A heat dissipation control method and system based on an AI model and a robot

Abstract

This invention provides a heat dissipation control method and system based on AI models and robots. The invention relates to the field of robot heat dissipation control technology. The method includes acquiring infrared thermal imaging video of the stator winding operation of a large steam turbine generator and the generator stator winding operation data; based on the infrared thermal imaging video of the large steam turbine generator stator winding operation and the generator stator winding operation data, determining the winding edge adjustment zone and multiple core high-heat control zones of the winding; based on the winding edge adjustment zone and the multiple core high-heat control zones of the winding, determining multiple control point information for the winding edge adjustment zone and multiple control point information for each core high-heat control zone of the winding; and controlling the robot to dissipate heat from the core high-heat control zones of the winding using a second motor control scheme based on the six-axis heat dissipation robot. This method can efficiently and accurately achieve dynamic heat dissipation of the stator winding of a large steam turbine generator.

Description

Technical Field

[0001] This invention relates to the field of robot heat dissipation control technology, specifically to a heat dissipation control method and system based on an AI model and a robot. Background Technology

[0002] During the operation of large steam turbine generators, the stator windings are prone to localized overheating due to prolonged high-load operation, severely impacting generator efficiency and lifespan. Traditional heat dissipation control methods rely primarily on fixed cooling systems and manual inspections, making it difficult to accurately perceive and dynamically adjust winding temperature distribution. While existing heat dissipation robot technology enables mobile heat dissipation, it lacks intelligent analysis capabilities of the generator's thermal field characteristics and cannot autonomously adjust heat dissipation strategies based on real-time temperature changes. Furthermore, due to the complex internal structure and uneven heat distribution of generators, conventional heat dissipation solutions often have blind spots, resulting in ineffective solutions to localized overheating problems. Traditional heat dissipation methods struggle to accurately identify the specific thermal characteristics of different core high-heat control areas in various windings, such as temperature distribution patterns and differences in heat generation causes. This leads to a lack of targeted operation in heat dissipation solutions, resulting in either insufficient heat dissipation failing to effectively control high-heat areas or excessive heat dissipation causing energy waste.

[0003] Therefore, how to efficiently and accurately cool the stator windings of large steam turbine generators is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem solved by this invention is how to efficiently and accurately cool the stator windings of large steam turbine generators.

[0005] According to a first aspect, the present invention provides a heat dissipation control method based on an AI model and a robot, comprising: acquiring infrared thermal imaging video of the operation of a large steam turbine generator stator winding and operating data of the generator stator winding; determining a winding edge adjustment zone and multiple winding core high-heat control zones based on the infrared thermal imaging video of the operation of the large steam turbine generator stator winding and the operating data of the generator stator winding; determining multiple control point information of the winding edge adjustment zone and multiple control point information of each winding core high-heat control zone based on the winding edge adjustment zone and the multiple winding core high-heat control zones; generating a first motor control scheme for a six-axis heat dissipation robot based on the multiple control point information of the winding edge adjustment zone; controlling the robot to carry a semiconductor cooling chip to dissipate heat from the generator stator winding based on the first motor control scheme of the six-axis heat dissipation robot, and acquiring infrared video of the heat dissipation operation of the six-axis heat dissipation robot; determining a second motor control scheme for the six-axis heat dissipation robot based on the infrared video of the heat dissipation operation of the six-axis heat dissipation robot and the multiple control point information of each winding core high-heat control zone; and controlling the robot to dissipate heat from the winding core high-heat control zone based on the second motor control scheme of the six-axis heat dissipation robot.

[0006] In one possible implementation, determining the second motor control scheme for the six-axis cooling robot based on the infrared video of the cooling operation of the six-axis cooling robot and the information of multiple control points in the high-heat control zone of each winding core includes: constructing a cooling optimization map, which contains control point nodes in multiple high-heat control zones and multiple edges between the control point nodes. The node features of each control point node are control point information, and the edges between nodes are the distance and orientation between the control points; processing the cooling optimization map based on a graph convolutional network to determine the cooling path for each high-heat control zone of the winding core; and determining the cooling path based on the high-heat control zone of each winding core. The heat dissipation path of the controlled area and the infrared video of the heat dissipation operation of the six-axis heat dissipation robot are used to generate simulated heat dissipation infrared videos of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of each winding core high-heat controlled area; based on the simulated heat dissipation infrared videos of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of each winding core high-heat controlled area, the target contact dwell time of the six-axis heat dissipation robot in the heat dissipation path of each winding core high-heat controlled area is determined; based on the target contact dwell time of the six-axis heat dissipation robot in the heat dissipation path of each winding core high-heat controlled area, a second motor control scheme for the six-axis heat dissipation robot is generated.

[0007] In one possible implementation, determining the target contact dwell time of the six-axis cooling robot in the high-heat control zone of each winding core based on simulated heat dissipation infrared video of the six-axis cooling robot at different contact dwell times in the heat dissipation path of each winding core high-heat control zone includes: determining winding heat dissipation information at different contact dwell times in the heat dissipation path of each winding core high-heat control zone based on simulated heat dissipation infrared video of the six-axis cooling robot at different contact dwell times in the heat dissipation path of each winding core high-heat control zone; and determining the target contact dwell time of the six-axis cooling robot in the heat dissipation path of each winding core high-heat control zone based on the winding heat dissipation information at different contact dwell times in the heat dissipation path of each winding core high-heat control zone.

[0008] In one possible implementation, determining the winding edge adjustment zone and multiple core high-heat control zones based on the infrared thermal imaging video of the large steam turbine generator stator winding operation and the operating data of the generator stator winding includes: determining multiple high-significance feature heat point information, multiple medium-significance feature heat point information, and multiple low-significance feature heat point information based on the infrared thermal imaging video of the large steam turbine generator stator winding operation and the operating data of the generator stator winding; clustering the multiple high-significance feature heat point information, the multiple medium-significance feature heat point information, and the multiple low-significance feature heat point information to obtain K clusters; and processing the K clusters based on a thermal feature partitioning model to determine the winding edge adjustment zone and multiple core high-heat control zones of the winding.

[0009] According to a second aspect, the present invention provides a heat dissipation control system based on an AI model and a robot, comprising: an acquisition module for acquiring infrared thermal imaging video of the operation of a large steam turbine generator stator winding and operating data of the generator stator winding; a region determination module for determining a winding edge adjustment zone and multiple winding core high-heat control zones based on the infrared thermal imaging video of the operation of the large steam turbine generator stator winding and the operating data of the generator stator winding; a control point determination module for determining multiple control point information of the winding edge adjustment zone and multiple control point information of each winding core high-heat control zone based on the winding edge adjustment zone and the multiple winding core high-heat control zones; and a first control scheme generation module for... A first motor control scheme for a six-axis heat dissipation robot is generated based on multiple control point information of the winding edge adjustment area; a first heat dissipation control module is used to control the robot to carry a semiconductor cooling chip to dissipate heat from the generator stator winding based on the first motor control scheme of the six-axis heat dissipation robot, and to acquire infrared video of the heat dissipation operation of the six-axis heat dissipation robot; a second control scheme determination module is used to determine a second motor control scheme for the six-axis heat dissipation robot based on the infrared video of the heat dissipation operation of the six-axis heat dissipation robot and multiple control point information of each winding core high-heat control area; a second heat dissipation control module is used to control the robot to dissipate heat from the winding core high-heat control area based on the second motor control scheme of the six-axis heat dissipation robot.

