52results about How to "Avoid response delays" patented technology

The invention relates to the technical field of resource allocation management, in particular to a first-aid medical resource allocation management method and system, a medium and a product, and the method comprises the steps: obtaining call-for-help appeal information, and determining first-aid preparation information based on the call-for-help appeal information, the first-aid preparation information comprising a first-aid route and first-aid vehicle-mounted resources; carrying out patient receiving and transporting based on the first-aid preparation information, and collecting real-time road condition information, patient sign information and patient medical history information in the receiving and transporting process; adjusting a first-aid route based on the real-time road condition information and the physical sign information of the patient, and recording receiving and transporting driving information; real-time hospital rescue resource information is acquired, and on-site medical resources and resource preparation time are determined based on the real-time hospital rescue resource information, the patient sign information and the patient medical history information; on-site preparation information is determined based on the on-site medical resources and the resource preparation time. According to the invention, rationality and accuracy in a first-aid medical resource allocation process can be improved, so that a patient can be ensured to be timely and effectively rescued.

The invention belongs to the technical field of power generation, and discloses a power generation gas turbine grid-connected power allocation technology and device which comprises a power detection module, a data processing module and a power regulation and control module. And the output power of the gas turbine is detected in real time through the power detection module, so that the real-time performance of power allocation is ensured. The power detection has the advantages of single current detection, simplicity, linearity and high efficiency. And accuracy: through the data processing module and a preset algorithm, a power regulation and control instruction can be accurately calculated, and accurate allocation of the grid-connected power of the gas turbine is realized. The power allocation effect is that any proportion can be adopted for adjustment, and the power error is reduced; and stability: the output power of the gas turbine is regulated and controlled through the power regulation and control module, so that the stability of the power grid can be effectively improved. According to the power allocation effect, the problem that power oscillates between the two gas turbines does not occur.

The invention provides a method and system for determining reply verbal skills of an outbound robot, and relates to the technical field of conversation robotics.The method comprises the steps that tone, volume and voice activity rate features of user voice signals are extracted, and voice feature vectors are generated; converting the voice signal into a text, generating a text semantic vector, and combining to generate a user state vector; generating a context semantic vector and an intention semantic vector based on a RoBERTa language model, calculating an intention correlation score, calculating an intention completion rate through a prediction model, generating a priority score in combination with an emergency degree, selecting a user intention according to a priority score function, training a reply model based on historical verbal skill data, and obtaining a reply result. The most appropriate recovery verbal skill is generated in real time, and the intelligence and adaptability of robot recovery are improved.

The invention discloses a multi-person interaction method based on a large-space computing technology, and relates to the technical field of multi-person interaction. In order to solve the technical problem that in the prior art, most of existing multi-user virtual interaction systems depend on a single interaction area, so that the user experience is poor when multiple users interact at the same time, the technical scheme provided by the invention is as follows: a multi-user interaction method based on a large-space computing technology is provided. Comprising the following steps: dividing a virtual space into a plurality of functional areas; controlling the temperature of the interaction area and the non-interaction area in real time; sharing of virtual roles is realized through the cloud computing platform; a sharing conflict coordination mechanism is introduced, an operation request of a user for a virtual role is managed, and conflicts are coordinated; creating a virtual computing space and dynamically allocating computing resources by combining distributed computing and a cloud platform; a digital co-construction and sharing platform is built through a 5G network and a holographic image technology, and multi-user interaction and co-construction and sharing of virtual articles are achieved. The method is suitable for being applied to virtual reality scene building work of multi-person interaction.

The embodiment of the invention provides a printing control method, electronic equipment, a storage medium and a program product. The method comprises the steps that a task queue sent by a non-real-time application kernel is received through a communication protocol, and the task queue comprises a plurality of to-be-processed tasks corresponding to the five-axis laser printing equipment; determining a task type and a real-time requirement level corresponding to each to-be-processed task; according to the task type and the real-time requirement level corresponding to each to-be-processed task, adjusting the task priorities of the plurality of to-be-processed tasks to obtain a task priority queue; determining distributed system resources corresponding to the to-be-processed tasks according to the task priority queue and the current resource state; and executing each to-be-processed task according to the task priority queue and the allocated system resources corresponding to each to-be-processed task. The method is used for achieving the effect of improving the reliability of the printing system.

