Accurate internet-of-things irrigation control system based on soil moisture content prediction

Abstract

The invention relates to the field of agricultural intelligent management, and discloses a precise Internet of Things irrigation control system based on soil moisture content prediction, and the system comprises a data collection module, a soil moisture modeling module, an optimal irrigation decision module, a reinforcement learning optimization module, and an intelligent irrigation execution module. The method comprises the steps that environment data are collected through an Internet of Things sensor, a soil moisture model is established through an Ito stochastic differential equation, transpiration and evaporation losses are calculated in combination with a Penman-Monteith equation, an optimal irrigation strategy is solved based on a stochastic HJB equation, deep reinforcement learning optimization is carried out, and finally irrigation is precisely controlled and the strategy is dynamically adjusted through an intelligent irrigation execution module. By combining the Internet of Things, edge calculation and deep reinforcement learning, precise irrigation control is realized, water resource utilization is optimized, network delay is reduced, response speed is improved, closed-loop optimization is formed, crop requirements are precisely matched, irrigation adaptability and flexibility are improved, excessive or insufficient irrigation is avoided, and irrigation efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent agricultural management, specifically to a precision IoT irrigation control system based on soil moisture prediction. Background Technology

[0002] With the increasing demand for water resources in agricultural production, traditional irrigation methods face problems such as water waste and uneven crop growth. Most existing irrigation systems rely on preset rules and simple sensor feedback, which cannot be flexibly adjusted according to real-time changes in soil moisture and different environmental factors. For example, some traditional irrigation control systems can only irrigate according to the predetermined irrigation time without taking into account the specific soil moisture conditions. This often leads to over-irrigation or under-irrigation, wasting a lot of water resources and failing to ensure the healthy growth of crops.

[0003] Furthermore, many existing technologies employ centralized data processing methods, requiring all data to be transmitted to a central server for computation and decision-making. This approach not only leads to data transmission delays but may also affect the real-time nature of irrigation decisions due to network instability. In large-scale agricultural production, timely response is crucial. If the system struggles to react quickly to changes in soil moisture, irrigation strategies may not be effectively adjusted, potentially causing crop water shortages or overwatering, impacting yield and quality.

[0004] Finally, traditional irrigation systems struggle to achieve comprehensive adaptive adjustments. While some systems can acquire and provide feedback on soil moisture data through sensors, these systems often lack robust data processing capabilities, making it difficult to intelligently optimize based on this feedback. Furthermore, many existing solutions fail to effectively integrate deep learning and reinforcement learning technologies, resulting in an inability to autonomously learn and optimize irrigation strategies in complex environments. Compared to the variability of the natural environment and crop growth needs, traditional systems typically cannot make precise and long-term effective decision adjustments. Therefore, this invention proposes a precise IoT-based irrigation control system based on soil moisture prediction to address the shortcomings of existing technologies. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a precise IoT irrigation control system based on soil moisture prediction, which solves the problems of slow response, inaccurate irrigation decisions, water waste, and difficulty in adapting to environmental changes in traditional irrigation methods.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a precision IoT irrigation control system based on soil moisture prediction, comprising: The data acquisition module is used to obtain soil moisture, ambient temperature, rainfall, and wind speed parameters, and transmit the collected data to the soil moisture modeling module. The soil moisture modeling module is used to construct a dynamic soil moisture model and predict soil moisture based on the parameters obtained by the data acquisition module, and provide the prediction results to the optimal irrigation decision module. The optimal irrigation decision module is used to calculate the optimal irrigation strategy based on the prediction results of the soil moisture modeling module, and then provide the calculation results to the reinforcement learning optimization module. The reinforcement learning optimization module is used to optimize irrigation decisions based on the calculation results of the optimal irrigation decision module using deep reinforcement learning methods, and then transmits the optimized decisions to the intelligent irrigation execution module. The intelligent irrigation execution module receives the optimization results from the reinforcement learning optimization module, controls the irrigation equipment to perform irrigation operations, and feeds back soil moisture change data after execution to the data acquisition module.

[0007] Preferably, the data acquisition module includes: A soil moisture acquisition unit is used to acquire soil moisture data at different depths, including the surface layer, middle layer and deep layer. The soil moisture acquisition unit includes a soil moisture sensor for measuring the moisture content of each soil layer. An environmental monitoring unit is used to acquire environmental meteorological parameters, and the environmental monitoring unit includes: An air temperature and humidity sensor is used to detect ambient temperature and relative humidity. Wind speed and direction sensors are used to detect wind speed and direction; Rain gauges are used to detect rainfall. A solar radiation sensor used to detect the intensity of solar radiation; The data transmission unit is used to send collected data to a remote server via LoRa, WiFi, or 5G wireless communication technologies.

