131 results about "Nonlinear modelling" patented technology

The invention relates to the technical field of intelligent water affairs, in particular to an intelligent water affairs management and control system based on the Internet of Things, and aims to solve the problems that in the prior art, multi-source water quality data and user behavior characteristics can be fused to construct a risk assessment model based on fuzzy logic and a Bayesian network, dynamic assessment of a complex water quality state cannot be achieved, and the risk assessment accuracy is poor. Abnormity cannot be quickly recognized through a similarity matching mechanism, and the response speed and accuracy of early warning are reduced; multi-source water quality data and user behavior characteristics are fused through the water quality intelligent early warning module, a risk assessment model based on fuzzy logic and a Bayesian network is constructed, the method has high nonlinear modeling and uncertainty processing capacity, dynamic assessment of a complex water quality state is achieved, abnormity is rapidly recognized through a similarity matching mechanism, and the risk assessment efficiency is improved. The method improves the early warning response speed and accuracy, combines the countercurrent tracking and GIS technology, accurately locates the pollution source, and enhances the emergency disposal and decision support capability.

The invention discloses a deep learning and machine learning fused soil moisture inversion construction method, and relates to the technical field of measurement of physical properties of materials, and the method comprises the steps: capturing complementary information and spatial context of multi-source data through a multi-source heterogeneous data space-time adaptive fusion step by using a cross-modal attention mechanism and a graph neural network; through a deep learning and machine learning dual-path collaborative inversion step, advantage complementation is realized by combining data-driven nonlinear modeling and a physical constraint interpretable model; according to the method, the defects of single data source, insufficient model generalization ability and incomplete physical mechanism consideration in the prior art are overcome, the inversion precision is improved by 12%-18% under the complex earth surface condition, and the method has the advantages that the method is suitable for large-scale popularization and application. And a high-precision, strong-generalization and reliable technical means is provided for precise monitoring of soil moisture.

The invention discloses an oil reservoir production dynamic prediction method fusing discrete gradient information, and belongs to the technical field of oil reservoir development and artificial intelligence crossing, and the method comprises the steps: building a heterogeneous oil reservoir oil-water two-phase flow numerical simulation data set based on a numerical simulation method; designing a double-branch network structure and extracting spatial and physical characteristics of input field data in parallel, wherein the spatial and physical characteristics comprise a main characteristic coding branch and a differential operator branch; designing a backbone network to carry out deep nonlinear modeling; an efficient pressure and saturation field prediction neural network model is constructed based on a double-branch network structure and a backbone network, in a model training stage, spatial region observation points of part of time steps are used to participate in data item loss calculation, and meanwhile, physical control equation residuals are introduced into all time steps and a whole space to serve as physical loss items; a trained efficient pressure and saturation field prediction neural network model is obtained, and high-precision prediction of a full-time-sequence pressure field and a saturation field is achieved.

The invention relates to the technical field of biological feature recognition, and particularly provides an anti-migration PPG recognition method based on rate perception and a state space model. The method comprises the following steps: performing physiological feature front-end extraction on an original single-channel PPG signal to obtain a high-dimensional shallow feature sequence; performing double-flow cooperative processing on the high-dimensional shallow-layer feature sequence, and distributing the high-dimensional shallow-layer feature sequence to two parallel branches, namely a control flow branch and a data flow branch; in the control flow branch, an amplitude spectrum and an instantaneous physiological rate curve are obtained; in the data stream branch, acquiring a deep global feature sequence with rate invariance; obtaining multi-scale refinement features based on the high-dimensional shallow feature sequence and the deep global feature sequence; according to the multi-scale refinement features, a final biological feature recognition result is obtained, the method can actively sense the physiological rate change, and efficient nonlinear modeling can be achieved with the extremely low parameter quantity.

The invention discloses a multi-energy system closed-loop prediction control method, system, equipment and medium, and the method comprises the steps: carrying out the decomposition, dimension reduction and nonlinear modeling of the time sequence characteristics and environmental influence factors of photovoltaic output through a photovoltaic power prediction model, and outputting a future multi-period photovoltaic power prediction sequence; establishing a state equation and an output equation, integrating equipment operation constraints, and constructing a hydrogen-containing energy storage state space model; inputting local load power and a future multi-period photovoltaic power prediction sequence into the hydrogen-containing energy storage state space model, and solving an objective function through a rolling optimization algorithm to obtain a future multi-period optimal scheduling scheme; and applying a first hour control instruction of the optimal scheduling scheme to an actual system, and proportionally superposing the deviation between an actual measurement value and a historical prediction value to a photovoltaic power prediction sequence of a next period through a feedback correction item to realize closed-loop control. According to the method, the photovoltaic local consumption rate can be improved, and the power grid fluctuation rate is reduced.

