Intelligent greenhouse temperature control method based on multi-source sensor

Abstract

The invention discloses an intelligent greenhouse temperature control method based on a multi-source sensor, and relates to the technical field of seed and seedling cultivation. Performing preliminary temperature prediction on the acquired data; designing a temperature control strategy according to a temperature prediction result; adjusting the environment in the greenhouse according to the temperature control strategy; performing growth prediction on the crops according to the temperature and humidity data; performing intelligent irrigation management based on a crop growth prediction result; performing greenhouse energy efficiency optimization according to an irrigation adjustment result; and carrying out greenhouse comprehensive management decision support based on an energy efficiency optimization result. According to the invention, through improving a growth predicted value conversion method output by a deep learning model and reward function design of a reinforcement learning algorithm and adopting an innovative reinforcement learning reward function design method, the innovative method ensures accurate connection of a crop growth state and greenhouse environment control; and healthy growth of crops is ensured while the energy efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of seed and seedling cultivation, and specifically provides an intelligent greenhouse temperature control method based on multi-source sensors. Background Art

[0002] In recent years, the application of advanced artificial intelligence technologies such as deep learning and reinforcement learning in greenhouse management for seed and seedling cultivation has gradually increased. Especially in crop growth prediction and optimization of environmental control strategies, it has shown great potential. With the rapid development of agricultural intelligence, the adjustment of greenhouse environment (such as temperature, humidity, light intensity, etc.) increasingly relies on the collaborative application of sensor data and intelligent algorithms.

[0003] In order to ensure that crops can obtain the best growth environment in the greenhouse, researchers have proposed a method of multi-source sensor data fusion, which conducts real-time monitoring and adjustment through information such as temperature, humidity, air flow rate, light intensity, and soil humidity, and uses the reward function of the reinforcement learning algorithm to achieve intelligent control of the greenhouse environment.

[0004] In the prior art, the design of the reward function of the reinforcement learning algorithm is insufficient to fully consider the complexity of multi-source sensor data. Existing reinforcement learning algorithms often rely on simple reward function designs and fail to fully consider the complex non-linear relationship between crop growth and greenhouse environment, resulting in optimization strategies being difficult to cope with changing environmental conditions and unable to achieve long-term and stable energy optimization and crop growth improvement in the greenhouse. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides an intelligent greenhouse temperature control method based on multi-source sensors to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides an intelligent greenhouse temperature control method based on multi-source sensors, including the following steps: S1. Data collection; Collect environmental data inside the greenhouse through multi-source sensors and use data preprocessing technology to clean the environmental data to obtain collected data; S2. Conduct preliminary temperature prediction on the collected data; Conduct preliminary temperature prediction on the collected data through prediction software to obtain a temperature prediction result; S3. Obtain a temperature control strategy according to the temperature prediction result; Design a temperature control strategy based on fuzzy control for the temperature prediction result to obtain a temperature control strategy; S4. Adjust the environment inside the greenhouse according to the temperature control strategy; Compare the real-time collected data from multi-source sensors with the temperature control strategy, adjust the temperature and humidity data, and obtain the temperature and humidity data; S5. Predict the growth of crops based on the temperature and humidity data; Combine the temperature and humidity data, and use a deep learning model to predict the growth of crops to obtain the crop growth prediction result; S6. Conduct intelligent irrigation management based on the crop growth prediction result; Based on the crop growth prediction result, ensure that the water required by the crops is met by dynamically adjusting the irrigation amount and irrigation time, and then obtain the irrigation adjustment result; S7. Optimize the greenhouse energy efficiency according to the irrigation adjustment result; According to the irrigation adjustment result, optimize the greenhouse energy efficiency through a reinforcement learning algorithm to obtain the energy efficiency optimization result; S8. Provide decision support for greenhouse comprehensive management based on the energy efficiency optimization result; Based on the energy efficiency optimization result, provide decision support for the comprehensive management of the greenhouse and generate a comprehensive report.

[0007] Further optimize this technical solution. The environmental data inside the greenhouse in step S1 includes: Temperature, humidity, air velocity, light intensity, soil humidity. These data are collected by multi-source sensors.

[0008] Further optimize this technical solution. The deep learning model in step S5 is based on: A multi-input neural network that uses a weighted fusion mechanism to process and fuse five environmental parameters: temperature, humidity, air velocity, light intensity, and soil humidity.

