885results about How to "Prevent overload" patented technology

The invention discloses a charging pile power regulation and control method and system, and particularly relates to the technical field of charging pile management, data is collected from a multi-dimensional data source and summarized to form a preliminary information set, and a standard information set is obtained through format standardization processing; feature data of the charging pile are extracted based on the standard information set, and charging task analysis and resource occupation analysis are carried out respectively; inputting the analysis result into a pre-trained machine learning model to generate a charging priority coefficient; constructing a priority matrix, and dividing the charging piles into high priorities and low priorities according to the charging priority coefficients; and finally, dynamically adjusting a power distribution strategy according to the priority type, optimizing low-priority task power distribution while meeting high-priority task requirements, and improving the operation efficiency and the resource utilization rate of the charging station. According to the method, the real-time operation state and the charging task characteristics of the charging piles are combined, the power distribution strategy is dynamically adjusted, the priority division and power distribution proportion of the charging piles are adjusted in real time, and the self-adaptive capacity of operation of the charging station is improved.

An energy scheduling system of a high-power liquid cooling charging pile belongs to the technical field of liquid cooling charging piles, and comprises a power scheduling module, a liquid cooling management module, a data acquisition and monitoring module, an energy storage and energy recovery module, a communication and interaction module, an intelligent scheduling module and a remote management and analysis module. Through the intelligent scheduling and energy optimization technology, power distribution can be adjusted in real time according to the vehicle demand and the power grid state, overload is avoided, and the charging efficiency is improved; and secondly, an advanced liquid cooling management technology is adopted, and temperature monitoring and cooling liquid flow rate adjustment are combined, so that heat balance and stable operation of the equipment under a high-load condition are ensured, the service life of the equipment is prolonged, and energy waste is reduced. In addition, the system can store redundant electric energy and release the electric energy in the peak load period of the power grid through the energy storage and energy recovery module, so that the power grid pressure is effectively reduced, and the energy utilization rate is also improved.

The invention relates to the technical field of low-power-consumption AI processors, and discloses a load balancing method of a low-power-consumption AI processor, a chip and a storage medium, the method comprises the following steps: performing cache resource division on a plurality of neural mimicry cores in the low-power-consumption AI processor to obtain a double-layer cache structure; performing characteristic analysis on the target calculation task, and constructing a calculation task directed acyclic graph and a dependency intensity matrix; performing task segmentation to obtain a subtask division scheme; creating a task-core mapping table; organizing the neural mimicry core into a plurality of concentric ring layers based on the task-core mapping table to obtain a multi-layer ring cooperative computing network; and calculating a plurality of performance indexes of each neural mimicry core based on the multilayer annular cooperative computing network, and performing dynamic load optimization to obtain a dynamic load optimization strategy. According to the invention, the load balancing of the low-power AI processor is realized, and the calculation efficiency and the energy utilization rate are improved.

The invention discloses a big data dynamic allocation and optimal scheduling method based on reinforcement learning. The method comprises the following steps: S1, constructing a real-time operation data set; s2, preprocessing the real-time operation data set, and constructing a reinforcement learning environment model; s3, initializing a hierarchical reinforcement learning model; s4, generating an optimal data distribution and scheduling strategy based on the hierarchical reinforcement learning model; s5, adjusting data distribution and task scheduling of each computing node in real time to balance the system load and maximize the resource utilization rate; s6, the hierarchical reinforcement learning model continuously monitors the system state and dynamically updates the strategy, and self-adaptive adjustment is carried out on resource allocation and scheduling in time when the system load or environment changes; and S7, through multi-round iterative optimization, gradually converging the hierarchical reinforcement learning model to an optimal strategy. Through application of reinforcement learning, automatic optimization of resource allocation and scheduling strategies is realized, and the resource utilization rate, the scheduling efficiency and the adaptability of the system are remarkably improved.

