14 results about "Model composition" patented technology

The invention belongs to the technical field of electric power automation, and particularly relates to an automatic modeling method of power grid dispatching knowledge based on a large model and a related system.The electric power knowledge is decomposed into an explainable intermediate reasoning path through the thinking chain technology, and adaptive small models are selected according to the path to form a modeling link; a modeling link is represented as a triple form, edges and nodes are complemented, a complete knowledge graph is obtained, problems in the knowledge graph are further decomposed to construct a reasoning link, meanwhile, input data are monitored in real time, newly added entities are associated with the reasoning link, and unified representation and processing of heterogeneous power system data are achieved. Due to the fact that data from different sources are different in format, precision and semantics, according to the method, through thinking chain decomposition and small model link processing, sub-module optimization and dynamic adaptation aiming at different data types are achieved, and precision loss and adaptation difficulty caused by data heterogeneity are effectively solved.

The invention discloses a polymer property prediction method and system, and relates to the technical field of high polymer material information, the polymer property prediction method comprises the following steps: S1, constructing a polymer multi-property standardized data set containing an active end group label; s2, preprocessing the data set and extracting multi-modal features; s3, constructing an integrated prediction model composed of a geometric graph neural network and a GPR model; and S4, optimizing model parameters and quantizing prediction uncertainty, and fusing multiple GPR model outputs through an integration strategy to obtain a final prediction result. According to the method, the characterization accuracy is improved through data standardization and feature fusion, polymerization sites and connection chemical environments are considered through active end group labeling and multi-modal feature screening fusion, the structural characterization discrimination is enhanced, the sensitivity of a model to different main chain configurations and end group substitution is improved, and the stability of the model is improved. And through integrated modeling and uncertainty quantification, the prediction reliability is enhanced, and the prediction precision and robustness are improved.

The invention relates to a target-driven digital model orthogonal analysis and dual-adaptation fusion method and system, and the method comprises the steps: obtaining an overall demand, and combining the overall demand with empirical data and a knowledge graph to generate a model demand; obtaining a model composition file and a model description file which is determined based on a model demand and is used for dividing model categories, generating a meta-model description file according to the model composition file and the model description file, and generating a meta-model based on the meta-model description file; analyzing input and output data logic of the meta-model, scheduling parameters according to meta-model interface specifications, and redefining an interface by adopting a nonlinear iteration and aggregation optimization method; correlation adaptation and proxy packaging are carried out on a result of transmission logic numerical value fitting, and numerical value verification is completed in combination with a precision constraint condition; and adapting the incidence relation of the meta-agent model, generating a business-level numeralization agent model, scheduling parameters to perform precision verification on the business-level numeralization agent model, and generating a unified and operable system model.

This invention relates to a variable-structure interactive multi-model estimation method based on rolling time-domain estimation, belonging to the field of vehicle control technology. The purpose of this invention is to design a novel classification method using residual information, thereby reducing the dependence on model probability accuracy. The steps of this invention are: determining the system model set; in each iteration cycle, for all models in the model set, given a fixed time domain, if there is only one model with a certain time domain length, it is directly output as the interaction value; the interactive multi-model estimation from model set to model set based on rolling time-domain estimation is defined as the model set composed of all models adjacent to the main model. This invention reduces the dependence on model probability accuracy, better ensures the accuracy of model classification, has higher estimation precision, and solves the problem of non-interaction caused by different start times of the corresponding time domains of each model.

The application discloses a kind of multi-model integration post-processing methods and systems for Chinese regional subseasonal forecast, the method includes steps: S1, obtaining multi-mode multi-set data;S2, multi-mode multi-set data is handled, and the coarse resolution input sample, high-resolution input sample and corresponding target output sample that can be called by model layer are obtained;S3, after training FUNet post-processing model constructed, the dynamic weighted fusion module composed of multiple models is constructed;S4, dynamic weighted fusion module uses comprehensive verification error, calculates the comprehensive error of each model in the partition position;Then according to the comprehensive error updates fusion parameter, based on comprehensive verification error using adaptive anti-error weighting mode determines fusion weight, outputs final fusion prediction result.The application can improve the accuracy and spatial resolution of prediction result simultaneously, better retain high-frequency spatial structure, reduce blur and artifact phenomenon.