[0010] In one possible implementation, the second control scheme determination module is further configured to: construct a heat dissipation optimization map, the heat dissipation optimization map including multiple control point nodes of multiple core high-heat control zones and multiple edges between the multiple control point nodes, the node feature of each control point node being control point information, and the edges between nodes being the distance and orientation between control points; process the heat dissipation optimization map based on a graph convolutional network to determine the heat dissipation path of each winding core high-heat control zone; generate simulated heat dissipation infrared videos of the six-axis heat dissipation robot under different contact dwell times under the heat dissipation path of each winding core high-heat control zone based on the heat dissipation path of each winding core high-heat control zone and the infrared video of the heat dissipation operation of the six-axis heat dissipation robot; determine the target contact dwell time of the six-axis heat dissipation robot under the heat dissipation path of each winding core high-heat control zone based on the simulated heat dissipation infrared videos of the six-axis heat dissipation robot under different contact dwell times under the heat dissipation path of each winding core high-heat control zone; and generate a second motor control scheme for the six-axis heat dissipation robot based on the target contact dwell time of the six-axis heat dissipation robot under the heat dissipation path of each winding core high-heat control zone.

[0011] In one possible implementation, the second control scheme determination module is further configured to: determine the winding heat dissipation information of the six-axis heat dissipation robot at different contact dwell times under the heat dissipation path of the high-heat control zone of each winding core based on the simulated heat dissipation infrared video of the six-axis heat dissipation robot at different contact dwell times under the heat dissipation path of the high-heat control zone of each winding core; and determine the target contact dwell time of the six-axis heat dissipation robot on the heat dissipation path of the six-axis heat dissipation robot at different contact dwell times under the heat dissipation path of the high-heat control zone of each winding core based on the winding heat dissipation information of the six-axis heat dissipation robot at different contact dwell times under the heat dissipation path of the high-heat control zone of each winding core.

[0012] In one possible implementation, the region determination module is further configured to: determine multiple high-significance feature heat source information, multiple medium-significance feature heat source information, and multiple low-significance feature heat source information based on the infrared thermal imaging video of the large steam turbine generator stator winding operation and the operating data of the generator stator winding; cluster the multiple high-significance feature heat source information, the multiple medium-significance feature heat source information, and the multiple low-significance feature heat source information to obtain K clusters; and process the K clusters based on the thermal feature partitioning model to determine the winding edge adjustment zone and multiple winding core high-heat control zones.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method including: acquiring infrared thermal imaging video of the operation of a large steam turbine generator stator winding and operating data of the generator stator winding; determining a winding edge adjustment zone and multiple winding core high-heat control zones based on the infrared thermal imaging video of the operation of the large steam turbine generator stator winding and the operating data of the generator stator winding; and determining multiple adjustment zones of the winding edge adjustment zone based on the winding edge adjustment zone and the multiple winding core high-heat control zones. The system generates a first motor control scheme for a six-axis cooling robot based on control point information and multiple control point information for each core high-heat control zone of the winding. It then controls the robot to carry a semiconductor cooling chip to cool the generator stator winding based on the first motor control scheme, and acquires infrared video of the cooling robot's operation. Finally, it determines a second motor control scheme for the six-axis cooling robot based on the infrared video of the cooling robot's operation and the multiple control point information for each core high-heat control zone of the winding. Finally, it controls the robot to cool the core high-heat control zone of the winding based on the second motor control scheme.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the aforementioned heat dissipation control method based on an AI model and a robot. The method includes: acquiring infrared thermal imaging video of the operation of a large steam turbine generator stator winding and operating data of the generator stator winding; determining a winding edge adjustment zone and multiple winding core high-heat control zones based on the infrared thermal imaging video of the operation of the large steam turbine generator stator winding and the operating data of the generator stator winding; determining multiple control point information of the winding edge adjustment zone and each... Information on multiple control points in a high-heat control zone of a winding core; a first motor control scheme for a six-axis cooling robot is generated based on the information on multiple control points in the winding edge adjustment zone; the robot is controlled to carry a semiconductor cooling chip to cool the generator stator winding based on the first motor control scheme, and infrared video of the cooling operation of the six-axis cooling robot is acquired; a second motor control scheme for the six-axis cooling robot is determined based on the infrared video of the cooling operation and the information on multiple control points in each high-heat control zone of the winding core; the robot is controlled to cool the high-heat control zone of the winding core based on the second motor control scheme.

[0015] This invention provides a heat dissipation control method and system based on AI models and robots. The method includes acquiring infrared thermal imaging video of the stator winding operation of a large steam turbine generator and operating data of the generator stator winding; determining a winding edge adjustment zone and multiple core high-heat control zones based on the infrared thermal imaging video and the operating data of the generator stator winding; determining multiple control point information for the winding edge adjustment zone and multiple control point information for each core high-heat control zone based on the winding edge adjustment zone and the multiple core high-heat control zones; and determining multiple control point information for each core high-heat control zone based on the multiple control points of the winding edge adjustment zone. A first motor control scheme for a six-axis cooling robot is generated based on point information. Based on this scheme, the robot is controlled to carry a semiconductor cooling chip to cool the generator stator windings, and infrared video of the cooling operation is acquired. A second motor control scheme for the six-axis cooling robot is determined based on the infrared video of the cooling operation and multiple control point information of each winding's core high-heat control zone. The robot is then controlled to cool the winding's core high-heat control zone using this second motor control scheme. This method enables efficient and precise dynamic cooling of the stator windings of a large steam turbine generator. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a heat dissipation control method based on an AI model and a robot, provided in an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of a large steam turbine generator stator winding provided in an embodiment of the present invention;

[0018] Figure 3 This is a schematic diagram of an infrared thermal imaging device provided in an embodiment of the present invention;

[0019] Figure 4 This is a schematic diagram of a process for determining the winding edge adjustment zone and multiple winding core high-heat control zones according to an embodiment of the present invention;

[0020] Figure 5 A schematic diagram of a six-axis heat dissipation robot provided in an embodiment of the present invention;

[0021] Figure 6 A flowchart illustrating a second motor control scheme for a six-axis heat dissipation robot provided in an embodiment of the present invention;

[0022] Figure 7 A flowchart illustrating the process of determining the target contact dwell time of a six-axis heat dissipation robot in the heat dissipation path of each winding core high-heat control zone, as provided in an embodiment of the present invention.

[0023] Figure 8 This is a schematic diagram of a heat dissipation control system based on an AI model and a robot, provided as an embodiment of the present invention. Detailed Implementation

[0024] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0025] In this embodiment of the invention, the following are provided: Figure 1 The diagram illustrates a heat dissipation control method based on an AI model and a robot, comprising steps S1 to S7:

[0026] Step S1: Obtain infrared thermal imaging video of the operation of the stator winding of the large steam turbine generator and the operating data of the generator stator winding.