The invention discloses a cloud computing storage, calculation and training resource allocation method based on reinforcement learning and dynamic resource scheduling. The method comprises the following steps: S1, establishing a unified resource scheduling framework; s2, constructing a state space and an action space; s3, generating a state representation for predicting future resource demands; s4, initializing a reward function of key performance index weighted comprehensive evaluation; s5, generating an initial storage resource, computing resource and training resource allocation strategy; s6, continuously adjusting a reward function of the reinforcement learning model based on the self-supervised learning feedback signal; s7, dynamically adjusting a distribution scheme of the storage resources, the computing resources and the training resources through quick response of the resource scheduling framework; s8, optimizing resource allocation among a plurality of cloud computing nodes or platforms; s9, automatically identifying different stages of the current task through a self-supervised learning model; and S10, finally generating an optimal allocation scheme. According to the staged resource scheduling method, the condition of overload or insufficiency of resources is effectively avoided, and the task execution time is greatly shortened.

The invention discloses a data analysis method, system and device based on distributed communication and a medium, and relates to the technical field of data processing, and the method comprises the steps: monitoring a node state in real time based on a distributed communication network, obtaining a node resource state data set, building a node performance evaluation model through a machine learning algorithm, and obtaining a node performance evaluation result; computing resources of the data analysis tasks are dynamically allocated to optimize resource allocation, a task allocation strategy is designed according to requirements and node performance of the data analysis tasks, and a distributed data processing framework is established; the method comprises the following steps: performing distributed decomposition and executing a calculation data analysis task, aggregating and collecting distributed calculation results, and dynamically adjusting and optimizing resource task allocation to obtain a final data analysis result, so that a node state can be monitored in real time, and a resource allocation scheme can be intelligently adjusted; according to the advantages of dynamically optimizing task allocation according to the communication condition and the computing power of the node, the problems of insufficient dynamic resource management and poor task scheduling flexibility in the prior art are solved.

The invention discloses a three-degree-of-freedom series-parallel hybrid humanoid robot waist structure, and relates to the technical field of robots. The defects that an existing multi-degree-of-freedom waist structure is not compact enough, limited in precision, low in rigidity and poor in stability are overcome. The robot comprises a yaw joint mechanism and a pitching-sidesway joint mechanism which are connected in series, upper limbs and legs of the robot are connected through the yaw joint mechanism and the pitching-sidesway joint mechanism, and the yaw joint mechanism can drive the upper limbs of the robot to do yaw motion; the pitching-sidesway joint mechanism comprises a movable platform, a static platform, a joint universal joint and two joint driving assemblies, the movable platform is fixedly connected with the upper limbs of the robot, the static platform is fixedly connected with the legs of the robot, the joint universal joint is connected with the movable platform and the static platform, and the two joint driving assemblies are symmetrically arranged on the left side and the right side of the joint universal joint; and the two joint driving assemblies are connected with the movable platform and the static platform correspondingly, and when the two joint driving assemblies drive the movable platform, the movable platform can do pitching motion and / or side-sway motion around the joint universal joint.

The invention relates to a transformer area self-healing optimization control method and system based on flexible interconnection. The method comprises the following steps: S1, constructing an operation monitoring network based on equipment ledgers in a transformer area; s2, deploying flexible interconnection equipment at key nodes of a transformer area to realize fault isolation and power failure transfer, dynamic load transfer and voltage regulation, and flexible cooperative control of a distributed power supply and energy storage equipment; s3, introducing an equipment state evaluation model based on the transformer area operation data obtained by the operation monitoring network, and obtaining state evaluation and risk early warning of each equipment in the transformer area; s4, constructing a flexible interconnection dynamic topology management mechanism between the stations, and dynamically adjusting the interconnection topology between the stations according to the load transmission demand and the power supply condition; and S5, based on state evaluation and risk early warning of each device, performing load regulation and control and fault recovery according to a dynamic topology management mechanism of flexible interconnection. According to the invention, efficient and intelligent transformer area operation and power supply reliability improvement are realized.