The embodiment of the invention provides a subprogram instance creating method and device, equipment, a medium and a program product. The method comprises the following steps: in response to a calling request of a target subprogram in a session between a client and a database, searching a target shared instance from a shared memory; multiplexing the public data in the searched target shared instance, and creating a private instance structure; and generating a private instance stored in the private memory of the session according to the private instance structure and the private data. According to the method, when the session calls the subprogram, the shared instance is searched from the shared memory, and the private instance for the session is created by multiplexing the public data in the shared instance, so that multiple sessions do not need to repeatedly compile and store the same public data, and the memory overhead is reduced, and therefore, in a high-concurrency scene, the session can be quickly and efficiently called. The risk of insufficient memory is avoided, and the stability of database operation is improved.

The invention discloses a data deletion method, device and equipment for dynamically adjusting single-batch deletion quantity. Comprising the steps of averagely dividing to-be-deleted data in a data table into M1 batches; in the process of deleting the to-be-deleted data in batches, after N batches of to-be-deleted data are deleted every time, the single-batch deletion amount is updated based on the current CPU usage rate and the current memory usage rate of the server, batches of the remaining to-be-deleted data are divided again based on the latest single-batch deletion amount, and batch deletion operation is executed. The data to be deleted are deleted completely; specifically, based on the maximum index value in the current CPU usage rate and the current memory usage rate, a pressure state interval corresponding to the server is determined, and a pressure adjustment coefficient is determined based on an interval adjustment coefficient associated with the pressure state interval; and multiplying the pressure adjustment coefficient by the single-batch deletion quantity of the latest deletion operation to obtain the latest single-batch deletion quantity.

The invention discloses a data reporting method of a photovoltaic cleaning robot, the photovoltaic cleaning robot and a system, and belongs to the technical field of robot communication. The data reporting method of the photovoltaic cleaning robot comprises the steps of determining a relay robot in a plurality of adjacent robots under the condition that a communication link with a server is detected to meet a triggering condition, and establishing a near field communication link with the relay robot; determining a relative position relation with the relay robot; based on the relative position relation, the motion state of the robot itself is adjusted, so that the robot itself moves along with the relay robot in the mobile operation environment, and the communication stability of a near field communication link is kept; and sending the to-be-reported data to the relay robot through the near field communication link, so that the relay robot uploads the to-be-reported data to a server. Through stable near field communication between the photovoltaic cleaning robots, the data of the photovoltaic cleaning robots can be timely and completely reported to the server.

The invention relates to the technical field of data processing, and discloses a multi-modal data processing system and method supporting user-defined processing, and the system comprises a collection module which is configured to collect query information; the data processing module is configured to summarize query data types of the query entries, count query times and query interval durations of the query data types, and judge whether to add data types or not based on query results of the query data types and in combination with the query times and the query interval durations; the data storage module is configured to judge whether to adjust the storage mode of the data type or not according to the query times and the query interval duration of each data type; and the data response module is configured to dynamically adjust the response time of the query result according to the query interval duration of the data type and the interval duration influence coefficient. According to the method, the query efficiency can be effectively improved, the resource management is optimized, the user experience is improved, and high expansibility and flexibility are achieved.

The application relates to the technical field of geothermal energy utilization, and discloses a middle-deep geothermal energy intelligent control platform, a collection unit is used for a data unit, an energy unit is used for energy supply, an analysis unit is used for calculating a comprehensive temperature ZW, a mode unit is used for mode switching, an adjusting unit adjusts the opening degree of a valve in a room, a health unit is used for calculating a health value JK of equipment, an interactive unit is used for man-machine interaction, and a control unit is used for controlling the operation of the collection unit, the energy unit and the mode unit; through the arrangement of the collection unit and the analysis unit, solar radiation, outdoor temperature and other parameters are collected, and the comprehensive temperature ZW is calculated; the mode unit automatically switches a high-temperature module, a medium-temperature module, a low-temperature module and a refrigeration module according to the ZW value, the dynamic collaborative operation effect of a multi-energy system of solar energy, a ground source heat pump and an air source heat pump is realized, the self-adaptive capability to the environment and the energy utilization efficiency are improved, and energy waste and response delay are avoided.

