Intelligent irrigation strategy formulation method and system

Abstract

The present invention relates to the technical field of agricultural irrigation. Disclosed are an intelligent irrigation strategy formulation method and system. The method comprises: determining the current irrigation strategy on the basis of a preset irrigation decision-making model and a Q-learning intelligent agent and according to the current agricultural state; on the basis of the current irrigation strategy, performing irrigation simulation on a target growing region to obtain a corresponding irrigation reward value and the next agricultural state, wherein the irrigation reward value is determined on the basis of the crop yield, the crop water demand, and the annual economic cost after irrigation, the current agricultural state, the current irrigation strategy, the irrigation reward value, and the next agricultural state form an empirical four-tuple, and a plurality of empirical four-tuples form an experience pool; and randomly sampling the empirical four-tuple from the experience pool as a training sample to train the Q-learning intelligent agent so as to obtain an optimal irrigation Q-value table which is used for determining a corresponding optimal irrigation strategy on the basis of any agricultural state of the target growing region. The present invention can efficiently obtain an accurate irrigation strategy.

Description

A smart irrigation strategy formulation method and system Technical Field

[0001] The present invention relates to the field of agricultural irrigation technology, and in particular to a method and system for formulating an intelligent irrigation strategy. Background Art

[0002] In the field of agricultural irrigation information technology, intelligent algorithms such as radial basis function (RBF) neural networks, decision trees, and model predictive control (MPC) have facilitated the design and optimization of irrigation systems. However, given the inherent complexity of irrigation system control, developing a simplified yet accurate system model requires the reliance on a rich historical dataset. While simplified models may be more convenient to apply, oversimplification can reduce the model's ability to accurately describe system dynamics, leading to a discrepancy between control outputs and established objectives. As a typical model-free learning strategy, reinforcement learning (RL) algorithms offer a methodology for solving complex irrigation system optimization problems. Given RL's strong adaptive capabilities in adapting to complex environments and learning control, this algorithm exhibits significant potential for complex models such as irrigation systems. Therefore, further research on RL algorithms is necessary to optimize intelligent decision-making in irrigation systems and promote their integration and collaborative work in practical applications.

[0003] Summary of the Invention

[0004] The purpose of the present invention is to provide a smart irrigation strategy formulation method and system, which can efficiently obtain accurate irrigation strategies.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] In a first aspect, the present invention provides a method for formulating a smart irrigation strategy, comprising:

[0007] Obtaining the current agricultural status of the target planting area; the current agricultural status includes meteorological data, soil data, crop parameters and field management data;

[0008] Determine the current irrigation strategy based on the current agricultural status based on the preset irrigation decision model and the Q-learning agent;

[0009] performing an irrigation simulation on the target planting area based on the current irrigation strategy to obtain a corresponding irrigation reward value and a next agricultural state; the irrigation reward value is determined based on the crop yield after irrigation, the crop water requirement, and the annual economic cost; the current agricultural state, the current irrigation strategy, the irrigation reward value, and the next agricultural state constitute an experience quadruple; and a plurality of the experience quadruples constitute an experience pool;

[0010] Randomly sampling experience quadruples from the experience pool as training samples, the Q-learning agent is trained to obtain an optimal irrigation Q-value table; the optimal irrigation Q-value table is used to determine the corresponding optimal irrigation strategy according to any agricultural state of the target planting area.

[0011] In a second aspect, the present invention provides a smart irrigation strategy formulation system, comprising:

[0012] A current state acquisition module is used to obtain the current agricultural state of the target planting area; the current agricultural state includes meteorological data, soil data, crop parameters and field management data;

[0013] a current irrigation strategy determination module, configured to determine the current irrigation strategy according to the current agricultural status based on a preset irrigation decision model and a Q-learning agent;

[0014] An experience pool construction module is configured to perform irrigation simulation on the target planting area based on the current irrigation strategy to obtain a corresponding irrigation reward value and a next agricultural state; the irrigation reward value is determined based on the crop yield after irrigation, the crop water requirement, and the annual economic cost; the current agricultural state, the current irrigation strategy, the irrigation reward value, and the next agricultural state constitute an experience quadruple; and a plurality of such experience quadruples constitute an experience pool;

[0015] An optimal Q-value table determination module is used to randomly sample experience quadruples from the experience pool as training samples to train the Q-learning agent to obtain an optimal irrigation Q-value table; the optimal irrigation Q-value table is used to determine the corresponding optimal irrigation strategy according to any agricultural state of the target planting area.