The invention relates to the field of cloud computing and distributed computing, and discloses a cloud data processing system based on an artificial intelligence algorithm, which comprises a data input module, a task decomposition module, a resource scheduling module, a parallel computing module and a result reconstruction module, the data input module receives high-dimensional data and converts the high-dimensional data into a tensor structure; the task decomposition module decomposes the tensor, generates a core tensor and an orthogonal matrix, and generates a plurality of sub-tasks; the resource scheduling module is used for carrying out resource allocation and scheduling on the sub-tasks, and task execution time and resource consumption are minimized; the parallel computing module distributes subtasks to the computing nodes for execution, and the task execution sequence is guaranteed; and the result reconstruction module combines and verifies the execution results of the subtasks to generate a final calculation result. The cloud data processing efficiency and the resource utilization rate are effectively improved, the accuracy and consistency of calculation results are ensured, and the method is suitable for the fields of artificial intelligence training, large-scale data analysis, scientific calculation and the like.

The invention provides a self-adaptive computing power scheduling system for large model training, and relates to the field of electric digital data processing, the self-adaptive computing power scheduling system comprises a computing power management module, a task analysis module, a self-adaptive scheduling module and a data interaction module, the computing power management module is used for managing and monitoring overall computing power resources, and the task analysis module is used for analyzing the overall computing power resources; the task analysis module is used for analyzing requirements and characteristics of training tasks, the self-adaptive scheduling module performs computing power scheduling according to real-time resource conditions and task requirements, and the data interaction module is used for managing data interaction among different computing power nodes; according to the system, the training efficiency can be effectively improved by assigning the training tasks to different nodes and adjusting the allocated resources in the task execution process.

The invention relates to task scheduling, in particular to an intelligent task optimization method based on a self-adaptive multi-dimensional dependency graph and dynamic scheduling, which comprises the following steps of: inputting task information and resource information; constructing a self-adaptive multi-dimensional dependency graph according to the task information; in the task execution process, the task state of each task is monitored in real time, and the self-adaptive multi-dimensional dependency graph is dynamically updated; calculating the effective priority of the current task, and updating the effective priority of the current task according to the updated adaptive multi-dimensional dependency graph; dynamically scheduling the current task by adopting a dynamic scheduling algorithm; monitoring a task execution state and a resource use condition of each task in real time; according to the task execution state and the resource use condition, whether dynamic scheduling needs to be adjusted or not is judged until all tasks are executed; according to the technical scheme, the defect that in the prior art, accurate and efficient task optimization scheduling is difficult to carry out under the multi-dimensional dependency relationship between the tasks can be effectively overcome.

The invention relates to the technical field of power grid planning, in particular to a charging pile planning method based on dynamic analysis. The method comprises the following steps: obtaining electric vehicle use data in a target area, and carrying out charging demand prediction analysis according to the electric vehicle use data to obtain a charging demand prediction model; based on the charging demand prediction model, performing multi-scene simulation to obtain charging demand distribution map data, multiple scenes including different traffic flows, different driving modes and different charging demand scenes; obtaining geographic information data of the target area; and constructing a spatial analysis model according to the geographic information data, and evaluating the candidate position of the charging pile according to the charging demand distribution map data, thereby obtaining position suitability score data. According to the method, different scenes are simulated, so that the adaptability and robustness of planning are enhanced.

The invention belongs to the technical field of computing power scheduling, and particularly relates to a computing power distribution scheduling method and system for an edge computing platform, and the method comprises the steps: generating a business pool based on a historical resource request task; constructing a business computing power demand prediction model based on XGBoost, adding the influence factors into the prediction model, generating resource request business prediction output of each node device, updating the business pool according to the prediction output, evaluating the computing power scale required by the resource request business in the updated business pool, and obtaining the computing power demand; according to the computing power demand obtained by each node device, taking energy consumption minimization and response time minimization as targets, carrying out multi-target solution based on an improved firefly algorithm, and obtaining a computing power distribution scheme which enables the energy consumption and the response time to reach the optimal; and allocating schedulable computing power resources in the edge computing platform based on the computing power allocation scheme. According to the method, overload of part of the data centers can be avoided, meanwhile, the overall calculation efficiency is improved, and calculation power resources are reasonably utilized.

