56results about How to "Good modeling" patented technology

The invention provides a method, a device and a product for predicting the state of health of a battery based on a language model, and the method comprises the steps: obtaining the time-varying data of the electrical parameters of a target battery at an application side, carrying out the processing, and obtaining the input data according with the input characteristics of the language model, thereby achieving the prediction of the state of health of the battery. The battery health state corresponding to the target battery is obtained through the output characteristics of the target language model, and the pre-trained model backbone in the basic language model is frozen and the adjustable component is trained at the training side by taking the physical principle of the battery as a physical constraint condition, so that the target language model is obtained. By utilizing the generalization ability of the pre-trained model backbone and the generalization ability brought by the interpretability of the physical constraint, the demand for training data in the model training process is reduced, the modeling process is optimized, and the modeling efficiency is improved, so that the effectiveness of the battery health state prediction is improved.

The invention relates to the technical field of temperature regulation and control, in particular to an intelligent temperature regulation and control method and system based on an intelligent building. The method comprises the following steps: performing functional area air flow simulation on intelligent building structure design data to obtain functional area air conditioning flow simulation data; performing air circulation regulation of different functional areas on the functional area air regulation flow simulation data to obtain temperature coupling air circulation regulation data; indoor temperature circulation adjustment compensation is conducted on the temperature coupling air circulation adjustment data according to the human flow real-time monitoring data, and indoor temperature circulation adjustment compensation data are obtained; performing regulation and control logic training optimization on the indoor temperature circulation regulation and compensation data to obtain temperature regulation and control logic optimization data; and performing automatic firmware design based on the temperature regulation and control logic optimization data to obtain temperature regulation and control logic firmware. According to the invention, the temperature regulation and control technology is optimized, so that the temperature regulation and control technology is more perfect.

The invention relates to the field of intelligent control, in particular to an electric vehicle charging infrastructure intelligent management system which comprises a data acquisition and transmission module, a data processing and analysis module, a charging pile management module, a remote monitoring module, a fault diagnosis and early warning module, a billing and payment module and a safety certification module. According to the scheme, the HDCSA algorithm is provided to optimize the scheduling of the charging piles, multiple optimization objectives are considered at the same time, the method is suitable for the discrete problem, discrete decision variables in the scheduling of the charging piles can be processed, the situation that multiple objectives are involved in the scheduling of the charging piles is effectively solved, and compared with a traditional algorithm, the method has higher convergence speed and search efficiency, and the method is suitable for large-scale popularization and application. The scheduling accuracy and adaptability are improved; meanwhile, multi-fuzzy granulation processing is used, abnormal modes in the data are captured more comprehensively, more types of anomalies can be recognized, and a complex, fuzzy and high-dimensional data anomaly recognition method is provided for a system.

The invention provides a scientific research project expert matching method based on dynamic academic cooperation network graph representation learning, and aims to improve expert matching efficiency and quality in scientific research project management. According to the method, a dynamic complex network diagram is constructed, scholars serve as nodes, cooperation relations serve as directed edges, and the previous research history and cooperation experience of the scholars are comprehensively considered. And feature vector learning is carried out on scholar nodes by utilizing graph structure learning and a Transform architecture, so that accurate matching between scientific research projects and experts and scholars is realized. According to the method, more than 80% of matching precision and more than 89% of standardized discount cumulative earnings are obtained from data sets in the fields of artificial intelligence and data mining, the limitation of the prior art in the aspects of operation cost and scholar feature information utilization is effectively solved, and an innovative solution is provided for recommending appropriate partners for scientific research projects.

The invention discloses a water supply network flow prediction method based on time-invariant feature extraction, relates to the technical field of water supply network flow prediction, and provides a water supply network flow prediction model based on time-invariant feature extraction. Acquiring a time-invariant signal and a time-varying signal based on the high-frequency signal and the low-frequency signal; then a feature extraction module obtains time-invariant features and time-varying features based on the time-invariant signals and the time-varying signals; the flow prediction module obtains predicted water supply network flow based on the time-invariant features and the time-varying features; the time-varying features and the time-invariant features are processed separately, so that the respective features of the time-varying features and the time-invariant features can be captured more accurately; according to the method, the frequency domain signals are extracted from the time domain signals by using fast Fourier transform and inverse transform thereof, and are divided into the time-varying signals and the time-invariant signals, so that the performance of the model is improved, and the accuracy of flow prediction of the water supply network is improved.