The utility model discloses a remote switch system for a cable branch box, and relates to the technical field of power equipment. The problems that an existing system is insufficient in anti-interference capacity, unstable in wireless signal and high in power supply dependence are solved. In the scheme, the control module integrates the dual-core processor and the nonvolatile storage unit, and high-frequency noise conduction is blocked through a four-stage isolation driving circuit; the communication module is fused with a power line carrier and a dual-frequency wireless transceiver circuit to realize signal blind area automatic switching and multi-channel complementary transmission; the multi-source power supply module adopts a main power supply, a standby power supply and a photovoltaic unit to cooperatively supply power, and combines a bidirectional DC / DC module to dynamically deploy energy; the split type execution module doubly guarantees the reliability of the switch based on a magnetic latching relay and a redundant drive circuit; according to the scheme, the anti-interference capability in a complex electromagnetic environment is remarkably improved, control failure caused by signal disconnection is effectively avoided, and the continuous operation capability in a power-off scene is ensured through a multi-source energy supply mechanism.

In a drive device, a camera module, and a camera mounting device, the drive device includes: a drive holder that supports a lens unit; a housing accommodating the driving bracket; a shaft part for supporting the driving bracket; and a drive unit that moves the drive bracket, the drive bracket having a moving unit supported by the shaft unit, and a metal member embedded in the drive bracket, the metal member having a shape that follows the moving unit.

The application provides an energy-saving method and device, a terminal, network equipment, a chip and a storage medium, and relates to the field of communication. The method comprises the following steps: in response to the number of time units in which the terminal has not been continuously scheduled by a physical downlink shared channel (PDSCH) reaching a threshold, the terminal is controlled to operate in an energy-saving mode. Thus, the number of time units in which the terminal has not been continuously scheduled by a PDSCH is monitored, and when the number of time units reaches the threshold, the terminal is automatically triggered to operate in the energy-saving mode, thereby realizing an implicit energy-saving mechanism without the need for the network equipment to send explicit signaling. The additional control overhead and response delay caused by the transmission and analysis of the explicit signaling are effectively avoided, the energy-saving operation can be started immediately after the terminal actually enters an idle state, the invalid energy consumption caused by the terminal continuously monitoring a downlink control channel during the waiting period for the signaling is avoided, and the power consumption of the terminal is significantly reduced and the endurance time is prolonged.

The utility model discloses a kind of overflow collection drainage structure for water plant membrane pool, more specifically, water treatment facility safety protection technical field, including membrane pool mechanism, the side of the membrane pool mechanism is provided with water inlet mechanism, the other side of the membrane pool mechanism is provided with drainage mechanism, characterized by: the side of the membrane pool mechanism is below water inlet mechanism and is provided with cleaning drainage mechanism.Overflow collection drainage structure for water plant membrane pool described in the utility model, by setting overflow hole one, overflow hole two and overflow pipe and overflow discharge pipe located on the same horizontal line, multiple membrane pool units between overflow water horizontal flow and concentrated discharge are realized, in combination with liquid level monitoring mechanism real-time monitoring water level, once water level is abnormal, overflow discharge can be started, avoid the response delay or false alarm problem caused by traditional design in dependence on single alarm system, solve the safety hidden danger of equipment between submergence, electrical short circuit etc. caused by overflow.

The invention provides an early warning method and system for running performance abnormity of a graphics processor, and the method comprises the steps: calling a temperature monitoring interface and a video memory calculation interface based on any pre-obtained monitoring period; configuring the temperature monitoring interface and the video memory calculation interface to the graphic processor, and obtaining current temperature monitoring data and current video memory monitoring data; obtaining a temperature monitoring result based on the current temperature monitoring data and a preset temperature threshold; obtaining a video memory monitoring result based on the current video memory monitoring data and a preset video memory residual threshold value; and if the temperature monitoring result and / or the video memory monitoring result do / does not meet the preset operation requirement, generating an abnormity monitoring early warning signal so as to carry out early warning on the operation performance abnormity of the graphics processor. According to the invention, the monitoring serial port is periodically called to early warn the running performance abnormity of the graphics processor, and the running reliability of the graphics processor is improved.

The utility model relates to material processing technical field, and a kind of civil aviation material polishing equipment of easy to hold;Sliding platform is slidably arranged on the upper portion of support platform by slide rail pair, and two multi-point suction disc flexible tooling are fixedly arranged on the upper portion of sliding platform, support shaft is fixedly arranged on the rear side of support frame, lifting mechanism connected with support shaft is equipped on support platform, overturning mechanism connected with lifting mechanism is equipped on support shaft, linkage switching mechanism connected with overturning mechanism is equipped on sliding platform, and one end of electric telescopic rod is equipped with clamp, by setting two independent multi-point suction disc flexible tooling, the different curvature of workpiece front and back two sides is respectively adapted, the profiling adsorption support of "one side one tooling" is realized, and the pure mechanical linkage of lifting mechanism, overturning mechanism and linkage switching mechanism is matched, automatic completion is taken piece, overturning, shift and reposition in one clamping cycle, without manual intervention, improve polishing precision and efficiency, reduce labor intensity and workpiece damage risk.

