Optimization energy-saving control method of deep learning-based drying oven drying system

Abstract

The present application relates to the technical field of industrial drying control, and in particular to an optimization energy-saving control method of a drying oven drying system based on deep learning, which comprises the following steps: first, dividing the load area of the drying oven space and collecting pretreated multi-source data; then, identifying the local heat island area through a time series convolutional neural network model; combining the drying process requirements and the heat island negative compensation to generate the time series structure of the regional heat demand; then, according to the height layering air flow delay characteristics, linkage controlling the circulating fan speed and the drying furnace gas flow; finally, completing the closed-loop correction of the control parameters based on the PUE value and the spatial difference contribution value.

Description

Technical Field

[0001] This invention relates to the field of industrial drying control technology, and more specifically, to an optimized energy-saving control method for drying oven systems based on deep learning. Background Technology

[0002] Drying in drying ovens is a widely used material handling process in industrial production. Its operating energy consumption and drying quality directly affect production efficiency and product quality. Therefore, high requirements are placed on the control precision and energy efficiency of the drying oven system.

[0003] Existing drying oven systems mostly adopt an integrated and extensive control mode, which makes it difficult to accurately perceive the differences in temperature and humidity fields in different areas inside the drying oven, and cannot effectively identify local heat island areas, easily leading to problems of excess heat in some areas and insufficient heat in others. At the same time, the drying oven and circulating fan are mostly controlled independently, without fully considering the airflow delay characteristics and spatial load differences in the vertical direction of the drying oven, which easily generates a lot of ineffective air power and increases system energy consumption. In addition, existing control methods lack a dynamic closed-loop correction mechanism based on real-time energy consumption data, making it difficult to continuously optimize control parameters according to the system operating status.

[0004] To address the problems existing in the prior art, this invention proposes an optimized energy-saving control method for a drying oven system based on deep learning. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of this invention is to provide an optimized energy-saving control method for drying oven systems based on deep learning.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The method for optimizing energy-saving control of drying oven systems based on deep learning includes the following steps: Step 1: Based on the distribution of drying ovens, circulating fans and supply and return air ducts in the drying room, divide the drying room space into load areas, collect temperature, humidity and equipment operation data of each load area and perform preprocessing. Step 2: Input the preprocessed temperature and humidity time series data and equipment operation data into the trained time series convolutional neural network model to obtain the heat island probability of each load area, and determine the local heat island area based on the heat island probability. Step 3: Generate the initial heat demand of each load area based on the drying process requirements and real-time operating status, configure negative heat compensation for the local heat island area, and generate the time sequence structure of the heat demand of the area in chronological order. Step 4: Based on the temporal structure of the regional heat demand and the airflow delay characteristics of the drying room's height stratification, determine the operating speed of the circulating fan and the gas flow rate of the drying oven, and generate a regional heating energy consumption sequence. Step 5: Generate PUE value based on the regional heating energy consumption sequence, total drying room energy consumption and effective drying energy consumption, and correct the ineffective wind power component according to the spatial difference contribution value, and then perform closed-loop correction on the control parameters.

[0007] Furthermore, the load area division includes: obtaining the distribution coordinates of all drying ovens in the drying room, the installation position and air supply angle of the circulating fan, and the path of the supply and return air ducts; dividing the drying room space into multiple rectangular load areas according to the air supply coverage of the drying ovens, and ensuring that the side length of each load area is matched with the maximum effective air supply distance of a single drying oven; arranging temperature sensors and humidity sensors at the center and corners of each load area, arranging flow sensors on the gas pipelines of the drying ovens, and arranging speed sensors and power sensors on the motors of the circulating fans.

[0008] Furthermore, the preprocessing includes outlier removal and moving average filtering of the collected raw data, and the filtered data is used as the input of the temporal convolutional neural network model. The temporal convolutional neural network model includes an input layer, a temporal convolutional layer, a global average pooling layer, and an output layer. The input layer receives the average temperature, average humidity, total heating power of the drying oven, and total power of the circulating fan within the historical data time window. The output layer outputs the probability value of each load area being a local heat island area at the current moment.