The invention relates to the technical field of fault detection, and discloses a fault detection method and system for a charger, and the method comprises the steps: obtaining the charging state of a battery, and the charging state of the battery comprises the charging current, the charging power, the state of charge, and the temperature of the battery; performing feature extraction on the battery charging state, and constructing a charging feature matrix; inputting the charging characteristic matrix into a battery health prediction model to obtain a battery charging demand prediction value; performing state analysis and calculation according to the battery charging state to obtain a charger state; performing correlation analysis according to the charger state and the battery charging demand prediction value to obtain fault prediction information; and obtaining a charger fault detection result according to the fault prediction information. The method can improve the fault detection efficiency of the charger.

A multi-target interaction flight trajectory prediction method based on an SCINet network belongs to the field of traffic trajectory prediction, and comprises the following steps: step 1, extracting trajectory data to form a multi-target interaction scene; step 2, designing a model framework of multi-target interactive flight trajectory prediction based on an SCINet network; the model captures long-term time dependence, multi-level characteristics and multi-target space characteristics of the trajectory; the method for capturing the long-term time dependence of the trajectory and the multilevel features is a time feature extraction method based on SCIcross; and step 3, inputting the tested observation trajectory information into the trained model to obtain a prediction trajectory. According to the invention, the flight path prediction accuracy under the condition of multi-target interaction of the same scene is improved.

The invention discloses an improved YOLOv7 human body key point detection method. According to the method, manual labeling is carried out through LabelImage to obtain abundant training samples; replacing an original MP module in the YOLOv7 with an L-MP module to improve feature extraction, and adding a LiftDown Pool branch and residual connection; an EDH-Conv module is newly added between a Backbone module and a Head module of the YOLOv7 network, and the output of the EDH-Conv module is used as spatial weight distribution to correct an activated region, so that the detection capability of the network on a small target is further improved; an improved RSwin Transformer encoder is added to improve the detection effect of small targets and overlapped targets in a complex scene; a loss function is changed into an improved key point detection loss based on L2 loss by referring to Wise-IoU LOSS, and a dynamic focusing mechanism used for adjusting a key point error weight is introduced to accelerate convergence of the model. According to the method, the human body key point detection and identification capability in a complex scene is enhanced, and the method has higher robustness, accuracy and efficiency for human body key point detection, identification and counting tasks in sports.

The invention discloses a method for predicting SOH based on an HS-MH-MoE model and current, voltage and temperature of a battery. The method comprises the following steps: S1, testing and acquiring current, voltage, temperature and capacity data of the battery; s2, constructing an HS-MH-MoE model, wherein the HS-MH-MoE model comprises a multi-layer multi-head sparse hybrid expert sub-model A two-layer structure is used, each layer comprises three expert sub-models, and each sub-model is a CNN-LSTM model; the CNN automatically learns the spatial correlation of input data, and extracts high-level abstract features; and S3, training and verifying the model. S4, predicting the SOH of the battery: deploying and predicting the model after training; and connecting the three SOH predicted values to a linear regression layer, and outputting a final value. The SOH is predicted based on the CNN-LSTM hybrid expert model HS-MH-MoE and the battery data, the expression ability and prediction precision of the model are improved, the robustness and generalization ability of the model are improved, and the over-fitting risk is reduced.

The invention provides a method for predicting protein-protein binding affinity based on a graph convolutional network, and relates to the field of bioinformatics. The basic principle is as follows: respectively converting a protein-protein compound and an interaction region into atomic-level graphs; intramolecular and intermolecular node features and edge features of a compound are learned through a model constructed on the basis of a graph convolutional network, graph structure features of protein are fully utilized, the protein-protein interaction relationship is deeply excavated, then the precision of predicting a protein-protein binding affinity value is improved, and the prediction precision of the protein-protein binding affinity value is improved. And the function of predicting the affinity of the multi-chain compound is realized by processing the data set.