The invention relates to the technical field of engineering machinery safety monitoring, and discloses a crane risk event handling method, terminal equipment and a storage medium, and the crane risk event handling method comprises the following steps: carrying out alarm identification on original operation data of each crane, and generating a crane risk event; calling an instruction template according to the risk characteristics of the event, and filling a parameter matched with a template content placeholder in the event into a corresponding placeholder to generate a business processing instruction; the instructions are distributed to the corresponding disposal objects; and if instruction confirmation information of the disposal object is received, retrieving a disposal plan in a risk case database based on the risk characteristics and the identifier of the business disposal instruction, so as to create a risk event disposal task and send the task to the disposal object for execution. Automation of risk disposal is achieved, and response accuracy and execution consistency are improved; state traceability, behavior traceability and responsibility definition of each link are ensured; and the intelligence, the robustness and the sustainable evolution capability of the hoisting operation safety risk control system are obviously enhanced.

The invention provides an edge cloud collaborative data processing system and method based on deep learning, and relates to the technical field of edge cloud collaboration, and the method comprises the steps: extracting a task feature vector of edge terminal equipment from a data task flow; constructing a task dependency graph, and determining an allocation label during sub-task processing through the task dependency graph and the resource allocation granularity of the sub-tasks in the data task flow; carrying out confidence adjustment on each distribution label to obtain execution confidence of the subtasks on the edge terminal equipment, and further carrying out load transfer analysis on the edge terminal equipment to obtain load transfer quantity of the edge terminal equipment; when the load transfer amount is smaller than a preset load transfer threshold value, the subtasks in the data task flow are asynchronously executed in the edge terminal device, and otherwise, the data task flow in the edge terminal device is sent to the cloud server through the edge cache server for batch processing. The task load between the edge terminal device and the cloud server can be divided under the condition that heterogeneous computing resources are limited.

The invention discloses a data migration method and device, a storage medium and electronic equipment, and relates to the technical field of data processing.The method comprises the steps that state information of different storage layers of the storage equipment is determined; the state information at least comprises the file average temperature of the storage layer and the storage cost of the storage layer; under the condition that the access frequency or size of a target file is changed, taking the state information of different storage layers as input, and outputting a file storage strategy through a decision model taking response cost and storage cost variation as optimization targets; the file storage strategy is used for indicating whether a target file executes migration operation or not, and indicating a target storage layer corresponding to the migration operation under the condition that the file storage strategy indicates that the migration operation is executed. According to the method, the storage strategy can be dynamically adapted through the decision model according to the storage layer state and the target file attribute change, the response and storage cost is balanced, the storage resource utilization rate is improved, and the total ownership cost is reduced.

The utility model relates to a kind of gas monitoring devices with quick-release filter core and back blow ash cleaning structure, belong to environmental monitoring equipment technical field, to solve the problem of existing gas monitoring device filter core easy to block, replacement maintenance difficulty.This gas monitoring device with quick-release filter core and back blow ash cleaning structure includes detector main body and filter component, gas sensor, main control board, main air inlet channel and back blow ash cleaning component are equipped in detector main body;Filter component is detachably connected in the air inlet end of detector main body by quick-release mechanism;The back blow ash cleaning component includes back blow air pump and gas path switch valve, can be under the control of main control board, realize the reverse blow ash cleaning of filter core in filter component.This gas monitoring device with quick-release filter core and back blow ash cleaning structure realizes the automatic maintenance of device by integrated back blow ash cleaning function, prolongs filter core life;Meanwhile, quick-release mechanism simplifies the replacement operation of filter core, significantly improves the reliability of monitoring and the convenience of maintenance.

