Intelligent premixing control method and device based on machine learning, equipment and medium

Abstract

The invention provides an intelligent premixing control method and device based on machine learning, equipment and a medium, and relates to the technical field of machine learning, and the method comprises the steps: constructing a machine learning model obtained through training based on historical environment control data, plant growth data and environment disturbance response data, and embedding the machine learning model into a control system, intelligent preliminary adjustment and closed-loop feedback control of the environment of the premixing box are realized; after the environmental parameters are stable, stable gas is conveyed to the plant box body through the air duct, and dynamic compensation is conducted on environmental changes in the conveying path based on the model; a sensor is deployed in the plant box body to monitor disturbance data and input the disturbance data to the disturbance adaptation sub-module, the disturbance trend is predicted, and feed-forward compensation data is output; when the environment in the plant box body is abnormal, the system automatically exhausts air and supplements stable gas, meanwhile, the control strategy is dynamically adjusted, and a whole-process self-adaptive regulation and control mechanism is constructed. According to the invention, self-adaptive pre-regulation and control of the breeding environment can be realized.

Description

Technical Field

[0001] The present application relates to the technical field of machine learning, and in particular to an intelligent premixing control method, device, equipment and medium based on machine learning. Background Art

[0002] In the process of breeding new varieties, the traditional premixing control method mainly sets target values ​​for the temperature, humidity and carbon dioxide concentration in the environmental parameters, directly imports the set target parameters into the breeding plant box, and relies on the environmental sensors installed in the box to collect and feedback the actual environmental data in real time. The control system continuously executes the closed-loop adjustment strategy based on the deviation between the actual measured value of the box environment and the set target value to drive the control component to correct the premixed gas, thereby forming a traditional control closed loop that relies on the combination of static set parameters and passive feedback regulation. In the face of frequent environmental disturbances and high sensitivity of the breeding process, this method has significant deficiencies in response speed and stability, making it difficult to achieve high-precision adaptive control of complex and changeable breeding environments.

[0003] Because plants are highly sensitive to changes in environmental conditions, when the temperature, humidity or carbon dioxide concentration in the box fluctuates, even if the amplitude is small, it may have a stimulating effect on the plant's growth process, resulting in inconsistent stress responses of the plant during different growth periods, thereby causing physiological and metabolic rhythm disorders. If such environmental disturbances are not promptly and accurately regulated, it will not only cause abnormalities in the growth rate and direction of the plant, but will also directly affect the balanced development of its morphological structure, manifesting as a decline in plant appearance, obvious differences in plant height, uneven leaf distribution, and different fruit ripening times. These problems seriously restrict the stability of breeding work and the consistency of the appearance quality and commercialization potential of new varieties. Therefore, a method is needed to achieve adaptive pre-regulation of the breeding environment. Summary of the Invention

[0004] The present application provides an intelligent premixing control method, device, equipment and medium based on machine learning, which can achieve adaptive pre-regulation of the breeding environment.

[0005] In a first aspect of the present application, a method for intelligent premixing control based on machine learning is provided, the method comprising: Build machine learning models based on historical environmental control data, plant growth data, and environmental disturbance response data; The machine learning model is embedded in a control system, and the control system controls the regulating component to perform preliminary regulation on the environment in the premix box according to the target environmental parameters; Acquire first real-time environmental data through sensors deployed in the premixing tank and feed it back to the control system to form a closed-loop regulation; When the environmental parameters in the premix box are stable within the target range, the air duct is opened to deliver the stable environment corresponding to the environmental parameters to the interior of the plant box at a constant flow rate, and at the same time, the changes in the environmental parameters in the delivery path are compensated and corrected in real time based on the machine learning model; Inputting the disturbance data monitored in real time by the sensor in the plant box into the disturbance adaptation submodule in the machine learning model to predict the disturbance trend and output environmental disturbance compensation data; When the second real-time environmental parameter inside the plant box exceeds a set threshold, the control system drives the exhaust mechanism to exhaust the air from the plant box, while controlling the premixing box to output a new round of stable environmental gas, and adaptively adjusts the control strategy of the premixing box based on the environmental disturbance compensation data.

[0006] On the basis of the above technical solutions, preferably, the machine learning model is constructed based on historical environmental control data, plant growth data and environmental disturbance response data, specifically including: Collecting historical environmental control data of plants over multiple growth cycles, the historical environmental control data including environmental parameter set values, environmental parameter actual values, environmental parameter adjustment amplitudes, environmental parameter adjustment response times, environmental parameter adjustment frequencies, and environmental parameter stability indicators; Collecting plant growth data within a time period corresponding to the historical environmental control data, the plant growth data including plant species, growth stage, plant height, leaf area index, chlorophyll content, dry matter mass, and growth rate; Collecting environmental disturbance response data recorded during plant growth, the environmental disturbance response data including disturbance type, disturbance amplitude, disturbance duration, disturbance recovery time, and disturbance regulation feedback characteristics; Time-aligning the historical environmental control data, plant growth data, and environmental disturbance response data to construct a unified training sample data set; A deep neural network is used to train the training sample data set to form a machine learning model for outputting target environmental parameters and environmental disturbance compensation data, wherein a multi-layer feedforward neural network is constructed including an input layer, multiple hidden layers and an output layer. The input layer receives plant species, growth stages, historical sequences of environmental parameters and phenotypic indicator characteristics, and the output layer generates the target environmental parameters and disturbance compensation strategy. The training process of the machine learning model adopts a supervised learning mechanism, and the objective function is set to minimize the mean square error between the predicted environmental parameters and the actual adjustment results. At the same time, a regularization term is introduced to constrain the complexity of controller parameters to prevent model overfitting. A cross-validation method is used to evaluate the generalization ability of the model during the training process, and the model structure and control parameters are optimized through hyperparameter grid search.

[0007] Based on the above technical solution, preferably, the machine learning model is embedded in a control system, and the control system controls the adjustment component to perform preliminary adjustment on the environment in the premix box according to the target environmental parameters, specifically including: Deploy the machine learning model in the core control unit of the control system, wherein the core control unit includes a data receiving module, a model parsing module, a control signal generating module, and an execution feedback module; The data receiving module is used to receive the target environmental parameters output by the machine learning model, and the model parsing module generates corresponding adjustment instructions according to the target environmental parameters, wherein the adjustment instructions include temperature adjustment instructions, humidity adjustment instructions, and carbon dioxide concentration adjustment instructions; The control signal generating module outputs control signals to the temperature adjustment component, the humidity adjustment component and the carbon dioxide concentration adjustment component according to the adjustment instruction, so as to drive the adjustment components to perform adjustment operations; The execution feedback module collects current environmental parameters according to the environmental parameter sensor deployed inside the premix box, and the current environmental parameters include current temperature value, current humidity value and current carbon dioxide concentration value; The error between the current environmental parameter and the target environmental parameter is calculated, and the control signal is corrected in real time based on the error to form a closed-loop regulation control logic, so that the environmental parameter in the premix box is stabilized to the target environmental parameter.

[0008] On the basis of the above technical solution, preferably, the first real-time environmental data is obtained by a sensor deployed in the premixing tank and fed back to the control system to form a closed-loop regulation, specifically including: Deploying environmental parameter sensor modules at multiple spatial locations inside the premix box, the environmental parameter sensor modules including a temperature sensor, a humidity sensor, and a carbon dioxide concentration sensor; The environmental parameter sensor module collects first real-time environmental parameters at a preset sampling period, wherein the first real-time environmental parameters include a first real-time temperature value, a first real-time humidity value, and a first real-time carbon dioxide concentration value; Packing the first real-time environmental parameters into an environmental parameter data frame, and sending the frame to the control system via a data transmission module; The control system calculates the error between the environmental parameter data and the target environmental parameter to obtain a temperature error, a humidity error, and a carbon dioxide concentration error, and generates a control signal based on a proportional-integral-differential control algorithm to drive the temperature adjustment component, the humidity adjustment component, and the carbon dioxide concentration adjustment component to perform adjustment operations, thereby forming a closed-loop adjustment path with the environmental parameter sensor module as the perception end, the control system as the decision-making end, and the adjustment component as the execution end.