The invention provides an abnormal data processing method and an abnormal data processing device for a wind power plant. The abnormal data processing method comprises the steps that wind speed-power data of a wind power plant are acquired, and the wind speed-power data comprise wind speed and power corresponding to the wind speed; constructing a data set for fitting a curve based on the acquired wind speed-power data; based on the data set, fitting a preset fitting function to determine parameters of the fitting function, and generating the fitting curve corresponding to the fitting function with the parameters; identifying the wind speed-power data within the credible data range of the fitting curve as normal data, and identifying the wind speed-power data outside the credible data range of the fitting curve as abnormal data; in response to identifying the anomalous data, the data set is updated by rejecting the anomalous data, and the step of generating the fitted curve is iteratively performed until no anomalous data is identified.

This invention discloses a fine-grained image classification method based on transductive semi-supervised deep learning, comprising the following steps: preparing a semi-supervised training sample set for fine-grained images, dividing the training sample set into a labeled training sample set and an unlabeled training sample set; constructing a tree-like class hierarchy structure; building a deep convolutional neural network model based on the class hierarchy structure; training the built network model on the labeled training sample set; inferring the corresponding hierarchical class label vector set for all training samples; calculating the corresponding confidence score for each training sample; training the built network model using the entire training sample set from random initialization; repeating steps 5 to 7 until the fine-grained class label vectors inferred by the network model for unlabeled samples in the two rounds are basically consistent; inputting the fine-grained image to be classified into the trained network model, and obtaining the predicted category of the image in the fine-grained classification layer.

The invention discloses a visual question and answer method and device, computer equipment and a computer readable storage medium, and aims to realize cross-modal interaction by using a question guide image, so that generated image features are integrated with question information, and the intra-modal and inter-modal interaction capability is improved. The method comprises the steps of determining an adaptive attention feature map of a sample image and question feature representation of a sample question based on a visual question and answer model; performing cross-modal feature fusion on the adaptive attention feature map and the question feature representation, and determining a sample prediction answer; according to the loss between the sample prediction answer and the sample actual answer, optimizing model parameters of a visual question-answering model, and re-obtaining a training sample to optimize the visual question-answering model until the visual question-answering model reaches a standard of stopping model training, and obtaining a target model; and in response to the visual question and answer instruction, inputting a to-be-predicted image and a to-be-predicted question into the target model for prediction to obtain a predicted answer corresponding to the to-be-predicted question.

The invention discloses an optimal scheduling method of a comprehensive energy system containing an electricity-hydrogen-electricity energy hub, relates to the technical field of energy conversion efficiency analysis, and solves the problem that the existing scheduling mode of the comprehensive energy system containing the electricity-hydrogen-electricity energy hub is difficult to ensure the balance of supply and demand of a power grid. The method comprises the following steps: constructing a target function of an integrated energy system optimization scheduling model by taking maximization of the total operation profit of an integrated energy system containing an electricity-hydrogen-electricity conversion hub as a target; constructing constraint conditions of the integrated energy system optimization scheduling model by taking a power balance constraint, an electric hydrogen production power constraint, a power upper and lower limit constraint, a flexible load demand response constraint and a hydrogen energy flow balance constraint of the electric hydrogen coupling system as constraints; and solving the optimal scheduling model of the integrated energy system to obtain an optimal scheduling strategy of the integrated energy system.

The invention discloses a large model three-dimensional modeling system and method based on artificial intelligence, particularly relates to the technical field of three-dimensional modeling, and comprises a large model three-dimensional modeling data acquisition module, a large model three-dimensional modeling data analysis module, a large model three-dimensional modeling data evaluation module and a large model three-dimensional modeling comprehensive analysis module. And a large model three-dimensional modeling man-machine interaction module. The large-model three-dimensional modeling data acquisition module is used for acquiring performance data for analyzing large-model three-dimensional modeling based on artificial intelligence to obtain various parameters for evaluating performance indexes of large-model three-dimensional modeling, and outputting the various parameters to the large-model three-dimensional modeling data analysis module; according to the method, the modeling efficiency parameters are analyzed, an index is obtained to measure the efficiency of the three-dimensional modeling process, various indexes in the modeling process are monitored in real time, an administrator can adjust a modeling strategy and optimize the modeling process according to the indexes, the modeling speed is increased, and resources are optimized.