[0008] Preferably, the soil moisture modeling module includes: The water dynamic modeling unit is used to construct a soil water evolution model based on Ito stochastic differential equations based on the acquired soil moisture data and environmental factors. The evaporation calculation unit is used to calculate the evaporation loss based on the Penman-Monteith equation. The evaporation calculation unit performs the calculation based on the temperature, humidity, wind speed, and radiation data collected by the data acquisition module. The data preprocessing unit is used to filter, remove outliers, and interpolate the soil moisture data acquired by the data acquisition module, and to normalize the data.

[0009] Preferably, the optimal irrigation decision module includes: The objective optimization calculation unit is used to construct the optimal control objective function to minimize irrigation costs and maintain soil moisture within the target range; The stochastic optimal control unit is used to calculate the optimal irrigation strategy based on the stochastic Hamilton-Jacobi-Bellman equation and provide the calculation results to the reinforcement learning optimization module.

[0010] Preferably, the reinforcement learning optimization module includes: The reinforcement learning environment modeling unit is used to define the state space, action space, and reward function of reinforcement learning based on the calculation results of the optimal irrigation decision module. The strategy optimization unit is used to optimize the optimal irrigation strategy using a deep deterministic strategy gradient method and adjust the strategy based on historical data and environmental feedback. The value function approximation unit is used to approximate the value function solution of the HJB equation based on a neural network and optimize the calculated optimal strategy based on reinforcement learning.

[0011] Preferably, the intelligent irrigation execution module includes: An intelligent irrigation control unit is used to control irrigation equipment to perform irrigation based on an optimized irrigation strategy, the irrigation equipment including a drip irrigation system, a sprinkler irrigation system, or a seepage irrigation system; The real-time monitoring and feedback unit is used to acquire soil moisture change data after execution and feed it back to the reinforcement learning optimization module through the data acquisition module to adjust subsequent decision-making strategies. Edge computing units are used to process part of the computing tasks locally, reducing data transmission latency and improving response speed.

[0012] Preferably, the soil moisture modeling module uses the following form of soil moisture evolution model for calculation: ; in, Soil moisture; For irrigation input, it depends on the control variables. ; Evaporation loss is affected by temperature, humidity, wind speed, and radiation. The input is rainfall, which follows a probability distribution. ; For the standard Wiener process, model environmental stochasticity; Noise intensity reflects the degree of random fluctuation.

[0013] Preferably, the optimal irrigation decision module calculates the optimal value function. Satisfying the following stochastic HJB equations: ; in, This is the partial derivative of the value function with respect to time; For irrigation input, it depends on the control variables. ; Evaporation loss is affected by temperature, humidity, wind speed, and radiation. The input is rainfall, which follows a probability distribution. ; It is the optimal value function; The function is the irrigation cost function; For soil moisture loss function; The partial derivatives of the value function; The second derivative of the value function; This is a coefficient related to system noise.

[0014] Preferably, the reinforcement learning environment modeling unit of the reinforcement learning optimization module uses a reward function of the following form for policy optimization: ; in, As a reward value; For irrigation costs; The soil moisture loss function minimizes water resource consumption and keeps soil moisture close to the target value.

[0015] This invention also provides a precise IoT irrigation control method based on soil moisture prediction, comprising the following steps: S1. Collect environmental data such as soil moisture, air temperature, wind speed, and rainfall through IoT sensors; S2. A dynamic model of soil moisture was established based on the Ito stochastic differential equation, and evaporation loss was calculated based on the Penman-Monteith equation. S3. Calculate the optimal irrigation strategy based on the stochastic HJB equation; S4. Optimal irrigation strategy is calculated using deep reinforcement learning methods; S5. The intelligent irrigation execution module controls the irrigation equipment to perform irrigation, and optimizes subsequent decision-making strategies through real-time monitoring and feedback.

[0016] This invention provides a precision IoT irrigation control system based on soil moisture prediction. It has the following beneficial effects: 1. This invention employs a precision IoT irrigation control system, combining an optimal irrigation decision-making module with a reinforcement learning optimization module to dynamically adjust irrigation strategies based on soil moisture, meteorological data, and historical feedback. This technical solution achieves the effects of optimizing water resource utilization and improving the accuracy of irrigation decisions. Compared to existing irrigation control methods based on rules or preset models, this invention can more flexibly adapt to environmental changes and crop needs, avoiding water waste and crop growth problems caused by fixed rules in traditional methods.

[0017] 2. This invention introduces an edge computing unit to perform local computing tasks, reducing data transmission latency and significantly improving system response speed. By performing calculations locally, the edge computing unit can analyze soil moisture changes more quickly and adjust irrigation volume in real time. Compared with existing systems that rely solely on remote computing and centralized processing, this invention effectively avoids response delays caused by network latency, thereby improving the real-time performance and accuracy of irrigation execution.

[0018] 3. This invention feeds back soil moisture data after irrigation to the reinforcement learning optimization module through a real-time monitoring and feedback unit, promoting continuous optimization of the irrigation strategy. This technical solution achieves closed-loop control, which can continuously adjust subsequent decisions based on the actual irrigation effect, ensuring that the soil moisture is always kept within the most suitable range. Compared with the traditional irrigation method of manual monitoring and adjustment, this invention significantly improves irrigation efficiency and reduces the impact of human error.