[0009] Further optimize this technical solution. The construction steps of the multi-input neural network are as follows: First, obtain five environmental factor data from the environmental data and temperature and humidity data collected by multi-source sensors: is the temperature, with the unit of degree Celsius; is the humidity, with the unit of %; is the air velocity, with the unit of m / s; is the light intensity, with the unit of lux; is the soil humidity, with the unit of %; Then construct a multi-input neural network including an input layer, a hidden layer, and an output layer; Input layer: The dynamic weight of each input environmental factor Dynamically adjust according to the current environmental status, and then construct the input layer formula: , ; Among them, represents the processing function of each environmental factor; is the dynamic weight of each environmental factor; is the original input data of the environmental factor; is the weighted input of each environmental factor; Hidden layer: Fuse and process the input features obtained from the input layer through multiple hidden layers. The formula is: ; Among them, is the weight of the node of each environmental factor in each hidden layer; is the preset bias term, the translation amount of the model on the x-axis; is the output of the hidden layer; represents the number of nodes in the hidden layer; Output layer: Map the output of the hidden layer to the growth prediction value of the crop The formula is: ; Among them, is the final output of the model, that is, the growth prediction value; is the activation function sigmoid function of the output layer, and its mathematical expression is ; is the weight of the node of each hidden layer; is the bias term of the output layer, the translation amount of the model on the x-axis, which is set manually.

[0010] Further optimize this technical solution. The reinforcement learning algorithm in step S7 includes: State space and action space; Reward function; Policy function; Update policy.

[0011] Further optimize this technical solution. The state space is: Multiple environmental factors of the greenhouse and the growth status of the crops are represented as a vector at each time step as follows: ; wherein, is the temperature at time , in degrees Celsius; is the humidity at time , in %; is the air velocity at time , in m / s; is the light intensity at time , in lux; is the soil humidity at time , in %; is the crop growth status at time ; is the state space vector.

[0012] To further optimize this technical solution, the action space represents: all executable operations, including temperature adjustment, humidity adjustment, and irrigation volume adjustment; At each time step , a combined control strategy is constructed and represented as a vector: ; wherein, is the temperature control action, heating or cooling; is the humidity control action, increasing or decreasing moisture; is the irrigation control action, starting or stopping irrigation; is the action space vector.

[0013] To further optimize this technical solution, the reward function includes: the impact of the state space and the action space on the greenhouse energy efficiency and crop growth; the reward function At each time step has the formula: ; wherein, is the current time step The energy efficiency gain, i.e., the amount of energy saved; is the change in the growth state of the crop at the current time step ; are the preset target temperature and humidity; is the preset weight coefficient.

[0014] To further optimize this technical solution, the policy function is: Determine the optimal action space taken in the state space; the policy function is equal to , The value update formula is: ; where is the current state and of value, representing the value of this state-action pair; is the preset learning rate, controlling the update step size; is the preset discount factor, balancing the importance of the current reward and future rewards, ranging from 0 to 1; represents the optimal action's value in the next state

[0015] To further optimize this technical solution, the update strategy includes: As the training progresses, the value is continuously updated, and the policy function tends to the optimal policy. In each interaction, the reinforcement learning model selects the optimal action space to adjust the operating state of the greenhouse equipment and updates the decision-making policy according to the immediate reward function .

[0016] In a second aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program instructions are executed by the processor, the steps of an intelligent greenhouse temperature control method based on multi-source sensors as described in the first aspect of the present invention are implemented.

[0017] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, wherein: when the computer program instructions are executed by a processor, the steps of an intelligent greenhouse temperature control method based on multi-source sensors as described in the first aspect of the present invention are implemented.

[0018] Compared with the prior art, the present invention provides an intelligent greenhouse temperature control method based on multi-source sensors, having the following beneficial effects: In this intelligent greenhouse temperature control method based on multi-source sensors, by setting a conversion method for converting the growth prediction value output by an improved deep learning model and designing a reward function of a reinforcement learning algorithm, the lack of a conversion mechanism with physical meaning in the prior art is effectively overcome.

[0019] By normalizing, targetizing or weighted fusing the growth prediction value output by the deep learning model, the accuracy of docking the crop growth state with the greenhouse environment control is improved compared with the prior art.

[0020] By adopting an innovative design of the reinforcement learning reward function and fully considering the non-linear relationship between multi-source sensor data and crop growth, the greenhouse control strategy is optimized compared with the prior art, so as to ensure the healthy growth of crops while improving energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings according to these drawings without creative efforts.

