Efficient warm air supply device for greenhouse

Abstract

The invention provides an efficient warm air supply device for a greenhouse, and relates to the technical field of greenhouses, and the device comprises a heat source module used for generating warm air; the air supply module is used for conveying warm air to different areas in the greenhouse shed; the data acquisition module is used for acquiring environment data of different areas in the greenhouse; the temperature regulation module is used for constructing a DQN-based temperature control model and optimizing the DQN temperature control model by using a multi-target elk swarm optimization algorithm to obtain a temperature control model; inputting the processed environment data into a temperature control model to obtain a regulation and control quantity; the control module is used for generating a control instruction according to the regulation quantity and transmitting the control instruction to the heat source module and the air supply module. According to the method, the DQN temperature control model is optimized and put based on the multi-target elk swarm algorithm, so that the temperature of different areas in the greenhouse is accurately controlled, and the energy waste is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of greenhouses, and in particular to a high-efficiency warm air supply device for a greenhouse. Background Art

[0002] With the rapid development of facility agriculture, greenhouses are becoming increasingly popular, and more and more farmers and agricultural enterprises are adopting them for agricultural production. This has created a broad market for efficient greenhouse warm air supply devices. As an integral part of modern agriculture, greenhouse internal environmental regulation is crucial for crop growth. Maintaining a suitable temperature within the greenhouse, especially in colder months, requires the supply of warm air to raise the temperature. However, traditional warm air supply devices often suffer from low efficiency, high energy consumption, and uneven air supply. This results in uneven temperature distribution within the greenhouse, with some areas being too high or too low, impacting the normal growth of crops.

[0003] Furthermore, greenhouses are typically large and contain a variety of crops, each with varying temperature requirements. Consequently, traditional warm air supply systems struggle to meet the temperature requirements of different areas within the greenhouse, leading to energy waste and poor crop growth.

[0004] In order to improve the efficiency of the greenhouse warm air supply device, reduce energy consumption, and achieve uniform distribution of temperature inside the greenhouse, it is necessary to develop a greenhouse warm air supply device with high efficiency, energy saving and uniform air supply. Summary of the Invention

[0005] The present invention provides a high-efficiency warm air supply device for a greenhouse, which is used to solve the problem that the existing technology cannot meet the temperature requirements of different areas inside the greenhouse, resulting in energy waste, poor crop growth and low efficiency.

[0006] In one aspect, the present invention provides a greenhouse high-efficiency warm air supply device, comprising: The heat source module is used to heat the air to generate warm air.

[0007] The air supply module is used to deliver warm air to different areas in the greenhouse through multiple air supply outlets.

[0008] The data acquisition module is used to obtain environmental data of different areas in the greenhouse and pre-process the environmental data to obtain processed environmental data. The environmental data includes: the area of different areas, actual temperature, and preset temperature.

[0009] The temperature control module is used to build a DQN-based temperature control model. This model is optimized using the multi-objective elk swarm optimization algorithm to generate the desired temperature control model. Processed environmental data is input into the temperature control model to generate control variables. These control variables include the minimum power of the heat source module required to achieve the preset temperature for each zone and the corresponding air outlet size.

[0010] The control module is used to generate control instructions based on the control quantity and transmit the control instructions to the heat source module and the air supply module.

[0011] According to the present invention, a high-efficiency warm air supply device for a greenhouse includes an air supply module comprising a filter unit and a regulator unit. The filter unit is used to filter dust and impurities in the air. The regulator unit is used to adjust the wind speed at the air supply outlet.

[0012] According to a greenhouse high-efficiency warm air supply device provided by the present invention, the data acquisition module preprocesses environmental data including: Remove outliers in environmental data and fill missing values in environmental data using the mean method.

[0013] The moving average method is used to smooth the environmental data.

[0014] According to the greenhouse high-efficiency warm air supply device provided by the present invention, the DQN temperature control model construction process includes: Construct the main network structure including input layer, hidden layer and output layer, and initialize the network weights and hyperparameters.