[0009] On the basis of the above technical solution, preferably, when the environmental parameters in the premix box are stable within the target range, the air duct is opened to deliver the stable environment corresponding to the environmental parameters to the interior of the plant box at a constant flow rate, and at the same time, the changes in the environmental parameters in the delivery path are compensated and corrected in real time based on the machine learning model, specifically including: When the control system determines that the first real-time temperature value, the first real-time humidity value, and the first real-time carbon dioxide concentration value are all within the error range of the target temperature parameter, the target humidity parameter, and the target carbon dioxide concentration parameter, the control system outputs an air duct opening instruction to drive the air duct gas delivery mechanism to start, the air duct gas delivery mechanism including an air supply fan, an air flow control valve, and a pressure stabilization module; A conveying path environmental parameter monitoring module is arranged along the inside of the air duct. The conveying path environmental parameter monitoring module includes a conveying path temperature sensor, a conveying path humidity sensor, and a conveying path carbon dioxide concentration sensor, which are used to obtain environmental change data within the conveying path; The environmental change data is input into the environmental transmission prediction sub-model in the machine learning model, and the environmental transmission prediction sub-model generates an environmental compensation instruction. The control system dynamically adjusts the output amplitude of the temperature adjustment component, the humidity adjustment component and the carbon dioxide concentration adjustment component according to the environmental compensation instruction to achieve real-time compensation correction of the environmental parameters of the terminal of the transmission path.

[0010] On the basis of the above technical solution, preferably, the step of inputting the disturbance data monitored in real time by the sensor in the plant box into the disturbance adaptation submodule in the machine learning model to perform disturbance trend prediction and output environmental disturbance compensation data specifically includes: Environmental disturbance monitoring sensor modules are deployed at multiple locations inside the plant box. The environmental disturbance monitoring sensor modules include a plant box temperature sensor, a plant box humidity sensor, and a plant box carbon dioxide concentration sensor, which are used to collect real-time temperature disturbance values, real-time humidity disturbance values, and real-time carbon dioxide concentration disturbance values ​​inside the plant box to form a disturbance data sequence. The disturbance data sequence is packaged into a disturbance data input frame and transmitted to the disturbance adaptation submodule of the machine learning model, wherein the disturbance adaptation submodule performs disturbance trend prediction based on the neural network structure of time series modeling and outputs the disturbance trend prediction result; Environmental disturbance compensation data is generated according to the disturbance trend prediction result. The environmental disturbance compensation data includes a pre-adjustment amplitude, an adjustment rate, and a duration for controlling a temperature adjustment component, a humidity adjustment component, and a carbon dioxide concentration adjustment component, so as to achieve feedforward compensation correction of the disturbance.

[0011] On the basis of the above technical solution, preferably, when the second real-time environmental parameter in the plant box exceeds a set threshold, the control system drives the exhaust mechanism to exhaust the air in the plant box, and at the same time controls the premixing box to output a new round of stable environmental gas, and adaptively adjusts the control strategy of the premixing box based on the environmental disturbance compensation data, specifically including: The environmental parameter judgment module receives a second real-time environmental parameter collected by the plant box temperature sensor, the plant box humidity sensor, and the plant box carbon dioxide concentration sensor, wherein the second real-time environmental parameter includes a second real-time temperature value, a second real-time humidity value, and a second real-time carbon dioxide concentration value, and compares the real-time temperature value, the real-time humidity value, and the real-time carbon dioxide concentration value with an error threshold value of a target temperature parameter, a target humidity parameter, and a target carbon dioxide concentration parameter; When any of the second real-time environmental parameters exceeds the corresponding error threshold range, the control system outputs an exhaust instruction to drive the exhaust fan to exhaust the non-standard environmental gas in the plant box through the exhaust channel; At the same time, the control system outputs an air supply instruction to drive the premixing box to generate a new round of stable ambient gas based on the current target temperature parameter, target humidity parameter and target carbon dioxide concentration parameter, and transport it to the plant box through the air duct; The control system further inputs the environmental disturbance compensation data into the strategy adjustment module to adaptively optimize the parameters of the current premixing box, such as the adjustment period, adjustment amplitude, sensor sampling density and control signal gain, to form a control strategy with enhanced dynamic response.

[0012] In a second aspect of the present application, a machine learning-based intelligent premixing control device is provided. The device is configured to execute any one of the machine learning-based intelligent premixing control methods described above. The device includes an acquisition module, a processing module, and an output module, wherein: The acquisition module is used to build a machine learning model based on historical environmental control data, plant growth data and environmental disturbance response data; The processing module is used to embed the machine learning model into a control system, and the control system controls the adjustment component to perform preliminary adjustment on the environment in the premix box according to the target environmental parameters; The processing module is configured to obtain first real-time environmental data through sensors deployed in the premixing tank and feed the data back to the control system to form a closed-loop regulation; The processing module is configured to, when the environmental parameters in the premix box are stable within a target range, open the air duct to deliver the stable environment corresponding to the environmental parameters to the interior of the plant box at a constant flow rate, and simultaneously perform real-time compensation and correction for changes in the environmental parameters in the delivery path based on the machine learning model; The processing module is used to input the disturbance data monitored in real time by the sensor in the plant box into the disturbance adaptation submodule in the machine learning model to perform disturbance trend prediction and output environmental disturbance compensation data; The output module is used to drive the exhaust mechanism to exhaust the air from the plant box when the second real-time environmental parameter in the plant box exceeds a set threshold, and at the same time control the premixing box to output a new round of stable environmental gas, and adaptively adjust the control strategy of the premixing box based on the environmental disturbance compensation data.

[0013] In the third aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device performs any of the methods described above.

[0014] In a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores instructions. When the instructions are executed, any one of the methods described above is executed.

[0015] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. This application constructs a machine learning model trained based on historical environmental control data, plant growth data and environmental disturbance response data, and embeds it into the control system, thereby realizing dynamic adjustment of the premix box environment, real-time compensation for changes in the conveying path environment, and predictive response to disturbance trends in the plant box; when the plant box environment is abnormal, the control system can adaptively adjust the regulation strategy in combination with the disturbance compensation data, forming a closed-loop mechanism of perception-prediction-compensation-regulation, so that the system has feedforward regulation capability and adaptive response capability, effectively solving the problems of delayed response to environmental disturbances and insufficient regulation accuracy in the traditional mode, and ultimately realizing adaptive pre-regulation of the breeding environment.

[0016] 2. By collecting multi-period historical environmental control data, plant growth data, and environmental disturbance response data, and training them with deep neural networks, a machine learning model with multi-factor joint reasoning capabilities was constructed, which achieved accurate prediction of target environmental parameters and disturbance compensation strategies, effectively improving the personalized adaptability and forward-looking response capabilities of environmental regulation, and providing a highly reliable data-driven foundation for subsequent environmental control.

[0017] 3. By embedding the machine learning model into the core control unit of the control system, a closed-loop control path is established from receiving target environmental parameters, command parsing, control signal generation to executing feedback correction, ensuring that the environmental regulation in the premix box has full-process automated control capabilities, significantly improving the regulation response efficiency, control accuracy and environmental stability maintenance capabilities.

[0018] 4. By deploying multi-point environmental parameter sensor modules inside the premix box and implementing feedback regulation based on the proportional-integral-differential control algorithm, a local closed-loop control chain of perception-decision-execution is constructed, which can identify small disturbances at high frequency and accurately adjust the control output, greatly improving the environmental stability of the premix box and establishing a high-consistency gas foundation for subsequent gas transportation.

[0019] 5. By determining that the environmental parameters in the premix box are stable, the air duct is opened, and the environmental parameter monitoring module is deployed in the conveying path. The environmental transmission prediction sub-model is introduced for real-time compensation and correction to achieve consistent control of the entire process from the source to the plant box. This effectively avoids problems such as temperature attenuation, humidity loss, and gas concentration offset during the conveying process, ensuring that the actual receiving environment of the plant highly matches the set value.

[0020] 6. By placing environmental disturbance monitoring sensor modules in the plant box and using the disturbance adaptation submodule in the neural network model to predict the disturbance trend, the system can realize dynamic perception and trend modeling before the disturbance occurs, and generate environmental disturbance compensation data based on the prediction results, thereby realizing feedforward regulation and enhancing the timeliness and initiative of environmental disturbance response.

[0021] 7. When abnormal fluctuations occur in the environmental parameters inside the plant box, the non-standard environmental gas is discharged through the exhaust mechanism, and the premix box is controlled to replenish a new round of stable environmental gas. The control strategy parameters are dynamically optimized based on the disturbance compensation data to form a disturbance closed-loop response and adaptive strategy reconstruction mechanism, which significantly improves the system's steady-state maintenance ability and anti-disturbance recovery ability, and effectively ensures that the plant growth environment is continuously in an optimized state. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of an intelligent premixing control method based on machine learning disclosed in an embodiment of the present application; Figure 2 This is a module schematic diagram of an intelligent premixing control device based on machine learning disclosed in an embodiment of the present application; Figure 3 This is a structural diagram of an electronic device disclosed in an embodiment of the present application.