[0009] Furthermore, the temporal convolutional layer adopts a causal convolutional structure, and each temporal convolutional layer is followed by a batch normalization layer and an activation function layer; the temporal convolutional neural network model is trained in a supervised manner using actual operating data of the drying room collected under different load conditions, and the label data is determined based on the temperature distribution cloud map and drying quality detection results. During actual operation, continuous preprocessed data is input into the trained temporal convolutional neural network model, and the load area whose heat island probability meets the preset judgment condition is marked as a local heat island area.

[0010] Furthermore, generating the time-series structure of the regional heat demand includes: determining the target drying curve for each load area based on the type of material to be dried, the initial moisture content, and the corresponding drying process specifications; determining the theoretical heat demand based on the target drying curve and the current temperature and current moisture content of each load area; generating an air supply coverage matrix based on the air supply angle, air supply speed, and relative position of each drying oven to each load area, and determining the initial heat output demand corresponding to the drying oven based on the air supply coverage matrix and the theoretical heat demand; for load areas marked as local heat island areas, configuring a negative heat compensation amount based on the deviation between the actual average temperature and the target temperature, combining the initial heat demand and the negative heat compensation amount into a regional enhanced heat demand, and integrating them into a time-series structure of regional heat demand in chronological order.

[0011] Furthermore, determining the operating speed of the circulating fan includes: dividing the drying room space vertically into multiple height layers and determining the airflow delay coefficient for each height layer; breaking down the temporal structure of the heat demand in the area into heat demands for each height layer according to the material mass ratio within each height layer; calculating the effective operating intensity of the circulating fan at each moment based on the heat demand of each height layer and the airflow delay coefficient, and determining the actual operating speed of the circulating fan based on the effective operating intensity and the rated speed of the circulating fan.

[0012] Furthermore, determining the gas flow rate of the drying oven and the regional heating energy consumption sequence includes: determining the initial gas flow rate of the drying oven based on the initial heat output demand of the drying oven; adjusting the initial gas flow rate in conjunction with the average effective operating intensity of each circulating fan to obtain the actual gas flow rate of the drying oven; calculating the fan power component based on the actual operating speed of the circulating fan; calculating the burner power component based on the actual gas flow rate of the drying oven; and determining the ineffective air power component based on the area ratio of the local heat island region and the total power of the circulating fan; combining the fan power component, burner power component, and ineffective air power component into the regional heating energy consumption at the corresponding time, and generating the regional heating energy consumption sequence in chronological order.

[0013] Furthermore, the generation of the PUE value and the closed-loop correction of the control parameters include: performing time alignment processing on the regional heating energy consumption sequence, and generating the PUE value by combining the total energy consumption of the drying room and the effective drying energy consumption at each time point; calculating and summing the spatial difference contribution values ​​of each load area, and using the sum of the spatial difference contribution values ​​to correct the ineffective wind power component; recalculating the total heating energy consumption of the drying room and the corrected PUE value based on the corrected ineffective wind power component; comparing the corrected PUE value with the target PUE value, and adjusting the heat compensation coefficient and the ineffective wind coefficient online according to the comparison results, and applying the adjusted control parameters to the next control cycle.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. To address the problem that existing technologies struggle to accurately identify local heat island areas and uneven heat distribution, this invention employs a load area division method based on equipment distribution and airflow characteristics. Multiple types of sensors are deployed in each area to collect operational data, and a temporal convolutional neural network model is constructed to process the time-series temperature and humidity data and equipment operational data. This enables dynamic identification of local heat island areas. Simultaneously, negative heat compensation is configured for the heat island areas, and a temporal structure of regional heat demand is generated. This helps improve the uniformity of heat distribution within the drying chamber and reduces energy waste caused by localized heat excess. 2. To address the problems of poor coordination between the drying oven and the circulating fan and insufficient energy consumption control accuracy in existing technologies, this invention divides the drying chamber space into multiple height layers along the vertical direction and determines the airflow delay coefficient. Based on the temporal structure of regional heat demand, the operating speed of the circulating fan and the gas flow rate of the drying oven are determined in a coordinated manner. At the same time, the ineffective air power component is corrected by calculating the PUE value and the spatial difference contribution value, thereby realizing the online closed-loop adjustment of control parameters. This can improve the energy consumption control accuracy of the system and optimize the energy-saving operation effect of the system. Attached Figure Description