The invention relates to the field of mechanical fatigue strength optimization design, and discloses a transmission shaft fatigue strength optimization method and system, a storage medium and computing device.The transmission shaft fatigue strength optimization method comprises the steps that the contribution degree of key structure parameters to transmission shaft output response is calculated, and the structure parameter with the highest contribution degree serves as a design variable of transmission shaft fatigue strength optimization design; establishing a mathematical model of transmission shaft fatigue strength optimization design to calculate and obtain a performance response value of a transmission shaft model; constructing an initial Kriging agent model based on the selected design variables and the corresponding performance response values, setting a control function based on Bayesian posterior characteristics, and introducing the control function into a classical expectation point adding strategy to establish a Kriging agent model of Bayesian expectation improvement control; the maximum service life of the transmission shaft is taken as a target, the fatigue strength of the transmission shaft is optimally designed on the basis of a Kriging agent model of Bayesian expectation improvement control, global search is carried out by adopting a genetic algorithm, and the optimal structure parameters and fatigue life of the transmission shaft meeting constraint conditions are solved.

The invention relates to a road scene modeling method based on real road surface data, belongs to the technical field of road modeling, and solves the problems that existing road scene modeling is low in efficiency and cannot fit real road surface conditions. Comprising the steps of collecting continuous road scene data according to sampling frequency, and obtaining attributes of each sampling point; according to the attribute of each sampling point, selecting a segmentation point to segment the road scene data to obtain a plurality of scene primitives; performing geometric feature classification on each scene element according to the pavement type and the average attribute value of each scene element to obtain a plurality of geometric element sets; performing multi-dimensional feature clustering on each geometric primitive set according to the multi-dimensional attribute features of each scene primitive to obtain a plurality of subdivided primitive sets; and selecting a plurality of scene primitives from each subdivided scene primitive set according to received modeling demand information, and connecting the selected scene primitives to obtain a road scene model. And the road scene rapid modeling of highly restoring a real road surface is realized.

The invention discloses an unmanned vehicle behavior decision-making method for a pedestrian-vehicle mixed road environment, and belongs to the technical field of intelligent decision-making of unmanned vehicles. Comprising the following steps: acquiring historical data of different types of traffic participants; establishing graph representation of a multi-agent interaction environment, wherein the graph representation comprises a node feature matrix, an adjacent matrix and a supplementary adjacent matrix; performing traffic scene interaction feature extraction by using the supplementary adjacent matrix, and fusing the traffic scene interaction feature with the node feature matrix to obtain an updated node feature matrix; and inputting the adjacent matrix and the updated node feature matrix into a decision model to complete a behavior decision. According to the method, the edge features and the edge types are projected to the same feature space, the feature matrix containing the edges is formed after splicing, traffic scene interaction features are extracted and fused with node features for information enhancement, the problem that GAT cannot process the edge features in an interaction graph is solved, understanding of vehicle behaviors is helped, and the prediction capacity is improved.

The invention discloses a power dispatching strategy determination method and device and electronic equipment. The method comprises the following steps: acquiring region parameters of a target power region; according to the region parameters, determining a target function with the purpose of minimizing the power consumption cost of the target power region and a first constraint condition for constraining the target function; and determining a power dispatching strategy corresponding to the target power region according to the target function under the first constraint condition. According to the method and the device, the technical problems that only the adjustable load equipment is adjusted when the power dispatching strategy of the target power area is determined, the number of adjustable resources is small, and the power operation cost is high in the prior art are solved.

The invention provides a neural cognitive function evaluation method and device based on artificial intelligence, and relates to the technical field of neural cognitive function evaluation. The method comprises the following steps: collecting neuroimaging data, physiological signal data and behavior data; the collected multiple kinds of neurological function cognition evaluation data are preprocessed; performing feature extraction on the preprocessed data to obtain corresponding feature data; constructing a convolutional neural network evaluation model based on a capsule network combination graph, training the model by using a training set, and optimizing model hyper-parameters by using an ant lion algorithm; and evaluating the model through the evaluation index. According to the method, various data related to the neurocognitive function are collected, the convolutional neural network evaluation model based on the capsule network and the graph is constructed, and the model hyper-parameters are optimized through the ant lion algorithm, so that a hidden mode in the neuroimage data can be better mined, comprehensive evaluation on the cognitive function from multiple aspects is realized, and the evaluation efficiency is improved. And the accuracy and applicability of neurocognitive function evaluation are improved.