[0022] Figure 1 It is a schematic flowchart of an intelligent greenhouse temperature control method based on multi-source sensors proposed by the present invention; Figure 2 It is a schematic flowchart of a reinforcement learning algorithm of an intelligent greenhouse temperature control method based on multi-source sensors proposed by the present invention; Figure 3 It is a schematic flowchart of a weighted fusion mechanism of an intelligent greenhouse temperature control method based on multi-source sensors proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the drawings in the specification.

[0024] In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0025] Secondly, as used herein, "an embodiment" or "embodiment" refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in an embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selectively exclusive embodiment from other embodiments.

[0026] Embodiment 1: Referring to Figures 1 to 3 , which is the first embodiment of the present invention. This embodiment provides an intelligent greenhouse temperature control method based on multi-source sensors, including the following steps: S1. Data collection; Collect environmental data inside the greenhouse through multi-source sensors and use data preprocessing technology to clean the environmental data to obtain the collected data. In this embodiment, environmental data inside the greenhouse, including temperature, humidity, air velocity, light intensity, and soil humidity, etc., is collected through multi-source sensors. The data collection is carried out simultaneously by multiple sensors to ensure the diversity and accuracy of the data. The data collected by the sensors will be input into the preprocessing module for preliminary data cleaning, missing value filling, and noise removal. For the original sensor data, time series synchronization processing is performed so that subsequent analysis can accurately correspond to the environmental state at each moment. This step relies on data preprocessing technology, and during the data cleaning process, methods such as interpolation or Kalman filtering are used to remove sensor errors and noise.

[0027] The key to data collection and preprocessing lies in ensuring that the data collected from multiple sensors can be seamlessly integrated so that the modeling and analysis in subsequent steps can be accurately performed. In practical applications, different types of sensors may have different sampling frequencies and data accuracies. Therefore, through time series synchronization processing, it can be ensured that all data can be aligned along the same time axis. In addition, invalid data or outliers are eliminated through the preprocessing step, which is crucial for the training and accurate prediction of the model. Finally, the preprocessed data will serve as the basis for further analysis and modeling in subsequent steps.

[0028] S2. Perform preliminary temperature prediction on the collected data; Based on time series analysis technology, perform preliminary temperature prediction on the collected data to obtain the temperature prediction result.

[0029] In this embodiment, time series analysis technology is used for preliminary temperature prediction. Time series data is indispensable in greenhouse environment monitoring. In particular, the temperature fluctuations often have periodicity and regularity. By analyzing the trend of historical temperature data, the model can predict future temperature changes and provide a reference for subsequent temperature control. The ARIMA (AutoRegressive Integrated Moving Average) model is used for temperature prediction, and its advantage lies in being able to capture the time dependence and seasonal characteristics of temperature.

[0030] The ARIMA model is a very mature and commonly used technology in time series prediction. It can analyze the trend and seasonal components in temperature data through methods such as autoregression, differencing, and moving average. In practical applications, the temperature in the greenhouse is affected by factors such as external weather changes and air flow in the greenhouse, and has strong periodic characteristics. Through the ARIMA model, relatively accurate short-term temperature prediction can be carried out, thus providing a reference basis for temperature regulation and avoiding the adverse effects of temperature fluctuations on crop growth.

[0031] S3. Obtain a temperature control strategy according to the temperature prediction result; Based on fuzzy control, design a temperature control strategy for the temperature prediction result to obtain a temperature control strategy.

[0032] Fuzzy control technology is a control method based on fuzzy logic and is widely used in dealing with complex, nonlinear, and uncertain systems. Different from traditional precise control methods (such as PID control), fuzzy control realizes the adjustment and control of the system by processing fuzzy and inaccurate information, and is especially suitable for systems that cannot be simply described by mathematical models. Its basic idea is to simulate the expert's empirical judgment through the use of fuzzy sets and fuzzy rules, so as to realize the intelligent control of the system.

[0033] Temperature control strategy: In the greenhouse environment, the task of the strategy is to maintain an appropriate temperature for crop growth. Since the temperature is affected by various factors (such as external weather, greenhouse equipment, heat release of the crops themselves, etc.), and the changes of these factors often cannot be accurately modeled, fuzzy control becomes a very effective temperature control strategy.

[0034] The temperature control strategy adjusts the heating, ventilation, shading and other equipment in the greenhouse according to the deviation between the real-time temperature and the set target temperature and the rate of change of the temperature. The specific steps of the temperature control strategy are as follows: Temperature deviation: The temperature control system first calculates the deviation between the current temperature and the target temperature , where is the target temperature, is the current temperature.