[0015] Figure 1 A flowchart illustrating an optimized energy-saving control method for a deep learning-based drying oven system; Figure 2 This is a flowchart illustrating the generation and dynamic compensation of the time-series structure of regional heat demand in this invention. Detailed Implementation

[0016] Example, refer to Figure 1 The optimized energy-saving control method for the drying oven system based on deep learning in this embodiment specifically includes the following steps: Step 1: Divide the drying room space into load zones and collect multi-source data.

[0017] Based on the physical distribution of equipment and airflow organization characteristics within the drying chamber, spatial gridding is performed, and multi-dimensional operational data is collected and preprocessed simultaneously to provide standardized input for subsequent heat island identification and energy consumption calculation.

[0018] S11. Load Area Division: Obtain the distribution coordinates of all drying ovens within the drying chamber, the installation location and air supply angle of the circulating fans, and the routing of the supply and return air ducts. The maximum effective air supply distance for a single drying oven can be obtained through the oven's factory parameters or on-site air supply testing. During testing, the wind speed is measured at different distances directly in front of the drying oven; the distance at which the wind speed drops to a preset percentage of the rated wind speed is the maximum effective air supply distance. Using the drying oven's air supply coverage area as the basic unit, the drying chamber space is divided into... OK Lieutenant General A rectangular load area, in which The side length of each load area does not exceed half of the maximum effective air supply distance of a single drying oven, ensuring that the airflow and temperature fields in each area are relatively uniform. S12, Sensor Deployment: A temperature sensor and a humidity sensor are installed at the center and four corners of each load area. A flow sensor is installed on the gas pipeline of each drying oven, and a speed sensor and a power sensor are installed on the motor of each circulating fan. The acquisition frequency of all sensors can be set according to the system response requirements. S13, Data Preprocessing: The collected raw data underwent outlier removal and moving average filtering. Outlier removal employed... The criterion is to delete data points that exceed three standard deviations from the mean; the window size of the moving average filter can be adjusted according to the data noise level, and the filtered data is used as input for subsequent steps.

[0019] Step 2: Dynamic identification of local heat island regions based on deep learning.

[0020] A lightweight temporal convolutional neural network model is constructed, which takes into account the time-series temperature and humidity data of each load area and the equipment operation data, and identifies local heat island areas with excess heat in real time.

[0021] S21. Model Building: Construct a lightweight temporal convolutional neural network model consisting of an input layer, three temporal convolutional layers, one global average pooling layer, and one output layer. The dimension of the input layer is [missing information]. ,in The length of the historical data time window. The characteristic dimensions include the average temperature, average humidity, total heating power of the drying oven, and total power of the circulating fan for each load area; The temporal convolutional layers adopt a causal convolutional structure with a kernel size of 3 and dilation coefficients of 1, 2, and 4 respectively. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function. The number of output channels can be set according to the model complexity requirements. The global average pooling layer compresses the feature map output by the convolutional layer into a one-dimensional vector. The output layer uses the Sigmoid activation function to output the probability value of each load region being a heat island region at the current time step. S22, Model Training: The model training employs a supervised learning approach. The training dataset is obtained through actual operating data of the drying chamber under different load conditions. The label data is annotated by professional technicians based on temperature distribution cloud maps and drying quality test results. The collected dataset is divided into training, validation, and test sets according to a preset ratio. The model loss function uses the binary cross-entropy loss function, the optimizer uses the Adam algorithm, the learning rate can be adjusted according to the training convergence, and the training batch size can be set according to the hardware computing power. An early stopping mechanism is adopted during training. Training stops when the validation set loss no longer decreases for several consecutive rounds to prevent overfitting. S23, Urban Heat Island Identification: In actual operation, the pre-processed continuous The duration data is input into the trained model to obtain the heat island probability value for each load region. ;when When the load exceeds a preset threshold, the load area is marked as a local heat island area. The preset threshold can be adjusted based on test results under different operating conditions to obtain an identification accuracy that meets actual needs.