The invention discloses a vehicle power supply distribution control management system and method, and relates to the technical field of vehicle power supplies, and the method comprises the steps: collecting the historical driving data of a vehicle and the driving behavior data of a driver, analyzing the driving habit of the driver, and monitoring the driving state of the vehicle in real time through combining with a vehicle-mounted sensor; comprehensively judging whether the vehicle enters a high-load state or not; when the vehicle enters a high-load state, the charging and discharging power of each battery module is dynamically adjusted according to the temperature distribution and load requirements of each battery module in the battery pack; evaluating the power supply condition of each battery module according to a power adjustment result, sending an early warning prompt to a driver if the power supply is insufficient, and continuously monitoring and analyzing the temperature difference between the battery modules in the battery pack and predicting the temperature change of each battery module if the power supply is sufficient, and the cooling strategy of the vehicle system is intelligently adjusted, and temperature balance of all battery modules of the vehicle in the high-load state is achieved.

The invention relates to the technical field of signal analysis and processing, in particular to an underground flow switch control method based on vibration signal analysis. The method comprises the steps that 1, vibration signals of a pump are collected through a three-axis vibration sensor, and acceleration time domain signals in the three-axis direction are obtained; performing multi-dimensional vibration spectrum component decoupling on the acceleration time-domain signal to obtain a time-frequency characteristic matrix of three-axis vibration; 2, calculating a Reynolds stress tensor based on the turbulence intensity; calculating a flow state transition criterion value according to the turbulence intensity and the Reynolds stress tensor, and establishing a flow estimation model; 3, according to the flow estimation model, dynamic response of underground flow switch control is generated; and according to the dynamic response, a self-adaptive control law for controlling the underground flow switch is generated by using a self-adaptive control method. According to the invention, the limitation of a traditional pressure sensing and flow sensing technology in a complex underground flow state is effectively overcome, and the real-time performance, the accuracy and the environmental adaptability of flow control are remarkably improved.

The invention discloses a method and a system for establishing a comprehensive load model containing a distributed power supply, and relates to the technical field of load model establishment, after the establishment system obtains power system information through an API interface, a topological graph model of a power system is established, each component in the power system is abstracted as a node and an edge in the topological graph model, and the topological graph model is established; after simulation software is used for conducting simulation detection on the topological graph model, a target value of the topological graph model is calculated through an objective function, an optimal path between nodes is updated through a path algorithm, current distribution of edges is adjusted through a graph cut algorithm, power loss is reduced, overload is avoided, the steps are repeated for multiple iterations, and when convergence conditions are met, the optimal path is obtained. And outputting the final topological graph model as a comprehensive load model. The system can optimize the targets at the same time through a graph optimization algorithm, improves the overall benefits of the system, and overcomes the defect that the complexity of the power system cannot be comprehensively considered in single target optimization.

The invention provides a solar-powered motor protection method and system, and is applied to the field of equipment data processing. The output torque of the motor is adjusted in real time according to load change by using a dynamic torque compensation algorithm, normal operation of the motor can still be maintained under the condition of low electric quantity, sudden power failure caused by electric quantity fluctuation or load change is avoided, work interruption caused by power mismatching of the motor can be effectively reduced, and the service life of the motor is prolonged. Meanwhile, loads are dynamically distributed, power management is optimized, it is ensured that all parts of the system operate under proper voltage, current and torque, overload or long-time low-efficiency work is avoided, a multi-layer protection strategy is applied to carry out delay protection on the motor, harmonic interference can be effectively restrained, temperature rise can be effectively controlled, and the service life of the system is prolonged. And standby power supply support is provided, so that safe transition of the motor can be ensured under the condition of current fluctuation, and unnecessary shutdown is reduced.