[0035] Rate of change of temperature: Then calculate the rate of change of temperature , i.e., the change in temperature between the previous time step and the current time step.

[0036] Among them, represents the temperature at the previous time step.

[0037] Fuzzy rules: Based on the temperature deviation and the rate of change, the control system performs inference through a predefined fuzzy rule base. For example: If the temperature deviation is large and the rate of change of temperature is fast, then stronger heating or stronger cooling is required.

[0038] If the temperature deviation is small and the rate of change is also small, then the current temperature is maintained.

[0039] These rules will dynamically adjust the operation strategy of the device according to the specific situation of the temperature change.

[0040] Control output: According to the result of the fuzzy inference, the defuzzification process converts the output fuzzy control quantity into specific device control instructions (such as the power of the heater, the rotation speed of the fan, the opening and closing degree of the sunshade net, etc.).

[0041] Fuzzy control technology converts the temperature deviation and the rate of change into control instructions through the processes of fuzzification, inference, and defuzzification, and is used to adjust the environmental equipment in the greenhouse. This method is very suitable for the complex temperature regulation problems in the greenhouse environment, can handle uncertainty and hysteresis, and achieve efficient automatic control.

[0042] In this embodiment, considering that there is a certain hysteresis and uncertainty in the temperature change inside the greenhouse, fuzzy control can flexibly adjust the control strategy. According to the predicted temperature change trend, by adjusting facilities such as heating equipment, fans, and sunshade nets, precise temperature control is achieved.

[0043] Fuzzy control is a control method widely used in complex and nonlinear systems. The temperature control system of the greenhouse usually needs to respond to the changes of multiple environmental parameters, and these changes often do not fully conform to the linear law. Therefore, fuzzy control provides an effective non-precise control method. By setting fuzzy linguistic variables such as "high", "medium", and "low" for temperature, as well as fuzzy rules for temperature deviation and rate of change, the system can dynamically adjust the device operation strategy to achieve stable temperature control.

[0044] S4. Adjust the environment inside the greenhouse according to the temperature control strategy; Compare the real-time collected data of the multi-source sensors with the temperature control strategy, adjust the temperature and humidity data, and obtain the temperature and humidity data; In this embodiment, the temperature and humidity in the greenhouse are precisely controlled by driving devices (such as fans, heaters, humidity regulators, etc.). The system compares the real-time sensor data (such as the feedback data from temperature and humidity sensors) with the set temperature control strategy, and then adjusts the operating state of the devices to ensure that the environment in the greenhouse is maintained within the optimal range.

[0045] The core task is to implement the temperature control strategy on specific hardware devices to ensure that the temperature and humidity in the greenhouse are maintained within the ideal range. Through a real-time feedback control mechanism, the system can dynamically adjust the operating mode of the devices according to the environmental parameters detected by the sensors. For example, when the temperature in the greenhouse is too high, the fan will automatically speed up to help dissipate heat; when the humidity is too low, the humidity regulation device will start to maintain the appropriate air humidity. The entire adjustment process is based on real-time data for closed-loop control, thus ensuring the stability of the internal environment of the greenhouse.

[0046] S5. Predict the growth of crops based on the temperature and humidity data; Combined with the temperature and humidity data, use a deep learning model to predict the growth of crops and obtain the crop growth prediction result.

[0047] In this embodiment, combined with the obtained temperature and humidity data, use a deep learning model to predict the growth of crops. The growth of crops is affected by various environmental factors such as temperature, humidity, and light, and the environmental control in the greenhouse can optimize these factors. By training a multi-input deep neural network (DNN), the model can predict the growth rate, health status, and yield of crops under different environmental conditions. Deep learning not only predicts a single variable (such as temperature), but combines multiple environmental factors for comprehensive prediction.

[0048] The deep learning model learns the complex relationship between environmental conditions and crop growth by inputting data from multiple sensors. In this process, the network can automatically extract the non-linear features in the data and perform efficient feature learning and prediction through the combination of multiple layers of neural network hierarchies. Temperature and humidity are two important factors affecting crop growth, but the prediction of a single variable may not be sufficient to comprehensively reflect the growth of crops. Therefore, deep learning can more accurately evaluate the growth state of crops by integrating multiple factors and provide scientific decision-making support for growers.

[0049] The deep learning model is based on a Multivariate Deep Neural Network (MDNN), which can effectively process and fuse multiple environmental parameters (temperature, humidity, air velocity, light intensity, soil humidity), and extract the complex relationships between these factors and crop growth through feature fusion and weighting mechanisms. The core of this model is the weighted fusion mechanism, which can dynamically adjust the weights of each input factor to better adapt to the crop growth patterns under different environmental conditions.