[0019] 4. The intelligent irrigation execution module of the present invention can precisely control drip irrigation, sprinkler irrigation and seepage irrigation equipment, ensuring that the irrigation volume is highly matched with the actual needs, avoiding the problems of over-irrigation or under-irrigation; compared with the general irrigation control system in the prior art, the present invention adopts a refined control technical solution, which greatly improves the adaptability and flexibility of irrigation equipment, meets the irrigation needs of different crops and different environmental conditions, and optimizes the efficiency of water resource use. Attached Figure Description

[0020] Figure 1 This is a system architecture diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 This invention provides a precision IoT irrigation control system based on soil moisture prediction, comprising: The data acquisition module is used to obtain soil moisture, ambient temperature, rainfall, and wind speed parameters, and transmit the collected data to the soil moisture modeling module. In the precision IoT irrigation control system of this invention, the data acquisition module plays a crucial role, providing the system with basic data on various environmental and soil information. This data not only helps to monitor soil moisture and meteorological conditions in real time, but also provides reliable input for subsequent soil moisture modeling and irrigation decision-making modules. Therefore, the accuracy, real-time performance, and effective integration with other modules of the data acquisition module determine the efficiency and sustainability of the entire irrigation system.

[0023] In this embodiment, the data acquisition module consists of multiple sub-units, including a soil moisture acquisition unit, an environmental monitoring unit, and a data transmission unit. Each unit has a unique function, and they work together to ensure that the system can accurately and in real time acquire the required data. The following is a detailed description of each unit and its function.

[0024] Generally, the soil moisture acquisition unit is the core component of the data acquisition module, responsible for monitoring soil moisture at different depths. This soil moisture data has a significant impact on subsequent soil moisture modeling and irrigation decisions. Specifically, the soil moisture acquisition unit includes multiple soil moisture sensors, which are deployed in the surface, middle, and deep layers to obtain multi-dimensional soil moisture data. Through this setup, the system can comprehensively assess the soil moisture status, especially when there are significant differences in moisture distribution across multiple soil layers, providing more accurate information.

[0025] In one possible implementation, the soil moisture sensor employs a capacitive sensor, which reflects moisture content by measuring changes in the dielectric constant of the soil; changes in soil moisture affect the dielectric constant of the soil, thereby changing the capacitance value of the sensor; based on the change in capacitance value, the system can calculate the change in soil moisture.

[0026] Specifically, soil moisture The calculation can be performed as follows: ; in, This represents the percentage of soil moisture. The measured capacitance value; This represents the capacitance value of dry soil. This represents the maximum capacitance value of saturated soil.

[0027] These soil moisture data will be used to build a dynamic model of soil moisture, providing real-time feedback for irrigation decisions.

[0028] The environmental monitoring unit is used to acquire environmental meteorological parameters, including data on temperature, humidity, wind speed, wind direction, precipitation, and solar radiation. This environmental data is crucial for irrigation decisions and soil moisture change prediction. This unit includes the following sub-sensors: Air temperature and humidity sensor: used to detect ambient temperature and relative humidity These data directly affect evaporation and crop transpiration, thus influencing changes in soil moisture; specifically, evaporation... The following formula can be used for calculation: ; in, Evaporation rate, in mm / d; Temperature, in °C; The reference temperature is expressed in °C. This is relative humidity, expressed as a percentage. This is the evaporation coefficient, which depends on climatic conditions.

[0029] Wind speed and direction sensor: used to monitor wind speed Wind direction and wind speed are important factors in calculating evaporation rates. Higher wind speeds generally result in higher evaporation rates, so irrigation amounts need to be adjusted based on wind speed.

[0030] Rain star sensor: used to detect precipitation. This sensor helps the system determine whether irrigation is needed and can automatically adjust the irrigation plan when rainfall reaches a preset threshold, thus avoiding water waste; the output of the rainfall sensor The relationship with actual precipitation can be expressed as: ; in, In time Rainfall at any given time; For the first Rainfall amount within a time period, in mm; This represents the number of precipitation measurement points.

[0031] Solar radiation sensor: used to monitor solar radiation intensity Solar radiation affects evaporation rates, therefore this data is crucial for optimizing irrigation strategies; the relationship between solar radiation and evaporation rate can be expressed by the following formula: ; in, Evaporation rate, in mm / d; Solar radiation intensity, expressed in W / m² 2 ; RH is a constant that depends on local climate conditions; RH is relative humidity, expressed as a percentage.

[0032] In this embodiment, the data transmission unit uses wireless communication technology (such as LoRa, WiFi, or 5G) to transmit all collected data to a remote server for processing. The data transmission unit can support large-scale real-time data collection and transmission, ensuring efficient system operation and real-time data updates. Specifically, the collected data will be sent to the cloud or edge computing unit in real time via wireless network. Subsequently, through data processing and analysis, it will provide the necessary real-time data input for the soil moisture modeling module and irrigation decision module.