[0022] Step 3: Generation and dynamic compensation of the temporal structure of regional heat demand.

[0023] like Figure 2 As shown, based on the drying process requirements and real-time operating status of each load area, the initial heat demand of the area is calculated, and negative heat compensation is performed on the local heat island area to generate the time sequence structure of the regional heat demand.

[0024] S31. Calculation of theoretical heat demand: Based on the type and initial moisture content of the material to be dried, a target drying curve is determined for each load zone. The target drying curve is defined as a function of the material's moisture content over time and is determined by the corresponding material's drying process specifications. The current moisture content of each load zone at each moment can be measured in real time by moisture content sensors installed within the load zone, or estimated based on the target drying curve and the drying time. Based on the target drying curve, the theoretical heat requirement for each load zone at each moment is calculated. : ; in, For the first The mass of the material to be dried in each load area, in kg; The specific heat capacity at constant pressure of the material to be dried, in units of... It is determined by the type of material; For the first Each load area in The difference between the target temperature and the current temperature at any given time, expressed in Kelvin (K). The latent heat of vaporization of water is taken as 2260 kJ / kg; For the first Each load area in The difference between the target moisture content and the current moisture content at any given time; S32. Generation of air supply coverage matrix: Based on the air supply angle, air supply velocity, and relative position to each load area of ​​each drying oven, the air supply coverage coefficient of each drying oven to each load area is calculated using an airflow diffusion model, generating an air supply coverage matrix corresponding to each load area. ,in Indicates the first The drying oven for the first The air supply coverage coefficient for each load area ranges from 0 to 1; when the drying oven Completely does not cover the load area hour, When the drying oven Fully cover the load area hour, ; S33. Calculation of initial heat output requirements: Based on air supply coverage matrix Compared with theoretical heat demand Calculate the initial heat output requirement for each drying oven. : ; S34. Heat compensation in urban heat island areas: For load areas designated as localized heat islands, a negative heat compensation amount is added to their initial heat demand. The amount of heat compensation is directly proportional to the temperature deviation in the heat island region: ; in, It is the heat compensation coefficient, with units of W / K and a value range of 100~500 W / K, which is determined based on the heat preservation performance and airflow organization characteristics of the drying room; For the first Each load area in The actual average temperature at any given time, in Kelvin (K). For the first Each load area in The target temperature at any given time, in Kelvin (K). S35, Generation of temporal structures: The initial heat demand of each load zone is added to the heat compensation amount to obtain the zone-enhanced heat demand. The enhanced heat demand of all load areas is integrated in chronological order to generate a time-series structure of regional heat demand.

[0025] Step 4: Optimize the intensity of coordinated operation between the drying oven and the circulating fan.

[0026] Based on the temporal structure of regional heat demand and combined with the airflow characteristics of the stratified drying room, the gas flow rate of each drying oven and the operating speed of each circulating fan are optimized to achieve coordinated control of heating and ventilation.