[0016] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0017] This paper discloses a smart irrigation strategy formulation method and system that comprehensively considers meteorological factors, soil data, crop parameters, and field management information, and incorporates them into the computational framework of an irrigation decision-making model. By calculating and analyzing the irrigation system's crop yield, water requirements, and annual economic costs, it optimizes irrigation strategies, thereby promoting overall improvements in irrigation system efficiency and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] FIG1 is a flow chart of a method for formulating a smart irrigation strategy according to the present invention;

[0019] Figure 2 is a diagram showing the principle of reinforcement learning;

[0020] Figure 3 is a diagram showing the principle of the Q-learning algorithm;

[0021] Figure 4 is a simulation diagram of irrigation decision-making based on Q-learning;

[0022] Figure 5 is a diagram of the Python-AquaCrop joint simulation based on Q-learning;

[0023] Figure 6 is a diagram of the collaborative simulation process based on Q-learning;

[0024] Figure 7 is a diagram of the iterative process based on Q-learning. DETAILED DESCRIPTION

[0025] By deeply analyzing the control decision-making issues of irrigation systems and the impact of different irrigation decisions on crop yield, water demand and economic benefits, this paper provides a method and system for formulating intelligent irrigation strategies, and develops a Python-AquaCrop collaborative simulation model that integrates an efficient Q-learning algorithm framework, providing new research perspectives and technical paths for the scientific research and development of large-scale irrigation systems.

[0026] Example 1

[0027] As shown in FIG1 , the present invention provides a method for formulating a smart irrigation strategy, comprising:

[0028] S1: Obtain the current agricultural status of the target planting area; the current agricultural status includes meteorological data, soil data, crop parameters, and field management data. The meteorological data includes precipitation, temperature, and sunshine duration; the crop parameters include growth period parameters and yield data. These parameters are established to accurately model the water demand and response during agricultural production.

[0029] First, weather station data for the target planting area was collected from the China Meteorological Data Network (http: / / data.cma.cn / ), including key meteorological variables such as precipitation, temperature, and sunshine duration. This provided the necessary basic information for assessing meteorological conditions. Then, the Penman-Monteith formula was applied to estimate the potential evapotranspiration (ET0) of the reference crop:

[0030] Among them, R n represents the net radiation of the crop surface; G represents the soil heat flux; T refers to the average air temperature; u2 is the wind speed at a height of 2 meters; γ represents the hygrometer constant; e s Indicates the saturated water vapor pressure of air; e a Represents the actual water vapor pressure of the air; Δ is the slope of the curve of the relationship between saturated water vapor pressure and temperature.

[0031] This equation considers the combined influence of multiple meteorological factors and provides a preliminary assessment of crop water requirements. It should be noted that during the physiological process of crop growth, water consumption through evaporation and transpiration directly affects crop water requirements, and water supply is closely related to crop yield, which in turn affects agricultural economic benefits. Therefore, the potential evaporation and transpiration calculated by this formula can serve as a reference for reward values ​​in weight determination. Furthermore, crop yield, crop water requirements, and annual economic costs are mutually constrained and influenced, and potential evaporation and transpiration underpins these three constraints.

[0032] This method extracts soil data from the cropping area, which is crucial for understanding the soil's ability to retain water and the crop's ability to absorb it. Crop parameters and field management information, which directly impact irrigation needs and yield forecasts, are also considered. Through data aggregation and analysis, a comprehensive and accurate agricultural model is established, providing a scientific basis for developing optimal irrigation strategies.

[0033] S2: Based on the preset irrigation decision model and the Q-learning agent, determine the current irrigation strategy according to the current agricultural status. The preset irrigation decision model includes six irrigation strategies, which together constitute the parameter set for irrigation strategy selection in the model.

[0034] The first irrigation strategy: rain-fed irrigation (no irrigation).

[0035] The second irrigation strategy: irrigation is triggered when the soil moisture content in the crop root zone is below a preset threshold.

[0036] The third irrigation strategy is to implement irrigation at preset intervals of days, which is called periodic irrigation.

[0037] The fourth irrigation strategy: predefined irrigation schedules.

[0038] The fifth irrigation strategy: daily irrigation to fill all soil gaps and maintain soil moisture at a preset moisture value.

[0039] Sixth irrigation strategy: irrigate daily to a preset soil depth.

[0040] Regarding Q-learning agents: As shown in Figures 2 and 3, starting with a random policy, the agent iteratively updates its Q-value table through interaction with the environment. A balance is sought between exploration (trying new actions) and exploitation (selecting the best action based on known information). The agent executes actions in the environment and accumulates experience based on observed rewards and new states, gradually optimizing the Q-table and converging towards the optimal policy by reducing the discrepancy between estimated Q-values ​​and actual rewards.