[0023] Explanation of the reference numerals: 201, acquisition module; 202, processing module; 203, output module; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0025] In the description of the embodiments of this application, words such as "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "for example" or "for instance" in the embodiments of this application should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "for example" or "for instance" is intended to present the relevant concepts in a concrete manner.

[0026] In the description of the embodiments of the present application, the term "multiple" means two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0027] In the breeding process of new varieties, the traditional premixing control method relies on a closed-loop mechanism of static settings and passive feedback regulation, making it difficult to achieve rapid response and precise control in complex and multi-disturbance environments. In particular, in the context of plants being highly sensitive to temperature, humidity and carbon dioxide concentration, any slight fluctuation may cause plant physiological rhythm disorders and asynchrony of stress response, resulting in significant differences in plant morphology, developmental rhythm and fruit ripening, thereby affecting the phenotypic consistency and commercialization potential of breeding work. Therefore, there is an urgent need for an adaptive environmental control method with dynamic prediction and feedforward control capabilities to achieve precise and stable control in highly sensitive breeding environments.

[0028] This embodiment discloses an intelligent premixing control method based on machine learning, referring to Figure 1 , including the following steps S110-S160: S110, building a machine learning model based on historical environmental control data, plant growth data, and environmental disturbance response data.

[0029] The embodiments of the present application disclose a machine learning-based intelligent premixing control method applied to a server. The server includes, but is not limited to, electronic devices such as mobile phones, tablet computers, wearable devices, and personal computers (PCs), and may also be a background server running the machine learning-based intelligent premixing control method. The server may be implemented as a standalone server or a server cluster consisting of multiple servers.

[0030] In one possible implementation, a machine learning model is constructed based on historical environmental control data, plant growth data, and environmental disturbance response data, specifically including: collecting historical environmental control data of plants in multiple growth cycles, the historical environmental control data including environmental parameter set values, environmental parameter actual values, environmental parameter adjustment amplitudes, environmental parameter adjustment response times, environmental parameter adjustment frequencies, and environmental parameter stability indices; collecting plant growth data within a time period corresponding to the historical environmental control data, the plant growth data including plant species, growth stages, plant heights, leaf area indexes, chlorophyll content, dry matter mass, and growth rates; collecting environmental disturbance response data recorded during plant growth, the environmental disturbance response data including disturbance type, disturbance amplitude, disturbance duration, disturbance recovery time, and disturbance adjustment feedback characteristics; and integrating historical environmental control data, plant growth data, and environmental disturbance response data into a machine learning model. The plant growth data and the environmental disturbance response data are time-aligned to construct a unified training sample data set; the training sample data set is trained using a deep neural network to form a machine learning model for outputting target environmental parameters and environmental disturbance compensation data, wherein a multi-layer feedforward neural network is constructed including an input layer, multiple hidden layers and an output layer. The input layer receives plant species, growth stage, historical sequence of environmental parameters and phenotypic indicator characteristics, and the output layer generates target environmental parameters and disturbance compensation strategies. The training process of the machine learning model adopts a supervised learning mechanism, and the objective function is set to minimize the mean square error between the predicted environmental parameters and the actual adjustment results. At the same time, a regularization term is introduced to constrain the complexity of the controller parameters to prevent model overfitting. The cross-validation method is used to evaluate the generalization ability of the model during training, and the model structure and control parameters are optimized through hyperparameter grid search.

[0031] Specifically, first, historical environmental control data for plants over multiple growth cycles is collected. This data is obtained by recording the entire process of temperature, humidity, and carbon dioxide concentration parameters in the control system over a long period of time. This data includes the set value, actual value, adjustment amplitude, response time, and frequency of adjustment, as well as stability indicators at each moment. This historical environmental control data must be exported from the environmental control log system via a standardized data interface and uniformly timestamped for subsequent alignment with plant growth data.

[0032] Next, plant growth data for the time period corresponding to the historical environmental control data was collected. This data, acquired using high-frequency growth monitoring equipment, includes information on plant species, growth stage, plant height, leaf area index, chlorophyll content, dry matter mass, and growth rate. All plant growth data was recorded in segments using timestamps and annotated with structured data using a phenotyping platform. To ensure data consistency, plant phenotypic changes at different growth stages were strictly time-matched with the corresponding historical environmental control data to avoid model training errors caused by offsets in recording timing.

[0033] Next, we collected environmental disturbance response data recorded during plant growth. This data primarily comes from an automatic identification system for environmental disturbance events, including disturbance type, disturbance amplitude, disturbance duration, disturbance recovery time, and disturbance adjustment feedback characteristics. The system automatically classifies and identifies disturbance types based on the slope and amplitude of environmental parameter changes. The disturbance amplitude and duration are used to calculate the maximum deviation and duration of the disturbance event, respectively. The disturbance recovery time is the time it takes the control system to readjust the environmental parameters to a stable range. The disturbance adjustment feedback characteristics record the control strategy response path and the trajectory of control signal changes.

[0034] Subsequently, historical environmental control data, plant growth data, and environmental disturbance response data were aligned along the time dimension to construct a training sample dataset with consistent structure and synchronized timestamps. Each record in the training sample dataset corresponds to a time segment and contains the target environmental state, plant growth state, and its corresponding disturbance response. These records are organized into multidimensional input samples using a unified data structure framework. All data features are normalized and missing values ​​are processed to ensure numerical stability during training.

[0035] Finally, a deep neural network was used to train the training sample dataset, constructing a multi-layer feedforward neural network as the core architecture of the machine learning model. The neural network consists of an input layer, multiple hidden layers, and an output layer. The input layer receives features such as plant species, growth stage, historical sequences of environmental parameters, and plant phenotypic indicators. The output layer generates target environmental parameters and a disturbance compensation strategy. The model training process uses a supervised learning mechanism, constructing an objective function to minimize the mean squared error between the predicted environmental parameters and the historical actual adjustment results. The specific expression is as follows:

[0036] in, represents the total number of training samples, Indicates the The target environment parameter output obtained by the neural network prediction of the samples is Indicates the historical actual adjustment result corresponding to the sample. is the model's prediction loss. This loss function measures the deviation between the model output and the actual historical control results. By averaging the squared errors of all samples, it effectively amplifies large errors and smoothes small errors, enabling the model to more accurately learn the numerical mapping relationship of historical control behaviors.

[0037] At the same time, the L2 regularization term is introduced into the model loss function to suppress the complexity of the controller parameters to prevent the model from overfitting. The specific expression is as follows:

[0038] in, Indicates the neural network connection weights, is the total weight, is the regularization coefficient, A penalty term that represents the complexity of the controller. By adding this regularization term to the total loss function, it can effectively suppress excessive weight values ​​and prevent the model from overfitting to the training data, thereby improving the model's generalization ability on unknown data and ensuring the model's adaptability under dynamic environmental disturbances.

[0039] To improve the generalization ability of the model, a cross-validation mechanism was introduced into the training process, and a hyperparameter grid search strategy was used to systematically optimize the neural network structure parameters such as the number of hidden layers, the number of nodes, the type of activation function, and the learning rate. This resulted in a machine learning model with stable convergence and high generalization ability. The final model output prediction formula is expressed as follows:

[0040] in, Represents the input feature vector, including plant species, growth stages, historical sequences of environmental parameters, and plant phenotypic indicators, and Respectively represent The weight matrix and bias vector of the layer, Indicates the The activation function of the layer, Indicates the number of network layers, The target environmental parameters and environmental disturbance compensation strategies output by the model are described in this formula. This formula fully describes the nonlinear mapping process from input to output of a deep neural network. Through multi-layer nonlinear combinations, the model can capture the complex interactions between the environment and the plant's physiological state, enabling decision support for environmental control in complex disturbance scenarios.

[0041] S120, embedding the machine learning model into the control system, and the control system controls the adjustment component to perform preliminary adjustment on the environment in the premix box according to the target environmental parameters.

