Method for controlling a fan, method for training a model, and intelligent temperature control device

Abstract

The disclosure provides a fan control method, a model training method and an intelligent temperature control device, and relates to the technical field of computers, in particular to the technical fields of artificial intelligence, deep learning, big data and device temperature control. The specific implementation scheme is: using the classification task layer of the target model, based on the first feature vector corresponding to the associated feature data of the target object, the starting state information of the multiple fans arranged at different positions of the conveying section is determined; according to the starting state information, a fan group is determined from the multiple fans, wherein the fans in the fan group are used for temperature regulation of the target object; using the regression task layer of the target model, based on the first feature vector, the recommended rotating speed of the fans in the fan group is determined, so as to realize fan control of the fans in the fan group according to the recommended rotating speed. According to the technology of the disclosure, accurate operation control of the fans in the fan group can be realized, and then the product stability and the qualified rate of the target objects in production are improved.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to the fields of artificial intelligence, deep learning, big data, and equipment temperature control technology. Background Technology

[0002] In the production process of some products, it is necessary to use fans to quickly regulate the temperature of the product itself in order to change the product's performance or make the product more suitable for use in subsequent production and processing. Summary of the Invention

[0003] This disclosure provides a method for fan control, a method for model training, an intelligent temperature control device, apparatus, equipment, and storage medium.

[0004] According to one aspect of this disclosure, a method for controlling a wind turbine is provided, comprising:

[0005] By utilizing the classification task layer of the target model, and based on the first feature vector corresponding to the associated feature data of the target object, the start-up status information of multiple fans set at different locations in the conveying section is determined.

[0006] Based on the startup status information, a fan group is determined from multiple fans, wherein the fans in the fan group are used to regulate the temperature of the target object;

[0007] By utilizing the regression task layer of the target model, the recommended speed of the wind turbines in the wind turbine unit is determined based on the first feature vector, so as to realize the control of the wind turbines in the wind turbine unit according to the recommended speed.

[0008] According to another aspect of this disclosure, a method for model training is provided, comprising:

[0009] Using the classification task layer of the initial model, the startup status information of multiple wind turbines located at different positions in the transport section is determined based on the second feature vector corresponding to the historical sample data of the target object.

[0010] Based on the startup status information, a fan group is determined from multiple fans, wherein the fans in the fan group are used to regulate the temperature of the target object;

[0011] Using the regression task layer of the initial model, the recommended speed of the wind turbines in the wind turbine unit is determined based on the second feature vector;

[0012] Using a preset loss function, the parameters of the initial model are tuned based on the startup state information and recommended rotation speed to train the target model of any embodiment of this disclosure.

[0013] According to another aspect of this disclosure, an intelligent temperature control device is provided, comprising:

[0014] Furnace body, inlet end and outlet end;

[0015] The conveyor section, running through both the inlet and outlet ends, is used to transport the target object;

[0016] Multiple fans are spaced apart inside the furnace body, and the positions of the multiple fans correspond to the positions of the conveying sections;

[0017] A controller, connected to multiple wind turbines, is used to control the multiple wind turbines according to the method of any embodiment of the present disclosure.

[0018] According to another aspect of this disclosure, a fan control device is provided, comprising:

[0019] The first determining module is used to determine the start-up status information of multiple fans set at different positions in the conveying section based on the first feature vector corresponding to the associated feature data of the target object by utilizing the classification task layer of the target model.

[0020] The second determining module is used to determine the fan group from multiple fans based on the start-up status information, wherein the fans in the fan group are used to regulate the temperature of the target object;

[0021] The third determination module is used to determine the recommended speed of the wind turbines in the wind turbine unit based on the first feature vector using the regression task layer of the target model, so as to realize the control of the wind turbines in the wind turbine unit according to the recommended speed.

[0022] According to another aspect of this disclosure, an apparatus for model training is provided, comprising:

[0023] The fourth determination module is used to determine the start-up status information of multiple wind turbines located at different positions in the transport section based on the classification task layer of the initial model and the second feature vector corresponding to the historical sample data of the target object.

[0024] The fifth determination module is used to determine the fan group from multiple fans based on the start-up status information. The fans in the fan group are used to regulate the temperature of the target object.

[0025] The sixth determination module is used to determine the recommended rotational speed of the wind turbines in the wind turbine unit based on the second feature vector using the regression task layer of the initial model.

[0026] The parameter tuning module is used to tune the parameters of the initial model using a preset loss function, based on the startup state information and recommended rotation speed, so as to train the target model of any embodiment of this disclosure.

[0027] According to another aspect of this disclosure, an electronic device is provided, comprising:

[0028] At least one processor; and

[0029] The memory is communicatively connected to the at least one processor; wherein,

[0030] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform any of the methods described in the present disclosure.

[0031] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to perform any of the methods according to embodiments of this disclosure.

[0032] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements any of the methods according to embodiments of this disclosure.

[0033] According to the technology disclosed herein, precise operation control of the fans in the fan unit can be achieved, thereby improving the product stability and qualification rate of the target production object.