[0015] Set up the target network according to the main network structure.

[0016] Create an experience replay pool, set its maximum capacity based on the model's input and output, and define the structure for storing data.

[0017] According to the present invention, a high-efficiency warm air supply device for a greenhouse comprises a hidden layer comprising a first hidden layer and a second hidden layer. The first hidden layer comprises 128 nodes and introduces nonlinear features. The second hidden layer comprises 64 nodes and uses a ReLU activation function to maintain nonlinear features.

[0018] According to a greenhouse high-efficiency warm air supply device provided by the present invention, the steps of optimizing the DQN temperature control model include: Initialize the population and randomly generate elk individuals, each of which represents a network model hyperparameter.

[0019] The mean square error is used as the fitness function of the model.

[0020] The fitness values of individuals in the population are calculated and sorted. The number of male elk is set according to their behavior during the estrus season. Individuals with higher fitness are selected as male elk, and the remaining individuals are selected as female elk. A roulette wheel selection method is used to assign each male elk a number of female elk proportional to its fitness, and multiple new populations are formed.

[0021] Multiple new populations obtain offspring individuals based on their calving season behavior.

[0022] Merge all new populations and offspring individuals to obtain population one.

[0023] The fitness value of population one is calculated. If the output fitness value is higher than the preset fitness threshold, the optimal hyperparameter is output. Otherwise, the iteration is continued until the fitness value is higher than the preset fitness threshold or the maximum number of iterations is reached to obtain the optimal hyperparameter.

[0024] According to the greenhouse high-efficiency warm air supply device provided by the present invention, the formula for expressing the number of male elk in the estrus season is: in, is the number of bull elk, is the proportion of bull elk in the population, is the total number of elk.

[0025] According to the greenhouse high-efficiency warm air supply device provided by the present invention, the calving season behavior expression formula is: in, is the position of the new individual, It is a male elk individual. It is a female elk individual, is a parameter that controls the mating weight, is the factor of variation, It is a random factor.

[0026] According to the present invention, a high-efficiency warm air supply device for a greenhouse includes a control module comprising a processor, a digital-to-analog converter, and a communication device. The processor is configured to generate control instructions based on a control variable. The digital-to-analog converter is configured to convert the control instructions into analog signals. The communication device is configured to transmit the analog signals to the heat source module and the air supply module.

[0027] According to the present invention, a high-efficiency warm air supply device for a greenhouse comprises a control module comprising a maintenance reminder unit and a statistical analysis unit. The maintenance reminder unit is configured to obtain the operating hours of the heat source module and the air supply module and generate maintenance reminders based on the operating hours. The statistical analysis unit is configured to obtain the energy consumption of the heat source module and the air supply module and generate energy consumption reports.

[0028] The present invention provides a high-efficiency warm air supply device for a greenhouse. This device builds a DQN-based temperature control model and optimizes it using a multi-objective elk herd optimization algorithm to obtain a temperature control model. Processed environmental data is input into the temperature control model to generate a control variable. This control variable transmits instructions to the heat source module and air supply module, achieving precise temperature control of different areas within the greenhouse. This addresses the temperature requirements of different areas within the greenhouse, improves greenhouse temperature control efficiency, and reduces energy waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0030] Figure 1 The present invention provides a flow chart of a high-efficiency warm air supply device for a greenhouse. DETAILED DESCRIPTION

[0031] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0032] See also Figure 1 This embodiment provides a greenhouse high-efficiency warm air supply device including a heat source module, an air supply module, a data acquisition module, a temperature adjustment module, and a control module.

[0033] The heat source module is used to heat the air to generate warm air.