[0042] In one possible embodiment, a machine learning model is embedded in a control system, and the control system controls the adjustment component to perform preliminary adjustment on the environment inside the premix box according to the target environmental parameters, specifically including: deploying the machine learning model in the core control unit of the control system, the core control unit including a data receiving module, a model parsing module, a control signal generating module and an execution feedback module; the data receiving module is used to receive the target environmental parameters output by the machine learning model, and the model parsing module generates corresponding adjustment instructions according to the target environmental parameters, and the adjustment instructions include temperature adjustment instructions, humidity adjustment instructions and carbon dioxide concentration adjustment instructions; the control signal generating module outputs control signals to the temperature adjustment component, humidity adjustment component and carbon dioxide concentration adjustment component according to the adjustment instructions to drive the adjustment component to perform the adjustment operation; the execution feedback module collects current environmental parameters according to the environmental parameter sensors deployed inside the premix box, and the current environmental parameters include current temperature value, current humidity value and current carbon dioxide concentration value; calculates the error between the current environmental parameters and the target environmental parameters, and corrects the control signal in real time based on the error to form a closed-loop adjustment control logic to stabilize the environmental parameters in the premix box to the target environmental parameters.

[0043] Specifically, the machine learning model is embedded within the control system's core control unit. The core control unit serves as the central module for environmental control execution, integrating a data reception module, a model parsing module, a control signal generation module, and an execution feedback module. The machine learning model, serving as the target environmental parameter calculation unit, resides in the computing resource area where the model parsing module resides. The control system triggers the model inference process through an embedded logic call mechanism to obtain dynamic environmental control decision outputs.

[0044] Next, the data receiving module is responsible for receiving the target environmental parameters output by the machine learning model. These parameters are generated by the model based on a combination of plant species, growth stage, historical environmental parameter sequences, and plant phenotypic indicators. Specifically, they include target temperature, target humidity, and target carbon dioxide concentration. The data receiving module parses and verifies the target environmental parameters to ensure their structure is legal and their values ​​fall within the set control range. These parameters serve as the basic input for subsequent instruction generation.

[0045] Next, the model parsing module constructs standardized adjustment instructions based on the received target environmental parameters. The adjustment instructions are clearly divided into three categories: temperature adjustment instructions, humidity adjustment instructions, and carbon dioxide concentration adjustment instructions. Each type of adjustment instruction encapsulates the control target, execution cycle, and response level of the corresponding environmental adjustment component, and is used to drive the control signal generation module to generate specific execution control quantities.

[0046] Then, the control signal generation module converts the control signal that can directly drive the physical device according to the adjustment instruction. The control signal is sent to the temperature adjustment component, humidity adjustment component and carbon dioxide concentration adjustment component in the premixing box in the form of analog voltage, PWM pulse width modulation or digital communication protocol, driving the refrigeration module, humidification module, gas injection module and other equipment to perform precise environmental adjustment operations according to the target parameters.

[0047] The execution feedback module also receives real-time data from environmental parameter sensors deployed inside the premix tank, including temperature, humidity, and carbon dioxide concentration. To ensure accurate environmental status, these sensors utilize a multi-point layout and high-precision samplers. Feedback data is transmitted in real time to the execution feedback module for error analysis.

[0048] Finally, the execution feedback module calculates the errors between the current environmental parameters and the target environmental parameters item by item, generates temperature error, humidity error and carbon dioxide concentration error, and corrects the original control signal in real time based on the proportional integral differential control algorithm; the corrected control signal is re-output to the adjustment component, forming a continuously iterative closed-loop adjustment control logic until the environmental parameters in the premix box are stably maintained within the error tolerance range of the target environmental parameters, thereby achieving accurate preliminary adjustment of the environment.

[0049] S130, obtaining first real-time environmental data through sensors deployed in the premixing tank and feeding it back to the control system to form a closed-loop regulation.

[0050] In one possible embodiment, first real-time environmental data is obtained by sensors deployed in the premix box and fed back to the control system to form a closed-loop regulation, specifically including: deploying environmental parameter sensor modules at multiple spatial locations inside the premix box, the environmental parameter sensor modules including a temperature sensor, a humidity sensor, and a carbon dioxide concentration sensor; the environmental parameter sensor modules collect first real-time environmental parameters at a preset sampling period, the first real-time environmental parameters including a first real-time temperature value, a first real-time humidity value, and a first real-time carbon dioxide concentration value; packaging the first real-time environmental parameters into environmental parameter data frames, and sending them to the control system through a data transmission module; the control system calculates the errors between the environmental parameter data and the target environmental parameters to obtain temperature errors, humidity errors, and carbon dioxide concentration errors, and generates control signals based on a proportional-integral-differential control algorithm to drive the temperature regulation component, the humidity regulation component, and the carbon dioxide concentration regulation component to perform regulation operations, thereby forming a closed-loop regulation path with the environmental parameter sensor module as the perception end, the control system as the decision-making end, and the regulation component as the execution end.

[0051] Specifically, environmental parameter sensor modules, consisting of temperature, humidity, and CO2 concentration sensors, are deployed at multiple representative locations within the premixing tank. These locations are optimized based on the premixing tank's internal airflow path, mixing uniformity, and heat exchange efficiency to ensure representativeness, spatial coverage, and responsiveness of the collected environmental parameters to support high-precision control decisions.

[0052] Next, the environmental parameter sensor module periodically collects first real-time environmental parameters at a sampling interval set by the control system. These first real-time environmental parameters include a first real-time temperature value, a first real-time humidity value, and a first real-time carbon dioxide concentration value. Each first real-time environmental parameter is timestamped to form a continuous time series for subsequent error calculation and trend assessment. All first real-time environmental parameters are immediately submitted to the data transmission module after collection.

[0053] Next, the data transmission module packages the first real-time environmental parameters output by each sensor module into a structured environmental parameter data frame. This frame contains fields such as sensor identification, environmental parameter type, parameter value, and acquisition timestamp. This frame is sent to the control system's data receiving channel using a standard communication protocol, ensuring complete, accurate, and high-speed data frame transmission even in complex communication environments.

[0054] The control system then parses the received environmental parameter data frame and extracts the first real-time temperature value, the first real-time humidity value, and the first real-time CO2 concentration value. These values ​​are then compared with the target temperature parameter, target humidity parameter, and target CO2 concentration parameter output by the machine learning model. This error calculation uses a difference model to calculate the temperature error, humidity error, and CO2 concentration error, which serve as the basic input for generating the control variable.

[0055] Finally, based on the proportional-integral-differential control algorithm, the control system uses the calculated temperature error, humidity error, and carbon dioxide concentration error to generate control signals. These signals are output in real time by the control signal generation module to the temperature, humidity, and carbon dioxide concentration adjustment components, driving each adjustment component to correct and adjust the environmental parameters within the premix tank. This process constitutes a complete closed-loop regulation path, with the environmental parameter sensor module as the sensing end, the control system as the decision-making end, and the adjustment component as the execution end. Through continuous high-frequency sensing, dynamic decision-making, and precise execution, the closed-loop regulation path achieves dynamic and precise control of the premix tank's environmental parameters, ensuring that they are stably maintained within the set target environmental parameter range.

[0056] S140, when the environmental parameters in the premix box are stable within the target range, the air duct is opened to deliver the stable environment corresponding to the environmental parameters to the interior of the plant box at a constant flow rate, and at the same time, the changes in the environmental parameters in the delivery path are compensated and corrected in real time based on the machine learning model.

[0057] In one possible embodiment, when the environmental parameters in the premix box are stable within a target range, the air duct is opened, and the stable environment corresponding to the environmental parameters is delivered to the interior of the plant box at a constant flow rate. At the same time, based on the machine learning model, real-time compensation and correction are performed on the changes in the environmental parameters in the delivery path, specifically including: when the control system determines that the first real-time temperature value, the first real-time humidity value and the first real-time carbon dioxide concentration value are all within the error range of the target temperature parameter, the target humidity parameter and the target carbon dioxide concentration parameter, the air duct opening instruction is output to drive the air duct gas delivery mechanism to start, the air duct gas delivery mechanism includes an air supply fan, an air flow control valve and a pressure stabilization module; a delivery path environmental parameter monitoring module is arranged along the inside of the air duct, the delivery path environmental parameter monitoring module includes a delivery path temperature sensor, a delivery path humidity sensor and a delivery path carbon dioxide concentration sensor, which is used to obtain environmental change data in the delivery path; the environmental change data is input into the environmental transmission prediction sub-model in the machine learning model, the environmental transmission prediction sub-model generates an environmental compensation instruction, and the control system dynamically adjusts the output amplitude of the temperature adjustment component, the humidity adjustment component and the carbon dioxide concentration adjustment component according to the environmental compensation instruction to achieve real-time compensation and correction of the environmental parameters at the terminal of the delivery path.