[0034] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0035] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0036] Figure 1 This is a schematic flowchart of a fan control method according to an embodiment of the present disclosure;

[0037] Figure 2 This is a schematic diagram illustrating an application scenario of the fan control method according to an embodiment of the present disclosure;

[0038] Figure 3 This is a schematic diagram illustrating an application scenario of the fan control method according to an embodiment of the present disclosure;

[0039] Figure 4 This is a schematic diagram illustrating an application scenario of the fan control method according to an embodiment of the present disclosure;

[0040] Figure 5 This is a flowchart illustrating a model training method according to an embodiment of the present disclosure;

[0041] Figure 6 This is a schematic diagram of the structure of an intelligent temperature control device according to an embodiment of the present disclosure;

[0042] Figure 7This is a schematic diagram of the structure of a fan control device according to an embodiment of the present disclosure;

[0043] Figure 8 This is a schematic diagram of the structure of a model training apparatus according to an embodiment of the present disclosure;

[0044] Figure 9 This is a block diagram of an electronic device used to implement the wind turbine control method and / or model training method of the embodiments of this disclosure. Detailed Implementation

[0045] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0046] like Figure 1 As shown in the embodiments of this disclosure, a method for controlling a wind turbine is provided, including:

[0047] Step S101: Using the classification task layer of the target model, based on the first feature vector corresponding to the associated feature data of the target object, determine the start-up status information of multiple wind turbines set at different locations in the conveying section.

[0048] Step S102: Based on the startup status information, determine the fan group from multiple fans, wherein the fans in the fan group are used to regulate the temperature of the target object.

[0049] Step S103: Using the regression task layer of the target model, based on the first feature vector, determine the recommended speed of the wind turbines in the wind turbine unit, so as to realize the control of the wind turbines in the wind turbine unit according to the recommended speed.

[0050] According to the embodiments of this disclosure, it should be noted that:

[0051] The specific structure of the neural network model used in the classification task layer is not limited here, as long as it can achieve classification. For example, existing classification models can be used.

[0052] The specific structure of the neural network model used in the regression task layer is not limited here, as long as it can predict the recommended parameters based on the input feature data.

[0053] The target object can be understood as any object whose temperature needs to be regulated by a fan during the production and processing process; no specific limitation is made here. For example, the target object can be metal profiles, semiconductor materials, tobacco, tea, chemical materials, etc., without specific restrictions.

[0054] Related feature data can be understood as data related to the target object itself or external factors. Specifically, it can be selected and adjusted according to the category of the target object, without making specific limitations here.

[0055] The first feature vector can be understood as a high-dimensional vector obtained by extracting features from associated feature data.

[0056] The conveying section is used to transport the target object so that the target object can be regulated by passing through at least one of the multiple fans.

[0057] Multiple fans positioned at different locations within the conveying section can be understood as multiple fans being installed inside the furnace body enclosing the conveying section, with these fans spaced out at different locations within the furnace body. For example, some fans can be located at the furnace inlet to regulate the temperature of the target object just arriving at the furnace inlet from the conveying section. Some fans can be located at the furnace outlet to regulate the temperature of the target object conveyed from the furnace inlet to the furnace outlet from the conveying section. Some fans can be located between the furnace inlet and outlet to regulate the temperature of the target object conveyed between the inlet and outlet.

[0058] Startup status information is used to determine whether each of the multiple fans needs to be on or off when the target object is being conditioned.

[0059] Determining the startup status information of multiple fans located at different positions in the conveying section can be understood as determining the startup status information of each of the multiple fans. That is, the fan is classified as either on or off by a classification task layer, and the classification results of on or off status are converted into startup status information.

[0060] Based on the startup status information, determining the fan group from multiple fans can be understood as: determining whether each fan in different locations within the conveying section needs to be on or off when regulating the temperature of the target object. Then, the fans that need to be on are identified as a fan group. The purpose of identifying the fan group is to determine exactly which fans, at which locations, are needed to regulate the temperature of the target object.

[0061] A blower unit includes at least one of multiple blowers. The blowers in the unit can be located at different positions within the conveying section. The specific location and number of blowers selected depend on the associated characteristic data of the target object. For example, if the current temperature of the target object is 220℃ and it needs to be cooled to 180℃, then two blowers can be selected from ten blowers, and these two blowers can be located near the furnace outlet in the conveying section. Alternatively, if the current temperature of the target object is 750℃ and it needs to be cooled to 450℃, then eight blowers can be selected from ten blowers, and these eight blowers need to be located at various positions in the conveying section to ensure that the target object's temperature reaches the target temperature of 450℃ after cooling.

[0062] Determining the recommended operating speed of the fans in the fan unit can be understood as determining the recommended operating speed of each fan in the fan unit based on the first feature vector corresponding to the associated feature data of the target object. This allows the fans in the fan unit to operate at the recommended speed, thereby achieving temperature regulation of the target object conveyed in the conveying section.