The invention relates to the technical field of data communication, and discloses a network flow deep packet inspection multi-task classification method and system, and the method comprises the steps: firstly obtaining a target network flow data flow, recognizing the corresponding protocol and load content of the target network flow data flow, and generating a flow topological graph related to the target network flow data flow and the load content; collecting a topological data sequence in the atlas, analyzing a corresponding data transmission type, analyzing a type guiding mode, and fitting a flow transmission curve; identifying an abnormal traffic peak value and a core transmission hotspot in the curve, calculating a traffic throughput threshold and analyzing a network cache state; then, a deep packet detection framework is constructed based on the cache state, a processing delay value and a classification misjudgment rate of a real-time classification path in the framework are collected, and a traffic coordination index is calculated; and finally, reconstructing a data coordination task according to the index, identifying a task classification protocol, and generating a multi-task-oriented classification result of the flow data flow. According to the invention, the detection and classification efficiency of multi-task traffic in a complex network environment can be improved.

The invention relates to the technical field of cloud computing and data center operation and maintenance, and provides a data center resource scheduling optimization method based on artificial intelligence, and the method comprises the following steps: collecting multi-dimensional data; preprocessing the data; constructing and training a hybrid intelligent scheduling model; and real-time scheduling decision generation, scheduling execution and closed-loop optimization are carried out. According to the artificial intelligence-based hybrid intelligent model-driven data center resource scheduling optimization method provided by the invention, through multi-dimensional data acquisition, feature engineering optimization, hybrid intelligent model construction and closed-loop iterative optimization, multi-target collaborative optimization of a resource utilization rate, an SLA standard-reaching rate and energy consumption is realized; and the intelligent level and the dynamic adaptability of resource scheduling of the data center are improved.

The invention relates to the technical field of industrial automation control and vehicle-mounted equipment integration, and discloses a vehicle-mounted remote control terminal and method, and the vehicle-mounted remote control terminal comprises a multi-protocol communication module; a vehicle-mounted bus adaptation module; the core control module is internally provided with a task scheduling unit and a protocol analysis unit, and the task scheduling unit adopts a priority scheduling algorithm to preferentially process an emergency control instruction; the protocol analysis unit pre-stores analysis rules of various communication protocols and vehicle-mounted bus protocols and is used for identifying and analyzing various types of instructions; and the man-machine interaction module carries keys and a touch screen, is connected with the core control module and is used for providing a user interaction function. The method can guarantee the priority execution of the emergency control instruction, avoids response delay caused by instruction congestion, can directly adapt to a vehicle modification scene, does not need to greatly change an original vehicle bus system, is high in adaptability, and can be widely applied to the remote control of various vehicle-mounted equipment in logistics, engineering, special operation and the like.

The invention discloses a municipal engineering traffic accident emergency processing method, which collects accident related information in real time through equipment such as a sensor, an automobile data recorder and a traffic camera, and transmits the accident related information to an emergency resource scheduling platform in real time through a wireless network or an Internet of Things platform to ensure rapid and accurate transmission of the information. Therefore, the time delay of manual information confirmation and transmission is greatly reduced; the AI algorithm and the real-time traffic data are utilized, the driving speed and the optimal driving time of the emergency resources are calculated, and the optimal distribution strategy of the emergency resources is further calculated, so that the high efficiency and the reasonability of resource distribution are ensured, and the rescue response time can be remarkably shortened; the emergency resource scheduling platform can perform information sharing and real-time linkage with a plurality of departments such as police, a medical emergency center, a traffic management department and a fire brigade, so that efficient coordination among the departments is ensured, and the problems of response delay and unreasonable resource allocation caused by information asymmetry in a traditional system are avoided.

The application relates to a composite liquid carrier with a built-in volatile channel and a heating assembly, and belongs to the technical field of electronic cigarettes or aromatherapy equipment. The assembly comprises a liquid carrier integrally formed with the heating assembly, and a sealing coating covering the surface of the carrier and having a thermally induced phase change characteristic. At normal temperature, the sealing coating is in a solidified state, closes the volatile channel in the interior of the carrier, and inhibits the self-volatilization of the liquid; when the heating assembly is powered on to work, the coating is melted or softened by heat, automatically opens the volatile channel, and makes the liquid smoothly volatilize; after the heating is stopped, the coating is solidified again, and the channel is closed again. Through the synergistic effect of the integrally formed structure and the reversible thermally induced phase change coating, the on-demand and dynamic control of liquid volatilization is realized, the problem of serious static volatilization in the non-working state in the prior art is effectively solved, the invalid loss is significantly reduced, the service life of the product is prolonged, and the thermal response efficiency and the energy utilization rate are improved.