[0050] In conventional deep learning models, usually all input features are treated equally, and the model automatically learns the relationships between these features. However, in practical applications, the degrees of influence of different environmental factors on crop growth are different. For example, in certain seasons or climate conditions, light intensity may be more important than temperature, and vice versa. Therefore, we introduce a dynamic weighting mechanism based on reinforcement learning, which allows the model to automatically adjust the weights of different input features according to the current environmental state (such as the current combination of temperature and humidity), thereby improving the accuracy of model prediction. This is the environmental factor weight adjustment mechanism.

[0051] In this model, some concepts from the Convolutional Neural Network (CNN) are also adopted and embedded into the deep learning network to extract local features from the input data. By using convolutional operations in the input layer, we can capture the local correlations between environmental variables (such as the mutual relationship between temperature and humidity), and these local features will be further learned and fused in subsequent layers, thereby optimizing the overall prediction of crop growth. This is the multi-layer convolutional structure.

[0052] The steps to construct the Multivariate Deep Neural Network are as follows: First, obtain five environmental factor data from the environmental data inside the greenhouse, including: is the temperature, with the unit of degrees Celsius; is the humidity, with the unit of %; is the air velocity, with the unit of m / s; is the light intensity, with the unit of lux; is the soil humidity, with the unit of %; Then construct a Multivariate Deep Neural Network including an input layer, a hidden layer, and an output layer; Input layer: Define a weighted fusion mechanism to make the dynamic weights of each input environmental factor dynamically adjusted according to the current environmental state. Then construct the input layer formula: , ; Among them, represents the REUL function, which is the processing function for each environmental factor; when is greater than 0, the output value is equal to the input value, and when is less than or equal to 0, the output value is 0.

[0053] is the dynamic weight of each environmental factor; is the original input data of the environmental factor; is the weighted input of each environmental factor; Hidden layer: Each input feature obtained from the input layer is fused through multiple hidden layers. The formula is: ; Among them, is the weight of the node of each environmental factor in each hidden layer; is the bias term, which is the translation amount of the model on the x-axis and is set artificially; is the output of the hidden layer; represents the number of nodes in the hidden layer; Output layer: Maps the output of the hidden layer to the growth prediction value of the crop The formula is: ; Among them, is the final output of the model, that is, the growth prediction value; is the activation function of the output layer, the sigmoid function, and its mathematical expression is ; is the weight of the node of each hidden layer; is the bias term of the output layer, which allows the model to be translated arbitrarily on the x-axis to better fit the data.

[0054] The application of this model is explained as: Input data at the input end: The input end of the model includes environmental factors such as temperature, humidity, air velocity, light intensity, and soil humidity. These data are collected in real time by sensors and used as the input to the deep neural network. Each input environmental factor is processed through a dynamic weighting mechanism according to the real-time environmental state and historical data, and different weights are assigned. In this way, the model can automatically adjust the importance of input features according to the actual situation.

[0055] Weighting mechanism: At each time step, the reinforcement learning model automatically adjusts the value so that the system can optimize the prediction ability of the model according to the changing environmental conditions. In this way, the model can dynamically adapt to the impact of factors such as seasonal changes and weather changes on crop growth.

[0056] Model training: The training process of the deep learning model uses labeled crop growth data for supervised learning. By optimizing the objective function (such as minimizing the mean square error), the model will learn the optimal parameter and weight values. During the training process, the reinforcement learning module will automatically adjust the weights of the input features according to the prediction error.

[0057] Prediction output: The trained model can predict the growth state of the crop (such as growth rate, health status, or yield) based on real-time sensor data. This prediction result can provide decision-making support for greenhouse managers and help them adjust the environmental conditions of the greenhouse according to the needs of the crop.

[0058] S6. Perform intelligent irrigation management based on the crop growth prediction results; Based on the crop growth prediction results, the fuzzy logic control technology FLC is used to further optimize the intelligent irrigation management strategy to obtain the irrigation adjustment result.

[0059] In this embodiment, by predicting the growth requirements of the crop and the trend of environmental changes, the irrigation amount and irrigation time are dynamically adjusted to ensure that the water requirements of the crop in different growth stages are met. Using the FLC technology, the working state of the irrigation system is adjusted according to the real-time feedback of the crop growth situation and soil humidity.