[0063] According to the technology of this disclosure, precise and intelligent operation control of the fans in the wind turbine unit can be achieved through the target model, thereby improving the product stability and qualification rate of the target object in production, and realizing the control of fan operation without relying on human experience. The technology of this disclosure can also achieve complex linkage control of multiple fans. Because the technology of this disclosure only predicts the recommended speed of fans in the on state, it saves computing resources and avoids introducing useless data as much as possible, thereby improving the prediction accuracy of the regression task layer and thus improving the prediction accuracy of the target model. The technology of this disclosure is data-driven (related feature data), based on actual production needs, and achieves end-to-end optimization, that is, it enables the direct acquisition of the recommended speed of the fan from the feature data, so that the fan can be directly controlled according to the recommended speed without secondary calculations.

[0064] In one example, the fan control method of any embodiment of this disclosure can be applied to any intelligent temperature control device, mainly depending on the category of the target object. For example, devices applying the fan control method of the embodiments of this disclosure may include: intelligent temperature control for continuous hot-dip galvanizing annealing furnaces in the metallurgical industry, intelligent temperature control for industrial furnaces and kilns (smelting furnaces, melting furnaces, calcining furnaces, heating furnaces, heat treatment furnaces, drying furnaces, coke ovens, gas generators, etc.), intelligent temperature control for petrochemical reactors, intelligent temperature control for power plant heating equipment, intelligent thermal management for spacecraft equipment, and re-drying equipment in the tobacco industry, etc.

[0065] In one example, the fan control method of any embodiment of this disclosure can be applied to temperature regulation of strip steel (the target object). Galvanized strip steel is widely used in industries such as automobiles, home appliances, construction, and agricultural machinery. Continuous hot-dip galvanizing technology is the main production process for galvanized strip steel. In the continuous hot-dip galvanizing production process, cold-rolled or hot-rolled steel strip runs continuously on the production line at a certain speed, undergoing two main processes: annealing and hot-dip galvanizing. The purpose of annealing is to improve the mechanical properties of the steel strip, and the purpose of hot-dip galvanizing is to give the steel plate better corrosion resistance. The strip steel passes through the annealing furnace at a certain speed via a conveyor section to reach the temperature required for the annealing process. Since the temperature of the strip steel coming directly from the annealing furnace is too high (generally around 650-750℃), it needs to be cooled to around 450℃ in a rapid cooling section before entering the hot-dip galvanizing zinc bath for zinc coating. The cooling effect of the rapid cooling section directly determines the temperature of the strip steel entering the zinc bath, and directly determines the thickness and quality of the zinc layer of the strip steel. Therefore, the precise control of the rapid cooling section is crucial. The rapid cooling section mainly consists of a cooling furnace and rapid cooling fans. Cooling gas (generally a mixture of nitrogen and hydrogen) is directly sprayed onto the strip surface by the rapid cooling fans to cool the strip. The rapid cooling fans (usually around 10) are continuously arranged inside the furnace along the strip conveying direction. According to the fan control method of any embodiment of this disclosure, by reasonably controlling the speed of the rapid cooling fan unit (including multiple rapid cooling fans), the temperature of the strip delivered from the furnace outlet of the rapid cooling section can be precisely controlled to 450℃. In actual production, due to the continuous changes in strip specifications (thickness, width), strip movement speed, strip inlet temperature in the rapid cooling section, and ambient temperature (affecting the inlet temperature of the cooling gas in the rapid cooling section), the speed of the rapid cooling fan assembly needs to be continuously adjusted. However, currently, the adjustment of the rapid cooling fan speed relies heavily on manual experience, resulting in poor adjustment effects and causing the strip temperature at the rapid cooling section outlet to frequently deviate from the target temperature value. This is partly because it is impossible to accurately calculate the thermal process parameters of the rapid cooling section in real time during production; and partly because the complexity of controlling multiple fans in the rapid cooling section has exceeded the capabilities of manual adjustment.

[0066] The fan control method of any embodiment of this disclosure can effectively solve the fan control problem of the rapid cooling section of the continuous hot-dip galvanizing annealing furnace. The embodiments of this disclosure propose a fan group intelligent optimization control method based on multi-task classification regression. Based on a deep neural network (target model), the target model sequentially gives the on / off state of each fan in the fan group and the speed of the fan in the on state, so as to realize the real-time and accurate control of the speed of each fan in the fan group required for the rapid cooling section, thereby getting rid of the dependence on manual experience and improving the product qualification rate and product performance stability.

[0067] In one application example, such as Figure 2As shown, multiple target objects 2 are conveyed on conveyor section 1. A furnace body 3 is located outside conveyor section 1, and multiple fans 4 are spaced apart within the furnace body 3. The target objects 2 are conveyed from... Figure 2 The left side is conveyed into the furnace body 3, and the target object 2 is temperature-regulated by at least one of the multiple blowers 4, and then from... Figure 2 The right side is conveyed out of the furnace body 3. When adjusting the temperature of the target object 2, the specific control of the multiple fans 4 can be achieved using the fan control method of any embodiment of this disclosure.

[0068] In one embodiment, the wind turbine control method of this disclosure includes steps S101 to S103, wherein, before step S101, it further includes:

[0069] By utilizing the shared layer of the target model, feature extraction is performed on the associated feature data of the target object to obtain the first feature vector corresponding to the associated feature data of the target object.