[0058] Specifically, the control system continuously monitors first real-time environmental parameters, including the first real-time temperature, humidity, and CO2 concentration, and compares them item by item with the target temperature, humidity, and CO2 concentration parameters. It then uses absolute error calculation to determine whether all current environmental parameters within the premix tank are within a set error tolerance. If the control system determines that all first real-time environmental parameters meet the tolerance conditions, it deems the premix tank's internal environment to be stable. The system then issues an air duct opening command, and the delivery phase begins.

[0059] Secondly, the duct gas delivery mechanism activates upon receiving the duct opening command. It comprises a supply blower, an airflow control valve, and a pressure stabilization module. The supply blower provides constant power to propel ambient gas flow. The airflow control valve dynamically adjusts its opening according to the system's set constant flow parameters to maintain a stable gas flow rate. The pressure stabilization module suppresses transient pressure differential fluctuations caused by long-distance pipeline transport or gas disturbances, ensuring a controlled, orderly laminar flow of stable ambient gas in the duct.

[0060] Next, to accurately monitor the environmental parameters of the gas during duct transport, a conveyor path environmental parameter monitoring module was deployed at key locations within the duct. This module includes a temperature sensor, a humidity sensor, and a carbon dioxide concentration sensor. These sensors collect real-time data on temperature, humidity, and carbon dioxide concentration along the conveyor path, generating environmental change data. The environmental change data output by all sensors is organized as a time series and sent through a data channel to the control system for unified processing.

[0061] The control system then inputs this environmental change data into the environmental transmission prediction submodel within the machine learning model. This submodel compares the environmental parameters at the premix tank outlet with the real-time environmental parameters at the end of the delivery path. It then combines delivery time delay, pipeline length, heat transfer characteristics, and gas diffusion patterns to infer the environmental attenuation trend of the gas during delivery. Using a feedforward compensation mechanism, the submodel generates environmental compensation instructions, including temperature compensation, humidity correction, and CO2 concentration correction targets.

[0062] Finally, the control system adjusts the output amplitudes of the temperature, humidity, and carbon dioxide concentration control components in real time based on the environmental compensation instructions, ensuring that the ambient gas output in the next cycle has advanced regulation characteristics, thereby accurately reconstructing the target environmental parameters at the end of the air duct, that is, the entrance to the plant box. This compensation correction process dynamically responds to environmental changes within the delivery path, accurately restoring the final environmental parameters of the gas after delivery, ensuring that the environment received by the plant box always remains consistent with the target premix box, forming a high-precision and high-consistency environmental control mechanism.

[0063] S150, input the disturbance data monitored in real time by the sensor in the plant box into the disturbance adaptation submodule in the machine learning model to predict the disturbance trend and output the environmental disturbance compensation data.

[0064] In one possible embodiment, the disturbance data monitored in real time by the sensor in the plant box is input into the disturbance adaptation submodule in the machine learning model for disturbance trend prediction, and the environmental disturbance compensation data is output, which specifically includes: deploying environmental disturbance monitoring sensor modules at multiple locations inside the plant box, the environmental disturbance monitoring sensor module including a plant box temperature sensor, a plant box humidity sensor and a plant box carbon dioxide concentration sensor, for collecting real-time temperature disturbance values, real-time humidity disturbance values ​​and real-time carbon dioxide concentration disturbance values ​​in the plant box to form a disturbance data sequence; packaging the disturbance data sequence into a disturbance data input frame and transmitting it to the disturbance adaptation submodule of the machine learning model, the disturbance adaptation submodule performs disturbance trend prediction based on the neural network structure of time series modeling, and outputs the disturbance trend prediction result; generating environmental disturbance compensation data according to the disturbance trend prediction result, the environmental disturbance compensation data including the pre-adjustment amplitude, adjustment rate and duration for controlling the temperature adjustment component, the humidity adjustment component and the carbon dioxide concentration adjustment component to achieve feedforward compensation correction of the disturbance.

[0065] Specifically, first, environmental disturbance monitoring sensor modules are deployed at multiple locations within the plant box. These modules consist of a plant box temperature sensor, a plant box humidity sensor, and a plant box carbon dioxide concentration sensor. These sensors are used to collect real-time temperature disturbance values, real-time humidity disturbance values, and real-time carbon dioxide concentration disturbance values ​​within the plant box at high frequency. The deployment locations of the environmental disturbance monitoring sensor modules must cover the intersection of different vertical levels within the box and the airflow path to ensure that the collected environmental disturbance data is spatially representative and dynamically responsive, fully reflecting the local environmental disturbance characteristics caused by plant metabolic behavior.

[0066] Secondly, the control system organizes the real-time temperature, humidity, and carbon dioxide concentration disturbance values ​​continuously collected by the environmental disturbance monitoring sensor module into a disturbance data sequence in chronological order and packages it into a structured disturbance data input frame. The disturbance data input frame contains fields such as the disturbance type identifier, disturbance amplitude information, sampling timestamp, and corresponding sensor number. The disturbance data input frame serves as the time series input to the neural network model and, after data verification, is transmitted to the disturbance adaptation submodule of the machine learning model.

[0067] Next, the disturbance adaptation submodule predicts disturbance trends for the disturbance data input frames based on a neural network structure that models temporal patterns. This neural network, employing a gated recurrent unit network or a long short-term memory network, is capable of modeling the time-dependent characteristics of disturbances, accurately identifying disturbance development patterns and potential change paths. After receiving the disturbance data input frames, the model generates disturbance trend predictions through multi-layer recursive operations. These predictions include the changing trend of the environmental disturbance amplitude, the slope of the disturbance intensity change, and the range of possible disturbance thresholds over a period of time.

[0068] Then, based on the disturbance trend prediction results, the disturbance adaptation submodule outputs the corresponding environmental disturbance compensation data. This environmental disturbance compensation data includes three control dimensions: pre-adjustment amplitude, adjustment rate, and duration. These are used to define the pre-disturbance response strategies of the temperature control component, the humidity control component, and the carbon dioxide concentration control component, respectively. The pre-adjustment amplitude sets the control component's advance control target, the adjustment rate controls the execution rhythm of the adjustment process, and the duration defines the effective window of control action to achieve disturbance feedforward suppression.

[0069] Ultimately, the control system interprets the environmental disturbance compensation data as control instructions, driving the regulatory components into a pre-response state and completing parameter corrections before the actual environmental disturbance occurs, forming a control chain of "disturbance identification - trend prediction - feedforward regulation." This control chain enables dynamic perception and proactive response to environmental disturbances induced by plant metabolism, significantly improving the stability and predictability of environmental regulation within the plant chamber, ensuring that plants maintain a highly consistent and suitable target environment throughout their growth cycle.

[0070] S160, when the second real-time environmental parameter in the plant box exceeds the set threshold, the control system drives the exhaust mechanism to exhaust the air in the plant box, and at the same time controls the premixing box to output a new round of stable environmental gas, and adaptively adjusts the control strategy of the premixing box based on the environmental disturbance compensation data.

[0071] In one possible embodiment, when the second real-time environmental parameter in the plant box exceeds the set threshold, the control system drives the exhaust mechanism to exhaust the air in the plant box, and at the same time controls the premix box to output a new round of stable environmental gas, and adaptively adjusts the control strategy of the premix box based on the environmental disturbance compensation data, specifically including: the environmental parameter judgment module receives the second real-time environmental parameter collected by the plant box temperature sensor, the plant box humidity sensor and the plant box carbon dioxide concentration sensor, the second real-time environmental parameter includes the second real-time temperature value, the second real-time humidity value and the second real-time carbon dioxide concentration value, and compares the real-time temperature value, the real-time humidity value and the real-time carbon dioxide concentration value with the target temperature parameter, the target humidity parameter and the error threshold of the target carbon dioxide concentration parameter; when any second real-time environmental parameter exceeds the corresponding error threshold range, the control system outputs an exhaust instruction to drive the exhaust fan to discharge the non-standard environmental gas in the plant box through the exhaust duct; at the same time, the control system outputs an air supply instruction to drive the premix box to generate a new round of stable environmental gas based on the current target temperature parameter, target humidity parameter and target carbon dioxide concentration parameter, and transport it to the plant box through the air duct; the control system further inputs the environmental disturbance compensation data into the strategy adjustment module, and adaptively optimizes the parameters such as the current premix box adjustment period, adjustment amplitude, sensor sampling density and control signal gain to form a control strategy with enhanced dynamic response.