The application discloses an alumina substance content detection device and method, and belongs to the technical field of alumina substance content detection. The method of the application is as follows: 1, the alumina substance content detection device is used to configure the spectrum collection on the surface of a sample; 2, a spectrum matrix M of an nth sample is constructed n ; 3, the spectrum data obtained by the alumina sample is preprocessed; 4, a prediction model for detecting the alumina substance content of the in-model sample composed of 75% training spectrum data and 25% training spectrum data is constructed by a PCA-PLS linear regression sub-model and a preliminary model, the prediction model for detecting the alumina substance content is compared and verified by using out-of-model samples, and a trained prediction model for detecting the alumina substance content is obtained; 5, test data is input into the trained prediction model for detecting the alumina substance content, and a prediction result of the alumina substance content is obtained; compared with the prior art, the application improves the efficiency of the existing alumina detection.

The invention discloses a cold hydrogenation process parameter prediction method, an optimization method, a cold hydrogenation process parameter prediction device and cold hydrogenation process parameter optimization equipment. The cold hydrogenation process parameter prediction method comprises the following steps: acquiring initial process parameters of a cold hydrogenation process flow; obtaining a pre-created cold hydrogenation process parameter composite model, wherein the cold hydrogenation process parameter composite model comprises a front-section mechanism model, a reactor big data model and a rear-section mechanism model; and based on the initial process parameters, simulating the whole cold hydrogenation process flow by utilizing a cold hydrogenation process parameter composite model to obtain prediction parameters of the cold hydrogenation process flow. According to the technical scheme, the whole technological process can be simulated and predicted by obtaining the initial technological parameters and utilizing the composite model composed of the front-section mechanism model, the reactor big data model and the rear-section mechanism model, and accurate prediction and analysis of the technological performance are achieved.

The invention relates to the technical field of inventory management, and discloses an industrial product shopping mall inventory prediction algorithm and system based on machine learning, and a medium, and the prediction comprehensiveness is improved by fusing multi-dimensional data. The abnormal value is detected by combining the 3 sigma principle and the isolated forest algorithm, and the data quality is effectively improved by combining cause differentiation correction. Three types of features of time, commodities and influence factors are constructed, primary redundancy screening is carried out through Pearson's correlation coefficients, and key features are screened through attention weight optimization. And a mixed prediction model composed of an LSTM sub-model and an XGBoost sub-model is constructed, and the result is further optimized through weighted fusion, so that the multi-scene precision is ensured. By monitoring the deviation ratio of the actual sales volume and the predicted value, the supplier state and the policy factor in real time, the dynamic adjustment of the predicted value is triggered when the preset threshold is exceeded, so that the output result better meets the actual inventory demand, the decision is more accurate, and the operation cost is reduced.

This invention discloses a method and apparatus for predicting passenger cabin temperature based on a stacked regression model. The method includes: collecting key operating parameters affecting passenger cabin temperature and corresponding temperature values ​​for different areas of the passenger cabin; selecting a preset number of samples from the key operating parameters and corresponding temperature values ​​according to preset rules to obtain a training dataset; constructing a stacked regression model architecture, wherein the stacked regression model includes a base regressor layer composed of multiple regression models and a meta regressor layer for fusing the prediction results of multiple regression models; training the stacked regression model architecture using the training dataset to obtain a passenger cabin temperature prediction model; obtaining the current key operating parameters of the passenger cabin; and inputting the current key operating parameters of the passenger cabin into the passenger cabin temperature prediction model to obtain the corresponding temperature values ​​for different areas of the passenger cabin. This invention solves the problem of insufficient accuracy in passenger cabin temperature simulation and prediction in existing technologies.