[0070] According to the embodiments of this disclosure, it should be noted that:

[0071] The specific structure of the neural network model used in the shared layer is not limited here, as long as it can achieve feature extraction based on feature data.

[0072] The output of the shared layer is used to send the first feature vector to the classification task layer and the regression task layer, so that the classification task layer can determine the start-up status information of multiple wind turbines based on the first feature vector, and the regression task layer can determine the recommended speed of each wind turbine in the wind turbine group based on the first feature vector.

[0073] According to the technology of this disclosure, by extracting features from the associated feature data of the target object, the subsequent classification task layer can predict more accurate start-up status information of multiple wind turbines, and the subsequent regression task layer can predict more accurate recommended speed of the wind turbines.

[0074] In one application example, such as Figure 3As shown, the target model includes a shared layer, a classification task layer, and a regression task layer. When executing the wind turbine control method of this embodiment, the associated feature data of the target object is first input into the shared layer, and feature extraction is performed using the shared layer to obtain a first feature vector. The first feature vector is then input into the classification task layer, so that multiple classification tasks in the classification task layer determine the start-up status information of multiple wind turbines located at different positions. The classification task layer outputs the start-up status information of each wind turbine. For wind turbines whose start-up status information is "on," the first feature vector is input into the regression task layer, and the regression task layer predicts and outputs the recommended rotational speed of the on-state wind turbine. Finally, the on-state wind turbines are controlled to operate at the recommended rotational speed, and the wind turbines whose start-up status information is "off" are controlled not to operate (i.e., the rotational speed is 0).

[0075] In one embodiment, the wind turbine control method of this disclosure includes steps S101 to S103, wherein step S101: using the classification task layer of the target model, based on the first feature vector corresponding to the associated feature data of the target object, to determine the start-up status information of multiple wind turbines located at different positions in the conveying section, including:

[0076] Based on the first feature vector corresponding to the associated feature data of the target object and the preset start-up state rules, the start-up state information of multiple wind turbines set at different positions in the transport section is determined by using multiple classification tasks in the classification task layer of the target model.

[0077] The startup status information is used to determine whether each of the multiple fans is in the on or off state when the temperature of the target object is adjusted.

[0078] According to the embodiments of this disclosure, it should be noted that:

[0079] The classification task layer contains multiple classification tasks, each used to output the startup status information of a wind turbine, and the startup status information of each classification task is different for each wind turbine.

[0080] The pre-defined activation state rules can be represented as labels in the classification task layer during target model training. These rules determine which wind turbines at which locations should be activated when the values ​​of associated feature data fall within a certain threshold range. In other words, when there are multiple associated feature data points, based on the threshold range that each feature data point falls into, and according to the pre-defined activation state rules, corresponding labels are determined. These labels map to corresponding wind turbine groups (i.e., which wind turbines are located at which locations).

[0081] According to the technology of this disclosure, the on / off state of a wind turbine can be accurately predicted by performing calculations based on the first feature data through a classification task layer. The multi-task learning approach of the target model considers both the relationship between the speeds of multiple wind turbines and the unique operating patterns of each turbine, which is consistent with actual operational experience.

[0082] In one embodiment, the wind turbine control method of this disclosure includes steps S101 to S103, wherein step S102: determining a wind turbine group from a plurality of wind turbines based on startup status information, including:

[0083] Based on the startup status information, identify the wind turbines whose startup status information is "on" from among multiple wind turbines.

[0084] The fan unit is determined based on the fan that is in the "on" state according to the startup status information.

[0085] According to the technology of this disclosure embodiment, since the technology of this disclosure only predicts the recommended speed of the wind turbine when it is in the on state, it saves the occupation of computing resources and can avoid introducing useless data as much as possible, thereby improving the prediction accuracy of the regression task layer and thus improving the prediction accuracy of the target model.

[0086] In one embodiment, the wind turbine control method of this disclosure includes steps S101 to S103, wherein step S103: using the regression task layer of the target model, based on the first feature vector, to determine the recommended rotational speed of the wind turbines in the wind turbine group, including:

[0087] Based on the first feature vector and the location of the wind turbines in the transmission section, the recommended rotational speed of the wind turbines in the wind turbine group is determined by using multiple regression tasks in the regression task layer of the target model.

[0088] According to the embodiments of this disclosure, it should be noted that:

[0089] The location of the fan in the conveying section can be understood as: when adjusting the temperature of the target object, the fan is located at which position relative to the conveying section.

[0090] Based on the location of the fan in the conveying section, the target fan whose recommended speed needs to be calculated can be accurately determined.

[0091] According to the technology of this disclosure, since the recommended speed prediction is only performed on wind turbines in the on state, it saves computing resources and avoids introducing useless data as much as possible, thereby improving the prediction accuracy of the regression task layer and thus improving the prediction accuracy of the target model. The multi-task learning approach of the target model considers both the relationship between the speeds of multiple wind turbines and the unique operating patterns of each wind turbine, which is consistent with actual operating experience.