The invention provides a Mars mineral abundance inversion method and system. The Mars mineral abundance inversion method comprises the steps of firstly obtaining target Mars hyperspectral reflectivity data; and then generating simulated Mars hyperspectral image data based on the target Mars hyperspectral reflectivity data. Secondly, training the constructed mineral abundance inversion deep learning model based on the simulated Mars hyperspectral image data, and determining the trained mineral abundance inversion deep learning model; thirdly, acquiring to-be-processed Mars hyperspectral image data, and performing preprocessing; and finally, performing mineral abundance inversion on the preprocessed to-be-processed Mars hyperspectral image data through the trained mineral abundance inversion deep learning model to obtain a mineral abundance inversion result. Therefore, the weight distribution of each wave band of the hyperspectral image can be enhanced through the mineral abundance inversion deep learning model, and the non-linear modeling capability and the wave band characteristic response capability of abundance inversion are improved, so that the accuracy and the efficiency of Mars mineral abundance inversion can be effectively improved.

The invention provides a multi-dimensional data preferential analysis method and device based on a gradient boosting decision tree model, and relates to the technical field of data mining, in the method, a target gradient boosting decision tree model is adopted, the nonlinear modeling problem of strategy optimization in multi-dimensional data is solved, and the multi-dimensional data is optimized according to the target weight determined in the training process. According to the method, key features can be accurately identified, the processing efficiency and precision are high, compared with a traditional linear regression or single decision tree method, the accuracy and real-time performance of strategy optimization can be improved, high interpretability and robustness are achieved, an optimization strategy can be adjusted in a self-adaptive mode in a changeable environment, and the method is suitable for popularization and application. The method is widely applied to multiple fields of industrial optimization, intelligent decision making and the like.

The invention discloses a photovoltaic power generation power prediction and power system scheduling method and a system for realizing the method. The method comprises four steps of data acquisition and preprocessing, similar day selection and training set construction, GA (Genetic Algorithm)-fuzzy RBF (Radial Basis Function) neural network modeling and power scheduling plan generation. The method comprises the following steps: selecting a meteorological variable with high correlation degree through a Spearman rank correlation coefficient, and constructing a time sequence feature; similar days are utilized to construct training samples, and the generalization ability of the model is improved; a fuzzy RBF network optimized by GA is adopted to enhance the nonlinear modeling precision; and inputting a prediction result into the power system simulation model, and generating an optimal scheduling instruction by adopting dynamic programming. According to the method, the photovoltaic power prediction accuracy and scheduling efficiency can be remarkably improved, the power grid stability is enhanced, and the new energy consumption capability is promoted.

The invention discloses a split type three-box girder nonlinear vortex vibration steady-state amplitude intelligent prediction method, belongs to the field of bridge vortex-induced vibration control, and aims to solve the problem of low accuracy of three-box girder vortex vibration nonlinear modeling and amplitude prediction. The method comprises the following steps: acquiring displacement time sequence data through a reduced scale model wind tunnel test, preprocessing to obtain displacement, speed and acceleration dimensionless data, and dividing a data set; constructing a candidate function library containing high-order polynomial terms of the primary function; performing nonlinear system sparse recognition by taking acceleration data as a target item and combining an improved algorithm to obtain a preliminary feature set; a final feature set is obtained through energy-statistics-self-adaption three-stage screening; and finally, constructing a control equation to predict the steady-state amplitude. The method can accurately predict amplitude, coincide test and real bridge observation results, and is less in computing resource occupation and high in numerical value precision.

The invention relates to the technical field of vehicle thermal management, in particular to a prediction model construction method and device for vehicle thermal management, a control method and device and a vehicle, and the construction method comprises the steps: obtaining a historical sample data set of a vehicle thermal management system; preprocessing the historical sample data set; an initial prediction model is established based on a Transform network, the input of the initial prediction model is state data, the output of the initial prediction model is control data, and a loss function comprises a mean square error term and a physical constraint term; and taking the minimum loss function value as an optimization target, and performing iterative training optimization on the initial prediction model through the preprocessed historical sample data to obtain a prediction model of vehicle thermal management. According to the method, on the basis of strong nonlinear modeling capability of the Transform network, model parameters are optimized by introducing physical constraint terms, so that a model prediction result is ensured to be consistent with an actual physical rule, and a prediction error caused by violation of the physical rule is avoided, thereby improving the prediction precision.