[0033] To ensure system reliability and data accuracy, the data acquisition module is equipped with an anomaly detection mechanism. When a sensor malfunctions or outputs abnormally, the system can immediately detect it and automatically verify the data, ensuring that the system is not affected by a single sensor failure. For example, if the soil moisture sensor reads data outside the normal range, the system will activate a redundancy mechanism to estimate soil moisture using other soil moisture sensors or meteorological data, ensuring data integrity and the accuracy of irrigation decisions.

[0034] The soil moisture modeling module is used to construct a dynamic soil moisture model and predict soil moisture based on the parameters obtained by the data acquisition module, and provide the prediction results to the optimal irrigation decision module. In the precision IoT irrigation control system, the soil moisture modeling module plays a core role. It is responsible for dynamic modeling of soil moisture, calculation of transpiration and evaporation loss, and data preprocessing based on information such as soil moisture and environmental meteorological data obtained from the data acquisition module. These processing results provide accurate data support for the subsequent irrigation decision module, ensuring that the irrigation system can adjust the irrigation strategy in real time according to crop needs and environmental conditions, avoiding water waste and ensuring the healthy growth of crops.

[0035] In this embodiment, the soil moisture modeling module consists of multiple sub-units, including a moisture dynamic modeling unit, a transpiration and evaporation calculation unit, and a data preprocessing unit; each unit has different functions and works together to complete the task of predicting and calculating soil moisture.

[0036] Generally, soil moisture changes are affected by a variety of factors, including precipitation, evaporation, crop transpiration, and irrigation. In order to accurately predict soil moisture changes, the water dynamic modeling unit in this embodiment adopts a soil moisture evolution model based on the Ito stochastic differential equation. By considering the random fluctuations of soil moisture and changes in environmental factors, this model can simulate the stochastic evolution process of soil moisture.

[0037] In one possible implementation, the water dynamics modeling unit describes the process of soil moisture change through the following Ito stochastic differential equation: ; in, Indicates soil moisture over time Changes; The amount of irrigation input depends on the irrigation decisions made by the soil moisture control system. ; Evaporation and transpiration losses depend on soil moisture. ; This refers to precipitation input, expressed in mm / d. The intensity of moisture fluctuations; This represents standard Brownian motion (Wiener process), indicating the influence of random environmental factors.

[0038] This model can adjust based on soil moisture. By dynamically adjusting the evolution path of soil moisture in response to changes in external meteorological inputs (such as precipitation and evaporation), accurate prediction of soil moisture can be achieved.

[0039] As an alternative, the transpiration evaporation calculation unit uses the Penman-Monteith equation to calculate transpiration evaporation loss. The Penman-Monteith equation is widely used in meteorology and agricultural science and is an accurate model for calculating plant transpiration and soil moisture evaporation. This equation comprehensively considers various meteorological factors such as temperature, humidity, wind speed, and radiation. In this embodiment, the transpiration evaporation is calculated based on the temperature, humidity, wind speed, and radiation data collected by the data acquisition module.

[0040] Specifically, the Penman-Monteith equation is expressed as follows: in, Total evapotranspiration, in mm / d; The slope of the saturated water vapor pressure curve is expressed in kPa / ℃. Net radiation, measured in MJ / m² 2 d; Soil heat flux density, in MJ / m³ 2 ·d; Air density, unit: kg / m³ 3 ; Specific heat capacity, expressed in MJ / kg·K; The wind speed at a height of 2 meters is expressed in m / s. This is the saturated water vapor pressure, expressed in kPa. This is the actual water vapor pressure, in kPa. This is the dry-to-humid air ratio, expressed in kPa / ℃.

[0041] Using this equation, the evaporation calculation unit can accurately calculate the evaporation and transpiration loss of soil moisture, thereby providing the necessary evaporation loss data for subsequent soil moisture evolution models.

[0042] To ensure the accuracy and usability of the data, the data preprocessing unit processes the soil moisture data acquired from the data acquisition module. Specifically, this unit includes steps such as filtering, outlier removal, interpolation correction, and normalization. These steps effectively remove noise and outliers from the sensor data, ensuring the data quality used in subsequent modeling processes.

[0043] Filtering: Use smoothing algorithms (such as moving average or low-pass filter) to remove high-frequency noise from the data.

[0044] Outlier removal: Detects and removes extreme or erroneous values ​​from soil moisture data to ensure data accuracy.

[0045] Interpolation correction: For data missing due to sensor failure or other reasons, interpolation algorithms (such as linear interpolation, spline interpolation, etc.) are used to supplement the data.

[0046] Normalization: Normalizing data typically involves scaling the data to a standard range (such as 0 to 1) so that data from different sources can be processed on the same scale.

[0047] Through these preprocessing steps, the data preprocessing unit provides clean, standardized data input for subsequent soil moisture modeling and evaporation calculations.