[0092] In one example, such as Figure 3 As shown, when the fan control method of this embodiment is applied to the fan control of the rapid cooling section of a continuous hot-dip galvanizing annealing furnace, it is assumed that there are n fans in the rapid cooling section, and the target model has a total of n classification outputs and n regression outputs. The classification task layer and the regression task layer share a shared layer. The shared layer represents the common characteristics among different fans, while the classification task layer and the regression task layer represent the independent characteristics of each fan. After the associated feature data is input into the target model, it first passes through the shared layer and then enters the classification task layer to determine the on / off state of all n fans. When the classification task layer determines that the fan state is "on", the output of the shared layer enters the corresponding regression task layer, and the regression task layer outputs the recommended speed of the fan with the "on" state. When the classification task layer determines that the fan state is "off", the classification task layer directly outputs that the fan speed is 0. Finally, based on the index of the fans with different states obtained by the classification task layer, the outputs of the classification task layer and the regression task layer are summarized to obtain the speed of all n fans.

[0093] In one implementation, the associated feature data includes at least: the target object's own data, the target object's temperature data, and the operation data of the conveyor segment used to transport the target object.

[0094] According to the embodiments of this disclosure, it should be noted that:

[0095] The target object's intrinsic data can be understood as data used to describe the target object's own characteristics. For example, the target object's intrinsic data may include at least one of the target object's width, thickness, length, and density.

[0096] Temperature data of a target object can be understood as data describing the object's own temperature at different times. For example, the temperature data of a target object may include at least one of the following: specific heat capacity data, the target object's current temperature before temperature regulation by the fan unit, and the target object's own temperature after temperature regulation by the fan unit.

[0097] Operational data can be understood as data related to the operating parameters of the conveyor section. For example, the operational data of the conveyor section may include at least one of the length data of the conveyor section and the conveying speed data of the conveyor section.

[0098] According to the technology of the embodiments of this disclosure, by using the target object's own data, the target object's temperature data, and the operation data of the conveying section used to transport the target object as associated feature data, the predictive accuracy of the fan control method and the target model of the embodiments of this disclosure has better interpretability. Through these associated feature data, it can be accurately determined whether the recommended speed obtained by the fan control method is reasonable.

[0099] In one implementation, the associated feature data further includes: ambient temperature data of the space where the target object is located.

[0100] In one example, such as Figure 4 As shown, the characteristic data input to the target model include: the convective heat transfer coefficient h of the target object, the velocity v of the target object conveyed in the conveying section, the width w of the target object, the thickness δ of the target object, and the ambient temperature T. a The target object's current temperature data T before temperature regulation by the fan unit. si The target's own temperature data T after temperature regulation by the fan unit. so .in, Figure 4 f in i Let be the recommended operating speed for n fans. The convective heat transfer coefficient h is calculated using the following heat transfer mechanism formula, and the remaining input characteristics are sensor measurements.

[0101]

[0102] Where ρ is the density of the strip steel, and c p L represents the specific heat capacity of the strip steel, and L represents the length of the rapid cooling section (conveyor section).

[0103] The method of this disclosure, which proposes a multi-task classification and multi-task regression target model, can achieve high-precision recommendation of the rotational speed of multiple wind turbines. The heat transfer mechanism formula is adopted in the construction of model features, which improves the prediction accuracy of the classification task layer and the regression task layer, and at the same time improves the interpretability of the target model.

[0104] like Figure 5 As shown, this disclosure provides a method for model training, including:

[0105] Step S501: Using the classification task layer of the initial model, based on the second feature vector corresponding to the historical sample data of the target object, determine the start-up status information of multiple wind turbines set at different locations in the conveying section.

[0106] Step S502: Based on the startup status information, determine the fan group from multiple fans, wherein the fans in the fan group are used to regulate the temperature of the target object.

[0107] Step S503: Using the regression task layer of the initial model, determine the recommended rotational speed of the wind turbines in the wind turbine unit based on the second feature vector.

[0108] Step S504: Using a preset loss function, the parameters of the initial model are tuned according to the startup state information and recommended rotation speed to train the target model of any embodiment of this disclosure.

[0109] According to the technology of this disclosure, a large amount of actual production data is used as training data for the target model, enabling the target model to make real-time recommendations for different operating conditions. The model structure is the same for different production lines; it only needs to be retrained based on new data and parameters. The model has high replicability and can be easily migrated and applied across different production lines. The prediction of fan speed is first classified to determine the fan's start-up status information, and then the speed of the fan in the start-up state is predicted. This achieves very high classification accuracy in the regression task and improves the accuracy of the regression task, avoiding the impact of discontinuous historical sample data on the accuracy of the regression task.

[0110] In one example, when the method of this disclosure embodiment is applied to the fan control of the fast cooling section of a continuous hot-dip galvanizing annealing furnace, based on the production history big data of the fast cooling section of the annealing furnace production line, and coupled with the cooling mechanism and operating experience of the fast cooling section, an optimized control method for the fast cooling fan unit based on a multi-task classification and multi-task regression model is proposed.