[0041] In the field of reinforcement learning, especially in the framework of Q-learning algorithms, a policy is a decision guide that specifies an action for each possible state. The policy is usually represented by the symbol π, which is a mapping from state space to action space (π: S→A), which specifies an action a for each state s to guide the agent to make decisions in the environment, as shown in the following formula: π(a|s)=P(A=a t |S=S t ).

[0042] Among them, P is the conditional probability of outputting the control action, A is the control action, and S refers to the state.

[0043] Within the reinforcement learning framework, the quality of a policy π is determined by evaluating the expected value of its cumulative reward. Cumulative reward is the sum of future rewards obtained by taking actions according to a given policy, starting from an initial state. It reflects the potential benefits of the policy over the long term. To estimate the expected value of cumulative reward, researchers typically rely on empirical trajectories, or action-state-reward sequences, which record the chronological trajectory of following a particular policy in an environment, as shown in the following formula:

[0044] Among them, G t is the cumulative return, r is the immediate reward obtained at the current time step, and β is the discount coefficient.

[0045] The state value function defines the state S t The expected cumulative reward obtained by following a specific strategy π under π. This function measures the quality of the state under a given strategy and is expressed as the mathematical expectation of the future cumulative rewards that may be obtained from this state, as shown in the following formula:

[0046] The state-action value function is also called the action value function. The expected value of this function represents the state S t Next, select action a t And then follow the strategy π to perform the action and get the expected cumulative reward, as shown in the following formula:

[0047] The state-action value function provides an evaluation of the expected benefit of performing a specific action in a specific state, thereby guiding the selection of the optimal decision in the state space and action space.

[0048] In the field of reinforcement learning, the fundamental goal is to discover or approximate an optimal policy, that is, to find a decision rule that maximizes the expected cumulative reward. This goal is achieved by gradually improving the policy performance, generally based on an estimate of the expected reward associated with each state-action. In the process of policy improvement, the policy improvement usually relies on the advantage function, which measures the probability of taking a specific action a under a given policy π.t In state S t The advantage function can be calculated based on the difference between the state value function and the state-action value function. The specific expression is: π (s,a)=Q π (s,a)-v π (s).

[0049] Q-learning is a model-free reinforcement learning algorithm that relies on learning the optimal policy for decision-making by directly interacting with the environment. The optimal policy is a sequence of actions taken in a given state that optimizes the expected cumulative reward in the future. The pseudocode is as follows:

[0050] Initialize the Q table and initialize all Q(s,a) to 0, where s is the state and a is the action.

[0051] Set the learning rate alpha (usually a small positive number).

[0052] Set the discount factor beta (usually between 0 and 1).

[0053] Set the exploration probability epsilon (usually a small positive number used to explore new actions).

[0054] For episode: i=0,1,2…do.

[0055] Start from the initial state.

[0056] When the state is not a terminal state, repeat the following steps:

[0057] Action a is selected according to the epsilon-greedy strategy.

[0058] Randomly select an action with probability epsilon.

[0059] Otherwise, the action with the largest Q value is selected by taking a mini-batch of steps to calculate the loss function.

[0060] Perform action a, observe reward r and new state s t+1 .

[0061] Update Q table:

[0062] Update the state s to s t+1 .

[0063] End for.

[0064] The final learned Q-table contains the Q-value of each state-action pair and can be used to select the optimal policy.

[0065] Among them, Q(s t ,a t ) represents the current agricultural status s t and action a t The Q value under this condition, the action refers to the irrigation strategy; represents the learning rate; β represents the discount factor, which is used to measure the current value of future rewards; Indicates that in the next agricultural state s t+1 The maximum Q value of all possible actions represents the expected value of the optimal future action, r t+1 Indicates the next irrigation reward value.

[0066] Q-learning demonstrates its computational efficiency in multiple dimensions: model independence; policy orthogonality; efficient processing of delayed rewards; and offline learning capabilities. The Q-learning algorithm uses tables to store Q values ​​and is suitable for smaller discrete state spaces and action spaces. In this invention, the developed Q-learning algorithm framework consists of two parts: a decision-making loop and a value evaluation loop. The decision-making loop is responsible for executing the current policy and exploring the environment; the value evaluation loop adjusts and optimizes the policy based on the exploration results. The following are the detailed steps of the Q-learning algorithm:

[0067] Decision execution: Continuously monitor the environment state and derive action decisions based on the greedy strategy, while recording state transitions and reward feedback.

[0068] Experience replay: By randomly sampling stored experience data, the Q value of the state-action pair is evaluated and the value function is updated.

[0069] Policy update: Use the estimated Q value to adjust the execution strategy, and introduce the learning rate as a tuning parameter for policy optimization, aiming to optimize the long-term cumulative return.