[0034] In this embodiment, the heat source module can utilize a fan and an air heat source pump. The air heat source pump is used to heat the air. The fan draws in outside air and outputs heated air from the air heat source pump. Highly efficient and energy-saving fans and air heat source pumps can also be used to reduce energy consumption. The air heat source pump can utilize advanced heat pump technology to improve efficiency.

[0035] The air supply module is used to deliver warm air to different areas in the greenhouse through multiple air supply outlets.

[0036] The air supply module includes a filter unit and a regulator unit. The filter unit is used to filter dust and impurities in the air. The regulator unit is used to adjust the wind speed and direction at the air supply outlet.

[0037] In this embodiment, the air supply module may include a main air supply duct and branch air supply ducts. The main air supply duct is connected to the heat source module, and the branch air supply ducts branch off from the main air supply duct and are respectively installed in different areas of the greenhouse to achieve temperature control in different areas. The air supply outlet adjustment mechanism includes an adjustable air outlet and a drive device. The adjustable air outlet is installed at the end of the air supply module. The drive device is connected to the adjustable air outlet and is used to drive the adjustable air outlet to adjust the size and direction of the adjustable air outlet. The filter unit is responsible for filtering dust and other impurities in the air, ensuring that the air entering the greenhouse is clean and fresh, which is conducive to the healthy growth of crops. A filter screen can be provided. The filter screen can be made of non-woven fabric or synthetic fiber material supported by a metal grid. It can effectively block large particles of dust and impurities. This filter screen has the advantages of low wind resistance, large dust holding capacity, and easy replacement. The filter screen should be replaced or cleaned regularly according to the usage of the greenhouse and the air quality. A filter screen record sheet can also be established to record the time of each replacement or cleaning, the filter screen type, and the person who replaced or cleaned the filter screen to help track the maintenance of the filter screen and ensure the filtering effect.

[0038] The data acquisition module is used to obtain environmental data from different areas of the greenhouse and pre-process the environmental data to obtain processed environmental data. The environmental data includes: the area, actual temperature, and preset temperature of different areas.

[0039] In this embodiment, a flexible data collection strategy can be employed to accommodate the changing characteristics of the greenhouse environment at different times of day. During periods of significant temperature fluctuation, such as early morning and evening, the temperature collection frequency can be increased to every half hour to ensure that even the slightest temperature changes can be captured in real time. During periods of relatively stable temperature with less fluctuation, such as midday and late at night, the collection frequency can be appropriately reduced to once an hour to balance data accuracy and system resource consumption. Real-time data collection can also be employed. To accurately obtain the actual temperature within the greenhouse, highly sensitive temperature sensors can be used. These sensors are positioned in corresponding locations within the greenhouse to ensure a comprehensive and accurate reflection of the greenhouse's temperature conditions. Advanced communication technologies can also be utilized to enable real-time data transmission between the sensors and the data acquisition module, thereby ensuring data timeliness and accuracy. To ensure data continuity and availability, an efficient data storage solution can be designed. Environmental data collected daily is promptly stored in a local database. Even in the event of a poor or interrupted network connection, historical data can still be retrieved, providing a solid foundation for subsequent data analysis and intelligent decision-making. Given the potential for sensor errors or drift, the collected data can be rigorously calibrated. Regular sensor calibration tests are conducted to determine the sensor's error range by comparing the sensor output value with the standard value. Compensation corrections can also be made to the collected data based on the test results to ensure data accuracy and reliability.

[0040] The data acquisition module preprocesses the environmental data by removing abnormal values in the environmental data and filling in missing values in the environmental data using the mean method.

[0041] In this embodiment, the statistical characteristics of the data can be used to set a threshold, and data exceeding the threshold is considered an outlier. The IQR method can be used to effectively identify and remove obviously erroneous values, improving data quality. Alternatively, the isolation forest learning model can be used to learn normal data and then identify data with significant distribution differences from the normal data as outliers. Clearly erroneous or invalid outliers can be directly deleted. Deleting too much data may result in information loss, affecting the accuracy of subsequent analysis. Outliers that may be caused by sensor failure or data transmission errors can be replaced with data from adjacent time points or the average data for the area.