The invention discloses a flexible fresh air control method and system for an intelligent stadium, and the system comprises a data collection module which is used for detecting the visitor flow rate, the age and the somatosensory comfort degree, and collecting the operation parameters of a fan motor and environment parameters; the big data analysis and deep learning module comprises a somatosensory comfort model, a motor operation optimization model and an age and physiological difference adjustment model, and is used for dynamically outputting control parameters of the fresh air system and the air conditioning system according to the data acquired by the data acquisition module; and the control execution module is used for adjusting a fresh air system, an air conditioning system and motor frequency conversion speed regulation and load distribution of the stadium according to the control parameters output by the big data analysis and deep learning module. The invention relates to the technical field of artificial intelligence, and can solve the problems of low intelligent degree and low control refinement degree of a fan control system in the prior art.

The invention relates to a load regulation and control method and system based on reinforcement learning, and the method comprises the steps: 1, decomposing a multi-time scale optimization problem into a plurality of layers in a source network load storage management and control scene through employing a layered optimization frame, enabling each layer to correspond to a different time scale, and carrying out the optimization; a model prediction control algorithm is adopted in the short-term optimization layer, a mixed integer linear programming algorithm is adopted in the medium-term optimization layer, and a reinforcement learning method is adopted in the long-term optimization layer to solve a multi-objective optimization problem; and 2, long-term load regulation and control decision making: controlling the power grid to make different load regulation and control decisions through an actually adjustable reward mechanism based on reinforcement learning, so that the intelligent power grid load regulation and control system adapts to continuously changing environments and targets, and the flexibility and adaptability of the system are improved. According to the invention, the reliability and stability of power supply can be ensured, fine management and regulation of power loads are realized, and the operation efficiency, reliability and safety of a power system are improved.

The invention relates to the technical field of data detection, and discloses a real-time big data stream processing and anomaly detection system, which comprises a data acquisition module for collecting big data streams from a plurality of sources in real time, and the data acquisition module is equipped with an intelligent adapter for automatically identifying and adapting to data sources of different formats and protocols; the high-performance data preprocessing module is integrated with dynamic data cleaning, analysis and standardization functions and is used for processing the collected big data flow; the distributed stream computing platform is used for executing complex data analysis and computing tasks by real-time data streams; the anomaly detection module integrates a machine learning algorithm and a statistical analysis algorithm, and the algorithm is an unsupervised anomaly detection algorithm; the system can efficiently process real-time big data streams and accurately detect anomalies, and has intelligent resource management and strong safety protection capabilities. An advanced unsupervised anomaly detection algorithm is adopted, and a visual monitoring and alarm mechanism is provided.

The invention relates to the technical field of computing power optimization, and provides a computer and cloud computing power optimization method based on multi-task collaboration. The method comprises the following steps: receiving Q to-be-executed tasks of a local cloud collaborative network; performing priority constraint on the Q to-be-executed tasks according to the multivariate priority evaluation factors, and establishing a task chain; collecting callable computing resource information of a local cloud collaborative network, and establishing a resource space; based on the resource space, performing task chain resource allocation by utilizing a computing resource scheduling optimization algorithm to obtain a scheduling strategy chain; performing cloud computing power balance optimization according to the scheduling strategy chain to obtain an optimized chain; and executing the task chain according to the optimization chain based on the cloud collaborative network. The technical problems of unbalanced scheduling and low resource utilization rate between the cloud and the local computing resources are solved, and the technical effects of improving the cooperative work efficiency and the resource utilization rate of the cloud and the local computing resources through priority constraint and computing resource scheduling optimization are achieved.