[0072] Specifically, the environmental parameter determination module first receives, in real time, second real-time environmental parameters collected by the plant box temperature sensor, the plant box humidity sensor, and the plant box carbon dioxide concentration sensor. The second real-time environmental parameters include a second real-time temperature value, a second real-time humidity value, and a second real-time carbon dioxide concentration value. The environmental parameter determination module performs a data integrity check on the second real-time environmental parameters and compares them one-to-one with the target temperature parameter, target humidity parameter, and target carbon dioxide concentration parameter output by the machine learning model. Using absolute error threshold determination logic, the module identifies whether any environmental parameter deviates beyond the tolerance range.

[0073] If the deviation of any of the second real-time environmental parameters exceeds the corresponding error threshold, the system determines that the current environmental state within the plant box is unstable. The control system immediately outputs an exhaust command to the exhaust control unit, driving the exhaust fan to start. The exhaust fan is connected to the exhaust duct via a preset path, rapidly exhausting the non-standard ambient air inside the plant box to the outside of the box. A one-way valve is installed in the exhaust path to prevent external air from flowing back, ensuring the unidirectional and efficient exhaust operation and preventing the recirculation and accumulation of disturbed air within the box.

[0074] At the same time, the control system sends a re-airing command to the premix box control module, containing the current target temperature, humidity, and CO2 concentration parameters. The temperature, humidity, and CO2 concentration control components within the premix box rapidly complete a new round of environmental adjustments based on the target parameters, generating a stable ambient gas with the set parameters. This stable ambient gas is then delivered to the plant box at a constant flow rate through the air duct gas delivery mechanism, completing the internal environmental reconstruction process.

[0075] To prevent repeated disturbances and improve the environmental control system's response sensitivity and regulatory adaptability, the control system further inputs the environmental disturbance compensation data previously generated by the disturbance adaptation submodule into the strategy adjustment module. The strategy adjustment module dynamically optimizes and adjusts key control parameters in the current control strategy, including the adjustment period, adjustment amplitude, sensor sampling density, and control signal gain. The adjustment period influences the frequency of regulatory response, the adjustment amplitude controls the amplitude boundary of regulatory behavior, the sensor sampling density determines the accuracy of environmental perception, and the control signal gain influences the regulation rate and control stability. The strategy adjustment module adaptively reconfigures these parameters at the differential level, taking into account the current disturbance type and predicted trends, to form a control strategy with enhanced adaptability.

[0076] Ultimately, through the above-mentioned environmental parameter error judgment, non-standard environmental gas discharge, standard environmental gas reconstruction and adaptive adjustment of the control strategy, a closed-loop disturbance response mechanism with environmental parameters as the core control target was constructed, which significantly improved the environmental stability of the plant box and the overall disturbance carrying capacity of the system, ensuring that plants can still maintain a stable and consistent growth environment under fluctuating interference conditions.

[0077] This embodiment also discloses an intelligent premixing control device based on machine learning, referring to Figure 2 , comprising an acquisition module 201, a processing module 202 and an output module 203, the device is used to execute any of the above-mentioned intelligent premixing control methods based on machine learning, wherein: The acquisition module 201 is used to build a machine learning model based on historical environmental control data, plant growth data and environmental disturbance response data.

[0078] The processing module 202 is used to embed the machine learning model into the control system, and the control system controls the adjustment component to perform preliminary adjustment on the environment in the premix box according to the target environmental parameters.

[0079] The processing module 202 is configured to obtain first real-time environmental data through sensors deployed in the premixing tank and feed the data back to the control system to form a closed-loop regulation.

[0080] The processing module 202 is used to open the air duct when the environmental parameters in the premix box are stable within the target range, and deliver the stable environment corresponding to the environmental parameters to the interior of the plant box at a constant flow rate. At the same time, based on the machine learning model, real-time compensation and correction are performed on the changes in the environmental parameters in the delivery path.

[0081] The processing module 202 is used to input the disturbance data monitored in real time by the sensors in the plant box into the disturbance adaptation submodule in the machine learning model to predict the disturbance trend and output the environmental disturbance compensation data.

[0082] Output module 203 is used to control the exhaust mechanism to exhaust the air from the plant box when the second real-time environmental parameter in the plant box exceeds the set threshold, and at the same time control the premixing box to output a new round of stable environmental gas, and adaptively adjust the control strategy of the premixing box based on the environmental disturbance compensation data.

[0083] In one possible embodiment, the acquisition module 201 is used to collect historical environmental control data of plants during multiple growth cycles, and the historical environmental control data includes environmental parameter set values, environmental parameter actual values, environmental parameter adjustment amplitudes, environmental parameter adjustment response times, environmental parameter adjustment frequencies, and environmental parameter stability indicators.

[0084] The acquisition module 201 is used to collect plant growth data within a time period corresponding to the historical environmental control data. The plant growth data includes plant species, growth stage, plant height, leaf area index, chlorophyll content, dry matter mass and growth rate.

[0085] The acquisition module 201 is used to collect environmental disturbance response data recorded during plant growth. The environmental disturbance response data includes disturbance type, disturbance amplitude, disturbance duration, disturbance recovery time and disturbance regulation feedback characteristics.

[0086] The processing module 202 is used to time-align the historical environmental control data, plant growth data, and environmental disturbance response data to construct a unified training sample data set.

[0087] The output module 203 is used to train the training sample data set using a deep neural network to form a machine learning model for outputting target environmental parameters and environmental disturbance compensation data, wherein a multi-layer feedforward neural network including an input layer, multiple hidden layers and an output layer is constructed. The input layer receives plant species, growth stages, historical sequences of environmental parameters and phenotypic indicator characteristics, and the output layer generates target environmental parameters and disturbance compensation strategies. The training process of the machine learning model adopts a supervised learning mechanism, and the objective function is set to minimize the mean square error between the predicted environmental parameters and the actual adjustment results. At the same time, a regularization term is introduced to constrain the complexity of the controller parameters to prevent model overfitting. The cross-validation method is used to evaluate the generalization ability of the model during the training process, and the model structure and control parameters are optimized through hyperparameter grid search.

[0088] In one possible implementation, the processing module 202 is used to deploy the machine learning model in a core control unit of the control system, where the core control unit includes a data receiving module, a model parsing module, a control signal generating module, and an execution feedback module.

[0089] The processing module 202 is used to determine the target environmental parameters output by the machine learning model received by the data receiving module, and the model parsing module generates corresponding adjustment instructions based on the target environmental parameters. The adjustment instructions include temperature adjustment instructions, humidity adjustment instructions and carbon dioxide concentration adjustment instructions.

[0090] The output module 203 is used to control the control signal generation module to output control signals to the temperature adjustment component, the humidity adjustment component and the carbon dioxide concentration adjustment component according to the adjustment instruction, so as to drive the adjustment components to perform adjustment operations.

[0091] The processing module 202 is used to control the execution feedback module to collect current environmental parameters according to the environmental parameter sensors deployed inside the premixing box. The current environmental parameters include current temperature value, current humidity value and current carbon dioxide concentration value.

[0092] The processing module 202 is used to calculate the error between the current environmental parameters and the target environmental parameters, and to modify the control signal in real time based on the error to form a closed-loop regulation control logic to stabilize the environmental parameters in the premix box to the target environmental parameters.

[0093] In a possible implementation, the processing module 202 is configured to deploy environmental parameter sensor modules at multiple spatial locations inside the premixing box. The environmental parameter sensor modules include a temperature sensor, a humidity sensor, and a carbon dioxide concentration sensor.

[0094] The acquisition module 201 is used to control the environmental parameter sensor module to collect first real-time environmental parameters at a preset sampling period. The first real-time environmental parameters include a first real-time temperature value, a first real-time humidity value, and a first real-time carbon dioxide concentration value.

[0095] The output module 203 is used to package the first real-time environmental parameter into an environmental parameter data frame and send it to the control system through the data transmission module.

[0096] The processing module 202 is used to control the control system to calculate the error between the environmental parameter data and the target environmental parameters to obtain the temperature error, humidity error and carbon dioxide concentration error, and generate a control signal based on the proportional integral differential control algorithm to drive the temperature adjustment component, the humidity adjustment component and the carbon dioxide concentration adjustment component to perform adjustment operations, forming a closed-loop adjustment path with the environmental parameter sensor module as the perception end, the control system as the decision end, and the adjustment component as the execution end.

[0097] In one possible embodiment, the processing module 202 is used to output a duct opening instruction to drive the duct gas delivery mechanism to start when the control system determines that the first real-time temperature value, the first real-time humidity value and the first real-time carbon dioxide concentration value are all within the error range of the target temperature parameter, the target humidity parameter and the target carbon dioxide concentration parameter. The duct gas delivery mechanism includes an air supply fan, an air flow control valve and a pressure stabilization module.