[0027] The stator winding of a large steam turbine generator is a key component for generating induced electromotive force and realizing electrical energy conversion in a large steam turbine generator. The stator winding consists of multiple sets of coils arranged in a specific pattern and embedded in the stator core slots. Figure 2 This is a schematic diagram of a large steam turbine generator stator winding provided in an embodiment of the present invention.

[0028] Infrared thermal imaging videos of the stator windings of large steam turbine generators are obtained by capturing images of the stator windings of a large steam turbine generator in operation using infrared thermal imaging equipment. For example, an infrared thermal imager installed in the generator room captures a 10-second continuous video stream every 5 seconds. Figure 3 This is a schematic diagram of an infrared thermal imaging device provided in an embodiment of the present invention.

[0029] Infrared thermal imaging videos of the stator windings of large steam turbine generators record the temperature distribution of various parts of the generator stator windings and can show the heating areas and temperature differences of the generator stator windings.

[0030] The operating data of the generator stator winding is collected by sensor detection equipment during the operation of the generator stator winding, reflecting the operating load and basic operating conditions of the winding. The operating data of the generator stator winding includes the winding's operating current, voltage, power load, operating time, and ambient temperature.

[0031] Step S2: Based on the infrared thermal imaging video of the operation of the stator winding of the large steam turbine generator and the operating data of the generator stator winding, determine the winding edge adjustment zone and multiple core high-heat control zones of the winding.

[0032] In some embodiments, Figure 4 This is a schematic flowchart illustrating the process of determining a winding edge adjustment zone and multiple winding core high-heat control zones according to an embodiment of the present invention. The determination of the winding edge adjustment zone and multiple winding core high-heat control zones includes steps S21 to S23:

[0033] Step S21: Based on the infrared thermal imaging video of the operation of the stator winding of the large steam turbine generator and the operation data of the generator stator winding, determine multiple high-significance feature heat-generating point information, multiple medium-significance feature heat-generating point information, and multiple low-significance feature heat-generating point information.

[0034] In some embodiments, a hotspot determination model can be used to determine multiple high-significance hotspot information points, multiple medium-significance hotspot information points, and multiple low-significance hotspot information points. The hotspot determination model is a long short-term neural network model. The input to the hotspot determination model is an infrared thermal imaging video of the operation of the stator winding of the large steam turbine generator and the operating data of the generator stator winding. The output of the hotspot determination model is multiple high-significance hotspot information points, multiple medium-significance hotspot information points, and multiple low-significance hotspot information points.

[0035] Long Short-Term Neural Network (LSTM) models include LSTM networks. LSTM is a special type of recurrent neural network that can process time-series data through the synergistic action of forget gates, input gates, and output gates. The forget gate can selectively discard irrelevant data from historical information, the input gate can be used to filter and store new key information, and the output gate can regulate the output content of the current state. LSTM networks can effectively capture long-term and short-term dependencies in data.

[0036] Highly significant feature hot spot information is a data set of key attributes of the prominent hot spot location on the generator stator winding determined by the hot spot determination model. The highly significant feature hot spot information includes the three-dimensional spatial coordinates of the hot spot, real-time temperature value, temperature fluctuation range, abnormal feature type, abnormal feature duration, and associated winding operating parameters.

[0037] Highly significant heating points exhibit obvious abnormalities, such as localized heating points at winding joints caused by abnormally increased contact resistance.

[0038] The information on salient hot spots is a set of key attributes of hot spots in the abnormal features of generator stator windings determined by the hot spot determination model. The information on salient hot spots includes the three-dimensional spatial coordinates of the hot spot, real-time temperature value, temperature fluctuation range, abnormal feature type, abnormal feature duration, and associated winding operating parameters.

[0039] The abnormal characteristics of medium-significant heating points are weaker than those of highly significant heating points. For example, a slight aging of the insulation layer in a section of the winding leads to a decrease in heat dissipation efficiency, resulting in a heating point with a temperature slightly higher than the surrounding area.

[0040] Low-saliency feature heating point information is a data set of key attributes of weak heating points with abnormal features on the generator stator winding, determined by a heating point determination model. This information includes the heating point's three-dimensional spatial coordinates, real-time temperature value, temperature fluctuation range, abnormal feature type, abnormal feature duration, and associated winding operating parameters.

[0041] The temperature of low-significance hotspots is only slightly higher than that of the normal area, and the abnormal characteristics are not obvious. For example, hotspots formed at the edge of the winding due to slightly impaired airflow in the machine room, resulting in slightly poor heat dissipation.

[0042] Infrared thermal imaging videos of the stator windings of large steam turbine generators record the dynamic changes in winding temperature distribution in a time-series frame format. Each frame contains the visual temperature characteristics of various points on the winding, directly reflecting the location and intensity trends of heating areas. The operating data of the generator stator windings provides information on the causes of winding heating from an electrical parameter perspective, such as the correlation between increased current and temperature rise.

[0043] Long Short-Term Neural Network (LSTN) models can perform frame sequence analysis on infrared thermal imaging videos of large steam turbine generator stator windings to extract features such as spatial coordinates and temperature differences of temperature anomaly regions in each frame. They also combine this with video temporal data to capture dynamic features such as the rate of temperature change and duration of heat points. Simultaneously, the model can perform temporal encoding on the generator stator winding operating data, thereby associating changes in parameters such as current and voltage with heat features in the infrared video. The LSTN can filter out irrelevant normal temperature fluctuations using a forget gate, incorporate temporal and parameter data related to anomalies into the input gate, and classify heat points according to the significance of their anomalies based on preset anomaly feature evaluation criteria, outputting corresponding heat point information.

[0044] Step S22: Based on the multiple high-significance feature hot spot information, the multiple medium-significance feature hot spot information, and the multiple low-significance feature hot spot information, clustering is performed to obtain K clusters.

[0045] The clustering algorithm described is the K-means clustering algorithm. K-means clustering is an unsupervised learning algorithm that divides a dataset into K predetermined clusters, thereby achieving the aggregation and grouping of data with similar characteristics. Each cluster contains data points with similar features. The value of K in the K-means clustering algorithm can be predetermined manually.

[0046] The K clusters are sets of hotspot information obtained by clustering multiple hotspot information with high saliency, multiple hotspot information with medium saliency, and multiple hotspot information with low saliency using the K-means clustering algorithm. Each cluster represents a set of hotspot information with similar comprehensive characteristics, such as similarity in temperature fluctuation range, spatial clustering area, and anomalous feature association attributes.

[0047] For example, clustering multiple high-significance, medium-significance, and low-significance heat point information resulted in three clusters. Cluster 1 contained heat point information with temperatures between 110℃ and 140℃, concentrated in the stator winding end region, and whose anomalies were all related to insulation wear. This cluster included some high-significance, medium-significance, and low-significance heat point information. Cluster 2 contained heat point information with temperatures between 80℃ and 110℃, distributed in the winding coil gaps, and whose anomalies were all related to heat dissipation channel blockage. Cluster 2 encompassed heat point information of different significance levels and was clustered due to similar heat dissipation-related attributes. Cluster 3 contained heat point information scattered at the connection between the winding and the housing, and whose anomalies were all related to changes in contact resistance. The core commonality of Cluster 3 was the thermal characteristics related to contact resistance.