[0048] The optimal irrigation decision module is used to calculate the optimal irrigation strategy based on the prediction results of the soil moisture modeling module, and then provide the calculation results to the reinforcement learning optimization module. In a precision IoT irrigation control system, the optimal irrigation decision module is a core component, primarily responsible for formulating the optimal irrigation strategy based on dynamic changes in soil moisture and environmental factors. This module calculates the optimal irrigation amount based on soil moisture data, transpiration and evaporation calculations, and other environmental parameters provided by the soil moisture modeling module, thereby achieving the goals of water conservation and crop growth promotion. By combining control theory and reinforcement learning, the optimal irrigation decision module can dynamically adjust and make decisions based on different environmental and crop needs.

[0049] The optimal irrigation decision module in this embodiment mainly consists of an objective optimization calculation unit, a stochastic optimal control unit, and a reinforcement learning optimization module. The specific workflow is as follows: The goal of the optimal irrigation decision module is to optimize the cost-effectiveness of the irrigation process while maintaining soil moisture within the optimal range. To this end, the objective optimization calculation unit constructs a control objective function that comprehensively considers the impact of irrigation costs and soil moisture on crop growth. Specifically, the objective function... It can be represented as: ; in, Let be the objective function, representing the total cost over the entire irrigation cycle; Let be the irrigation cost function, representing the irrigation amount. The resulting costs; Let be the soil moisture loss function, representing soil moisture. The losses incurred when the target value is deviated from; For irrigation input (control variable); Soil moisture; This represents the total duration of the irrigation cycle.

[0050] In some embodiments, the irrigation cost function Modeling can be performed based on factors such as water resource usage costs and energy consumption, while the soil moisture loss function... This indicates the deviation between soil moisture and the target range, such as losses caused by excessively high or low soil moisture.

[0051] In this embodiment, the stochastic optimal control unit calculates the optimal irrigation strategy based on the stochastic Hamilton-Jacobi-Bellman (HJB) equations. The HJB equations are a type of control equations widely used in dynamic system optimization, which can calculate the optimal control strategy according to a given control objective.

[0052] Specifically, the optimal value function Satisfying the following stochastic HJB equations: ; in, This is the partial derivative of the value function with respect to time; For irrigation input, it depends on the control variables. ; Evaporation loss is affected by temperature, humidity, wind speed, and radiation. The input is rainfall, which follows a probability distribution. ; It is the optimal value function; The function is the irrigation cost function; For soil moisture loss function; The partial derivatives of the value function; The second derivative of the value function; This is a coefficient related to system noise.

[0053] By solving the HJB equations above, the optimal irrigation decision module can obtain an optimal irrigation strategy. This strategy provides an optimal irrigation input at each moment, minimizing the total cost over the entire irrigation cycle and ensuring that soil moisture remains within the target range.

[0054] To further improve the effectiveness of irrigation decisions, a reinforcement learning optimization module is introduced in this embodiment. Reinforcement learning is a machine learning method based on environmental interaction for optimization, which can continuously adjust the strategy to adapt to the dynamically changing environment. The reinforcement learning optimization module uses the solution of the HJB equation as the initial strategy and optimizes the strategy online based on the feedback of the actual environment (such as changes in soil moisture, evaporation and transpiration losses, etc.).

[0055] In this implementation, reinforcement learning algorithms update irrigation strategies through multiple rounds of interaction with the environment to achieve a more refined and efficient irrigation decision. For example, by employing Q-learning or deep reinforcement learning methods, the system can continuously adjust the irrigation amount. This allows for more precise irrigation control.

[0056] Specifically, the reinforcement learning optimization module optimizes based on the interaction between the agent and the environment; at each moment, the agent optimizes according to the current soil moisture. Based on environmental and meteorological data and historical decision-making, select an irrigation volume. As an action; then, the environment adjusts according to the irrigation volume. Soil moisture is updated and a reward is given in response to other external factors (such as precipitation, evaporation, etc.); this reward is usually defined based on the deviation between soil moisture and target moisture, for example: ; in, Let be the reward function, representing the effect of the current irrigation decision; Target soil moisture; and These are weighting coefficients, which control the relative importance of humidity deviation and irrigation amount, respectively.

[0057] Through multiple interactions, reinforcement learning algorithms can learn the optimal irrigation strategy to minimize irrigation costs and soil moisture loss.

[0058] The reinforcement learning optimization module is used to optimize irrigation decisions based on the calculation results of the optimal irrigation decision module using deep reinforcement learning methods, and then transmits the optimized decisions to the intelligent irrigation execution module. In the precision IoT irrigation control system, the reinforcement learning optimization module, as a core component, further optimizes irrigation decisions. This module uses reinforcement learning methods to continuously adjust irrigation strategies based on system feedback, ensuring that soil moisture is maintained within an ideal range while minimizing water waste. The reinforcement learning optimization module works closely with other modules (such as the optimal irrigation decision module) to achieve precise and sustainable irrigation control.