[0042] The moving average method is used to smooth the environmental data.

[0043] In this embodiment, when using the moving average method, a simple movement can be performed to calculate the mean of the data points in a fixed-size window. The window size can be adjusted according to the data characteristics, such as a 3-point, 5-point, or 7-point moving average. In addition to the simple moving average method, a weighted translation method can also be used to further improve the accuracy of data processing. The core of the weighted translation method is to assign a specific weight to each data point in the window. The most recent data point can be given a higher weight because such a design can better reflect the current change trend of the data. This weight distribution method makes the data analysis results closer to the actual situation and also enhances the timeliness of data prediction.

[0044] The temperature control module is used to build a DQN-based temperature control model. This model is optimized using the multi-objective elk swarm optimization algorithm to generate the desired temperature control model. Processed environmental data is input into the temperature control model to generate control variables. These control variables include the minimum power of the heat source module required to achieve the preset temperature for each zone and the corresponding air outlet size.

[0045] In this embodiment, the temperature adjustment module may use a temperature sensor and a computer. The temperature sensor is used to collect temperature data from each area in real time. The computer is used to run the DNQ network model and the multi-objective elk herd optimization algorithm.

[0046] The control module is used to generate control instructions based on the control quantity and transmit the control instructions to the heat source module and the air supply module.

[0047] The control module includes a processor, a digital-to-analog converter, and a communication device. The processor is used to generate control instructions based on the control variables. The digital-to-analog converter converts the control instructions into analog signals. The communication device transmits the analog signals to the heat source module and the air supply module.

[0048] In this embodiment, the choice of processor should be determined by the system's complexity and real-time requirements. A high-performance microprocessor or digital signal processor can be selected. High-performance microprocessors offer powerful computing and multitasking capabilities, enabling them to rapidly process large amounts of data and generate accurate control instructions. DSPs excel at processing digital signals and are particularly suitable for control tasks requiring high precision and real-time performance. By selecting the right processor, the system can maintain high efficiency and stability even when faced with complex control tasks. Secondly, the communication device should adhere to appropriate communication protocols to ensure accurate data transmission. For wired communication, protocols such as RS-485 and CAN bus can be selected. The RS-485 protocol offers advantages such as long transmission distance and strong anti-interference capabilities, making it suitable for long-distance, multi-node data transmission scenarios. The CAN bus, on the other hand, is widely used in industrial automation due to its high real-time performance, high reliability, and flexible network topology. For wireless communication, protocols such as Wi-Fi and Zigbee can be selected. Wi-Fi offers high transmission rates and wide coverage, making it suitable for scenarios requiring high-speed data transmission. Zigbee, on the other hand, is widely used in the Internet of Things (IoT) due to its low power consumption, self-organizing network, and ease of deployment. By selecting the appropriate communication protocol, data transmission can be ensured to be accurate and reliable. The control module can also incorporate a fault detection mechanism to continuously monitor the operating status of the communication device, including key indicators such as data transmission integrity, speed, and error rate. By comparing these indicators with preset thresholds, communication failures can be promptly detected. Machine learning or statistical methods can also be used to develop anomaly detection algorithms to analyze communication data in real time. When data fluctuates abnormally or deviates from normal ranges, the algorithm can issue an alarm and trigger a fault handling process. When a communication device failure is detected, the control module should be able to automatically switch to an alternate communication channel. This involves reconfiguring communication parameters, establishing a new connection, and ensuring data transmission continuity. In addition to automatic switching, the control module should also be able to issue an alert notification to the system administrator or monitoring center. The alert information should include key information such as the fault type, time of occurrence, and scope of impact, allowing administrators to quickly locate the problem and take corrective action.

[0049] The DQN temperature control model construction process includes: Construct the main network structure including input layer, hidden layer and output layer, and initialize the network weights and hyperparameters.