[0048] Information on multiple hotspots with high, medium, and low significance provides the foundational data for clustering, including the intensity of anomalous features. Each anomalous feature intensity level corresponds to anomalies of varying severity: high significance levels correlate with prominent faults, medium significance levels with potential hazards, and low significance levels with weak tendencies, thus assigning a severity attribute to each hotspot. Clustering, while preserving anomalous intensity information, integrates common features such as temperature fluctuation range, spatial distribution area, and anomalous triggers. This allows for the differentiation between low-intensity, concentrated similar problems and high-intensity, dispersed different problems, retaining the severity information of each hotspot while aggregating patterns of different intensities but with the same trigger.

[0049] The K-means clustering algorithm is used to cluster these hot spots as follows: First, K points are randomly selected from a dataset containing multiple hot spots with high, medium, and low saliency features as initial cluster centers. These initial centers must cover hot spots with different saliency types. Then, for each hot spot in the dataset, its temperature, spatial coordinates, and anomaly intensity are analyzed. Euclidean distance is used to calculate the distance between the hot spot and the K initial cluster centers, and the hot spot is assigned to the corresponding cluster based on the closest distance. After all hot spots have been divided, the average temperature, average spatial coordinates, and average anomaly intensity of each hot spot within each cluster are recalculated, and the cluster centers of each cluster are updated accordingly. This process of dividing hot spots and updating cluster centers is repeated until the differences in all features between two consecutive cluster centers are less than a preset threshold. At this point, the clustering process is considered converged, and the clustering of K clusters is complete.

[0050] Clustering allows for the integration of heat source information based on common characteristics such as temperature fluctuation range, spatial aggregation pattern, and anomalous feature correlation attributes, thereby forming aggregated units with clear comprehensive characteristics. This aggregation effectively simplifies the analytical dimensions of thermal feature division and avoids the inefficiency and chaos of directly analyzing individual heat source information one by one, instead extracting the overall thermal state pattern using clusters as the basic unit.

[0051] Step S23: Based on the thermal feature partitioning model, the K clusters are processed to determine the winding edge adjustment area and multiple winding core high-heat control areas.

[0052] The thermal feature partitioning model is a deep neural network model. The input of the thermal feature partitioning model is the K clusters, and the output of the thermal feature partitioning model is the winding edge adjustment area and multiple winding core high-heat control areas.

[0053] Deep neural network models include deep neural networks (DNNs), which are neural network architectures composed of multiple hidden layers. The layer structure of a deep neural network can include convolutional layers, fully connected layers, and activation layers. Convolutional layers are responsible for extracting local features from the data, fully connected layers enable global feature fusion, and activation layers can introduce complex feature mapping capabilities through nonlinear functions. Deep neural networks possess powerful feature learning and complex pattern recognition capabilities, and can mine potential correlations from data to perform classification or partitioning tasks.

[0054] The winding edge adjustment zone is a winding area identified by a thermal characteristic segmentation model. It is prioritized for heat dissipation, and the resulting data on heat dissipation effectiveness is collected for analysis. This zone is characterized by data representativeness and low operational difficulty. The winding edge adjustment zone includes its spatial extent, the dominant type of abnormal heating characteristics within the zone, the average temperature of the heating points, and their spatial distribution density.

[0055] Multiple core high-heat control zones for windings are identified by processing K clusters using a thermal feature segmentation model. These zones are characterized by distinct thermal properties and require independent, targeted heat dissipation operations. Each core high-heat control zone includes its spatial extent, the dominant type of abnormal heating feature within the zone, the average temperature and spatial distribution density of heating points, and the combination information of the clusters constituting that core high-heat control zone.

[0056] Multiple high-heat control zones at the core of the windings can be adapted to the types of thermal anomalies, thermal parameters, and spatial attributes of different locations in the stator windings, thereby achieving precise and differentiated heat dissipation. By dividing the system into multiple control zones, dedicated heat dissipation can be carried out for the core thermal characteristics of each zone, ensuring that the heat dissipation operation of each zone can accurately match its heat source and thermal characteristics.

[0057] Deep neural networks possess the ability to extract multi-layer nonlinear features and mine complex correlations. The model can analyze the dominant anomaly features and temperature range of each cluster, thereby extracting the core requirements of the heat dissipation strategy for each cluster and determining the core adaptation direction of each cluster in heat dissipation operations. When dividing the winding edge adjustment zone, the model can select clusters from K clusters that have low difficulty in performing heat dissipation operations and can reflect the thermal characteristics of the winding edge, and define their corresponding areas as edge adjustment zones. Then, the model can integrate the clusters based on their spatial correlation, grouping clusters with consistent core heat dissipation requirements into one category. Regardless of differences in temperature details or scale, as long as these clusters have the same core requirement for heat dissipation operations and are spatially adjacent or suitable for unified heat dissipation operations, the model will integrate these clusters into a single winding core high-temperature control zone.

[0058] Step S3: Based on the winding edge adjustment area and the multiple winding core high-heat control areas, determine multiple control point information of the winding edge adjustment area and multiple control point information of each winding core high-heat control area.

[0059] In some embodiments, a control information determination model can be used to determine multiple control point information of the winding edge adjustment zone and multiple control point information of each winding core high-heat control zone. The control information determination model is a Transformer model, the input of which is the winding edge adjustment zone and the multiple winding core high-heat control zones, and the output of which is the multiple control point information of the winding edge adjustment zone and the multiple control point information of each winding core high-heat control zone.

[0060] The Transformer model is a neural network model built on a self-attention mechanism, comprising an encoder and a decoder. The encoder in the Transformer model can process input data in parallel through a multi-head self-attention mechanism, capturing the spatial correlation and feature dependencies between the winding edge adjustment region and the core high-temperature control region. The decoder, based on the encoder output, can generate control point information in conjunction with the target task requirements. The Transformer model does not depend on sequence order and can efficiently mine the complex relationship between regional features and control point locations.

[0061] The multiple control point information of the winding edge adjustment zone is the control point information output by the model through control information to control the heat dissipation of the winding edge adjustment zone and analyze the heat dissipation effect.

[0062] The information on multiple control points in each core high-heat control zone of the winding is determined by the control information, which is the point information output by the model to achieve precise heat dissipation control in each core high-heat control zone of the winding.

[0063] Each control point information includes the three-dimensional coordinates of the control point, the corresponding heat source type, and the target heat dissipation temperature to be achieved.