[0059] The reinforcement learning optimization module in this embodiment includes a reinforcement learning environment modeling unit, a policy optimization unit, and a value function approximation unit; the following is a detailed description of each sub-unit of this module: Generally, the first step of the reinforcement learning optimization module is to model the environment through the reinforcement learning environment modeling unit so that it can make decisions in a given state space and action space. In this embodiment, the core task of the environment modeling unit is to define the state space, action space and reward function of reinforcement learning based on the calculation results of the optimal irrigation decision module.

[0060] state space Indicates the system at a certain moment The state; the state typically includes soil moisture. Environmental meteorological data (such as temperature) ,humidity Wind speed Parameters such as these affect crop growth and water requirements; specifically, the state space... It can be represented as: ; in, This represents the current soil moisture level. Temperature; Air humidity; This refers to wind speed.

[0061] Action space This indicates the amount of irrigation that can be carried out under the current conditions. This refers to the range of irrigation amounts that the system can select.

[0062] reward function Used to evaluate in a given state and the selected irrigation amount The effect of taking a certain action; in this embodiment, the reward function is designed as follows: ; in, As a reward value; For irrigation costs; The soil moisture loss function aims to minimize water resource consumption while maintaining soil moisture close to the target value. This loss can be defined as follows: ; in, Target soil moisture; This is the loss weighting coefficient, used to adjust the impact of humidity deviation on the loss.

[0063] As a core component of the reinforcement learning optimization module, the policy optimization unit uses the Deep Deterministic Policy Gradient (DDPG) method to optimize the optimal irrigation policy. The Deep Deterministic Policy Gradient method is a reinforcement learning algorithm based on an actor-critic architecture, which is suitable for handling optimization problems in continuous action spaces.

[0064] Specifically, during the policy optimization process, the policy network (i.e., the actors) adjusts the policy based on the current state. Output irrigation control quantity The value network (i.e., the commentator) evaluates the long-term reward of each state-action pair; in order to optimize irrigation decisions, the policy network continuously adjusts based on the feedback from the value network, thereby finding the optimal irrigation strategy.

[0065] In one possible implementation, the policy update is performed using the following formula: in, These are the parameters of the policy network; The learning rate; Let be the objective function, representing the performance of the current policy.

[0066] By continuously updating its strategies, the system can optimize irrigation volume. To meet the needs of crops and minimize water consumption.

[0067] The value function approximation unit uses a neural network to approximate the value function solution of the HJB equation; the HJB equation is an important equation in control theory, used to describe the optimal control strategy in dynamic optimization problems; in this embodiment, the value function... Indicates the soil moisture content at a given level. and time The optimal expected cost is determined by the HJB equation; since the HJB equation is generally difficult to solve directly, neural networks are used to approximate the solution of the value function.

[0068] The goal of a neural network is to minimize the following loss function to approximate the true value function: in, The value function estimated by the neural network; It is a true value function; These are the parameters of the neural network.

[0069] By minimizing this loss function, the neural network progressively optimizes its parameters, making the output value function closer to the true solution, thereby supporting the optimization of the optimal irrigation strategy.

[0070] The intelligent irrigation execution module is used to receive the optimization results of the reinforcement learning optimization module, control the irrigation equipment to perform irrigation operations, and at the same time feed back the soil moisture change data after execution to the data acquisition module. In the precision IoT irrigation control system, the intelligent irrigation execution module serves as the core component for implementing irrigation decisions. It ensures precise control of irrigation equipment based on optimized strategies and makes adjustments based on real-time monitoring data, ultimately achieving rational utilization of water resources and healthy crop growth. The design of this module not only considers the precision of irrigation control but also improves the system's response speed and computing efficiency through edge computing units.

[0071] The intelligent irrigation execution module in this embodiment mainly consists of an intelligent irrigation control unit, a real-time monitoring and feedback unit, and an edge computing unit. These units are described in detail below.

[0072] Generally, the main function of an intelligent irrigation control unit is to control the execution of irrigation equipment based on the irrigation strategy calculated by the reinforcement learning optimization module. Irrigation equipment includes drip irrigation systems, sprinkler irrigation systems, and seepage irrigation systems. These devices can precisely adjust the irrigation volume and time to ensure that the soil moisture is maintained within the target range.

[0073] Specifically, the irrigation control unit optimizes the irrigation strategy. The irrigation equipment is regulated, among which This indicates the optimal irrigation volume at each moment; the control unit needs to precisely control the on / off states of equipment such as water pumps and valves to achieve different irrigation volumes.

[0074] For example, in a drip irrigation system, the irrigation volume The irrigation volume can be controlled by adjusting the drip rate; for sprinkler systems, the irrigation volume can be achieved by adjusting the switching frequency of the sprinklers and the water flow rate; the output of the irrigation control unit determines the specific execution mode of the irrigation equipment, ensuring the accuracy and timeliness of the irrigation volume.

