73 results about "Multi objective model" patented technology

The invention provides a deep exploration AI reasoning method and system, and the method comprises the steps: generating a dynamic calculation graph based on a to-be-reasoned task, carrying out the risk assessment of the dynamic calculation graph, generating a risk prediction result, and carrying out the risk prediction of the to-be-reasoned task. The risk assessment at least comprises complexity assessment, memory occupancy assessment and data transmission quantity assessment; based on a risk prediction result, performing optimization processing on the dynamic calculation graph to generate an optimized calculation graph, the optimized calculation graph being used for constraining an inference calculation depth and an inference resource allocation range, the method converting natural language input into a semantic weighted dynamic calculation graph, and identifying a resource risk through node-level complexity assessment; performing self-adaptive pruning in combination with the permission level and reinforcement learning to generate an optimal calculation graph; elastic resource scheduling is implemented based on a multi-target model and real-time monitoring, so that the problem of low collaboration of reasoning depth and resource efficiency is solved.

Provided is a framework for fine-tuning pre-trained sequence processing models to human preferences and / or other objective(s). Instead of using reinforcement learning to fine-tune the LLM parameters towards the human preferences, example systems take a Bayesian approach which can preserve the learned prediction distributions of the pre-trained model, but adds explicit sequential preference tuned predictions in a multi-objective model fine-tuning training setup. The model can be tuned to predict posterior token probabilities conditioned on the human preferences.

The invention discloses a flexible control method for new energy grid connection, and relates to the technical field of new energy grid connection control. The method comprises the steps of obtaining grid-connected point electrical parameters through a self-measurement mode or a communication interface mode, performing harmonic feature extraction and compensation based on a deep learning network, and adopting a state estimation method to fuse multi-source data to estimate power grid parameters; and constructing a multi-objective optimization function including voltage quality, a power factor and harmonic suppression, dynamically adjusting a weight coefficient of the optimization function by using a swarm intelligence algorithm, and realizing inverter switching sequence optimization through multi-objective model prediction control. Through a dual-mode communication redundancy mechanism and local intelligent control, while a communication function of a traditional main link is reserved, a standby wireless communication link is introduced to realize real-time backup transmission of a control instruction, and through a local autonomous decision-making capability, the problem of control failure caused by a communication problem in a traditional scheme is fundamentally solved.

The invention relates to a multi-distribution-center open type vehicle path intelligent optimization method and system, and belongs to the technical field of logistics distribution optimization and intelligent transportation, and the method comprises the steps: firstly obtaining the input data of a multi-distribution-center vehicle path optimization problem, selecting a multi-distribution-center processing strategy according to the problem scale and constraint conditions, and carrying out the optimization of the multi-distribution-center vehicle path; a vehicle path optimization model is constructed, the vehicle path optimization model comprises a single-target model and a multi-target model, a multi-algorithm collaborative optimization framework is adopted for solving, and the multi-algorithm collaborative optimization framework comprises an ant colony algorithm, a variable neighborhood search optimization ant colony algorithm and a non-dominated sorting genetic algorithm; and outputting an optimal vehicle path scheme, wherein the optimal vehicle path scheme comprises a distribution route, a distribution sequence and a corresponding objective function value of each vehicle. According to the method, strategy adaptive selection and algorithm collaborative optimization are carried out, global exploration, local optimization and multi-target equalization are carried out by combining the advantages of the ant colony algorithm, the variable neighborhood search algorithm and the non-dominated sorting genetic algorithm, and the method is good in reproducibility, high in scene adaptability and high in decision support capability.

The invention discloses a power control method for a methanol reforming fuel cell power generation system, and relates to the technical field of power generation system power control, and the method achieves the precise representation of the internal dynamic coupling relation of the system through the construction of a digital twin model covering the whole methanol reforming process and the whole electrochemical conversion process of a fuel cell. Therefore, the problem of hydrogen supply lag caused by reforming reaction delay in the traditional control is effectively eliminated, and the output power of the fuel cell can be highly synchronized with the load demand; the Smith predictor is adopted to perform online estimation and compensation on methanol reforming inherent time delay, and the MPC is predictively controlled to perform real-time optimization scheduling under the constraints of economical efficiency and equipment health state in combination with a multi-target model, so that the power generation efficiency is improved, the fuel consumption and the operation cost are remarkably reduced, and the inhibition capability of the system on load fluctuation is enhanced.