[0050] In this embodiment, the actual temperature, area, and preset temperature of different zones within the greenhouse are considered. These features together constitute the dimensions of the input vector. For example, if the greenhouse is divided into n zones, each with 3 features, the number of nodes in the input layer should be 3n. The number of nodes in the input layer directly corresponds to the number of input features, ensuring that each feature can be correctly processed by the network. The activation function selected for the output layer should be based on the nature of the output value. Since the output is continuous, a linear activation function can be used to ensure that the linear relationship between the output value and the input features is preserved. If the output value needs to be constrained to a certain range, a sigmoid function can be used. Based on the network structure and activation function, an appropriate weight initialization method can be selected. For example, if the hidden layer uses the ReLU activation function, the He initialization method can be selected as it accounts for the nonlinear characteristics of ReLU and helps prevent the vanishing gradient problem. The learning rate determines the step size of the parameter update and is typically set to a small value, such as 0.001 or less. It determines the number of samples used for each parameter update and can be adjusted based on computing resources.

[0051] Set up the target network according to the main network structure.

[0052] Create an experience replay pool, set its maximum capacity based on the model's input and output, and define the structure for storing data.

[0053] The hidden layers include the first hidden layer and the second hidden layer. The first hidden layer has 128 nodes and introduces nonlinear features. The second hidden layer has 64 nodes and uses the ReLU activation function to maintain nonlinear characteristics.

[0054] Set up the target network according to the main network structure.

[0055] In this embodiment, the main network is responsible for multi-level processing of input data, ultimately outputting state values and action policies. To improve the stability of reinforcement learning, a target network is introduced. The target network's structure is highly consistent with the main network, but its parameters remain relatively fixed during the training process. The target network's parameters are not updated in real time with each round of training. Instead, the latest parameters are copied from the main network at regular intervals. This parameter update strategy is designed to mitigate the instability that can arise from overly frequent parameter updates. In reinforcement learning, if the target network's parameters change too rapidly, it may rely on an unstabilized policy output when estimating the value function, increasing uncertainty and volatility in the learning process. By adopting this strategy of slowly updating the target network, the policy output that the target network relies on during value function estimation is relatively stable. This stability not only helps to speed up the convergence of reinforcement learning but also significantly improves the overall learning effect. Experimental and practical experience have shown that this approach significantly reduces volatility during the learning process, enabling more stable and efficient model training.

[0056] Create an experience replay pool, set its maximum capacity based on the model's input and output, and define the structure for storing data.

[0057] In this embodiment, when setting the maximum capacity of the experience replay pool, full consideration should be given to the actual available memory resources. Ensure that the maximum capacity of the experience replay pool does not exceed the computing resource limits, as exceeding these limits may cause memory overflows, which in turn affect the model training process. On the other hand, the size of the experience replay pool should not be too small. If the experience pool capacity is too small, the experience data it stores may not be sufficient to fully reflect the various state and action combinations encountered by the intelligent agent during its interaction with the environment. This will greatly limit the model's ability to learn effective strategies from this data, potentially resulting in poor learning results or even failure to converge. To ensure effective model learning, the experience replay pool capacity is set to at least several thousand experience data points. For complex tasks that require extended training and adjustment, setting a larger experience replay pool capacity is particularly important. A larger capacity means that more experience data can be stored, thereby capturing a richer state space. This helps the model more comprehensively understand the environment and learn better strategies from diverse experiences.

[0058] The steps to optimize the DQN temperature control model include: Initialize the population and randomly generate elk individuals, each of which represents a network model hyperparameter.

[0059] In this embodiment, a hybrid strategy can also be used during initialization. Hyperparameter values for some elk individuals are generated completely randomly, while others are fine-tuned based on prior knowledge or experience. This ensures population diversity while also leveraging existing knowledge to a certain extent.