[0064] The Transformer model can convert the feature data of the winding edge adjustment area and multiple core high-heat control areas into a sequence, where the boundary coordinates, heat point distribution, and temperature data of each area are elements in the sequence. The encoder's multi-head self-attention mechanism can simultaneously calculate the self-attention weights of features within each area and the cross-attention weights between different areas. For example, it focuses on the area surrounding the highest-temperature heat point in the core high-heat control area, as well as the boundary positions adjacent to the edge adjustment area and the core area. The Transformer model can extract fused features such as heat point density and spatial accessibility by performing a nonlinear transformation on the attention-weighted features through a feedforward neural network. After receiving the feature representation output by the encoder, the decoder can combine the motion range constraints of the heat dissipation robot to generate a preliminary sequence of candidate control point positions. Through an autoregressive generation method, the model can verify the validity of each candidate position, such as determining whether the control point is within the area boundary and whether it can cover the target heat point, and eliminate invalid candidate points. Finally, it outputs multiple control point information for the winding edge adjustment area and multiple control point information for each core high-heat control area.

[0065] Step S4: Generate the first motor control scheme for the six-axis heat dissipation robot based on the information of multiple control points in the winding edge adjustment area.

[0066] In some embodiments, a first scheme generation model can be used to generate a first motor control scheme for a six-axis cooling robot. The first scheme generation model is a deep neural network model. The input to the first scheme generation model is information on multiple control points in the winding edge adjustment area, and the output of the first scheme generation model is the first motor control scheme for the six-axis cooling robot.

[0067] The first motor control scheme for the six-axis cooling robot is obtained by processing information from multiple control points in the winding edge adjustment area using a first scheme generation model. This scheme includes the rotation angle, speed, and time of each joint motor, as well as the start-up timing and power parameters of the semiconductor cooling chip.

[0068] The information from multiple control points in the winding edge adjustment area includes specific data such as the three-dimensional coordinates of the control points and the target temperature for heat dissipation. This data clarifies the position the robot needs to reach and the heat dissipation task requirements. A deep neural network model can sort the information from these multiple control points in spatial order, extract the three-dimensional coordinates of each control point as core input features, and combine this with inherent parameters of the six-axis heat dissipation robot, such as joint length and motor rated speed, to construct an input vector. The input layer can pass these vectors to hidden layers. The first hidden layer calculates the inverse kinematics solution from the robot's initial position to the first control point, obtaining the theoretical rotation angle of each joint motor through nonlinear transformation. The second hidden layer can combine the dynamic response characteristics of the motor to convert the rotation angle into corresponding rotation speed and time parameters. For example, when the spacing between control points is small, the motor rotation speed can be reduced to ensure positioning accuracy. Subsequent hidden layers can further integrate the heat dissipation requirements of the semiconductor cooling chip, determining the starting power of the cooling chip when the robot reaches each control point, for example, adjusting the power based on the type of heat source corresponding to the control point. The output layer can integrate the processing results of each hidden layer into a structured sequence of control instructions. Each instruction corresponds to the motion parameters of a motor and the working parameters of the cooling chip, ultimately generating the first motor control scheme for the six-axis cooling robot to ensure that the robot can accurately reach each control point and perform the initial cooling operation.

[0069] Step S5: Based on the first motor control scheme of the six-axis heat dissipation robot, control the robot to carry the semiconductor cooling chip to dissipate heat from the generator stator winding, and acquire infrared video of the heat dissipation operation of the six-axis heat dissipation robot.

[0070] The infrared video of the six-axis cooling robot's cooling operation is captured by an infrared thermal imaging device, showing the robot performing cooling operations on the generator stator windings according to the first motor control scheme. For example, it can be a real-time video stream synchronously captured by an infrared thermal imager mounted at the robot's end effector. The infrared video of the six-axis cooling robot's cooling operation can record the temperature changes in the winding edge adjustment area and the robot's movement trajectory during the cooling process. Figure 5 This is a schematic diagram of a six-axis heat dissipation robot provided in an embodiment of the present invention.

[0071] Step S6: Determine the second motor control scheme for the six-axis cooling robot based on the infrared video of the cooling operation of the six-axis cooling robot and the information of multiple control points in the high-heat control zone of each winding core.

[0072] In some embodiments, Figure 6 This is a flowchart illustrating a method for determining a second motor control scheme for a six-axis cooling robot, as provided in an embodiment of the present invention. The method for determining the second motor control scheme for the six-axis cooling robot includes steps S31 to S35:

[0073] Step S31: Construct a heat dissipation optimization map. The heat dissipation optimization map includes multiple control point nodes of multiple core high-heat control zones and multiple edges between multiple control point nodes. The node features of each control point node are control point information, and the edges between nodes are the distance and orientation between control points.

[0074] A graph is a data structure representing relationships between objects. A graph consists of vertices and edges. The heat dissipation optimization graph is a dataset constructed based on control point information from multiple core high-heat control zones, used to analyze the relationships between control points. In the heat dissipation optimization graph, each node represents a control point within a core high-heat control zone, and the node characteristics of each control point node are its control point information. Edges between nodes represent the spatial relationship between two control points, containing the linear distance and relative orientation between the two control points.

[0075] Step S32: Process the heat dissipation optimization map based on the graph convolutional network to determine the heat dissipation path of each winding core high-heat control zone.

[0076] Graph Convolutional Networks (GCNs) are neural network models specifically designed for processing graph-structured data. GCNs can effectively extract local and global structural information from nodes in a graph. The input to the GCN is the heat dissipation optimization graph, and the output is the heat dissipation path for each high-temperature control zone of the winding core.

[0077] The heat dissipation path of each winding core high-heat control zone is obtained by processing the heat dissipation optimization map through a graph convolutional network, which is the robot's motion trajectory within each winding core high-heat control zone.

[0078] Each heat dissipation path includes a sequence of control points arranged in order, the robot's movement path between adjacent control points, and the order in which it passes through each control point.

[0079] The thermal optimization map can integrate control points in multiple high-heat control zones of the winding core into a coherent whole, thus avoiding isolated information states for the control points. The thermal optimization map provides a unified analytical foundation for graph convolutional networks (GCNNs) and enables GCNNs to systematically analyze the distribution and relationships of all control points.

[0080] Node features are the basis for graph convolutional networks to determine the function of control points. By using control point information as node features, graph convolutional networks can extract information such as spatial location and thermal data from the control points to determine whether each control point is suitable for inclusion in the heat dissipation path. For example, if the thermal data of a certain control point shows that the temperature is too high, the graph convolutional network will prioritize this control point as the key target for path coverage.

[0081] Edges represent the distance and orientation between control points, providing spatial relationships for graph convolutional networks. Graph convolutional networks can determine whether two control points are suitable as connection points in a path based on distance; excessive distance increases the difficulty of heat dissipation operations. Simultaneously, orientation determines the direction of movement between control points, ensuring the path conforms to spatial logic.

[0082] Graph convolutional networks (GCNNs) possess the ability to perform deep feature analysis and pattern mining on structured data containing relationships. GCNNs can focus on the control point nodes corresponding to each core high-heat control zone of the winding core to extract control point information from node features, such as the spatial location and thermal data of the control points. Simultaneously, combining the distances and orientations represented by edges between nodes, GCNNs can capture the spatial proximity relationships and thermal influence relationships between different control point nodes layer by layer through multi-layer convolution operations. For example, it can determine whether the distance between a certain control point and other surrounding control points is suitable as a connection point in the heat dissipation path, and combine the thermal data of the control point to determine whether the heat transfer efficiency on the path is reasonable. Subsequently, based on these mined relationships, GCNNs can select connection logics that start from the core control point in the core high-heat control zone of the winding core, efficiently connect surrounding control points, and meet spatial distance and thermal requirements, and then transform this logic into actual movement and operation paths.