[0027] S41. Determination of Height Stratification and Airflow Delay Coefficient: The drying room space is divided vertically into... The material is stacked in several layers, each with a height that can be set according to the material stacking height. An airflow delay coefficient is defined based on the airflow delay characteristics of each layer. ,in The airflow delay coefficient is a number assigned to each height layer. It represents the ratio of the time required for heat to be transferred from the oven outlet to the corresponding height layer to the reference time. The reference time is the time required for heat to be transferred to the bottom height layer. The airflow delay coefficient can be obtained through CFD simulation or actual testing. During the test, temperature sensors are installed at the oven outlet and each height layer to record the time difference of temperature change under the same heat input. The airflow delay coefficient of the corresponding height layer is then calculated. S42. Calculation of effective operating intensity of circulating fan: The temporal structure of regional heat demand is divided into height strata, with the division ratio determined based on the material mass proportion within each height stratum, thus obtaining the heat demand for each height stratum at each time point. According to the heat demand of highly stratified With airflow delay coefficient Calculate the effective operating intensity of each circulating fan at each time point. : ; in, The density of air is taken as 1.2 kg / m³. The specific heat capacity of air at constant pressure is taken as 1005 J / (kg・K); For the first The rated air volume of each circulating fan, in m³ / s; This is the difference between the supply air temperature and the return air temperature, expressed in Kelvin (K). Based on effective operating intensity Calculate the actual operating speed of the circulating fan. : ; in, For the first The rated speed of each circulating fan, in r / min; S43. Adjustment of gas flow rate in drying oven: Based on the temporal structure of regional heat demand, the initial gas flow rate of each drying oven at each time point is calculated. : ; in, The combustion efficiency of the drying oven is defined as 0.85 to 0.95. The lower calorific value of the gas is expressed in J / m³ and is determined by the type of gas. The initial gas flow rate is adjusted in conjunction with the effective operating intensity of the circulating fan to obtain the actual gas flow rate of the drying oven. : ; in, for The average effective operating intensity of all circulating fans at any given time; The rated operating intensity of the circulating fan is 1; S44. Energy Consumption Component Calculation: Calculate the fan power component based on the actual operating speed of the circulating fan and the actual gas flow rate of the drying oven. With burner power component : ; ; in, For the first The rated power of each circulating fan is expressed in watts (W). Define invalid wind power components This refers to the portion of the total power of the circulating fan that is not effectively utilized. Due to the excessively high temperature in the local heat island area, some of the heat transported by the circulating fan cannot be absorbed by the material, resulting in energy waste. The amount of waste is directly proportional to the total area of ​​the local heat island area. ; in, The ineffective wind coefficient has a value range of 0.05 to 0.2, and is determined based on the airflow organization characteristics of the drying room. for The ratio of the total area of ​​the local heat island region to the total area of ​​the drying room at any given time; for The total power of all circulating fans at any given time, in watts (W). Drying room Total heating energy consumption at any time It is the sum of all fan power components, all burner power components, and ineffective wind power components.

[0028] Step 5: Real-time PUE calculation and closed-loop correction of control parameters.

[0029] The system calculates the PUE value of the drying room in real time and performs closed-loop correction of the control parameters based on the PUE deviation and the deviation of regional heating energy consumption to continuously optimize the energy-saving effect of the system.