The invention provides a multi-target RTA model training method and system and an automatic advertisement bidding method and system. The training method comprises the steps that embedded feature information between user equipment and a put advertisement is generated based on a heterogeneous graph constructed by the user equipment and the put advertisement, and the heterogeneous graph is constructed by adopting a graph neural network algorithm; obtaining user side features and advertisement side features after real-time processing; the user side features, the advertisement side features and the embedded feature information serve as input and are input into a DNN multi-target model for model training; and verifying the trained DNN multi-target model, and obtaining a prediction model after determining that the output index of the DNN multi-target model meets the requirement. According to the method and the device, the user equipment can be scored through the established multi-target model, and the advertisement bid can be automatically given, so that the advertisement bid efficiency and the bid effect are improved; and the advertisement cost is also reduced.

The invention relates to a flight control technology of an unmanned aerial vehicle suspension load system, and provides a multi-target nonlinear model prediction control method for an unmanned aerial vehicle-load suspension system, so as to realize accurate track tracking, active suppression of lifting rope swinging and collaborative optimization of flight smoothness. According to the multi-state multi-target model prediction control method for the unmanned aerial vehicle-load suspension system, firstly, a nonlinear model prediction control equation is constructed; discretizing the continuous kinetic model by adopting a fourth-order Runge-Kutta method to obtain a high-precision state transition equation under discrete time as a dynamic constraint; forming an optimal control function of a flight nonlinear model predictive control equation of the unmanned aerial vehicle-load suspension system; adding thrust amplitude constraint and moment boundary constraint as inequality constraint conditions of a nonlinear model predictive control equation; and finally, multi-target optimal control of the unmanned aerial vehicle-load suspension system is realized. The method is mainly applied to design and manufacturing occasions of unmanned aerial vehicle hanging load systems.

The invention discloses a county planting and breeding space collaborative optimization method, which comprises the following steps of: constructing an economic-ecological-planting and breeding coupled multi-target model, setting dynamic constraint conditions such as water resources and cultivated land, generating a Pareto non-inferior solution set by combining an NSGA-III algorithm, performing optimization such as chromosome coding and fitness calculation, and simulating a future land utilization pattern by utilizing a PLUS model. The method integrates multiple benefits, breaks through the limitation of a single target, and overcomes the defect that a traditional method is difficult to balance multiple targets; an optimization algorithm and a simulation model are deeply fused, a historical data trend is mined, and the problem of simulation and optimization disjunction is solved; through a fine constraint mechanism and optimization design, the method adapts to resource changes, improves accuracy, and provides scientific support for county planting and breeding space layout.

A method of maximizing coverage and minimizing interference in an Internet of Drones (IoD) network with multiple drones is provided. The method includes obtaining input parameters such as radio frequency parameters, drone locations, velocities, coverage areas, and the number of drones. A multi-objective model determines an initial interference and coverage based on these parameters. The iterative process continues for a predetermined time to identify the drone locations that achieve maximum coverage while minimizing interference. The approach accounts for beam angles, antenna configurations, and carrier frequencies, ensuring efficient IoD deployment. By leveraging a radio frequency-based determination model, the method dynamically adjusts drone positions to enhance network performance. The solution effectively balances coverage and interference, facilitating reliable communication in drone-based networks.

The invention belongs to the technical field of intelligent manufacturing and industrial scheduling, and provides a high-altitude platform multi-test task production scheduling method based on a clustering algorithm and multi-objective optimization, and the method comprises the steps: automatically calculating the exclusive resource demands of all working condition points in a plurality of aero-engine test tasks based on a multi-parameter coupling model; for each task, dividing a plurality of work packages by adopting a clustering algorithm based on the multi-dimensional features containing exclusive resource requirements in the task; a multi-objective model of multiple optimization objectives is constructed, all the work packages are jointly sorted globally, and the optimization objectives comprise the equipment utilization rate; and finally, under the condition that constraint of resource capacity, mutual exclusion of test cabins and task continuity is satisfied, allocating an available tester and a continuous time window to each work package, and generating an initial production scheduling plan. According to the method, the production scheduling efficiency and the resource utilization rate are effectively improved, the plan robustness is enhanced, and the method is suitable for high-complexity aero-engine test scheduling scenes.