[0083] Step S33: Based on the heat dissipation path of each winding core high-heat control zone and the infrared video of the six-axis heat dissipation robot's heat dissipation operation, generate simulated heat dissipation infrared video of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of each winding core high-heat control zone.

[0084] In some embodiments, a simulated heat dissipation model can be used to generate simulated infrared videos of a six-axis heat dissipation robot at different contact dwell times along its heat dissipation path in each high-heat control zone of the winding core. The simulated heat dissipation model is a generative adversarial network (GAN). The inputs to the simulated heat dissipation model are the heat dissipation path in each high-heat control zone of the winding core and the infrared video of the six-axis heat dissipation robot's heat dissipation operation. The output of the simulated heat dissipation model is the simulated infrared video of the six-axis heat dissipation robot at different contact dwell times along its heat dissipation path in each high-heat control zone of the winding core.

[0085] Generative Adversarial Networks (GANs) consist of two subnetworks: a generator and a discriminator. The generator and discriminator can be trained adversarially to improve performance. The generator produces realistic synthetic data based on the input feature data. The discriminator is responsible for distinguishing between real data and simulated data generated by the generator. Through continuous adversarial training, the generator can generate data that gradually approximates the real distribution.

[0086] The simulated heat dissipation infrared videos of the six-axis heat dissipation robot under different contact dwell times along the heat dissipation path in each winding core high-heat control zone are generated by a simulated heat dissipation model to simulate the robot's heat dissipation process. Each simulated heat dissipation infrared video corresponds to a specific heat dissipation path and a certain contact dwell time in a specific core high-heat control zone, and records the simulated temperature changes in the winding area under that condition.

[0087] The heat dissipation path of each winding core high-temperature control zone can clearly define the robot's motion trajectory and the sequence of control points, thus enabling the generative adversarial network to provide the robot's trajectory parameters. Infrared video of the six-axis cooling robot's heat dissipation operation can provide realistic heat dissipation temperature change patterns and video characteristics, ensuring that the simulation video generated by the model conforms to actual heat dissipation physics.

[0088] During training and generation, the generator of the generative adversarial network (GAN) can combine spatial information of the heat dissipation path to simulate the contact state between the robot and the winding under different contact dwell times. Simultaneously, it generates simulated heat dissipation infrared videos for corresponding dwell times, referencing the thermal state changes in real six-axis heat dissipation robot infrared videos. The discriminator compares the generated simulated video with the real infrared video to determine the realism of the simulated heat dissipation infrared video and transmits feedback information to the generator, thereby continuously optimizing the generator's simulation capabilities. Through this adversarial learning mechanism, the GAN can accurately generate simulated heat dissipation infrared videos that conform to the heat dissipation path constraints and whose thermal state changes under different contact dwell times are realistic and believable.

[0089] Step S34: Based on the simulated heat dissipation infrared video of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of the high heat control zone of each winding core, determine the target contact dwell time of the six-axis heat dissipation robot in the heat dissipation path of the high heat control zone of each winding core.

[0090] In some embodiments, Figure 7 This is a flowchart illustrating the process of determining the target contact dwell time of a six-axis heat dissipation robot in the high-heat control zone of each winding core, as provided in an embodiment of the present invention. The determination of the target contact dwell time of the six-axis heat dissipation robot in the high-heat control zone of each winding core includes steps S41-S42:

[0091] Step S41: Based on the simulated heat dissipation infrared video of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of each winding core high heat control area, determine the winding heat dissipation information under different contact dwell times in the heat dissipation path of each winding core high heat control area.

[0092] In some embodiments, a heat dissipation information analysis model can be used to determine the winding heat dissipation information under different contact dwell times along the heat dissipation path of each winding core high-heat control zone. The heat dissipation information analysis model is a long short-term neural network model. The input to the heat dissipation information analysis model is the simulated heat dissipation infrared video of the six-axis heat dissipation robot under different contact dwell times along the heat dissipation path of each winding core high-heat control zone, and the output of the heat dissipation information analysis model is the winding heat dissipation information under different contact dwell times along the heat dissipation path of each winding core high-heat control zone.

[0093] The heat dissipation information of each winding core high-heat control zone under different contact dwell times is determined by analyzing simulated heat dissipation infrared video through a heat dissipation information analysis model. This information reflects the heat dissipation effect of the winding core high-heat control zone under different dwell times. The heat dissipation effect information includes the area identifier of the corresponding winding core high-heat control zone, the sequence of control points along the heat dissipation path, the specific duration of different contact dwell times, the initial and final temperatures of each point on the winding at each dwell time, the temperature decrease magnitude, the temperature decrease rate, and the judgment result on whether the preset heat dissipation target temperature has been reached.

[0094] Simulated infrared video of a six-axis cooling robot at different contact dwell times along the heat dissipation path in the high-heat control zone of each winding core can simulate the dynamic temperature changes at various points in the high-heat control zone of the winding core at different dwell times on the corresponding heat dissipation path in the form of time-series frames. Each frame contains clear temperature visual features and spatial location information. This time-series thermal data can directly reflect the correlation between dwell time and heat dissipation effect. Long Short-Term Neural Network (LSTN) can perform frame sequence parsing on the input simulated infrared video of the six-axis cooling robot at different contact dwell times along the heat dissipation path in the high-heat control zone of each winding core, decomposing continuous video frames into independent time-series units in chronological order. At the same time, it extracts basic features such as the spatial coordinates of the high-heat control zone of the winding core and the temperature values ​​at each point in each frame, and labels the corresponding contact dwell time segment for each frame. Subsequently, the model can filter out irrelevant background interference information and redundant data of normal temperature fluctuations in the video frames through a forget gate, retaining only the temperature change features related to the heat dissipation effect. The input gate can associate and store the filtered temperature features with the corresponding dwell time information, forming a temporal correlation between dwell time and temperature change. The model integrates the temporal correlation pairs of each dwell time segment with preset heat dissipation effect evaluation dimensions, such as temperature drop rate and compliance status, through the output gate, and then calculates key indicators such as the temperature drop rate and whether the final temperature meets the standard.

[0095] Step S42: Based on the heat dissipation information of the winding core high-heat control zone under different bonding dwell times, determine the target bonding dwell time of the six-axis heat dissipation robot on the heat dissipation path of the high-heat control zone of each winding core.

[0096] In some embodiments, a target time determination model can be used to determine the target contact dwell time of the six-axis cooling robot on the heat dissipation path in each high-heat control zone of the winding core. The target time determination model is a Transformer model. The input to the target time determination model is the winding heat dissipation information at different contact dwell times along the heat dissipation path in each high-heat control zone of the winding core, and the output of the target time determination model is the target contact dwell time of the six-axis cooling robot on the heat dissipation path in each high-heat control zone of the winding core.