The invention provides an engineering drawing steel bar bulk sample annotation text detection method, system and device based on deep learning, and a storage medium. The method comprises the following steps: S1, preprocessing an engineering drawing image and labeling a steel bar bulk sample annotation text region data set; s2, constructing a staged multi-scale feature map extraction backbone network framework; s3, introducing compression excitation into each scale feature map; s4, adopting an ACON adaptive activation function in nonlinear modeling; s5, constructing a multi-scale feature pyramid structure based on the RSE-FPN, and superposing scales to reinforce fusion; and S6, outputting the approximate binary image to realize accurate prediction of the bounding box. According to the method, the OCR detection precision is remarkably improved, small character and complex background interference is effectively overcome, and missing detection and false detection are reduced; the model is lightweight to facilitate efficient deployment of equipment; the output boundary is compact and accurate, the semantics is reasonable, and subsequent recognition and analysis are facilitated; the method has excellent generalization ability, adapts to steel bar large sample drawings of different formats, definitions and styles, and is high in engineering practicability.

The invention relates to a Beidou / GNSS + 5G unmanned aerial vehicle cooperative positioning method based on a multi-mode neural network, and belongs to the technical field of unmanned aerial vehicle navigation and intelligent logistics. According to the invention, through nonlinear modeling, multi-source feature adaptive extraction and autonomous learning capabilities of the multi-modal neural network, Beidou, 5G, inertial navigation and visual environment perception data are deeply fused, and in combination with scene adaptive switching and cluster collaborative optimization, high-precision and high-robustness positioning support is provided for the low-altitude logistics unmanned aerial vehicle in a complex environment. According to the invention, through multi-source data deep fusion, scene adaptive switching and cluster collaborative optimization, the positioning precision and reliability of the low-altitude logistics unmanned aerial vehicle in a complex environment are effectively improved.

The invention relates to the technical field of intelligent detection, and discloses a T-SVAE feature extraction strategy and a method for improving near infrared spectrum soil rapidly available potassium measurement precision by using the T-SVAE feature extraction strategy, and the method comprises the following steps: step 1, collecting soil surface samples of different plots, obtaining near infrared spectrum data of the soil samples by using a Fourier transform near infrared spectrometer, and calculating the near infrared spectrum data of the soil samples; determining the actual content of quick-acting potassium in the soil sample by adopting a national standard method; the method comprises the following steps: 1, acquiring near infrared spectrum data, 2, preprocessing the acquired near infrared spectrum data, and removing impurity signals caused by instrument fluctuation, environmental interference and sample physical form difference, and 3, constructing a Transform and supervision constraint fused variational self-encoding model (T-SVAE). By constructing a variational self-encoding model fusing Transform and supervision constraint, the problem of feature blindness caused by high-dimensional data difficulty, nonlinear modeling limitation and unsupervised learning in near infrared spectrum data processing of a traditional feature extraction method is effectively solved.

The invention relates to the field of artificial intelligence and computer vision, in particular to an unmanned aerial vehicle multi-target tracking method based on a CNN-Transform-Mama network and a space-time Mama motion model, and the method comprises the steps: obtaining a video data set of multi-target tracking collected by an unmanned aerial vehicle; constructing a CNN-Transform-Mama network as a target detector to obtain bounding boxes and spatial position information of a plurality of targets on the ground in a single-frame image; constructing a multi-target space-time trajectory based on the space position information in the continuous space-time; a space-time Mama motion model is constructed to complete nonlinear modeling of a multi-target space-time trajectory, and multi-target tracking is achieved; and constructing a joint loss function training model, and evaluating the performance of the model through precision evaluation. According to the invention, by aggregating the advantages of the CNN, the Transform and the Mama network, the local-global-long-range dependency features are effectively fused, and the positioning precision of the multi-target position is improved; and the precision and the robustness of multi-target tracking are improved through nonlinear modeling of the space-time trajectory, so that the adaptive capacity to a complex scene is further improved.