[0060] Irrigation is an important task in greenhouse management. Too much or too little water will have a negative impact on crop growth. Traditional irrigation management often relies on fixed time or preset water volume, while the intelligent irrigation system can automatically adjust according to the environment and crop needs. In this step, based on the crop growth prediction, the system adjusts the irrigation amount through the fuzzy logic algorithm to ensure that the water supply of the crop is always in an ideal state. Through real-time feedback adjustment, the utilization efficiency of water resources can be significantly improved and waste can be reduced.

[0061] S7. Optimize the greenhouse energy efficiency according to the irrigation adjustment results; According to the irrigation adjustment results, the energy efficiency of the greenhouse is optimized through a reinforcement learning algorithm to obtain the energy efficiency optimization results.

[0062] In this embodiment, based on the irrigation adjustment results, a reinforcement learning algorithm is introduced to optimize the energy efficiency of the greenhouse. The energy efficiency optimization of the greenhouse involves temperature control and humidity regulation. Through reinforcement learning, the control strategy can be continuously adjusted to minimize energy consumption while maintaining the environmental conditions required for crop growth.

[0063] The reinforcement learning algorithm optimizes control decisions through interaction with the environment. In the optimization of greenhouse energy efficiency, the strategy is adjusted according to crop requirements. The system not only needs to consider the current environmental conditions but also make forward-looking adjustments based on future environmental predictions to achieve the lowest energy consumption and the best crop growth effect. The advantage of reinforcement learning lies in its ability to autonomously explore the optimal control strategy through continuous attempts and feedback, thereby significantly improving the energy use efficiency of greenhouse management while ensuring crop growth.

[0064] The reinforcement learning algorithm includes: State space. Multiple environmental factors of the greenhouse and the growth state of the crop are represented as a vector at each time step as follows: ; where is the temperature at time , in degrees Celsius; is the humidity at time , in %; is the air velocity at time , in m / s; is the light intensity at time , in lux; is the soil humidity at time , in %; is the crop growth state at time ; The conversion formula is: ; where is the output of the deep learning model, representing the growth prediction value of the crop at time step .

[0065] and is the current time step of the predicted crop growth value The minimum and maximum values can be obtained from historical data or experiments.

[0066] Then, the crop growth state is transformed. The normalized predicted growth value is used as to represent the growth state of the crop at a certain time step.

[0067] For the convenience of the reinforcement learning model to learn, it can be used as a scalar state value of crop growth for calculating rewards. For example, the following two methods can be defined to transform into the growth state value: Target processing (discretization): The growth state is mapped to high or low rewards, depending on whether the ideal crop growth state is reached. We set a threshold , when is greater than the threshold, it means the crop is in the ideal growth state and is given a high reward; otherwise, a low reward is given.

[0068] Reward function is defined as: ; This processing method helps the reinforcement learning model understand whether the crop is in the ideal growth state.

[0069] Weighted fusion processing (multi-dimensional fusion): If is a vector composed of multiple growth dimensions (such as growth rate, health status, expected yield, etc.), the weighted fusion method can be used to synthesize the growth prediction values of these dimensions into a comprehensive growth state value . The specific weighted fusion formula is: ; where respectively represent the growth rate, health status, and expected yield of the crop at time step .

[0070] are the weighting coefficients, indicating the contributions of different dimensions to the crop growth state.

[0071] is the state space vector.

[0072] Action space. It represents all executable operations, including temperature adjustment, humidity adjustment, and irrigation amount adjustment.

[0073] At each time step , construct a combined control strategy, expressed as a vector: ; where is the temperature control action, with heating being 1 and cooling being 0; is the humidity control action, with increasing being 1 and decreasing being 0; is the irrigation control action, with starting being 1 and stopping being 0; is the action space vector.

[0074] Reward function. The impact of the state space and action space on greenhouse energy efficiency and crop growth.

[0075] Reward function At each time step the formula is: ; where is the energy efficiency gain within the current time step i.e., the amount of energy saved; is the change in crop growth state within the current time step ; is the preset target temperature and humidity; is the preset weight coefficient.

[0076] Policy function. Determine the optimal action space taken in the state space; the policy function is equal to , the value update formula is: ; where is the current state and of value, representing the value of this state-action pair; is the preset learning rate, controlling the update step size, ranging from 0 to 1; is the preset discount factor, balancing the importance of current and future rewards, ranging from 0 to 1; represents the optimal action space value under the next state .