[0030] S51, Real-time PUE Calculation: Drying room The PUE value at any given time is defined as the ratio of the total energy consumption of the drying room to the effective drying energy consumption: ; in, for The total energy consumption of the drying room at all times includes heating energy consumption, lighting energy consumption, and auxiliary equipment energy consumption, in W; for The effective energy consumption used for material drying at any time, in W, is equal to the sum of the theoretical heat demand of all load areas multiplied by the insulation efficiency of the drying room. The insulation efficiency of the drying room can be obtained through on-site heat loss testing. S52. Calculation of Spatial Difference Contribution Value: Calculate the spatial difference contribution value of each load region This is used to quantify the impact of temperature differences between different load regions on PUE. ; in, for The average temperature of all load zones in the drying oven at any given time, in Kelvin (K). Calculate the sum of the spatial difference contribution values ​​of all load regions. : ; S53, Correction of Ineffective Wind Power Components: Sum of contributions based on spatial differences The ineffective wind power component is corrected to obtain the corrected ineffective wind power component. : ; The total heating energy consumption and PUE of the drying room are recalculated using the corrected ineffective wind power components, resulting in the corrected PUE value. ; S54. Closed-loop adjustment of control parameters: The corrected PUE value is compared with the preset target PUE value to calculate the PUE deviation. ,when A value greater than 0 indicates that the current system energy consumption is higher than the target value, and the control parameters need to be adjusted. A proportional-integral controller is used to adjust the thermal compensation coefficient. With ineffective wind coefficient Online adjustments are performed, with the controller input being the PUE deviation. The output is the adjustment amount of the control parameters; the proportional coefficient and integral coefficient are determined by the Ziegler-Nichols method or on-site debugging, and the adjustment step size can be set according to the system response speed. The control parameters are corrected once at a preset time interval, and the corrected parameters are immediately applied to the next control cycle. Through continuous closed-loop control, the PUE value of the drying room is always kept within the target range. Through the detailed description of the above embodiments, the energy-saving control method for the drying oven system based on deep learning of the present invention combines deep learning technology with the physical characteristics of the drying oven process to construct a complete control process from spatial area division, dynamic identification of heat islands, calculation of heat demand, coordinated control of equipment to closed-loop correction of energy consumption. This method uses the load area as the basic control unit, utilizes a temporal convolutional neural network to achieve accurate identification of local heat islands, combines airflow delay characteristics to achieve linkage control between the drying oven and the circulating fan, and continuously optimizes control parameters through feedback of PUE value and spatial difference contribution value. This achieves refined and dynamic control of the drying oven system, adapting to different load conditions and drying process requirements, helping to improve the operating efficiency of the drying oven system and reduce ineffective energy consumption during system operation.

[0031] The preset parameters in the above formulas shall be set by those skilled in the art according to the actual situation.

[0032] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0033] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0034] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0035] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0036] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0037] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0038] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An optimized energy-saving control method for a drying oven system based on deep learning, characterized in that, The method flow is as follows: Step 1: Based on the distribution of drying ovens, circulating fans and supply and return air ducts in the drying room, divide the drying room space into load areas, collect temperature, humidity and equipment operation data of each load area and perform preprocessing. Step 2: Input the preprocessed temperature and humidity time series data and equipment operation data into the trained time series convolutional neural network model to obtain the heat island probability of each load area, and determine the local heat island area based on the heat island probability. Step 3: Generate the initial heat demand of each load area based on the drying process requirements and real-time operating status, configure negative heat compensation for the local heat island area, and generate the time sequence structure of the heat demand of the area in chronological order. Step 4: Based on the temporal structure of the regional heat demand and the airflow delay characteristics of the drying room's height stratification, determine the operating speed of the circulating fan and the gas flow rate of the drying oven, and generate a regional heating energy consumption sequence. Step 5: Generate PUE value based on the regional heating energy consumption sequence, total drying room energy consumption and effective drying energy consumption, and correct the ineffective wind power component according to the spatial difference contribution value, and then perform closed-loop correction on the control parameters.

2. The optimized energy-saving control method for a drying oven system based on deep learning according to claim 1, characterized in that, The load area division includes: obtaining the distribution coordinates of all drying ovens in the drying room, the installation position and air supply angle of the circulating fan, and the path of the supply and return air ducts; dividing the drying room space into multiple rectangular load areas according to the air supply coverage of the drying ovens, and ensuring that the side length of each load area is matched with the maximum effective air supply distance of a single drying oven; arranging temperature sensors and humidity sensors at the center and corners of each load area, arranging flow sensors on the gas pipelines of the drying ovens, and arranging speed sensors and power sensors on the motors of the circulating fans.