The invention provides an intelligent control method, device and equipment for a steam crosslinking room and a medium, and relates to the technical field of intelligent control, and the method comprises the steps: collecting temperature, steam and humidity parameters in real time to form a multi-parameter observation vector with synchronous time; establishing a dynamic heat conduction model for describing the nonlinear coupling relation of the temperature, the steam flow and the thermal inertia; pID parameter output control quantity is adaptively adjusted based on the temperature deviation and the error change rate, a future temperature track is predicted by using the model, and an objective function is optimized under constraint conditions to obtain an optimal control increment; and combining the temperature deviation energy integral, the steam energy consumption and the temperature fluctuation variance to construct a multi-target optimization function to dynamically adjust the weight and output a target control signal. According to the method, the problems that a traditional control mode lacks nonlinear modeling and predictive compensation capability, temperature fluctuation is easily caused, a thermal field is uneven, and material performance is unstable can be solved.

The invention discloses a digital self-interference cancellation method based on transceiver nonlinear behavior modeling, and belongs to the technical field of electronic reconnaissance and interference. The digital self-interference cancellation method comprises the following steps of performing linear combination on memory components of a broadband signal, and generating cross terms with different phase structures by using a nonlinear effect so as to construct a Wiener model; establishing a multipath self-interference channel model; nonlinear modeling is carried out on linear components of interference signals received by a radio frequency front end at a receiver, and nonlinear component modeling is a linear model; for a nonlinear model constructed by a linear component in an interference signal received by a radio frequency front end, adopting dynamic reduction measurement to reduce the complexity of the model; parameter preliminary estimation is carried out by using the characteristics commonly represented by the intermediate variables, and parameter estimation is optimized by combining a two-step iterative algorithm; and reconstructing the transmitting signal in the digital domain according to a parameter estimation result to generate a reference signal, and realizing interference cancellation in the digital domain of the receiving end.

This invention discloses a method for dynamic prediction of reservoir production that integrates discrete gradient information, belonging to the interdisciplinary field of reservoir development and artificial intelligence. The steps are as follows: First, a numerical simulation dataset for oil-water two-phase flow in heterogeneous reservoirs is constructed based on numerical simulation methods. Second, a dual-branch network structure is designed to extract the spatial and physical features of the input field data in parallel, including a main feature encoding branch and a difference operator branch. Third, a backbone network is designed for deep nonlinear modeling. Fourth, an efficient pressure and saturation field prediction neural network model is constructed based on the dual-branch network structure and the backbone network. During the model training phase, spatial observation points from some time steps are used to participate in the calculation of data loss terms, while the residuals of the physical control equations are introduced as physical loss terms at all time steps and throughout the entire space. Finally, a well-trained efficient pressure and saturation field prediction neural network model is obtained, achieving high-precision prediction of pressure and saturation fields across the entire time series.

The application discloses a brake method and system of an automatic portal crane, and the method comprises the following steps: setting a target position of a grab bucket when braking is needed; and cyclically executing the following steps a to e until a difference between an actual position of the grab bucket and a stopping position is less than or equal to a threshold value, and outputting a frequency converter shutdown instruction and a brake starting instruction: a. collecting a real-time frequency of the frequency converter; b. calculating the actual position of the grab bucket; c. calculating a theoretical stopping distance under a current real-time frequency; d. compensating the theoretical stopping distance by considering a load; and e. calculating the stopping position based on the target position of the grab bucket and the compensated stopping distance. The application solves a long-existing problem of 'incompatibility between speed and accuracy' in the field of crane positioning by means of a trinity of non-linear modeling, dynamic calculation and data driving, and provides high-efficiency automatic operation for port automation.

The invention belongs to the technical field of intelligent driving and vehicle control, and discloses an EMB system closed-loop control method based on digital-analog double-drive modeling and KF-UI estimation, and the method comprises the following steps: building a physical model of an EMB system according to the structure of the EMB system; based on the physical model, a data-driven friction torque linear modeling method is adopted to convert the nonlinear friction torque into a linear system model; designing a separation strategy of the clamping force based on the linear system model, decomposing a linear part in the clamping force, and correcting the linear part; constructing a state space model containing process noise and measurement noise by combining a linear system model and a separation strategy; joint estimation of the system state and unknown input is carried out on the state space model, and online estimation of the clamping force is achieved; the problems that in the prior art, non-linear modeling precision is insufficient, and clamping force estimation lacks self-adaptability are effectively solved.