The invention discloses a task operation method and device of a vehicle, electronic equipment and a storage medium. The method comprises the steps that a to-be-operated target task in a cabin system of the vehicle is determined; based on the type of the target task, an operation unit matched with the cabin system is allocated to the target task, and a target operation unit is obtained; the power consumption mode of the target operation unit is determined, and the power consumption mode is used for representing the mode that the target operation unit consumes power when the cockpit system adopts the target operation unit to operate the target task; in the cabin system, according to the power consumption mode and the initial sequence information of the target task, the target operation unit is adopted to operate the target task, and the initial sequence information is used for representing the initial sequence of executing the target task. The technical problem that the task operation efficiency of the vehicle is low is solved.

The present invention discloses a method for dynamically adjusting a detection range and a method for detecting feed ingredients, and belongs to the field of automated detection technology. The method for dynamically adjusting the detection range divides a sample to be tested into a first component and a second component, wherein the first component enters a spectral channel and the second component enters a first mass spectrum channel. Spectral data is obtained from the first component, and mass spectrum data is obtained from the second component. The ion mass of the sample to be tested is predicted based on the spectral data, an effective mass-to-charge ratio interval is generated, and the maximum flight time of the ions is determined, and the sampling duration is adjusted accordingly. The present invention predicts the mass-to-charge ratio interval of the sample components based on spectral technology, and the mass-to-charge ratio interval reflects the interval where the main components of the current feed powder are located. The detection range is dynamically adjusted based on the mass-to-charge ratio interval to ensure that the main components are accurately measured, while shortening the sampling interval as much as possible to improve the accuracy of continuous detection.

The invention discloses a simple harmonic variable rotating speed generation mechanism which mainly comprises a main motor, an auxiliary motor, a planetary gear train and a Scotch yoke conversion mechanism. The main motor drives the sun gear; the auxiliary motor drives a crank disc with a radial guide rail, and the crank disc and the rack push rod form a Scotch yoke mechanism together through an adjusting rod assembly composed of a sliding block and a threaded rod, so that the rack push rod generates simple harmonic reciprocating rectilinear motion. The motion is converted into simple harmonic rotation of the internal and external tooth gear through the intermediate gear, then differential coupling is carried out on the simple harmonic rotation and the rotation speed of the sun gear in the planetary gear train, and finally the planetary gear output shaft outputs the simple harmonic changing rotation speed. The rotating speed fluctuation amplitude can be flexibly controlled by changing the eccentric distance of the adjusting rod assembly in the guide rail. The mechanism realizes dual-power synthesis in a pure mechanical mode, has the advantages of accurate waveform, convenience in adjustment, compact structure and the like, and is suitable for unsteady-state working condition simulation of rotating machinery.

The invention discloses an SATA (Serial Advanced Technology Attachment) bus anomaly detection method, system and device for local cache of an edge gateway, and relates to the technical field of edge computing and bus detection. According to the SATA bus anomaly detection method for the local cache of the edge gateway, the time sequence offset characteristic in the SATA bus link establishment stage and the statistical distribution characteristic in the data transmission stage are obtained, the health degree is calculated in a fusion mode and filtered, the protection priority is determined in combination with the data importance level and the service load, the cache writing strategy is matched, and the SATA bus anomaly detection method for the local cache of the edge gateway is achieved. According to the method, the characteristics of the SATA bus link establishment stage and the data transmission stage are obtained, a related model is constructed, the bus health degree is calculated in a fusion mode, and the adaptive cache protection strategy is matched, so that the safety of the SATA bus link establishment and data transmission stage is improved, and the safety of the SATA bus link establishment and data transmission stage is improved. Accurate detection of bus abnormity is realized, and security of local cache data of the edge gateway is guaranteed.

This invention discloses a multiphase transformer with a separate cooling circuit, comprising a dry-type transformer body and a cooling system. The cooling system includes a first cooling circuit located outside each phase's high-voltage coil and a second cooling circuit located within the main air passage between the high- and low-voltage coils. The second cooling circuit includes an insulating air duct frame and an axially extending cooling air duct inside, an air duct adjustment mechanism, and an independent air supply system. The independent air supply system is physically isolated from the first cooling circuit. A temperature detection system collects the temperature of each phase coil in real time, and a control unit performs differentiated control of the air duct adjustment mechanism based on the temperature data. This invention achieves precise differentiated cooling of the relatively high-temperature phases through a dual-circuit separate architecture of external air cooling and built-in controllable internal cooling air ducts, effectively suppressing inter-phase temperature differences.