[0077] Update strategy. As the training progresses, the Q-values are continuously updated, and the policy function tends towards the optimal policy. In each interaction, the reinforcement learning model selects the optimal action space according to the current state and the learned policy to adjust the operating state of the greenhouse equipment and updates the decision-making policy according to the immediate reward function

[0078] Steps for using the model: Initialization: First, the initial environmental state of the greenhouse is given by the data collected by the sensors. According to the initial state, the reinforcement learning algorithm randomly selects an action from the action space .

[0079] Interaction process: At each time step, the model selects an action according to the current state (based on the current policy . Then, according to the executed action, the environmental state transfers to , and the reward is calculated . .

[0080] Q-value update: According to the immediate reward and the maximum Q-value of the next state. As the training progresses, the Q-values tend towards the optimal, thus gradually optimizing the policy.

[0081] Energy efficiency optimization: Through repeated interactions, the model learns the optimal equipment operation policy, maximizing the greenhouse energy efficiency at each time step and ensuring the growth state of the crops.

[0082] In the model: Calculate the change in crop growth state , The change in crop growth state represents the change in crop growth between time step and . To calculate , use the difference between the current crop growth state value and the crop growth state of the previous time step : ; where is the change in crop growth, indicating the growth change of the crop between time step and .

[0083] and ​They are the crop growth status values at the current time step and the previous time step respectively.

[0084] This change can be used as a feedback signal for the reinforcement learning model to adjust the operation strategies of greenhouse equipment (such as heaters, irrigation systems, etc.), so as to achieve the goal of optimizing crop growth and greenhouse energy efficiency.

[0085] S8. Provide decision support for the comprehensive management of the greenhouse based on the energy efficiency optimization results; Based on the energy efficiency optimization results, provide decision support for the comprehensive management of the greenhouse and generate a comprehensive report.

[0086] In this embodiment, by combining temperature, humidity, light, soil humidity, crop growth conditions, and energy efficiency optimization results, the system generates a comprehensive report to help managers make more scientific management decisions. The report content includes energy efficiency indicators, crop growth predictions, temperature and humidity adjustment suggestions, etc., to help greenhouse managers achieve refined management.

[0087] The comprehensive management decision support system of the greenhouse provides a one-stop decision support tool by integrating various types of data and optimization results. In practical applications, managers often face complex decision-making problems and need to adjust the management strategies of the greenhouse based on multiple factors. This system generates a detailed report by summarizing the output results of each module and provides specific operation suggestions. Ultimately, the system helps managers make more scientific and efficient management decisions under different growth cycles and climate change conditions.

[0088] Embodiment 2: This embodiment also provides a computer device applicable to a situation of an intelligent greenhouse temperature control method based on multi-source sensors, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement an intelligent greenhouse temperature control method based on multi-source sensors as proposed in the above embodiment.

[0089] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements an intelligent greenhouse temperature control method based on multi-source sensors as proposed in the above embodiment.

[0090] The computer device can be a terminal, which includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be implemented through WIFI, carrier networks, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0091] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, etc., which can store program codes.

[0092] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device.

[0093] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection (electronic device) having one or more wirings, a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable media can even be paper or other suitable media on which a program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or otherwise processing it as appropriate, and then storing it in a computer memory.

[0094] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, the multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well-known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), and the like.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An intelligent greenhouse temperature control method based on multi-source sensors, characterized in that: The following steps are involved: S1, data collection; Collect environmental data inside the greenhouse through multi-source sensors to obtain collected data; S2. Make a preliminary temperature prediction based on the collected data; Conduct preliminary temperature prediction on the collected data to obtain temperature prediction results; S3, obtaining a temperature control strategy according to the temperature prediction result; Design the temperature control strategy based on the temperature prediction results and obtain the temperature control strategy; S4. Adjust the greenhouse environment according to the temperature control strategy; According to the real-time data collected by multi-source sensors, the temperature and humidity data are compared with the temperature control strategy, and the temperature and humidity data are adjusted to obtain the temperature and humidity data; S5. Predict crop growth based on temperature and humidity data; Combined with temperature and humidity data, the deep learning model is used to predict crop growth and obtain crop growth prediction results; S6. Intelligent irrigation management based on crop growth prediction results; Based on the crop growth prediction results, the irrigation amount and irrigation time are dynamically adjusted to ensure that the water required by the crops is met, thereby obtaining the irrigation adjustment results; S7. Optimize greenhouse energy efficiency based on irrigation adjustment results; According to the irrigation adjustment results, the greenhouse energy efficiency is optimized through the reinforcement learning algorithm to obtain the energy efficiency optimization results; S8. Provide greenhouse comprehensive management decision support based on energy efficiency optimization results; Based on the energy efficiency optimization results, it provides decision support for the comprehensive management of the greenhouse and generates comprehensive reports.