[0111] In one embodiment, the model training method of this disclosure includes steps S501 to S504, and further includes:

[0112] By using the shared layer of the initial model, features are extracted from the historical sample data of the target object to obtain the second feature vector of the historical sample data.

[0113] In one embodiment, the model training method of this disclosure includes steps S501 to S504, wherein the preset loss function consists of a binary cross-entropy loss function corresponding to the classification task layer and a mean squared error loss function corresponding to the regression task layer.

[0114] During model training, the losses from multiple classification tasks in the classification task layer and multiple regression tasks in the regression task layer are added together to form the model's total loss function. The loss function for the classification task layer is the binary cross-entropy loss, and the loss for the regression task layer is the mean squared error loss.

[0115] like Figure 6As shown, this disclosure provides an intelligent temperature control device, including:

[0116] Furnace body 100, inlet end and outlet end.

[0117] Conveying section 200 runs through the inlet and outlet ends and is used to transport the target object 500.

[0118] Multiple blowers 300 are spaced apart inside the furnace body 100, and the positions of the multiple blowers 300 correspond to the positions of the conveying section 200.

[0119] A controller 400 is connected to a plurality of fans 300 and is used to control the plurality of fans 300 according to a fan control method according to any embodiment of the present disclosure.

[0120] According to the technology of this disclosure, the controller, through the target model, can achieve precise and intelligent operation control of the fans in the wind turbine unit, thereby improving the product stability and pass rate of the target object in production, and realizing the control of fan operation without relying on human experience. According to the technology of this disclosure, complex linkage control of multiple fans can be achieved. Because the technology of this disclosure only predicts the recommended speed of fans in the on state, it saves computing resources and avoids introducing useless data as much as possible, thereby improving the prediction accuracy of the regression task layer and thus improving the prediction accuracy of the target model. The technology of this disclosure is data-driven (associative feature data), based on actual production needs, and achieves end-to-end optimization, that is, it enables the recommended speed of the fan to be directly obtained from the feature data, so that the fan can be directly controlled according to the recommended speed without secondary calculations.

[0121] like Figure 7 As shown, this disclosure provides a fan control device, including:

[0122] The first determining module 710 is used to determine the start-up status information of multiple fans located at different positions in the conveying section by utilizing the classification task layer of the target model and based on the first feature vector corresponding to the associated feature data of the target object.

[0123] The second determining module 720 is used to determine a fan group from multiple fans based on the startup status information, wherein the fans in the fan group are used to regulate the temperature of the target object.

[0124] The third determining module 730 is used to determine the recommended speed of the wind turbines in the wind turbine group based on the first feature vector using the regression task layer of the target model, so as to realize the wind turbine control in the wind turbine group according to the recommended speed.

[0125] In one embodiment, the fan control device further includes:

[0126] The first feature extraction module is used to extract features from the associated feature data of the target object using the shared layer of the target model, and obtain the first feature vector corresponding to the associated feature data of the target object.

[0127] In one implementation, the first determining module 710 is used to:

[0128] Based on the first feature vector corresponding to the associated feature data of the target object and the preset start-up state rules, the start-up state information of multiple wind turbines set at different positions in the transport section is determined by using multiple classification tasks in the classification task layer of the target model.

[0129] The startup status information is used to determine whether each of the multiple fans is in the on or off state when the temperature of the target object is adjusted.

[0130] In one implementation, the second determining module 720 is used to:

[0131] Based on the startup status information, identify the wind turbines whose startup status information is "on" from among multiple wind turbines.

[0132] The fan unit is determined based on the fan that is in the "on" state according to the startup status information.

[0133] In one implementation, the third determining module 730 is used to:

[0134] Based on the first feature vector and the location of the wind turbines in the transmission section, the recommended rotational speed of the wind turbines in the wind turbine group is determined by using multiple regression tasks in the regression task layer of the target model.

[0135] In one implementation, the associated feature data includes at least: the target object's own data, the target object's temperature data, and the operation data of the transport segment used to transport the target object.

[0136] In one implementation, the associated feature data further includes: ambient temperature data of the space where the target object is located.

[0137] In one embodiment, the target object's own data includes at least one of the following: the target object's width data, thickness data, and density data; the target object's temperature data includes at least one of the following: specific heat capacity data, the target object's current own temperature data before temperature adjustment by the fan unit, and the target object's own temperature data after temperature adjustment by the fan unit; the conveying section's operating data includes at least one of the following: the conveying section's length data and the conveying speed data.

[0138] The specific functions and examples of each module and submodule of the apparatus in this disclosure can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0139] like Figure 8 As shown, this disclosure provides an apparatus for model training, including:

[0140] The fourth determination module 810 is used to determine the start-up status information of multiple wind turbines located at different positions in the transport section based on the classification task layer of the initial model and the second feature vector corresponding to the historical sample data of the target object.

[0141] The fifth determining module 820 is used to determine the fan group from multiple fans based on the start-up status information, wherein the fans in the fan group are used to regulate the temperature of the target object.

[0142] The sixth determination module 830 is used to determine the recommended speed of the wind turbines in the wind turbine unit based on the second feature vector using the regression task layer of the initial model.