3. The optimized energy-saving control method for a drying oven system based on deep learning according to claim 2, characterized in that, The preprocessing includes outlier removal and moving average filtering of the collected raw data. The filtered data is used as the input of the temporal convolutional neural network model. The temporal convolutional neural network model includes an input layer, a temporal convolutional layer, a global average pooling layer, and an output layer. The input layer receives the average temperature, average humidity, total heating power of the drying oven, and total power of the circulating fan within the historical data time window. The output layer outputs the probability value of each load area being a local heat island area at the current moment.

4. The optimized energy-saving control method for a drying oven system based on deep learning according to claim 3, characterized in that, The temporal convolutional layer adopts a causal convolutional structure, and each temporal convolutional layer is followed by a batch normalization layer and an activation function layer. The temporal convolutional neural network model is trained in a supervised manner using actual operating data of the drying room collected under different load conditions. The label data is determined based on the temperature distribution cloud map and the drying quality detection results. During actual operation, continuous preprocessed data is input into the trained temporal convolutional neural network model, and the load area whose heat island probability meets the preset judgment condition is marked as a local heat island area.

5. The optimized energy-saving control method for a drying oven system based on deep learning according to claim 1, characterized in that, The generation of the regional heat demand time sequence structure includes: determining the target drying curve for each load area based on the type of material to be dried, the initial moisture content, and the corresponding drying process specifications; determining the theoretical heat demand based on the target drying curve and the current temperature and current moisture content of each load area; generating an air supply coverage matrix based on the air supply angle, air supply speed, and relative position of each drying oven to each load area, and determining the initial heat output demand corresponding to the drying oven based on the air supply coverage matrix and the theoretical heat demand; for load areas marked as local heat island areas, configuring a negative heat compensation amount based on the deviation between the actual average temperature and the target temperature, combining the initial heat demand and the negative heat compensation amount into a regional enhanced heat demand, and integrating them into a regional heat demand time sequence structure in chronological order.

6. The optimized energy-saving control method for a drying oven system based on deep learning according to claim 5, characterized in that, Determining the operating speed of the circulating fan includes: dividing the drying room space into multiple height layers along the vertical direction and determining the airflow delay coefficient for each height layer; breaking down the temporal structure of the heat demand in the area into heat demands for each height layer according to the material mass ratio within each height layer; calculating the effective operating intensity of the circulating fan at each moment based on the heat demand of each height layer and the airflow delay coefficient, and determining the actual operating speed of the circulating fan based on the effective operating intensity and the rated speed of the circulating fan.

7. The optimized energy-saving control method for a drying oven system based on deep learning according to claim 6, characterized in that, Determining the gas flow rate of the drying oven and the regional heating energy consumption sequence includes: determining the initial gas flow rate of the drying oven based on the initial heat output demand of the drying oven; adjusting the initial gas flow rate in conjunction with the average effective operating intensity of each circulating fan to obtain the actual gas flow rate of the drying oven; calculating the fan power component based on the actual operating speed of the circulating fan; calculating the burner power component based on the actual gas flow rate of the drying oven; and determining the ineffective air power component based on the area ratio of the local heat island region and the total power of the circulating fan; combining the fan power component, burner power component, and ineffective air power component to form the regional heating energy consumption at the corresponding time, and generating the regional heating energy consumption sequence in chronological order.

8. The optimized energy-saving control method for a drying oven system based on deep learning according to claim 7, characterized in that, The generation of the PUE value and the closed-loop correction of the control parameters include: performing time alignment processing on the regional heating energy consumption sequence, and generating the PUE value by combining the total energy consumption of the drying room and the effective drying energy consumption at each time point; calculating and summing the spatial difference contribution values ​​of each load area, and using the sum of the spatial difference contribution values ​​to correct the ineffective wind power component; recalculating the total heating energy consumption of the drying room and the corrected PUE value based on the corrected ineffective wind power component; comparing the corrected PUE value with the target PUE value, and adjusting the heat compensation coefficient and the ineffective wind coefficient online according to the comparison results, and applying the adjusted control parameters to the next control cycle.

Images