[0098] The processing module 202 is used to arrange a conveying path environmental parameter monitoring module along the inside of the air duct. The conveying path environmental parameter monitoring module includes a conveying path temperature sensor, a conveying path humidity sensor and a conveying path carbon dioxide concentration sensor, which are used to obtain environmental change data within the conveying path.

[0099] Processing module 202 is used to input environmental change data into the environmental transmission prediction sub-model in the machine learning model. The environmental transmission prediction sub-model generates environmental compensation instructions. The control system dynamically adjusts the output amplitudes of the temperature control component, the humidity control component and the carbon dioxide concentration control component according to the environmental compensation instructions to achieve real-time compensation and correction of the environmental parameters of the terminal of the transmission path.

[0100] In one possible embodiment, the processing module 202 is used to deploy environmental disturbance monitoring sensor modules at multiple locations inside the plant box. The environmental disturbance monitoring sensor modules include a plant box temperature sensor, a plant box humidity sensor, and a plant box carbon dioxide concentration sensor, which are used to collect real-time temperature disturbance values, real-time humidity disturbance values, and real-time carbon dioxide concentration disturbance values ​​inside the plant box to form a disturbance data sequence.

[0101] The processing module 202 is used to package the disturbance data sequence into a disturbance data input frame and transmit it to the disturbance adaptation submodule of the machine learning model. The disturbance adaptation submodule performs disturbance trend prediction based on the neural network structure of time series modeling and outputs the disturbance trend prediction result.

[0102] The processing module 202 is used to generate environmental disturbance compensation data based on the disturbance trend prediction results. The environmental disturbance compensation data includes the pre-adjustment amplitude, adjustment rate and duration for controlling the temperature adjustment component, the humidity adjustment component and the carbon dioxide concentration adjustment component to achieve feedforward compensation correction of the disturbance.

[0103] In one possible embodiment, the processing module 202 is used to control the environmental parameter judgment module to receive the second real-time environmental parameter collected by the plant box temperature sensor, the plant box humidity sensor and the plant box carbon dioxide concentration sensor, the second real-time environmental parameter including the second real-time temperature value, the second real-time humidity value and the second real-time carbon dioxide concentration value, and compare the real-time temperature value, the real-time humidity value and the real-time carbon dioxide concentration value with the error thresholds of the target temperature parameter, the target humidity parameter and the target carbon dioxide concentration parameter.

[0104] The processing module 202 is configured to control the control system to output an exhaust instruction and drive the exhaust fan to exhaust the non-standard environmental gas in the plant box through the exhaust channel when any second real-time environmental parameter exceeds the corresponding error threshold range.

[0105] The processing module 202 is used to control the control system to output the air supply instruction, drive the premixing box to generate a new round of stable ambient gas based on the current target temperature parameter, target humidity parameter and target carbon dioxide concentration parameter, and transport it to the plant box through the air duct.

[0106] The processing module 202 is used to control the control system to further input the environmental disturbance compensation data into the strategy adjustment module to adaptively optimize the parameters such as the adjustment cycle, adjustment amplitude, sensor sampling density and control signal gain of the current premixing box to form a control strategy with enhanced dynamic response.

[0107] It should be noted that the above embodiments provide devices that implement their functions using only the division of the above functional modules as examples. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiments are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0108] This embodiment also discloses an electronic device, referring to Figure 3The electronic device may include: at least one processor 301 , at least one communication bus 302 , a user interface 303 , a network interface 304 , and at least one memory 305 .

[0109] The communication bus 302 is used to implement the connection and communication between these components.

[0110] The user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0111] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0112] The processor 301 may include one or more processing cores. The processor 301 utilizes various interfaces and circuits to connect various components within the server. It executes instructions, programs, code sets, or instruction sets stored in the memory 305, as well as accesses data stored in the memory 305, to perform various server functions and process data. Optionally, the processor 301 may be implemented using at least one of the following hardware forms: a digital signal processing (DSP), a field-programmable gate array (FPGA), or a programmable logic array (PLA). The processor 301 may integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. The CPU primarily handles the operating system, user interface, and application programs. The GPU is responsible for rendering and drawing content displayed on the display screen. The modem handles wireless communications. It is understood that the modem may not be integrated into the processor 301 but implemented as a separate chip.

[0113] The memory 305 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory may include non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, codes, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, sound playback function, image playback function, etc.), instructions for implementing the above-mentioned method embodiments, etc. The data storage area may store data related to the above-mentioned method embodiments, etc. The memory 305 may also optionally be at least one storage device located remotely from the processor 301. The memory 305, as a computer storage medium, may include an operating system, a network communication module, a user interface 303 module, and an application program for an intelligent premixing control method based on machine learning.

[0114] exist Figure 3 In the electronic device shown, user interface 303 is primarily used to provide an input interface for the user and to obtain user input data. Processor 301 can be used to invoke an application stored in memory 305 that includes a machine learning-based intelligent premixing control method. When executed by one or more processors 301, the electronic device executes one or more of the methods described in the aforementioned embodiments.

[0115] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for this application.

[0116] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0117] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are merely schematic, such as the division of units, which is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interface, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0118] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0119] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0120] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of this application, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory 305 and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of this application. The aforementioned memory 305 includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a magnetic disk, or an optical disk.

[0121] The present application also discloses a computer-readable storage medium storing instructions, which, when executed by one or more processors 301 , enable an electronic device to execute one or more methods in the above embodiments.

[0122] The above are merely exemplary embodiments of the present disclosure and are not intended to limit the scope of the present disclosure. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure. This application is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not recorded in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. An intelligent premixing control method based on machine learning, characterized in that: The method comprises: Build machine learning models based on historical environmental control data, plant growth data, and environmental disturbance response data; The machine learning model is embedded in a control system, and the control system controls the regulating component to perform preliminary regulation on the environment in the premix box according to the target environmental parameters; Acquire first real-time environmental data through sensors deployed in the premixing tank and feed it back to the control system to form a closed-loop regulation; When the environmental parameters in the premix box are stable within the target range, the air duct is opened to deliver the stable environment corresponding to the environmental parameters to the interior of the plant box at a constant flow rate, and at the same time, the changes in the environmental parameters in the delivery path are compensated and corrected in real time based on the machine learning model; Inputting the disturbance data monitored in real time by the sensor in the plant box into the disturbance adaptation submodule in the machine learning model to predict the disturbance trend and output environmental disturbance compensation data; When the second real-time environmental parameter inside the plant box exceeds a set threshold, the control system drives the exhaust mechanism to exhaust the air from the plant box, while controlling the premixing box to output a new round of stable environmental gas, and adaptively adjusts the control strategy of the premixing box based on the environmental disturbance compensation data.

2. The intelligent premixing control method based on machine learning according to claim 1, characterized in that: The machine learning model is constructed based on historical environmental control data, plant growth data, and environmental disturbance response data, specifically including: Collecting historical environmental control data of plants over multiple growth cycles, the historical environmental control data including environmental parameter set values, environmental parameter actual values, environmental parameter adjustment amplitudes, environmental parameter adjustment response times, environmental parameter adjustment frequencies, and environmental parameter stability indicators; Collecting plant growth data within a time period corresponding to the historical environmental control data, the plant growth data including plant species, growth stage, plant height, leaf area index, chlorophyll content, dry matter mass, and growth rate; Collecting environmental disturbance response data recorded during plant growth, the environmental disturbance response data including disturbance type, disturbance amplitude, disturbance duration, disturbance recovery time, and disturbance regulation feedback characteristics; Time-aligning the historical environmental control data, plant growth data, and environmental disturbance response data to construct a unified training sample data set; A deep neural network is used to train the training sample data set to form a machine learning model for outputting target environmental parameters and environmental disturbance compensation data, wherein a multi-layer feedforward neural network is constructed including an input layer, multiple hidden layers and an output layer. The input layer receives plant species, growth stages, historical sequences of environmental parameters and phenotypic indicator characteristics, and the output layer generates the target environmental parameters and disturbance compensation strategy. The training process of the machine learning model adopts a supervised learning mechanism, and the objective function is set to minimize the mean square error between the predicted environmental parameters and the actual adjustment results. At the same time, a regularization term is introduced to constrain the complexity of controller parameters to prevent model overfitting. A cross-validation method is used to evaluate the generalization ability of the model during the training process, and the model structure and control parameters are optimized through hyperparameter grid search.