[0143] The parameter tuning module 840 is used to tune the parameters of the initial model using a preset loss function, based on the startup state information and the recommended rotation speed, so as to train the target model of any embodiment of this disclosure.

[0144] In one embodiment, the model training apparatus further includes:

[0145] The second feature extraction module is used to extract features from the historical sample data of the target object using the shared layer of the initial model, and obtain the second feature vector of the historical sample data.

[0146] In one implementation, the preset loss function consists of the binary cross-entropy loss function corresponding to the classification task layer and the mean squared error loss function corresponding to the regression task layer.

[0147] The specific functions and examples of each module and submodule of the apparatus in this disclosure can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0148] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0149] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0150] Figure 9A schematic block diagram of an example electronic device 900 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0151] like Figure 9 As shown, device 900 includes a computing unit 901, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 902 or a computer program loaded from storage unit 908 into random access memory (RAM) 903. RAM 903 may also store various programs and data required for the operation of device 900. The computing unit 901, ROM 902, and RAM 903 are interconnected via bus 904. Input / output (I / O) interface 905 is also connected to bus 904.

[0152] Multiple components in device 900 are connected to I / O interface 905, including: input unit 906, such as keyboard, mouse, etc.; output unit 907, such as various types of monitors, speakers, etc.; storage unit 908, such as disk, optical disk, etc.; and communication unit 909, such as network card, modem, wireless transceiver, etc. Communication unit 909 allows device 900 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0153] The computing unit 901 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as wind turbine control methods and / or model training methods. For example, in some embodiments, the wind turbine control methods and / or model training methods can be implemented as computer software programs tangibly contained in a machine-readable medium, such as storage unit 908. In some embodiments, part or all of the computer program can be loaded and / or installed on device 900 via ROM 902 and / or communication unit 909. When the computer program is loaded into RAM 903 and executed by the computing unit 901, one or more steps of the wind turbine control methods and / or model training methods described above can be performed. Alternatively, in other embodiments, the computing unit 901 may be configured by any other suitable means (e.g., by means of firmware) to perform methods of wind turbine control and / or methods of model training.

[0154] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0155] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0156] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0157] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0158] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0159] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0160] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0161] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for fan control, comprising: determining, by a classification task layer of a target model, start state information of a plurality of fans arranged at different positions of a conveying section based on a first feature vector corresponding to associated feature data of a target object and a preset start state rule; the start state information is used to determine that each fan in the plurality of fans is in an open state or a closed state in a case of temperature adjustment of the target object; the conveying section is used to convey the target object so that the target object can pass through at least one fan in the plurality of fans to achieve temperature adjustment of the target object itself; determining, from the plurality of fans, a fan group according to the start state information, wherein the fans in the fan group are used to adjust the temperature of the target object; determining, by a regression task layer of the target model, a recommended rotating speed of the fans in the fan group based on the first feature vector and the arranged positions of the fans in the fan group in the conveying section, to achieve fan control of the fans in the fan group according to the recommended rotating speed.

2. The method of claim 1, wherein, Before determining, by a classification task layer of a target model, start state information of a plurality of fans arranged at different positions of a conveying section based on a first feature vector corresponding to associated feature data of a target object, the method further comprises: extracting, by a shared layer of the target model, features of the associated feature data of the target object to obtain the first feature vector corresponding to the associated feature data of the target object.

3. The method of claim 1, wherein, Determining, by a classification task layer of a target model, start state information of a plurality of fans arranged at different positions of a conveying section based on a first feature vector corresponding to associated feature data of a target object and a preset start state rule, comprises: determining, by a plurality of classification tasks in the classification task layer of the target model, the start state information of the plurality of fans arranged at different positions of the conveying section based on the first feature vector corresponding to the associated feature data of the target object and the preset start state rule.

4. The method of claim 1, wherein, Determining, from the plurality of fans, a fan group according to the start state information, comprises: determining, from the plurality of fans, a fan whose start state information is in an open state according to the start state information; determining a fan group according to the fan whose start state information is in an open state.

5. The method of claim 1, wherein, Determining, by a regression task layer of the target model, a recommended rotating speed of the fans in the fan group based on the first feature vector and the arranged positions of the fans in the fan group in the conveying section, comprises: determining, by a plurality of regression tasks in the regression task layer of the target model, the recommended rotating speed of the fans in the fan group based on the first feature vector and the arranged positions of the fans in the fan group in the conveying section.

6. The method according to any one of claims 1 to 5, wherein, The associated feature data at least comprises: self data of the target object, temperature data of the target object, and operation data of the conveying section for conveying the target object.

7. The method of claim 6, wherein, The associated feature data further comprises: environmental temperature data of a space where the target object is located.

8. The method of claim 6, wherein, The self data of the target object includes at least one of width data, thickness data and density data of the target object; The temperature data of the target object includes at least one of specific heat capacity data, current self temperature data of the target object before temperature adjustment by the fan group and target self temperature data of the target object after temperature adjustment by the fan group; The operation data of the conveying section includes at least one of length data of the conveying section and conveying speed data of the conveying section.