The invention provides a photovoltaic array optimization generation method and device, electronic equipment and a storage medium, and the method comprises the steps: calculating a target inclination angle of a photovoltaic module unit in a target photovoltaic power station based on meteorological data and factory site geographic data of the target photovoltaic power station; obtaining an inverter type of a target photovoltaic power station, and determining a target number of photovoltaic module units connected with each inverter according to the inverter type; according to the factory site geographic data, calculating a target distance between the photovoltaic module units; and inputting the target inclination angle, the target number and the target spacing into an array optimization model for array optimization, and generating an optimized array of the target photovoltaic power station. Compared with the prior art, according to the embodiment of the invention, illumination absorption is maximized through the target inclination angle obtained through calculation, so that the power generation efficiency is improved; and the target inclination angle, the assembly number and the spacing are input into the array optimization model, so that a scientific basis can be brought to the design of the photovoltaic power station, and the overall performance and economic benefits of the system are further improved.

The invention discloses a self-adaptive resource scheduling method for improving cluster reconstruction efficiency, and relates to the field of resource scheduling distribution, and the method comprises the steps: monitoring the resource use data of each node in a cluster in real time through a monitoring agent, and obtaining the health state of the node through a heartbeat mechanism; collecting input task requests, predicting the resource demand quantity of each task, and performing preliminary queue processing on the tasks to obtain a task queue group; setting a resource competition constraint according to the task queue group and the resource use data and the health state of each node, obtaining dynamic resource scheduling corresponding to each task by using an optimization algorithm, and allocating the tasks to each node; a resource scheduling dynamic graph is constructed, a visual interface is set, and the resource scheduling dynamic graph is arranged on the visual interface and used for displaying resource scheduling, providing a user-friendly interface and realizing high-efficiency scheduling of adaptive resources in a cluster reconstruction process.

The invention relates to the technical field of industrial equipment fault diagnosis and intelligent control, and discloses an industrial equipment fault diagnosis system and method based on image recognition, and the system comprises a data collection module which is used for collecting the multi-modal data of industrial equipment, and the multi-modal data comprise equipment state data, environment data and image data; and the data preprocessing module is connected with the data acquisition module and is used for preprocessing the acquired multi-modal data, including smoothing, denoising and anomaly detection processing of the equipment state data, the environment data and the image data. The environment data, the equipment state data and the image data are subjected to weighted fusion through an attention mechanism, so that feature weights of different data sources can be dynamically adjusted, the actual state of the equipment can be reflected more accurately, a feedback control module is integrated, the running state of the equipment is dynamically adjusted according to a real-time fault diagnosis result, and the fault diagnosis accuracy is improved. And the closed-loop design realizes self-adaptive adjustment of the equipment state, so that the fault risk is effectively reduced.

The invention relates to a park equipment operation monitoring method and system, computer equipment and a storage medium, and the method comprises the steps: carrying out the real-time collection of operation parameters of energy nodes in a park, carrying out the denoising and standardization processing of the operation parameters through signal processing, and obtaining standard parameter data; obtaining a park task list, monitoring the standard parameter data in real time based on a preset operation rule, and generating a state evaluation report of the power distribution system; key operation data in the ground source heat pump system are collected, evaluation is carried out through a trend analysis algorithm in combination with historical operation data, and an equipment operation prediction report is generated; acquiring real-time environment monitoring data and video monitoring data of a preset important area, performing environment state evaluation and safety state evaluation on the park, and generating an environment safety monitoring report; and constructing a virtual three-dimensional model of the park, and integrating the monitoring parameters to the virtual three-dimensional model. The method and the device have the effect of improving the monitoring efficiency of the equipment in the park.

The invention discloses a task migration method and device and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining a calculation graph corresponding to a calculation task of a source device; through a linear programming model, target equipment is allocated to each computing node in the computing graph, and the target equipment is used for executing each computing node; and optimizing the operation of each computing node according to the device resource vector of the target device and the computing demand of each computing node. According to the method, the calculation graph is reconstructed, so that the task can be efficiently executed on the target equipment after being migrated, and the execution efficiency of the task is improved.