[0060] The mean square error is used as the fitness function of the model.

[0061] In this embodiment, the expression formula of the mean square error value is: in, is the mean square error, It is The true value of the sample, This is the model The predicted value of the sample.

[0062] The fitness values of individuals in the population are calculated and sorted. The number of male elk is set according to their behavior during the estrus season. Individuals with higher fitness are selected as male elk, and the remaining individuals are selected as female elk. A roulette wheel selection method is used to assign each male elk a number of female elk proportional to its fitness, and multiple new populations are formed.

[0063] In this embodiment, when sorting individuals, they are sorted from large to small according to their fitness values, and then the number of bull elk is set according to the rutting season, and they are selected from large to small.

[0064] The formula for expressing the number of bull elk in the rutting season behavior setting is: in, is the number of bull elk, is the proportion of bull elk in the population, is the total number of elk.

[0065] Multiple new populations obtain offspring individuals based on their calving season behavior.

[0066] The formula for expressing calving season behavior is: in, is the position of the new individual, It is a male elk individual. It is a female elk individual, is a parameter that controls the mating weight, is the factor of variation, It is a random factor.

[0067] Merge all new populations and offspring individuals to obtain population one.

[0068] The fitness value of population one is calculated. If the output fitness value is higher than the preset fitness threshold, the optimal hyperparameter is output. Otherwise, the iteration is continued until the fitness value is higher than the preset fitness threshold or the maximum number of iterations is reached to obtain the optimal hyperparameter.

[0069] The control module includes a maintenance reminder unit and a statistical analysis unit. The maintenance reminder unit is used to obtain the operating hours of the heat source module and the air supply module and generate maintenance reminders based on the operating hours. The statistical analysis unit is used to obtain the energy consumption of the heat source module and the air supply module and generate energy consumption reports.

[0070] In this embodiment, the maintenance reminder unit can collect operating hours data from connected heat source and air supply modules in real time or periodically via sensors or connected modules. Based on a preset maintenance cycle, such as every 500 hours of operation for the heat source module and every 300 hours for the air supply module, the maintenance reminder unit compares operating hours with the preset cycle. When the operating hours of any module reach or exceed the preset maintenance cycle, the maintenance reminder unit generates a maintenance reminder message, including the module requiring maintenance and the estimated maintenance time. This maintenance reminder message can be communicated to relevant personnel through various means, such as displaying it on the control unit's display screen, sending text messages or emails to designated maintenance personnel's mobile phones or email addresses, and so on. The maintenance reminder unit can also include a maintenance record management function, recording detailed information about each maintenance session, including the time, personnel, and details of the maintenance. The statistical analysis unit can summarize and calculate the collected energy consumption data by time period (daily, weekly, or monthly), to determine the total and average energy consumption of each module. Based on the energy consumption calculation and analysis results, the statistical analysis unit automatically generates an energy consumption report. The report should include the total energy consumption of each module, average energy consumption, energy consumption trend chart, etc. The report can also be customized according to actual needs, such as adding energy consumption comparison charts, energy consumption cost analysis, etc., to meet the needs of different users.

[0071] In this example, a temperature control model is constructed based on a DQN temperature control model and optimized using a multi-objective elk herd optimization algorithm. Processed environmental data is input into the temperature control model to generate a control variable. This control variable is used to transmit instructions to the heat source module and air supply module, achieving precise temperature control of different areas within the greenhouse. This addresses the temperature requirements of different areas within the greenhouse, improves efficiency, and reduces energy waste.