The invention relates to the technical field of artificial intelligence and digital twinning crossing, and discloses an equipment fault prediction method and system based on digital twinning. Comprising the steps of collecting multi-dimensional physical data and full life cycle data of equipment, constructing a digital twin model and synchronizing physical states in real time, quantifying implicit coupling interference through three-level linkage nonlinear modeling, dynamically correcting features, calculating fault evolution probability, and optimizing parameters through reinforcement learning. The system correspondingly comprises five functional modules. The method solves the problem of inaccurate prediction caused by neglect of multi-physics field coupling interference and feature processing lag in the prior art, realizes accurate pre-judgment and early warning of faults, and is suitable for a full-life-cycle monitoring scene of industrial equipment.

The application relates to the technical field of membrane pollution detection, in particular to a membrane pollution in-situ pollutant detection method based on front surface fluorescence data. The method adopts a machine learning model to perform nonlinear modeling on original fluorescence data containing multiple interferences, further introduces a symbolic regression method to analyze and process the prediction result of the machine learning model, obtains a membrane pollution in-situ pollutant relative concentration calculation formula, and corrects the relative concentration of a target pollutant, so that the collection of multiple fluorescence parameters and the correction of complex fluorescence data are not needed, the method has the comprehensive advantages of high prediction accuracy, few parameter dependencies and strong interference scene adaptability, and is suitable for in-situ fluorescence analysis of a membrane pollution process.

The application relates to a light wavelet enhancement fusion medical image segmentation method, and belongs to the technical field of medical image processing and computer-aided diagnosis. The core of the method is to construct a light WEF-Net network, which comprises an encoder, a bottleneck layer and a decoder. The encoder adopts a dual-domain perception module, extracts complementary features from the frequency domain and the time domain through wavelet transformation and convolution, and simultaneously extracts complementary features from the frequency domain and the time domain. The bottleneck layer is designed with a feature aggregation Kolmogorov-Arnold transformation module, which is used for efficiently fusing multi-scale semantic information and enhancing the nonlinear modeling capability. The network level also integrates a sawtooth rolling feature fusion module, which enhances the global continuity of the features through the channel rolling and spatial scanning mechanism to improve the integrity of the boundary segmentation. Experiments show that the method realizes excellent performance on multiple public medical image datasets, while maintaining low parameter quantity and low computational quantity, and significantly improves the segmentation precision.

The invention discloses an energy-based economic key element linkage prediction method and system, and relates to the technical field of economics and prediction modeling, and the method comprises the steps: determining and collecting a data source, and carrying out the preprocessing of the obtained data; exporting key business data based on the multi-dimensional data model, and constructing an energy consumption data and economic growth prediction model; training an energy consumption data and economic growth prediction model, and performing result evaluation and model optimization; and economic growth and energy prediction are carried out by using the optimized model, and analysis is carried out according to a prediction result and a corresponding decision is provided. By constructing the data cube model, data from different fields can be efficiently integrated, multi-dimensional comprehensive analysis is realized, and the comprehensiveness of the prediction model is improved; according to the method, the optimized model is utilized, a more flexible nonlinear modeling method is adopted, the complex relation between economic growth and energy consumption is better captured, and the prediction accuracy is improved.

The invention discloses a probability model analysis method for a secondary accident of an over-limit transport vehicle, and the method comprises the following steps: obtaining traffic accident related data, including numerical data and classification data; inputting the data into a trained over-limit transport vehicle secondary accident probability model, wherein the model is constructed by combining a random forest model after Bayesian optimization with a logistic regression model; and carrying out influence analysis on a model prediction result through an SHAP interpretation framework. According to the method, the interpretability of logistic regression and the nonlinear modeling capability of the random forest are fused, and the model precision is improved by combining Bayesian optimization, so that accurate prediction of the secondary accident probability of the over-limit transport vehicle and quantitative analysis of key influence factors are realized, the blank of model research in the field is filled up, and the method has a wide application prospect. And a scientific basis is provided for traffic safety management and accident prevention.

The invention provides a multi-stage feature processing method and related equipment, relates to the technical field of image processing, and realizes unified coding and channel adaptive compression of output features by introducing a rear feature coding module, reduces feature redundancy, and improves normalization and stability of feature representation. A global feature weight reconstruction module is introduced to explicitly model a global association relationship among different spatial positions, so that output features can be fused with global context information, and the overall consistency and structure expression ability of the features are significantly enhanced; and by setting a reconstruction feature decoding output module, the nonlinear modeling and structured decoding capabilities of an output stage are improved, so that a prediction result is more continuous, stable and fine in a complex background scene.