[0097] The target contact dwell time of the heat dissipation path of each winding core high-heat control zone is determined by the target time determination model. It is the optimal dwell time that enables the high-heat control zone of the winding core to achieve the preset heat dissipation target under its corresponding heat dissipation path.

[0098] Under the target contact dwell time of the heat dissipation path in each high-heat control zone of the winding core, the temperature at each point in the high-heat control zone of the winding core can be effectively reduced to the preset target temperature, while avoiding energy waste due to excessive dwell time or incomplete heat dissipation due to excessively short dwell time.

[0099] Winding heat dissipation information can record the heat dissipation effect corresponding to different dwell times, including data on the direct correlation between dwell time and heat dissipation effect. This data provides the model with a basis for judging the merits of different dwell times, and enables the model to select the optimal dwell time that satisfies both heat dissipation objectives and efficiency.

[0100] The Transformer model can convert the heat dissipation information of each winding core high-heat control zone under different contact dwell times into sequential data, and label the corresponding winding core high-heat control zone identifier and heat dissipation path information. Each element in the sequential data includes features such as dwell time value, corresponding temperature drop magnitude, final temperature, and compliance status. The encoder can simultaneously calculate the attention weight of elements with different dwell times under the same heat dissipation path through a multi-head self-attention mechanism. Higher weights are assigned to elements with large temperature drop magnitudes, compliant final temperatures, and short dwell times, while lower weights are assigned to elements with uncompliant temperatures or excessively long dwell times, to highlight the characteristics of high-quality dwell times. At the same time, the model can capture the correlation patterns of different dwell times, such as the nonlinear relationship between the increase in dwell time and the change in temperature drop magnitude. Based on the feature representation output by the encoder, the decoder combines preset heat dissipation targets such as target temperature and maximum allowable dwell time threshold to effectively evaluate the dwell time element under each heat dissipation path. First, dwell time elements that do not meet the temperature target are eliminated. Then, the element with the shortest dwell time is selected from the elements that meet the target. If multiple elements have similar dwell times and both meet the target, the temperature decrease rate is further compared, and the dwell time corresponding to the element with the faster rate is selected. Finally, the decoder can bind the selected optimal dwell time with the corresponding high-heat control zone identifier of the winding core and the heat dissipation path to generate the target fitting dwell time.

[0101] Step S35: Generate a second motor control scheme for the six-axis heat dissipation robot based on the target contact dwell time of the six-axis heat dissipation robot in the heat dissipation path of each winding core high heat control zone.

[0102] In some embodiments, a second scheme generation model can be used to generate a second motor control scheme for the six-axis cooling robot. The second scheme generation model is a deep neural network model. The input to the second scheme generation model is the target contact dwell time of the six-axis cooling robot on the heat dissipation path in each winding core high-temperature control zone, and the output of the second scheme generation model is the second motor control scheme for the six-axis cooling robot.

[0103] The second motor control scheme for the six-axis cooling robot is obtained by processing the target contact dwell time and heat dissipation path through the second scheme generation model. This scheme is used to control the robot to perform precise heat dissipation on the high-heat control area of ​​the winding core. The second motor control scheme includes the rotation angle, rotation speed, and rotation time of each joint motor of the six-axis cooling robot; the start-up timing, start-up power, and stop-up timing of the semiconductor cooling chip; the dwell sequence of each control point on the heat dissipation path of each high-heat control area of ​​the winding core; the target contact dwell time at each control point; and obstacle avoidance parameters during the robot's movement.

[0104] The convolutional layers of the deep neural network process the spatial information in the heat dissipation path, extracting distance features between adjacent control points. Combined with the physical parameters of the six-axis cooling robot joints, this calculates the theoretical rotation angle, speed, and time of each joint motor as the robot moves from one control point to the next. The fully connected layer fuses the initial motor motion parameters output from the convolutional layers with the target contact dwell time, determining the joint motor locking parameters for each control point to maintain stable contact between the robot and the windings. Simultaneously, it determines the starting power and timing of the thermoelectric cooler based on the initial winding temperature at that control point. The activation layer uses nonlinear transformations to incorporate constraints from actual operation, such as obstacle avoidance requirements to prevent collisions between the robot and other generator components, thereby correcting the parameters output from the fully connected layer to ensure they meet actual operational needs. Finally, the output layer integrates the processed motor motion parameters and thermoelectric cooler operating parameters into a structured sequence of control commands, ultimately generating the second motor control scheme for the six-axis cooling robot.

[0105] Step S7: Based on the second motor control scheme of the six-axis heat dissipation robot, control the robot to dissipate heat in the high-heat control area of ​​the winding core.

[0106] Once the second motor control scheme is determined, the six-axis cooling robot is controlled based on the second motor control scheme to perform cooling operations on the stator windings of the large steam turbine generator.

[0107] Based on the same inventive concept Figure 8 This invention provides a schematic diagram of a heat dissipation control system based on an AI model and a robot, comprising:

[0108] The acquisition module 81 is used to acquire infrared thermal imaging video of the operation of the stator winding of the large steam turbine generator and the operating data of the generator stator winding;

[0109] The region determination module 82 is used to determine the winding edge adjustment zone and multiple winding core high-heat control zones based on the infrared thermal imaging video of the operation of the large steam turbine generator stator winding and the operation data of the generator stator winding.

[0110] The control point determination module 83 is used to determine multiple control point information of the winding edge adjustment area and multiple control point information of each winding core high-heat control area based on the winding edge adjustment area and the multiple winding core high-heat control areas.

[0111] The first control scheme generation module 84 is used to generate a first motor control scheme for the six-axis heat dissipation robot based on the information of multiple control points in the winding edge adjustment area.

[0112] The first heat dissipation control module 85 is used to control the robot to carry the semiconductor cooling chip to dissipate heat from the generator stator winding based on the first motor control scheme of the six-axis heat dissipation robot, and to acquire infrared video of the heat dissipation operation of the six-axis heat dissipation robot.

[0113] The second control scheme determination module 86 is used to determine the second motor control scheme of the six-axis heat dissipation robot based on the infrared video of the heat dissipation operation of the six-axis heat dissipation robot and the information of multiple control points in the high-heat control zone of each winding core.

[0114] The second heat dissipation control module 87 is used to control the robot to dissipate heat in the high-heat control area of ​​the winding core based on the second motor control scheme of the six-axis heat dissipation robot.