The invention relates to an automobile thermal management system control method based on multi-target model predictive control, and belongs to the technical field of automobile thermal management systems. The method comprises the steps of establishing a discrete state space model of the thermal management system, setting a state variable, a control variable, a disturbance variable and a state space equation of the discrete state space model of the thermal management system, and then establishing a multi-target cost function and constraint conditions with stability, safety and low energy consumption of the thermal management system as targets; and then, based on fuzzy control logic, combining with an actual working condition, dynamically calculating a weight of each cost in the multi-target cost function, and finally solving an optimization problem according to the multi-target cost function under the determined weight in a discrete state space model of the thermal management system to obtain an optimal control input. And the optimal control input is transmitted to an actuator to regulate and control the automobile thermal management system. The system can adapt to various working conditions and environments, battery safety and comfort of a passenger compartment can be guaranteed, and energy efficiency and the service life of the actuator can be both considered.

The invention provides an urban environment performance prediction method and system based on multi-modal fusion and attention enhancement, and the method comprises the steps: collecting image data and numerical data, and generating a plurality of urban environment performance distribution truth value maps; preprocessing the multi-modal data to obtain an image map and a numerical value feature vector, and pairing the image map and the numerical value feature vector with the true value map to form a multi-modal data set; a multi-target model of a conditional generative adversarial network based on attention enhancement is constructed, a generator of the multi-target model comprises a two-way encoder, spatial features can be extracted based on an image map, and physical features can be extracted based on a numerical feature vector; an image coding path adopts a U-Net down-sampling structure containing a convolution block attention module, and a numerical value coding path adopts a multi-layer perceptron; and the multi-head output layer outputs a plurality of predicted urban environment performance distribution diagrams. After the model is trained, target area data are input, and three types of prediction distribution diagrams are output. According to the invention, the urban environment performance prediction effect is improved.

The application provides a multi-objective robust optimization method, system, storage medium and electronic equipment for production task planning of a multi-agent collaborative platform, and relates to the technical field of production task planning of a multi-agent collaborative platform. First, according to the production task planning resources of the multi-agent collaborative platform and a single-batch overseas order, a production task planning model with a manufacturing enterprise agent as the core is constructed under the premise of determining the sea transportation time. Then, an infinity norm ball uncertain set is designed for the sea transportation time of a large ship, a robust quadratic multi-objective model for cross-regional multi-agent production task planning is constructed, and the total delivery time deviation and total delivery cost multi-objective of the single-batch overseas order are optimized. The model considers the sea transportation plan prepared in advance by a shipping company agent and the long and uncertain sea transportation time. In addition, the model is converted into an equivalent model, a two-stage hybrid heuristic algorithm is designed for solving, and a reasonable and feasible production task planning suggestion is given for the single-batch overseas order.

The invention relates to a pulverized coal adjustable optimal distribution method based on a deep peak regulation state of a coal-fired unit. The method comprises the steps that pulverized coal conveying parameters, combustion state parameters and equipment abrasion parameters of a combustion system are collected in real time; constructing a multi-objective optimization model based on the parameters, wherein the model takes minimum equipment wear rate, maximum combustion efficiency and pulverized coal distribution balance as optimization objectives; solving the multi-target model through a dynamic iterative algorithm, generating a candidate allocation scheme, and selecting an optimal allocation strategy based on fuzzy logic evaluation; according to the optimal distribution strategy, the opening degree of a valve of a pulverized coal conveying pipeline, the angle of a distributor and an airflow guiding device are dynamically adjusted, and multi-area balanced distribution of pulverized coal in a combustion chamber is achieved; coordinated optimization control between combustion efficiency and equipment wear is achieved by collecting multi-dimensional parameters in real time and establishing a multi-target dynamic optimization model.