2. The intelligent greenhouse temperature control method based on multi-source sensors according to claim 1 is characterized in that: The environmental data inside the greenhouse in step S1 includes: Temperature, humidity, air velocity, light intensity, soil moisture; data is collected through multi-source sensors.

3. The intelligent greenhouse temperature control method based on multi-source sensors according to claim 1 is characterized in that: The deep learning model in step S5 is based on: A multi-input neural network uses a weighted fusion mechanism to process and fuse five environmental parameters: temperature, humidity, air velocity, light intensity, and soil moisture.

4. The intelligent greenhouse temperature control method based on multi-source sensors according to claim 3 is characterized in that: The multi-input neural network is constructed in the following steps: First, obtain five environmental factor data from the environmental data and temperature and humidity data collected by multi-source sensors, including: temperature ,humidity , air velocity , light intensity , soil moisture ; Then construct a multi-input neural network including input layer, hidden layer, and output layer; Input Layer: Dynamic weighting of each input environmental factor Dynamically adjust according to the current environment state, and then construct the input layer formula: ， ； in, Represents the processing function REUL function of each environmental factor; when When it is greater than 0, the output value is equal to the input value. When it is less than or equal to 0, the output value is 0; is the dynamic weight of each environmental factor; is the original input data of environmental factors; is the weighted input of each environmental factor; Hidden Layer: Each input feature obtained from the input layer is fused through multiple hidden layers. The formula is: ； in, is the weight of each environmental factor in each hidden layer node; is the preset bias term, the translation of the model on the x-axis; is the output of the hidden layer; Indicates the number of nodes in the hidden layer; Output layer: Mapping the output of the hidden layer to the predicted growth value of the crop In the formula: ； in, is the final output of the model, i.e., the growth prediction value; is the activation function sigmoid function of the output layer, and its mathematical expression is ; is the weight of each hidden layer node; It is the bias term of the output layer, which allows the model to perform arbitrary translation on the x-axis to better fit the data.

5. The intelligent greenhouse temperature control method based on multi-source sensors according to claim 1 is characterized in that: The reinforcement learning algorithm in step S7 includes: state space and action space, reward function, strategy function, and update strategy.

6. The intelligent greenhouse temperature control method based on multi-source sensors according to claim 5 is characterized in that: The state space includes: Multiple environmental factors of the greenhouse and the growth status of crops at each time step The following is represented as a vector: ； in, , , , , , , respectively, time Temperature, humidity, air velocity, light intensity, soil moisture, and crop growth status; is the state space vector.

7. The intelligent greenhouse temperature control method based on multi-source sensors according to claim 5 is characterized in that: The action space includes: All operations that can be performed include temperature adjustment, humidity adjustment, and irrigation amount adjustment; At each time step , construct a combined control strategy, expressed as a vector: ； in, For temperature control action, heating is 1 and cooling is 0; For humidity control action, increase is 1 and decrease is 0; For irrigation control action, start is 1 and stop is 0; is the action space vector.

8. The intelligent greenhouse temperature control method based on multi-source sensors according to claim 5 is characterized in that: The reward function includes: The impact of state space and action space on greenhouse energy efficiency and crop growth; Reward Function At each time step The formula is: ； in, is the current time step Internal energy efficiency gain is the amount of energy saved; is the crop current time step The amount of change in the endogenous growth state; The preset target temperature and humidity; is the preset weight coefficient.

9. The intelligent greenhouse temperature control method based on multi-source sensors according to claim 5, characterized in that: The policy function is: Determine the optimal action space to take in the state space; the policy function is equal to , The update formula of the value is: ； in, For the current state and of Value, which indicates the value of the state-action; is the preset learning rate, which controls the update step size and ranges from 0 to 1; It is a preset discount factor that balances the importance of current rewards and future rewards, ranging from 0 to 1; Indicates the next state Next, the optimal action value.

10. The intelligent greenhouse temperature control method based on multi-source sensors according to claim 5, characterized in that: The update strategy includes: As the training progresses, The value is constantly updated, and the policy function tends to the optimal policy. In each interaction, the reinforcement learning model selects the optimal action space based on the current state and the learned policy. To adjust the operating status of the greenhouse equipment and according to the immediate reward function Update decision-making strategy.

Images