[0075] During irrigation, the main function of the real-time monitoring and feedback unit is to monitor changes in soil moisture after irrigation and feed the monitoring data back to the reinforcement learning optimization module. In this way, the system can evaluate the effectiveness of the irrigation strategy in real time and adjust subsequent irrigation strategies based on the feedback.

[0076] Typically, the real-time monitoring feedback unit monitors soil moisture in real time through sensors (such as soil moisture sensors). The changes; the monitoring data can include multiple environmental parameters such as soil moisture, temperature, and humidity; these data will be transmitted to the system through the data acquisition module as input to the reinforcement learning optimization module.

[0077] For example, when the soil moisture is after irrigation Achieve target humidity When the humidity deviation is large, the feedback data will indicate that the current irrigation strategy is effective; if the humidity deviation is large, the feedback information will guide the adjustment of subsequent irrigation strategies and gradually optimize the efficiency of water resource use.

[0078] In some embodiments, the real-time monitoring feedback unit may monitor not only soil moisture but also meteorological data (such as temperature). ,humidity Wind speed This information, including data such as water quality, has a significant impact on irrigation decisions.

[0079] The main function of edge computing units is to reduce data transmission latency and improve system response speed. In the Internet of Things (IoT) environment, the real-time nature of data is crucial for irrigation control, especially in agricultural production, where rapid response can better cope with dynamic environmental changes.

[0080] Typically, edge computing units are located in local facilities within the irrigation system and can handle some computational tasks locally, such as real-time analysis of soil moisture data and adjustment of irrigation strategies. This reduces the latency of data transmission to the central server and improves response efficiency.

[0081] For example, after an irrigation, the edge computing unit can quickly analyze changes in soil moisture and send adjustment commands to the irrigation control unit in real time to control the amount or timing of the next irrigation. Through edge computing, the system can make rapid decisions without network latency, thereby improving overall irrigation efficiency.

[0082] Edge computing units can also process equipment data, such as analyzing operational data of water pumps and sprinklers, to ensure the normal operation of equipment and further optimize irrigation strategies.

[0083] Please see Figure 2 The present invention also provides a precise IoT irrigation control method based on soil moisture prediction, comprising the following steps: S1. Collect environmental data such as soil moisture, air temperature, wind speed, and rainfall through IoT sensors and upload them to the edge computing unit in real time to ensure the timeliness and completeness of the data. At the same time, combine historical data to perform trend analysis to improve the adaptability of irrigation strategies. S2. A dynamic soil moisture model is established based on the Ito stochastic differential equation, taking into account factors such as soil infiltration characteristics, precipitation recharge, and evaporation loss. The Penman-Monteith equation is also used to calculate evaporation loss, so as to accurately assess the trend of soil moisture change and provide a scientific basis for irrigation decisions. S3. Calculate the optimal irrigation strategy based on the stochastic HJB equation, and use dynamic programming to solve the long-term optimal water resource allocation scheme to ensure that crops achieve the best growth state under different environmental conditions and climate changes, while taking into account the balance between water conservation and efficient irrigation. S4. The optimal irrigation strategy is optimized by using deep reinforcement learning. Based on the interaction process between the reinforcement learning agent and the environment, the irrigation plan is dynamically adjusted by continuously learning historical data and real-time feedback, so that the irrigation strategy can adapt to different soil types, crop growth stages and external weather conditions, thereby improving the system's adaptability and generalization ability. S5. The intelligent irrigation execution module controls the irrigation equipment to perform irrigation, ensuring precise water delivery. The real-time monitoring and feedback unit continuously collects data on soil moisture changes and equipment operating status, compares the expected and actual results, optimizes subsequent decision-making strategies, and forms a closed-loop optimization of the entire irrigation process, improving the intelligence and precision of irrigation.

[0084] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A precision IoT irrigation control system based on soil moisture prediction, characterized in that, include: The data acquisition module is used to obtain soil moisture, ambient temperature, rainfall, and wind speed parameters, and transmit the collected data to the soil moisture modeling module. The soil moisture modeling module is used to construct a dynamic soil moisture model and predict soil moisture based on the parameters obtained by the data acquisition module, and provide the prediction results to the optimal irrigation decision module. The optimal irrigation decision module is used to calculate the optimal irrigation strategy based on the prediction results of the soil moisture modeling module, and then provide the calculation results to the reinforcement learning optimization module. The reinforcement learning optimization module is used to optimize irrigation decisions based on the calculation results of the optimal irrigation decision module using deep reinforcement learning methods, and then transmits the optimized decisions to the intelligent irrigation execution module. The intelligent irrigation execution module receives the optimization results from the reinforcement learning optimization module, controls the irrigation equipment to perform irrigation operations, and feeds back soil moisture change data after execution to the data acquisition module.