The invention relates to the technical field of cotton spinning production, in particular to a cotton spinning workshop autonomous scheduling method based on an improved genetic algorithm, and the method comprises the steps: taking four performance indexes of processing waiting time, processing load time, processing time and repairing time as optimization targets, and building a corresponding target function; forming a constraint condition according to the processing moment and the processing moment as a boundary condition; realizing conversion between individuals in the population and a scheduling scheme based on encoding and decoding operations in a genetic algorithm; according to the method, a multi-target model is established by adopting an optimization target, a traditional genetic algorithm is improved by means of operations such as crossover and variation, in the algorithm solving process, as the number of iterations is increased, the change of a population mean value tends to be stable, the change of the population mean value tends to be stable, comparative analysis on the algorithm solving performance is carried out through a simulation test case, and the algorithm solving efficiency is improved. The superiority and effectiveness of the improved genetic algorithm in solving the cotton spinning workshop production scheduling problem are verified through calculation.

The invention provides a control method and device of a braking system, electronic equipment and a storage medium, and belongs to the technical field of automobile braking, in the method, a collaborative architecture of AI dynamic prediction, 14-degree-of-freedom physical model constraint verification and multi-target model prediction control is adopted to carry out coupling analysis on road condition-vehicle-tire data, and a control result of the braking system is obtained. Wherein in the multi-objective model prediction control, optimization objectives comprise brake distance error, wheel cylinder pressure change rate, yaw velocity and energy recovery efficiency, the process retains the advantage of processing complex nonlinearity by AI, and the obtained four-wheel adaptive optimal brake pressure distribution scheme ensures physical rationality and also has the optimal brake pressure distribution effect. The electro-hydraulic servo braking system has the advantages that multi-dimensional optimal balance of braking distance, stability, comfort, energy efficiency and the like is achieved, the self-adaptive capacity is good, the response speed of electro-hydraulic servo braking is high, and the collaborative requirements of functional safety and energy recovery in an automatic driving scene can be met.

The application provides a robot dance automatic generation method and system based on music feature analysis, and relates to the technical field of robot control. First, a music feature spectrum covering rhythm, energy and other dimensions is constructed to convert abstract music into structured information understandable by machines, and then a dance semantic label sequence bound to a time axis is generated to provide accurate basis for action selection. By combining with a forward-looking window to predict future music features, the current and future candidate actions are filtered through a multi-objective model to realize the dual matching of action and music micro-rhythm and macro-structure. The finally generated control instruction sequence directly drives the physical robot, which breaks the dependence on video materials, realizes the automatic generation of dance with creativity and artistic expression, and improves the generation efficiency and adaptive flexibility.

The present invention provides a distribution network regulation method based on a multi-objective optimization dynamic aggregation strategy, which relates to the technical field of power system automation and control. The method includes: collecting in real time the voltage, current, power, active and reactive power, and harmonic operation data of each key node of the distribution network, and preprocessing the operation data to obtain processed data; constructing an initial optimization model including multiple optimization objectives according to the processed data; adjusting and determining the weights and priorities of each optimization objective according to the current operation state of the distribution network and each optimization objective to obtain a multi-objective optimization model optimized by weights. The present invention comprehensively improves the voltage quality of the distribution network, reduces network losses, promotes the consumption of renewable energy, enhances power supply reliability, and effectively reduces carbon emissions by dynamically constructing a multi-objective model optimized by weights, generating and executing precise regulation instructions, and real-time evaluation and closed-loop adjustment.

The invention discloses a permanent magnet synchronous motor multi-target model prediction control method based on parallel optimization, which relates to the field of motor control, and sequentially comprises the following steps: establishing a permanent magnet synchronous motor mathematical model, performing Euler discretization-based model prediction torque control, and selecting an optimal vector based on a comprehensive optimization mechanism. Compared with the prior art, the optimal vector is selected on the basis of a comprehensive optimization mechanism, so that the control system adaptively switches the priority of multiple targets under high-speed and low-speed working conditions so as to realize multi-dimensional variable global optimization and rapid dynamic coordination, and the control is enabled to meet the performance requirements of preliminary constraint conditions, so that the control efficiency is improved. Ripples of torque and flux linkage are reduced, unstable operation of the motor and extra noise interference are overcome, and the control precision and steady-state performance of a control system are improved.