The invention discloses an industrial silicon intelligent power distribution and control system and method. The system comprises a data acquisition and processing module, a shared feature extraction module, an intelligent control module and a power distribution module which are connected in sequence. Multi-mode operation data is collected in real time, a shared feature extraction module is adopted to perform feature extraction on the operation data, data input is provided for subsystems of the intelligent control module, and information isolated island is avoided; through cooperative work of all the modules, global optimization control over the industrial silicon production process is achieved. The intelligent control module comprises an intelligent industrial silicon submerged arc furnace baking system, an automatic electrode lifting system, an automatic submerged arc furnace transformer gear adjusting system and an automatic electric energy quality adjusting system, and each system performs task prediction based on corresponding characteristics by adopting a pre-trained model and outputs optimized parameters; and the power distribution module generates a power distribution instruction based on the optimization parameters, so that intelligent distribution of power resources is realized, and dynamic adjustment and optimal distribution of energy consumption are realized.

The invention discloses a method for crushing and recycling waste diapers, and relates to the technical field of resource recycling. The operation state of gravity separation equipment and the crushing condition of waste diapers are monitored in real time, and fault identification and risk analysis are performed on equipment power consumption fluctuation and material particle size distribution by utilizing feature extraction, so that early warning signals of corresponding levels are generated; the maintenance sequence is determined according to the risk level of the equipment, and the material putting amount and frequency are dynamically adjusted, so that the equipment is always operated in the optimal state; therefore, equipment overload and faults are effectively prevented, the service life of the equipment is prolonged, the maintenance cost is reduced, distribution of maintenance resources can be optimized, production stagnation caused by the equipment faults is reduced, the production efficiency and economic benefits are improved, and the continuity and high efficiency of the waste baby diaper treatment process are ensured.

The invention discloses a tree coding-based long video generation optimization method, which comprises the following steps of: constructing a video frame sequence, and obtaining frame features; clustering the frame features to obtain a correlation score between the key frame and the video query; calculating the number of high-correlation clusters, and outputting a corresponding frame feature set and a correlation score; key frames of all the high-correlation clusters serve as first-layer nodes of a video feature tree, the high-correlation clusters are clustered into w sub-clusters respectively, a second layer is constructed based on the sub-cluster key frames, the correlation score of the second-layer nodes is obtained by subtracting one from the correlation score of the first-layer nodes, each layer is constructed according to the method, and the first-layer nodes and the second-layer nodes are obtained; the relevance score of the last layer of nodes is 1; the large language model outputs an optimization reasoning result based on the speculation sampling principle and the draft model; and inputting the optimized reasoning result and the video frame sequence into a preset video generation model to obtain an answer video. According to the method, the reasoning accuracy and efficiency are improved by inquiring self-adaptive and hierarchical long video understanding.

The present invention provides a wind-driven direct-driven refrigeration device and a design method for protecting permafrost, which belongs to the field of permafrost engineering technology. The refrigeration device includes a cabinet and a wind turbine unit installed inside the cabinet, a transmission speed-changing mechanism unit and a refrigeration unit. The wind turbine unit is used to convert wind energy into mechanical energy, the transmission speed-changing mechanism unit is used to increase the speed of the wind turbine unit output and the output speed can be adjusted autonomously, the output speed of the transmission speed-changing mechanism unit is limited within a reasonable range, and the refrigeration unit is used to transport cold to the inside of the permafrost foundation to refrigerate the permafrost and protect the permafrost. The wind-driven direct-driven refrigeration device and a design method for protecting permafrost provided by the present invention solve the technical problem of low energy utilization rate of the equipment and methods used for protecting permafrost in the prior art, and have the technical effects of reducing energy loss, high energy utilization rate, high efficiency of the refrigeration system, and good practicality.