The present application relates to the field of three-dimensional tunnel geological model creation, and particularly relates to a method for creating a three-dimensional tunnel geological model based on Dynamo, which greatly improves modeling efficiency and model accuracy. The method for creating a three-dimensional tunnel geological model based on Dynamo comprises the following steps: analyzing design drawings, obtaining pile-by-pile coordinate data, and processing the pile-by-pile coordinate data; generating a road center line through a Dynamo program according to the processed pile-by-pile coordinate data; creating a geological entity model through the Dynamo program according to the road center line; importing the geological entity model into a project file, and assigning corresponding color materials to different geological entities. The present application is suitable for creating a three-dimensional tunnel geological model.

The present invention proposes a scientific research project expert matching method based on dynamic academic cooperation network graph representation learning, aiming to improve the efficiency and quality of expert matching in scientific research project management. By constructing a dynamic complex network graph, taking scholars as nodes and cooperation relationships as directed edges, this method comprehensively considers the scholars' past research history and cooperation experience. Using graph structure learning and Transformer architecture, feature vector learning is performed on scholar nodes to achieve accurate matching between scientific research projects and experts and scholars. This method has achieved a matching accuracy of more than 80% and a standardized discounted cumulative return of more than 89% on data sets in the fields of artificial intelligence and data mining, effectively solving the limitations of existing technologies in computing costs and the use of scholar feature information, and providing an innovative solution for recommending suitable collaborators for scientific research projects.

The present invention discloses a breeding environment control method and system based on environmental characteristics and pig behavior characteristics, and relates to the field of breeding environment control technology. The present invention can automatically control the temperature and humidity, carbon dioxide and ammonia concentrations in a pig house in real time and timely according to the needs of pigs, realize automated environmental control, and improve the health and welfare level of pigs; according to the environmental characteristics in the pig farm and the behavioral characteristics of pigs, the pig house is regulated to a suitable temperature and humidity environment to prevent pigs from having stress reactions, and can timely prevent the concentration of harmful gases in the pig house from exceeding the standard, and prevent the harm caused to pigs by excessive concentration of harmful gases; the environmental parameters of the pig house can be effectively predicted, and the pig behavior can be identified through the improved SlowFast network, effectively improving the identification accuracy; it has an early warning function, can timely detect abnormal conditions of equipment in the pig house and remind staff to carry out maintenance, and reduce manual labor.

The present invention discloses a model-based emotion-aware dialogue strategy learning method. The method comprises the following steps: a dialogue agent performs multiple rounds of interactions with a human user to obtain real dialogue experience, performs data cleaning on the real dialogue experience according to a task scenario, and trains a dialogue strategy model; an emotion-aware world model performs parameter adjustment based on the real dialogue experience to learn the user's actions and emotional changes; and strategy learning based on simulated experience, wherein the dialogue agent interacts with the emotion-aware world model to obtain simulated dialogue experience and optimize the parameters of the dialogue strategy model. The present invention transforms emotion-aware dialogue strategy learning from a model-free learning process into a model-based learning process, thereby eliminating the original method's reliance on a user simulator, while improving the problem of poor simulation performance in the existing deep Dyna‑Q learning framework, and improving the training efficiency of the dialogue strategy model.

This invention discloses a method for constructing a digital model of a high-energy beam pre-controlled projectile with random defects. The method includes: establishing a model of the high-energy beam pre-controlled projectile using TrueGrid software; outputting the node and element information of the mesh according to the format requirements of the finite element software; extracting node and element data from the mesh model file, dividing the element data into base element data and pre-controlled zone element data; obtaining element data of the defect-forming locations and the remaining element data in the pre-controlled zone according to a pre-defined ratio of defect formation locations and their volume in the pre-controlled zone; randomly deleting element data from the defect-forming locations; and merging the data to form a mesh model file that meets the requirements, thus obtaining the element data of the high-energy beam pre-controlled projectile with random defects. This construction method can generate controllable random defects in the pre-controlled zone, completing the defect characterization of the high-energy beam pre-controlled projectile.