The application discloses a kind of adaptive compensation method and related device for three-phase current imbalance and harmonic, belong to power system technical field;The method is based on the instantaneous reactive power theory and the quick separation and instruction generation of fundamental wave positive sequence component and unbalance compensation component of sliding discrete fourier transform;Multi-objective model predictive current control with embedded repeated learning correction mechanism to achieve high-precision tracking of compensation current instruction and system circulation suppression;Multi-objective weight factor online dynamic adjustment mechanism based on fuzzy reasoning, according to the adaptive balance between the control emphasis of tracking accuracy, circulation suppression and switching loss according to the real-time operating state of the system;And, adaptive closed-loop management architecture based on the above module cooperation, drive active power electronic converter to generate compensation current, realize the comprehensive management of three-phase imbalance and harmonic and global optimization of system performance.

The embodiment of the present invention discloses a multi-objective learning method, system, storage medium and terminal device, which are applied to the field of information processing technology based on artificial intelligence. The multi-objective learning system will determine that the multi-objective model includes a shared underlying sub-network, multiple underlying task sub-networks, multiple task feature modules and multiple task output modules, and determine that the training sample includes multiple sample information. The overall loss function related to the multi-objective initial model calculated based on the multi-objective initial model and the training sample includes: the first related information between the input features and output features of the shared underlying sub-network based on the sample information, and the second related information between the output features of the multiple underlying task sub-networks based on the sample information and the output features of the shared underlying sub-network, and then adjust the multi-objective initial model according to the overall loss function. The effect of the multi-objective model learned in the embodiment of the present invention is greatly improved.

The embodiments of the present disclosure disclose a method and device for training and predicting a response duration prediction model, a medium and a product. The method comprises: obtaining a set of sample data from a sample set as a training set, wherein the sample data comprises feature information of a sample service request and a real response duration of the sample service request; training a plurality of initial multi-objective models based on the sample data in the training set to obtain a plurality of multi-objective models, wherein the learning objectives of the multi-objective models comprise a probability that a service request response duration is within a predetermined range and a predicted response duration of a service request whose response duration is within the predetermined range, and each multi-objective model corresponds to a different predetermined range. The technical solution can more accurately predict the response duration of a service request.

The application discloses an intelligent scheduling multi-objective optimization method based on a pulse propagation algorithm, relates to the field of manufacturing scheduling, and comprises the following steps: a scheduling data preparation stage, which is used for making a pre-decision for the schedulable resources in a pool and generating data of an input model; a multi-objective model establishment stage, which describes scheduling rules and optimization targets in a mathematical modeling mode; and a pulse propagation algorithm stage, which solves the above multi-objective model by designing an epsilon-constraint and a pulse propagation algorithm. The application greatly reduces the work burden of a planner, and the batch decision and specific scheduling details can be automatically completed by an embedded advanced model and intelligent algorithm. The intelligent decision mechanism not only improves work efficiency, but also ensures the accuracy and rationality of the scheduling plan.

This invention belongs to the field of power grid investment optimization technology, and relates to a regional power grid investment optimization method based on a multi-objective programming model. It includes the following steps: Step S1: Considering the decoupling requirements between regional power grid investment and electricity generation, a multi-dimensional parameter system covering investment, electricity generation, safety, and green / low-carbon parameters is constructed; Step S2: Based on the multi-dimensional parameter system, a multi-objective programming model is constructed to maximize the net present value of investment, maximize the power grid safety margin, and maximize the investment-electricity decoupling degree; Step S3: An improved multi-objective particle swarm optimization algorithm is used to solve the model, and the Pareto optimal solution set is output to obtain the optimal investment scheme. This invention quantifies the decoupling relationship between investment and electricity generation through a multi-dimensional parameter system, and optimizes the balance between economic efficiency, power supply safety, and resource allocation efficiency using a multi-objective model, solving the problems of excessive coupling between investment and electricity generation and insufficient functional value mining in existing planning.

The present application provides a multi-objective RTA model training method, system and automatic advertising bidding method, system; the training method includes constructing a heterogeneous graph based on user devices and delivered advertisements to generate embedded feature information between the user devices and the delivered advertisements, and the heterogeneous graph is constructed using a graph neural network algorithm; obtaining user-side features and advertisement-side features after real-time processing; using the user-side features, the advertisement-side features and the embedded feature information as input, and inputting them into a DNN multi-objective model for model training; verifying the DNN multi-objective model after training, and obtaining a prediction model after determining that the output indicators of the DNN multi-objective model meet the requirements: the present application can score user devices and automatically give advertisement bids through the established multi-objective model, thereby improving the efficiency and effectiveness of advertisement bidding and reducing advertising costs.