The application relates to the technical field of precision timing instrument testing, in particular to an atomic clock aging prediction method based on feature fusion and quantum constraint learning. The atomic clock aging prediction method based on multi-modal feature fusion and quantum constraint learning provided by the application adopts a multi-modal tensor fusion means, simultaneously collects time domain microwave probe signals, space domain atomic cloud density distribution and frequency domain Ramsey fringe signals for feature extraction, and improves the completeness of the features; when a nonlinear model is established, Bloch equation constraint is adopted to ensure that the network model conforms to quantum physical laws. The method can improve the atomic clock aging feature expression effect through an artificial intelligence means, effectively avoids the overfitting defects caused by simply relying on training data, makes the atomic clock aging prediction model more conform to physical laws, and improves the atomic clock aging prediction precision and optimizes the prediction model complexity.

The invention discloses a braking method and system for an automatic portal crane, and the method comprises the following steps: setting a target position of a grab bucket when braking is needed; the following steps a to e are executed circularly until the difference value between the actual position of the grab bucket and the brake position is smaller than or equal to the threshold value, and a frequency converter stopping instruction and a brake starting instruction are output: a, the real-time frequency of the frequency converter is collected; b, calculating the actual position of the grab bucket; c, calculating a theoretical braking distance under the current real-time frequency; d, a load is considered to compensate the theoretical braking distance; and e, calculating the braking position based on the target position of the grab bucket and the compensated braking distance. Through three-in-one innovation of nonlinear modeling, dynamic calculation and data driving, the long-standing problem that speed and precision cannot be achieved at the same time in the field of crane positioning is fundamentally solved, and high-efficiency automatic operation is provided for port automation.

The invention discloses a bearing vibration signal feature extraction method, a bearing fault diagnosis model and a diagnosis method. The feature extraction method comprises the following steps: acquiring a vibration signal detected by a sensor; performing band-pass filtering and envelope demodulation on the obtained vibration signal, and dividing the vibration signal into a plurality of signal samples; after performing multi-scale differential processing on each signal sample, inputting an obtained normalized multi-channel differential matrix into a random recurrent neural network to obtain a final hidden state sequence matrix; and determining at least two of a variance matrix, an absolute deviation matrix, a kurtosis matrix and a spectrum entropy matrix according to the final hidden state sequence matrix, and splicing the matrixes to obtain a feature vector. According to the method, feature extraction can be completed on the premise that a large amount of annotation data or tedious parameter adjustment and optimization are not needed, and the method has the advantages of multi-scale sensitivity, nonlinear modeling capability and lightweight deployment. Compared with an existing method, the method is more stable, easy to migrate and lower in calculation overhead.

The invention discloses a radar image denoising method and device based on a deep network, and electronic equipment. The method comprises the steps of inputting a to-be-denoised image into an image denoising network; wherein the image to be denoised is a dual-channel ISAR image, and the image denoising network is constructed according to wavelet transform, KAN and UNet; the image denoising network comprises an encoder, a WKAN module and a decoder; the encoder is used for performing hierarchical feature extraction and wavelet embedding on an image to be denoised in sequence to obtain multi-scale input; the WKAN module is used for sequentially carrying out noise suppression, wavelet transformation and nonlinear modeling on the multi-scale input so as to obtain optimized multi-scale input; the decoder is used for processing the optimized multi-scale input through up-sampling and inverse wavelet transform so as to reconstruct a de-noised image; and the de-noised image is obtained. The method provided by the invention has good de-noising capability, and gives consideration to both calculation efficiency and network flexibility.

The present application relates to a kind of shale oil dessert prediction method and system based on multi-attribute fusion, the present application establishes the three-dimensional distribution model of porosity, permeability, oil saturation, total organic carbon content and brittle index under lithofacies constraint, constructs the feature set including single attribute and attribute interaction term;Nonlinear mapping relationship between multiple attributes and oil production is established based on machine learning model, and the contribution value of each attribute and interaction term is calculated using SHAP method, and then the corresponding weight coefficient is determined;On this basis, the three-dimensional prediction model of dessert coefficient is constructed, and the horizontal well trajectory is dynamically optimized according to the distribution of dessert coefficient.Multiple attribute nonlinear modeling and weight objective determination improve the accuracy and reliability of shale oil dessert prediction, can effectively guide horizontal well deployment and trajectory adjustment, significantly improve sand body drilling rate and yield level.