3. The intelligent premixing control method based on machine learning according to claim 1, characterized in that: The machine learning model is embedded in a control system, and the control system controls the regulating component to perform preliminary regulation on the environment in the premix box according to the target environmental parameters, specifically including: Deploy the machine learning model in the core control unit of the control system, wherein the core control unit includes a data receiving module, a model parsing module, a control signal generating module, and an execution feedback module; The data receiving module is used to receive the target environmental parameters output by the machine learning model, and the model parsing module generates corresponding adjustment instructions according to the target environmental parameters, wherein the adjustment instructions include temperature adjustment instructions, humidity adjustment instructions, and carbon dioxide concentration adjustment instructions; The control signal generating module outputs control signals to the temperature adjustment component, the humidity adjustment component and the carbon dioxide concentration adjustment component according to the adjustment instruction, so as to drive the adjustment components to perform adjustment operations; The execution feedback module collects current environmental parameters according to the environmental parameter sensor deployed inside the premix box, and the current environmental parameters include current temperature value, current humidity value and current carbon dioxide concentration value; The error between the current environmental parameter and the target environmental parameter is calculated, and the control signal is corrected in real time based on the error to form a closed-loop regulation control logic, so that the environmental parameter in the premix box is stabilized to the target environmental parameter.

4. The intelligent premixing control method based on machine learning according to claim 1, characterized in that: The first real-time environmental data is obtained by a sensor deployed in the premixing tank and fed back to the control system to form a closed-loop regulation, specifically including: Deploying environmental parameter sensor modules at multiple spatial locations inside the premix box, the environmental parameter sensor modules including a temperature sensor, a humidity sensor, and a carbon dioxide concentration sensor; The environmental parameter sensor module collects first real-time environmental parameters at a preset sampling period, wherein the first real-time environmental parameters include a first real-time temperature value, a first real-time humidity value, and a first real-time carbon dioxide concentration value; Packing the first real-time environmental parameters into an environmental parameter data frame, and sending the frame to the control system via a data transmission module; The control system calculates the error between the environmental parameter data and the target environmental parameter to obtain a temperature error, a humidity error, and a carbon dioxide concentration error, and generates a control signal based on a proportional-integral-differential control algorithm to drive the temperature adjustment component, the humidity adjustment component, and the carbon dioxide concentration adjustment component to perform adjustment operations, thereby forming a closed-loop adjustment path with the environmental parameter sensor module as the perception end, the control system as the decision-making end, and the adjustment component as the execution end.

5. The intelligent premixing control method based on machine learning according to claim 4, characterized in that: When the environmental parameters in the premix box are stable within the target range, the air duct is opened to deliver the stable environment corresponding to the environmental parameters to the interior of the plant box at a constant flow rate, and at the same time, the environmental parameter changes in the delivery path are compensated and corrected in real time based on the machine learning model, specifically including: When the control system determines that the first real-time temperature value, the first real-time humidity value, and the first real-time carbon dioxide concentration value are all within the error range of the target temperature parameter, the target humidity parameter, and the target carbon dioxide concentration parameter, the control system outputs an air duct opening instruction to drive the air duct gas delivery mechanism to start, the air duct gas delivery mechanism including an air supply fan, an air flow control valve, and a pressure stabilization module; A conveying path environmental parameter monitoring module is arranged along the inside of the air duct. The conveying path environmental parameter monitoring module includes a conveying path temperature sensor, a conveying path humidity sensor, and a conveying path carbon dioxide concentration sensor, which are used to obtain environmental change data within the conveying path; The environmental change data is input into the environmental transmission prediction sub-model in the machine learning model, and the environmental transmission prediction sub-model generates an environmental compensation instruction. The control system dynamically adjusts the output amplitude of the temperature adjustment component, the humidity adjustment component and the carbon dioxide concentration adjustment component according to the environmental compensation instruction to achieve real-time compensation correction of the environmental parameters of the terminal of the transmission path.

6. The intelligent premixing control method based on machine learning according to claim 1, characterized in that: The step of inputting the disturbance data monitored in real time by the sensor in the plant box into the disturbance adaptation submodule in the machine learning model to predict the disturbance trend and output the environmental disturbance compensation data specifically includes: Environmental disturbance monitoring sensor modules are deployed at multiple locations inside the plant box. The environmental disturbance monitoring sensor modules include a plant box temperature sensor, a plant box humidity sensor, and a plant box carbon dioxide concentration sensor, which are used to collect real-time temperature disturbance values, real-time humidity disturbance values, and real-time carbon dioxide concentration disturbance values ​​inside the plant box to form a disturbance data sequence. The disturbance data sequence is packaged into a disturbance data input frame and transmitted to the disturbance adaptation submodule of the machine learning model, wherein the disturbance adaptation submodule performs disturbance trend prediction based on the neural network structure of time series modeling and outputs the disturbance trend prediction result; Environmental disturbance compensation data is generated according to the disturbance trend prediction result. The environmental disturbance compensation data includes a pre-adjustment amplitude, an adjustment rate, and a duration for controlling a temperature adjustment component, a humidity adjustment component, and a carbon dioxide concentration adjustment component, so as to achieve feedforward compensation correction of the disturbance.

7. The intelligent premixing control method based on machine learning according to claim 1, characterized in that: When the second real-time environmental parameter in the plant box exceeds a set threshold, the control system drives the exhaust mechanism to exhaust the air in the plant box, and at the same time controls the premixing box to output a new round of stable environmental gas, and adaptively adjusts the control strategy of the premixing box based on the environmental disturbance compensation data, specifically including: The environmental parameter judgment module receives a second real-time environmental parameter collected by the plant box temperature sensor, the plant box humidity sensor, and the plant box carbon dioxide concentration sensor, wherein the second real-time environmental parameter includes a second real-time temperature value, a second real-time humidity value, and a second real-time carbon dioxide concentration value, and compares the real-time temperature value, the real-time humidity value, and the real-time carbon dioxide concentration value with an error threshold value of a target temperature parameter, a target humidity parameter, and a target carbon dioxide concentration parameter; When any of the second real-time environmental parameters exceeds the corresponding error threshold range, the control system outputs an exhaust instruction to drive the exhaust fan to exhaust the non-standard environmental gas in the plant box through the exhaust channel; At the same time, the control system outputs an air supply instruction to drive the premixing box to generate a new round of stable ambient gas based on the current target temperature parameter, target humidity parameter and target carbon dioxide concentration parameter, and transport it to the plant box through the air duct; The control system further inputs the environmental disturbance compensation data into the strategy adjustment module to adaptively optimize the parameters of the current premixing box, such as the adjustment period, adjustment amplitude, sensor sampling density and control signal gain, to form a control strategy with enhanced dynamic response.

8. An intelligent premixing control device based on machine learning, characterized in that: The device is used to execute an intelligent premixing control method based on machine learning as claimed in any one of claims 1 to 7, and the device comprises an acquisition module (201), a processing module (202) and an output module (203), wherein: The acquisition module (201) is used to construct a machine learning model based on historical environmental control data, plant growth data and environmental disturbance response data; The processing module (202) is used to embed the machine learning model into a control system, and the control system controls the regulating component to perform preliminary regulation on the environment in the premix box according to target environmental parameters; The processing module (202) is used to obtain first real-time environmental data through sensors deployed in the premixing tank and feed it back to the control system to form a closed-loop regulation; The processing module (202) is configured to, when the environmental parameters in the premix box are stable within a target range, open the air duct to deliver the stable environment corresponding to the environmental parameters to the interior of the plant box at a constant flow rate, and simultaneously perform real-time compensation and correction of changes in the environmental parameters in the delivery path based on the machine learning model; The processing module (202) is used to input the disturbance data monitored in real time by the sensor in the plant box into the disturbance adaptation submodule in the machine learning model to perform disturbance trend prediction and output environmental disturbance compensation data; The output module (203) is used for, when the second real-time environmental parameter in the plant box exceeds a set threshold, the control system drives the exhaust mechanism to exhaust the air in the plant box, and at the same time controls the premixing box to output a new round of stable environmental gas, and adaptively adjusts the control strategy of the premixing box based on the environmental disturbance compensation data.

9. An electronic device, characterized in that: The electronic device comprises a processor (301), a communication bus (302), a user interface (303), a network interface (304) and a memory (305), wherein the memory (305) is used to store instructions, the user interface (303) and the network interface (304) are both used to communicate with other devices, the communication bus (302) is used to realize connection and communication between components in the electronic device, and the processor (301) is used to execute the instructions stored in the memory (305) so that the electronic device executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions, and when the instructions are executed, the method according to any one of claims 1 to 7 is executed.

Images