector extracted by the application is more delicate, so that the speech feature is better preserved.The Fbank feature extracted reduces the amount of calculation in the speech preprocessing process, speeds up the feature construction speed.By setting a larger number of channels to the network, the fitting ability of the neural network model is enhanced.The scoring network uses the parameters obtained by training the PLDA algorithm to initialize the network parameters, which speeds up the network convergence speed and achieves better results.