9. A model training method, comprising: determining, by a classification task layer of an initial model, starting state information of a plurality of fans arranged at different positions of a conveying section based on a second feature vector corresponding to historical sample data of a target object; determining, from the plurality of fans, a fan group according to the starting state information, wherein the fans in the fan group are used for temperature adjustment of the target object; determining, by a regression task layer of the initial model, a recommended rotating speed of the fans in the fan group based on the second feature vector; optimizing, by a preset loss function, parameters of the initial model according to the starting state information and the recommended rotating speed to obtain the target model of any one of claims 1 to 8.

10. The method of claim 9, further comprising: extracting, by a shared layer of an initial model, features of historical sample data of a target object to obtain a second feature vector of the historical sample data.

11. The method of claim 9 or 10, wherein, The preset loss function is composed of a binary cross-entropy loss function corresponding to the classification task layer and a mean square error loss function corresponding to the regression task layer.

12. An intelligent temperature control device, comprising: a furnace body, an inlet end and an outlet end; a conveying section passing through the inlet end and the outlet end for conveying a target object; a plurality of fans arranged at intervals in the interior of the furnace body, and the arrangement positions of the plurality of fans correspond to the positions of the conveying section; a controller connected with the plurality of fans for controlling the plurality of fans according to the method of any one of claims 1 to 8.

13. A fan control device, comprising: a first determining module configured to determine, by a classification task layer of a target model, starting state information of a plurality of fans arranged at different positions of a conveying section based on a first feature vector corresponding to associated feature data of a target object and a preset starting state rule; The starting state information is used to determine that each fan in the plurality of fans is in an open state or a closed state in the case of temperature adjustment of the target object; the conveying section is used for conveying a target object, so that the target object can adjust its own temperature by at least one fan in the plurality of fans; a second determining module configured to determine, from the plurality of fans, a fan group according to the starting state information, wherein the fans in the fan group are used for temperature adjustment of the target object. The third determining module is configured to determine recommended rotating speeds of the fans in the fan group based on the first feature vector and positions of the fans in the fan group in the conveying section, by using regression task layers of the target model, so as to control the fans in the fan group to perform fan control according to the recommended rotating speeds.

14. The apparatus of claim 13, further comprising: The first feature extraction module is configured to perform feature extraction on the associated feature data of the target object by using shared layers of the target model, to obtain a first feature vector corresponding to the associated feature data of the target object.

15. The apparatus of claim 13, wherein, The first determining module is configured to: determine, by using multiple classification tasks in a classification task layer of the target model, starting state information of multiple fans arranged at different positions of the conveying section, based on the first feature vector corresponding to the associated feature data of the target object and a preset starting state rule.

16. The apparatus of claim 13, wherein, The second determining module is configured to: determine, from the multiple fans, a fan whose starting state information is in an open state, according to the starting state information. determine a fan group according to the fan whose starting state information is in the open state.

17. The apparatus of claim 13, wherein, The third determining module is configured to: determine recommended rotating speeds of the fans in the fan group, by using multiple regression tasks in a regression task layer of the target model, based on the first feature vector and positions of the fans in the fan group in the conveying section.

18. The apparatus of any one of claims 13 to 17, wherein, The associated feature data at least includes: self data of the target object, temperature data of the target object, and operation data of the conveying section for conveying the target object.

19. The apparatus of claim 18, wherein, The associated feature data further includes: environmental temperature data of a space where the target object is located.

20. The apparatus of claim 18, wherein The self data of the target object includes at least one of: width data of the target object, thickness data, and density data of the target object. The temperature data of the target object includes at least one of: specific heat capacity data, current self temperature data of the target object before temperature adjustment by the fan group, and target self temperature data of the target object after temperature adjustment by the fan group. The operation data of the conveying section includes at least one of: length data of the conveying section, and conveying speed data of the conveying section.

21. An apparatus for model training, comprising: A fourth determining module is configured to determine starting state information of multiple fans arranged at different positions of a conveying section, by using a classification task layer of an initial model, based on a second feature vector corresponding to historical sample data of a target object. A fifth determining module is configured to determine a fan group from the multiple fans according to the starting state information, wherein the fans in the fan group are used to perform temperature adjustment on the target object. A sixth determining module is configured to determine recommended rotating speeds of the fans in the fan group, by using a regression task layer of the initial model, based on the second feature vector. A parameter tuning module is configured to tune parameters of the initial model according to the start state information and the recommended rotating speed by using a preset loss function, so as to train a target model according to any one of claims 1 to 8.

22. The apparatus of claim 21, further comprising: A second feature extraction module is configured to extract features of historical sample data of the target object by using a shared layer of the initial model, so as to obtain a second feature vector of the historical sample data.

23. The apparatus of claim 21 or 22, wherein, The preset loss function is composed of a binary cross-entropy loss function corresponding to the classification task layer and a mean square error loss function corresponding to the regression task layer.

24. An electronic device, comprising: at least one processor; and a memory connected to the at least one processor in communication; wherein The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 11.

25. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to enable the computer to perform the method according to any one of claims 1 to 11.

26. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 11.

Images