2. The precision IoT irrigation control system based on soil moisture prediction according to claim 1, characterized in that, The data acquisition module includes: A soil moisture acquisition unit is used to acquire soil moisture data at different depths, including the surface layer, middle layer and deep layer. The soil moisture acquisition unit includes a soil moisture sensor for measuring the moisture content of each soil layer. An environmental monitoring unit is used to acquire environmental meteorological parameters, and the environmental monitoring unit includes: An air temperature and humidity sensor is used to detect ambient temperature and relative humidity. Wind speed and direction sensors are used to detect wind speed and direction; Rain gauges are used to detect rainfall. A solar radiation sensor used to detect the intensity of solar radiation; The data transmission unit is used to send collected data to a remote server via LoRa, WiFi, or 5G wireless communication technologies.

3. The precision IoT irrigation control system based on soil moisture prediction according to claim 1, characterized in that, The soil moisture modeling module includes: The water dynamic modeling unit is used to construct a soil water evolution model based on Ito stochastic differential equations based on the acquired soil moisture data and environmental factors. The evaporation calculation unit is used to calculate the evaporation loss based on the Penman-Monteith equation. The evaporation calculation unit performs the calculation based on the temperature, humidity, wind speed, and radiation data collected by the data acquisition module. The data preprocessing unit is used to filter, remove outliers, and interpolate the soil moisture data acquired by the data acquisition module, and to normalize the data.

4. The precision IoT irrigation control system based on soil moisture prediction according to claim 1, characterized in that, The optimal irrigation decision module includes: The objective optimization calculation unit is used to construct the optimal control objective function to minimize irrigation costs and maintain soil moisture within the target range; The stochastic optimal control unit is used to calculate the optimal irrigation strategy based on the stochastic Hamilton-Jacobi-Bellman equation and provide the calculation results to the reinforcement learning optimization module.

5. The precision IoT irrigation control system based on soil moisture prediction according to claim 1, characterized in that, The reinforcement learning optimization module includes: The reinforcement learning environment modeling unit is used to define the state space, action space, and reward function of reinforcement learning based on the calculation results of the optimal irrigation decision module. The strategy optimization unit is used to optimize the optimal irrigation strategy using a deep deterministic strategy gradient method and adjust the strategy based on historical data and environmental feedback. The value function approximation unit is used to approximate the value function solution of the HJB equation based on a neural network and optimize the calculated optimal strategy based on reinforcement learning.

6. The precision IoT irrigation control system based on soil moisture prediction according to claim 1, characterized in that, The intelligent irrigation execution module includes: An intelligent irrigation control unit is used to control irrigation equipment to perform irrigation based on an optimized irrigation strategy, the irrigation equipment including a drip irrigation system, a sprinkler irrigation system, or a seepage irrigation system; The real-time monitoring and feedback unit is used to acquire soil moisture change data after execution and feed it back to the reinforcement learning optimization module through the data acquisition module to adjust subsequent decision-making strategies. Edge computing units are used to process part of the computing tasks locally, reducing data transmission latency and improving response speed.

7. The precision IoT irrigation control system based on soil moisture prediction according to claim 1, characterized in that, The soil moisture modeling module uses the following form of soil moisture evolution model for calculation: ； in, Soil moisture; For irrigation input, it depends on the control variables. ; Evaporation loss is affected by temperature, humidity, wind speed, and radiation. The input is rainfall, which follows a probability distribution. ; For the standard Wiener process, model environmental stochasticity; Noise intensity reflects the degree of random fluctuation.

8. The precision IoT irrigation control system based on soil moisture prediction according to claim 1, characterized in that, The optimal irrigation decision module calculates the optimal value function. Satisfying the following stochastic HJB equations: ； in, This is the partial derivative of the value function with respect to time; For irrigation input, it depends on the control variables. ; Evaporation loss is affected by temperature, humidity, wind speed, and radiation. The input is rainfall, which follows a probability distribution. ; It is the optimal value function; The function is the irrigation cost function; For soil moisture loss function; The partial derivatives of the value function; The second derivative of the value function; This is a coefficient related to system noise.

9. The precision IoT irrigation control system based on soil moisture prediction according to claim 1, characterized in that, The reinforcement learning environment modeling unit of the reinforcement learning optimization module uses the following reward function for policy optimization: ； in, As a reward value; For irrigation costs; The soil moisture loss function minimizes water resource consumption and keeps soil moisture close to the target value.

10. A precise IoT irrigation control method based on soil moisture prediction, applied to the precise IoT irrigation control system based on soil moisture prediction as described in any one of claims 1-9, characterized in that, Includes the following steps: S1. Collect environmental data such as soil moisture, air temperature, wind speed, and rainfall through IoT sensors; S2. A dynamic model of soil moisture was established based on the Ito stochastic differential equation, and evaporation loss was calculated based on the Penman-Monteith equation. S3. Calculate the optimal irrigation strategy based on the stochastic HJB equation; S4. Optimal irrigation strategy is calculated using deep reinforcement learning methods; S5. The intelligent irrigation execution module controls the irrigation equipment to perform irrigation, and optimizes subsequent decision-making strategies through real-time monitoring and feedback.

Images