[0072] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0073] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A greenhouse high-efficiency warm air supply device, characterized in that: include: A heat source module, used for heating air to generate warm air; An air supply module, used for delivering the warm air to different areas in the greenhouse through a plurality of air supply ports; The data acquisition module is used to obtain environmental data of different areas in the greenhouse and pre-process the environmental data to obtain processed environmental data; the environmental data includes: the area and actual temperature of different areas; A temperature control module is configured to construct a DQN-based temperature control model and optimize the DQN temperature control model using a multi-objective elk herd optimization algorithm to obtain a temperature control model; the processed environmental data is input into the temperature control model to obtain a control variable; the control variable includes the minimum power of the heat source module to achieve the preset temperature of different areas and the size of the corresponding air outlet; The control module is used to generate a control instruction according to the control quantity and transmit the control instruction to the heat source module and the air supply module.

2. A greenhouse high-efficiency warm air supply device according to claim 1, characterized in that: The air supply module includes a filter unit and a regulating unit; the filter unit is used to filter dust and impurities in the air; and the regulating unit is used to adjust the wind speed of the air supply port.

3. A greenhouse high-efficiency warm air supply device according to claim 1, characterized in that: The data acquisition module preprocesses the environmental data in the following manner: Remove outliers in the environmental data and fill missing values in the environmental data using the mean method; The environmental data is smoothed using a moving average method.

4. A greenhouse high-efficiency warm air supply device according to claim 1, characterized in that: The DQN temperature control model construction process includes: Build the main network structure including input layer, hidden layer and output layer, and initialize the network weights and hyperparameters; Setting a target network according to the main network structure; Create an experience replay pool, set the maximum capacity of the experience replay pool based on the input and output of the model, and define the structure for storing data.

5. A greenhouse high-efficiency warm air supply device according to claim 4, characterized in that: The hidden layer includes: a first hidden layer and a second hidden layer; the first hidden layer is provided with 128 nodes and introduces nonlinear features; the second hidden layer is provided with 64 nodes and uses a ReLU activation function to maintain nonlinear characteristics.

6. A greenhouse high-efficiency warm air supply device according to claim 1, characterized in that: The multi-objective elk herd optimization algorithm includes: Initialize the population and randomly generate elk individuals, each of which represents a network model hyperparameter; Use the mean square error value as the fitness function of the model; Calculating the fitness values of the individuals in the population and ranking them, setting the number of male elk according to their behavior during the estrus season, selecting B individuals with the highest fitness as male elk, and the remaining individuals as female elk, using a roulette wheel selection method to assign each male elk a number of female elk proportional to its fitness, and forming multiple new populations; The plurality of new populations respectively obtain offspring individuals according to their calving season behaviors; Merging the plurality of new populations and the offspring individuals to obtain population one; The fitness value of each individual in population 1 is calculated. If the output fitness value is higher than the preset fitness threshold, the optimal hyperparameter is output. Otherwise, the iteration is continued until the fitness value is higher than the preset fitness threshold or the maximum number of iterations is reached to obtain the optimal hyperparameter.

7. A greenhouse high-efficiency warm air supply device according to claim 6, characterized in that: The calculation formula for the number of bull elk is: in, is the number of bull elk, is the proportion of bull elk in the population, is the total number of elk.

8. A greenhouse high-efficiency warm air supply device according to claim 6, characterized in that: The calving season behavior expression formula is: in, is the position of the new individual, This is the individual position of the bull elk. is the individual position of the female elk, is a parameter that controls the mating weight, is the factor of variation, It is a random factor.

9. A greenhouse high-efficiency warm air supply device according to claim 1, characterized in that: The control module includes: a processor, a digital-to-analog converter, and a communication device; the processor is used to generate control instructions from the control quantity; the digital-to-analog converter is used to convert the control instructions into analog signals; and the communication device is used to transmit the analog signals to the heat source module and the air supply module.

10. A greenhouse high-efficiency warm air supply device according to claim 9, characterized in that: The control module also includes a maintenance reminder unit and a statistical analysis unit; the maintenance reminder unit is used to obtain the working hours of the heat source module and the air supply module, and generate a maintenance reminder based on the working hours; the statistical analysis unit is used to obtain the energy consumption of the heat source module and the air supply module, and generate an energy consumption report.

Images