The invention discloses a complex deep space probe task planning method based on an event constraint network, and the method comprises the steps: obtaining a successive execution relation between flight events, and converting the successive execution relation between the flight events into a directed vector; connecting the flight events into an event constraint network according to the directed vector; converting the event constraint network into an event constraint network adjacency matrix; searching a multi-node event and a boundary node event through the event constraint network adjacency matrix, and listing the multi-node event and the boundary node event as key events; establishing an inter-event time constraint inequality; obtaining a time constraint inequality of the key event according to the time constraint inequality between the events; and judging whether the time constraint inequality of the key event is a null set, if not, determining that the intersection is the execution time range of the flight events, and arranging the flight events. According to the invention, the problem of planning difficulty caused by many flight events and complex constraints among the events in the flight control process is solved.

The application provides a simplex family network structure robustness analysis method and relates to the technical field of control and information. According to three simplex network generation algorithms, different initial networks are constructed, and based on a structure robustness index calculation formula, analytical expressions of the structure robustness of the three simplex networks are derived theoretically, and the influence mechanism of network topology and attack mode on the system structure robustness is revealed. The simplex family network is used to describe the structure of an actual complex system, can effectively reflect the high-order interaction relationship of the actual network, and is more suitable for modeling of a real network system; the node degree-based attack mode is used to simulate the deliberate attack suffered by the actual network system, and the influence mechanism of the high-order interaction and the attack mode on the structure robustness is revealed, which can provide scientific theoretical support and technical implementation guidance for reasonably formulating a network attack defense strategy in practical applications.

The invention discloses an intelligent bagged cement loading system which comprises a conveying belt, sliding rails are arranged on the two sides of the conveying belt, a support is installed on the sliding rails through sliding blocks, the support stretches across the conveying belt, a mechanical arm is installed on the support through an installation base, and a mechanical hand for taking and placing bagged cement is installed at the tail end of the mechanical arm. An inductive sensor is installed on the side, opposite to the conveying belt, of the support. A discharging groove is formed in the starting end of the conveying belt. The invention further relates to an intelligent bagged cement loading control method, which comprises the following steps of: reading a license plate number through the camera, inputting license plate information, establishing a vehicle transportation card according to purchase information, recording vehicle load information and releasing a vehicle; according to the system and the method, the collection work of the vehicle information can be completed, the loading work of a plurality of vehicles can be simultaneously carried out according to the established information, each operation system can operate at relatively high efficiency, the overall loading efficiency and the operation efficiency of the conveying belt are improved, and the condition that the transportation vehicles are queued to wait for loading is reduced.

The invention provides a fault detection, isolation and positioning method for a general battery system, and the method can achieve the detection, isolation and positioning of various types of faults, such as a voltage sensor fault, a current sensor fault, a poor contact, a battery cell fault, and the like, only through the utilization of voltage and current data which are commonly used at present. Modeling and calculation processes are simple and convenient, and additional disassembly and transformation and sensor installation of the battery system are not needed, so that multiple simple, efficient and reliable comprehensive fault diagnosis solutions are provided for a general battery system represented by a lithium ion battery.

The invention relates to the technical field of ships, in particular to a ship tail flow guide fin appearance drawing and modeling method which comprises the following steps: importing a stern outline and a stern rib molded line, and importing outlines of a rudder and a guide pipe; drawing inner and outer side steel pipes of a diversion section, drawing inner and outer side inclined tangent lines of the diversion section, drawing a channel arc section of the diversion section, and drawing a diversion section top intersecting line; drawing a guide pipe section top intersecting line, and drawing inner and outer side inclined tangent lines of the guide pipe section; a stern steel pipe is arranged, a top intersecting line of the plotting rudder platform section is arranged, inner and outer side oblique tangent lines of the plotting rudder platform section are arranged, and a contour line of the plotting rudder platform section is arranged; and obtaining control point coordinates of each cross section line of the flow guide fin, establishing each cross section line of the flow guide fin in a three-dimensional space, and adding a longitudinal curve to form a three-dimensional curved surface of the flow guide fin. According to the method, a two-dimensional drawing of fairing of the flow guide fins can be set and drawn through parameterized design of a standardized process according to different ship tail shapes and sizes and positions of guide pipes and rudders used by the ship tail shapes, and three-dimensional modeling of the flow guide fins is rapidly completed.