[0115] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0116] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A heat dissipation control method based on AI models and robots, characterized in that, include Acquire infrared thermal imaging video of the operation of the stator winding of a large steam turbine generator and operating data of the generator stator winding; Based on the infrared thermal imaging video of the operation of the stator winding of the large steam turbine generator and the operating data of the generator stator winding, the winding edge adjustment zone and multiple core high-heat control zones of the winding are determined. Based on the winding edge adjustment zone and the multiple winding core high-heat control zones, determine multiple control point information of the winding edge adjustment zone and multiple control point information of each winding core high-heat control zone; A first motor control scheme for a six-axis heat dissipation robot is generated based on multiple control point information of the winding edge adjustment area. The first motor control scheme for the six-axis heat dissipation robot is a scheme for controlling the robot's actions obtained by processing multiple control point information of the winding edge adjustment area through a first scheme generation model. The first scheme generation model is a deep neural network model. The first motor control scheme based on the six-axis cooling robot controls the robot to carry a semiconductor cooling chip to cool the generator stator windings, and acquires infrared video of the six-axis cooling robot's cooling operation. The second motor control scheme for the six-axis cooling robot is determined based on the infrared video of the cooling operation of the six-axis cooling robot and the information of multiple control points in the high-heat control zone of each winding core. This determination includes: A heat dissipation optimization map is constructed, which includes control point nodes of multiple core high-heat control zones and multiple edges between the control point nodes. The node features of each control point node are control point information, and the edges between nodes are the distance and orientation between the control points. The heat dissipation optimization map is processed based on graph convolutional networks to determine the heat dissipation path of each winding core high-heat control zone; Based on the heat dissipation path of each winding core high-heat control zone, and the infrared video of the heat dissipation operation of the six-axis heat dissipation robot, simulated heat dissipation infrared video of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of each winding core high-heat control zone is generated. Based on the simulated heat dissipation infrared video of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of each winding core high heat control area, the target contact dwell time of the six-axis heat dissipation robot in the heat dissipation path of each winding core high heat control area is determined. A second motor control scheme for the six-axis heat dissipation robot is generated based on the target contact dwell time of the six-axis heat dissipation robot in the heat dissipation path of each winding core high-heat control zone. The second motor control scheme based on the six-axis heat dissipation robot controls the robot to dissipate heat in the high-heat control area of ​​the winding core.

2. The heat dissipation control method based on AI model and robot as described in claim 1, characterized in that, The determination of the target contact dwell time of the six-axis heat dissipation robot in the heat dissipation path of each winding core high-heat control zone based on the simulated heat dissipation infrared video of different contact dwell times in the heat dissipation path of each winding core high-heat control zone includes: Based on the simulated heat dissipation infrared video of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of each winding core high heat control area, the heat dissipation information of the winding under different contact dwell times in the heat dissipation path of each winding core high heat control area is determined. Based on the heat dissipation information of the winding core high-heat control zone under different bonding dwell times, the target bonding dwell time of the six-axis heat dissipation robot in the heat dissipation path of each winding core high-heat control zone is determined.

3. The heat dissipation control method based on AI models and robots as described in claim 1, characterized in that, The determination of the winding edge adjustment zone and multiple core high-heat control zones based on the infrared thermal imaging video of the large steam turbine generator stator winding operation and the generator stator winding operation data includes: Based on the infrared thermal imaging video of the operation of the stator winding of the large steam turbine generator and the operation data of the generator stator winding, multiple high-significance feature heat point information, multiple medium-significance feature heat point information, and multiple low-significance feature heat point information are determined. Based on the information of multiple high-significance hot spots, multiple medium-significance hot spots, and multiple low-significance hot spots, K clusters are obtained; The K clusters are processed based on the thermal feature partitioning model to determine the winding edge adjustment zone and multiple winding core high-heat control zones.

4. A heat dissipation control system based on an AI model and a robot, used to execute the heat dissipation control method based on an AI model and a robot as described in any one of claims 1-3, characterized in that, include: The acquisition module is used to acquire infrared thermal imaging videos of the operation of the stator windings of large steam turbine generators and the operating data of the generator stator windings; The region determination module is used to determine the winding edge adjustment zone and multiple winding core high-heat control zones based on the infrared thermal imaging video of the operation of the large steam turbine generator stator winding and the operation data of the generator stator winding. The control point determination module is used to determine multiple control point information of the winding edge adjustment area and multiple control point information of each winding core high-heat control area based on the winding edge adjustment area and the multiple winding core high-heat control areas. The first control scheme generation module is used to generate a first motor control scheme for the six-axis heat dissipation robot based on multiple control point information of the winding edge adjustment area. The first motor control scheme for the six-axis heat dissipation robot is a scheme for controlling the robot's actions obtained by processing the multiple control point information of the winding edge adjustment area through the first scheme generation model. The first scheme generation model is a deep neural network model. The first heat dissipation control module is used to control the robot to carry the semiconductor cooling chip to dissipate heat from the generator stator winding based on the first motor control scheme of the six-axis heat dissipation robot, and to acquire infrared video of the heat dissipation operation of the six-axis heat dissipation robot. The second control scheme determination module is used to determine a second motor control scheme for the six-axis cooling robot based on the infrared video of the cooling operation of the six-axis cooling robot and the information of multiple control points in the high-heat control zone of each winding core. The second control scheme determination module is also used for: A heat dissipation optimization map is constructed, which includes control point nodes of multiple core high-heat control zones and multiple edges between the control point nodes. The node features of each control point node are control point information, and the edges between nodes are the distance and orientation between the control points. The heat dissipation optimization map is processed based on graph convolutional networks to determine the heat dissipation path of each winding core high-heat control zone; Based on the heat dissipation path of each winding core high-heat control zone, and the infrared video of the heat dissipation operation of the six-axis heat dissipation robot, simulated heat dissipation infrared video of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of each winding core high-heat control zone is generated. Based on the simulated heat dissipation infrared video of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of each winding core high heat control area, the target contact dwell time of the six-axis heat dissipation robot in the heat dissipation path of each winding core high heat control area is determined. A second motor control scheme for the six-axis heat dissipation robot is generated based on the target contact dwell time of the six-axis heat dissipation robot in the heat dissipation path of each winding core high-heat control zone. The second heat dissipation control module is used to control the robot to dissipate heat in the high-heat control area of ​​the winding core based on the second motor control scheme of the six-axis heat dissipation robot.

5. The heat dissipation control system based on AI models and robots as described in claim 4, characterized in that, The second control scheme determination module is also used for: Based on the simulated heat dissipation infrared video of the six-axis heat dissipation robot under different contact dwell times in the heat dissipation path of each winding core high heat control area, the heat dissipation information of the winding under different contact dwell times in the heat dissipation path of each winding core high heat control area is determined. Based on the heat dissipation information of the winding core high-heat control zone under different bonding dwell times, the target bonding dwell time of the six-axis heat dissipation robot in the heat dissipation path of each winding core high-heat control zone is determined.

6. The heat dissipation control system based on AI models and robots as described in claim 4, characterized in that, The region determination module is also used for: Based on the infrared thermal imaging video of the operation of the stator winding of the large steam turbine generator and the operation data of the generator stator winding, multiple high-significance feature heat point information, multiple medium-significance feature heat point information, and multiple low-significance feature heat point information are determined. Based on the information of multiple high-significance hot spots, multiple medium-significance hot spots, and multiple low-significance hot spots, K clusters are obtained; The K clusters are processed based on the thermal feature partitioning model to determine the winding edge adjustment zone and multiple winding core high-heat control zones.

7. An electronic device, characterized in that, include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the heat dissipation control method based on an AI model and a robot as described in any one of claims 1 to 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the heat dissipation control method based on the AI ​​model and robot as described in any one of claims 1 to 3.

Images