The present invention provides a multi-objective coupling optimization and decision-making method and device for the location and discharge intensity of a newly built sewage outlet. The method includes: Step 1, constructing a coupling optimization multi-objective model for the location and discharge amount of the sewage outlet considering three objectives of regional economy, regional water quality, and regional ecological environment impact; Step 2, using the adjoint method to derive the accuracy lossless replacement expressions of the average concentration of the research period at the center point of the water quality control section in the regional water quality objective function and the regional ecological environment impact objective function, and performing objective function calculation; Step 3, using the multi-objective genetic algorithm NSGA-II to solve the multi-objective model to obtain all feasible solutions; Step 4, using the trade-off rate method to screen the feasible solutions that meet the trade-off rate requirements, given multiple groups of trade-off rates, and obtaining the optimal solutions corresponding to each group of trade-off rates; Step 5, determining the risk values of the optimal solutions corresponding to each group of trade-off rates through risk analysis, and finally determining the best multi-objective optimization plan for the sewage outlet with the lowest risk value.

The invention relates to a continuous open caisson synchronous sinking construction method based on automatic monitoring. The method comprises the following steps that S1, open caisson attitude data and soil mass data are obtained based on an open caisson sensor network; s2, on the basis of the open caisson attitude data and the soil mass data, a multi-target model is constructed through a multi-target optimization algorithm, and the model is solved to obtain an optimal jacking force distribution scheme; s3, the open caisson is controlled to sink synchronously based on the optimal jacking force distribution scheme; and S3, constructing a deviation correction resource library, and in the process of executing S3, when deviation correction needs to be executed, adjusting the optimal jacking force distribution scheme and the earth cutting position and amount. Compared with the prior art, the method has the advantages that synchronous movement under the condition of multiple open caissons is controlled, uneven stress and position deviation due to speed difference are prevented, and the like.

The application provides a transformer area energy storage system operation regulation method and device, and belongs to the technical field of power system regulation. The method comprises the following steps: collecting transformer area historical data and constructing a historical database; training a prediction model based on the historical database, inputting real-time data of the transformer area into the prediction model for ultra-short-term prediction, and outputting future power output and load demand prediction values; taking the power output and load demand prediction values as the basis, taking the elimination of line overload, three-phase imbalance and voltage out-of-limit as the optimization target, considering the service life of the energy storage battery and the electricity economy, constructing a multi-objective model and solving the model to obtain an instruction sequence; issuing the instruction sequence to the transformer area energy storage system for execution; collecting real-time data; calculating the execution deviation according to the transformer area actual data, and adaptively updating and correcting the parameters of the prediction model and the multi-objective model. The application solves the transformer area power quality problem, prolongs the service life of the energy storage battery, and reduces the operation cost through intelligent prediction and optimization regulation.

The invention belongs to the technical field of intelligent control of a light storage system, and particularly relates to a light storage integrated intelligent operation method and system based on deep learning and multi-target model prediction control. And based on the photovoltaic output prediction value and the user load prediction value, a multi-target model prediction controller (M-MPC) is utilized to solve a multi-target optimization problem including economic cost calculation and battery attenuation cost calculation, and an optimal charging and discharging power sequence of the energy storage system is obtained. According to the technical scheme, through the synergistic effect of deep learning model prediction and multi-objective optimization control, the full life cycle operation cost of the energy storage system is reduced, the photovoltaic self-generation and self-use rate is improved, the impact on a power grid is reduced, a full-automatic closed-loop control system is realized, and the autonomy and intelligence of the optical storage system are improved.

Approximating a more complex multi-objective feed item scoring model using a less complex single objective feed item scoring model in a multistage feed ranking system of an online service. The disclosed techniques can facilitate multi-objective optimization for personalizing and ranking feeds including balancing personalizing a feed for viewer experience, downstream professional or social network effects, and upstream effects on content creators. The techniques can approximate the multi-objective model-that uses a rich set of machine learning features for scoring feed items at a second pass ranker in the ranking system-with the more lightweight, single objective model-that uses fewer machine learning features at a first pass ranker in the ranking system. The single objective model can more efficiently score a large set of feed items while maintaining much of the multi-objective model's richness and complexity and with high recall at the second pass ranking stage.