A system and processes address the anticipatory availability of complex, interdependent systems composed of multiple products and services related to the user and their body. These systems, known as mixed systems, utilize anticipatory digital modeling to facilitate user decision-making, aiming for both comfort and resource savings. A digital model for anticipatory modeling will consist of several partial models or sub-models, each with a fixed component modeling repetitive elements and a time-varying component specific to each availability model. Partial models of the user's body parts and their environment will be active within these models. The anticipatory modeling process, comprised of two phases—analysis and predictive modeling—will utilize these partial models.Variants of the Local Party Device (LPD) will be available to the user and, depending on their needs, interconnected by the central component of said device, the automatic Availability Supervisory Center (ASC), in communication with the interactive ASC and its human assistant to provide even more relevant advice. Figure to be published for the summary: [Fig. 5].

The invention discloses a multi-objective optimization method and system for an open thin-wall energy-absorbing stand column, and relates to the technical field of mechanical structure design, and the method comprises the steps: building an optimization model which aims at the minimization of peak compression force, the maximization of total energy absorption and the minimization of structural mass and comprises the minimum load threshold constraint; a high-precision agent model composed of a BP-Transform-LSTM auto-encoder hybrid neural network prediction model and a BP quality prediction model is constructed to replace expensive simulation to carry out millisecond-level performance evaluation; performing global optimization based on an NSGA-III algorithm, and applying punishment to a scheme violating constraints to obtain a Pareto non-dominated solution set; and finally, objectively screening out an optimal compromise design scheme from the solution set by adopting a multi-criterion decision strategy combining an entropy weight method and a TOPSIS method. Unification of optimization efficiency, prediction precision and engineering reliability is achieved, and the design quality and research and development speed of the energy absorption structure are remarkably improved.

The application relates to a vehicle lane-changing trajectory prediction method based on physical information deep learning, which comprises the following steps: S1: constructing a data set and dividing the data set into a training set, a verification set and a test set, extracting observation states and matching states I j from the training set, and extracting observation states S2: constructing a physical prediction model, combining the physical prediction model with an Attention-TCN model to form a physical information deep learning model, and training the physical information deep learning model, calculating a loss function and a loss on the verification set, and comparing the loss on the verification set with an optimal loss value and the size of the loss function to selectively update parameters of the Attention-TCN model and parameters of the physical prediction model; S3: using the trained physical information deep learning model to perform a test experiment, and comparing the prediction performance of different lane-changing trajectory prediction models. By combining the constructed physical prediction model with the Attention-TCN model to form the physical information deep learning model, the explainability and descriptiveness of the lane-changing trajectory prediction are improved.

This invention discloses an electricity load forecasting method based on a hybrid expert architecture, relating to the field of electricity load forecasting technology, and aimed at solving the problem of inaccurate forecasts in existing methods. The method includes the following steps: constructing a forecasting framework composed of multiple electricity load forecasting expert models; pre-training a general electricity load forecasting expert model and saving the model parameters obtained after training as initialization parameters; fine-tuning based on the initialization parameters; generating routing embeddings through class-aware contrastive learning, and dynamically assigning input electricity load samples to corresponding partial-specialized expert models based on the routing embeddings; and forecasting the electricity load time series using the assigned partial-specialized electricity load forecasting expert models. This invention also discloses an electricity load forecasting device, electronic device, and computer storage medium based on a hybrid expert architecture. This invention improves the forecasting accuracy of electricity load by introducing a partial-specialized expert architecture and a class-aware contrastive routing mechanism.

The embodiment of the invention provides a training method of a click-through rate estimation model and a click-through rate estimation method, and the training method of the click-through rate estimation model comprises the steps: obtaining training sample data and a reference click-through rate estimation model, and the training sample data comprises sample training information and sample click-through rate information; training an initial click rate estimation model according to the training sample data, and forming a teacher model according to the initial click rate estimation model and the reference click rate estimation model; training a teacher model according to the training sample data to obtain first hidden layer representation information and first click rate estimation information, and training a reference click rate estimation model according to the training sample data to obtain second hidden layer representation information and second click rate estimation information; calculating a loss value according to the first hidden layer representation information, the first click rate estimation information, the second hidden layer representation information and the second click rate estimation information; and adjusting model parameters of the reference click rate prediction model according to the loss value to obtain a target click rate prediction model.