[0070] S3: Based on the current irrigation strategy, an irrigation simulation is performed on the target planting area to obtain a corresponding irrigation reward value and a next agricultural state; the irrigation reward value is determined based on the crop yield, crop water requirement and annual economic cost after irrigation; the current agricultural state, the current irrigation strategy, the irrigation reward value and the next agricultural state constitute an experience quadruple; a plurality of the experience quadruple constitute an experience pool.

[0071] For an irrigation system using the Q-learning algorithm, four types of data closely related to optimizing the irrigation system's control strategy are used as state variables: meteorological factors, soil data, crop parameters, and field management information. The integration of these variables forms the foundation of the irrigation system's state space, providing information support for optimal control decisions. Crop water requirements are determined by the irrigation strategy, so the irrigation strategy is implemented as a reinforcement learning control action, with the Q-learning algorithm executing the optimization decision-making process.

[0072] The crop yield, crop water requirement and annual economic cost of the irrigation system are selected as the main variables of the reward function, aiming to guide the irrigation system to develop in the optimal direction. The calculation formula of the irrigation reward value is: Reward = -μ1QV-μ2GY-μ3CT.

[0073] Among them, Reward is the irrigation reward value, μ1, μ2, and μ3 are weight coefficients, which reflect the relative importance and priority in the optimization of irrigation control strategy. The values ​​of these three weight coefficients can be determined based on the potential evaporation and transpiration of crops; QV is the crop water requirement, GY is the crop yield, and CT is the annual economic cost.

[0074] S4: Randomly sample experience quadruplets from the experience pool as training samples, and train the Q-learning agent to obtain an optimal irrigation Q-value table; the optimal irrigation Q-value table is used to determine the corresponding optimal irrigation strategy according to any agricultural state of the target planting area.

[0075] The AquaCrop-OSPy extension module, developed in Python, provides a standardized interface for AquaCrop, allowing for model data exchange and co-simulation. This enhances the accessibility and integration of data structures between AquaCrop and Python, facilitating efficient model exchange and co-simulation. This integrated architecture not only supports AquaCrop's reinforcement learning variable setting and computation of irrigation system models in Python, but also facilitates the development and optimization of complex control strategies.

[0076] The present invention constructs an AquaCrop co-simulation model based on Python and with the Q-learning algorithm as the core. It not only enhances the optimization and control capabilities of the model, but also effectively outputs the optimal control strategy for a specific irrigation system, demonstrating the application potential of reinforcement learning in the field of agricultural water resources management. It can further improve the operating efficiency of the irrigation system and is expected to achieve a higher level of automation and intelligent management. The developed co-simulation model can perform performance evaluation of the control strategy in the OpenAI Gym environment, which is not only convenient for debugging but also shows excellent scalability. Using the Q-learning algorithm, the co-simulation platform achieves accurate output of the optimal control strategy for the irrigation system on the basis of ensuring the suitability of crop yield and water demand and maximizing economic benefits through continuous interaction and iterative learning with the environment.

[0077] The collaborative Q-learning simulation process involves three stages: 1) developing a simulation environment integrating Python and AquaCrop to simulate an irrigation decision-making system model; 2) adjusting the Q-learning algorithm parameter configuration and optimizing the algorithm's internal variables to ensure that the control system meets predetermined performance requirements; and 3) training the Q-learning algorithm model and evaluating the latest control strategy output by the Q-learning agent to verify the effectiveness and applicability of the Q-learning algorithm in optimizing irrigation system control strategies.

[0078] As shown in Figures 4, 5, 6 and 7, the method specifically includes:

[0079] 1) Use AquaCrop software to build a preset irrigation decision model and collect the current agricultural status of the target planting area.

[0080] 2) Build a Q-learning agent and the AquaCrop-OSPy module using Python. Create program routines within the Python environment that interface with the AquaCrop model and develop an integrated simulation framework that uses AquaCrop to dynamically monitor and analyze the effects of irrigation strategies and daily crop water fluxes under changing environmental conditions and management practices.

[0081] 3) A Q-learning agent written in Python continuously and dynamically interacts with the AquaCrop simulation environment, enabling real-time data synchronization and exchange. Initialization data from the Q-learning agent in Python is received via the AquaCrop-OSPy module. The agent then determines the current irrigation strategy (i.e., control action) based on the pre-defined irrigation decision model built within the AquaCrop software and the current agricultural status.

[0082] 4) Within the AquaCrop-OSPy module, an irrigation simulation is performed based on the current irrigation strategy. The simulation results are sent to the Q-learning agent in the Python software for one iteration. The irrigation reward and the next agricultural state are output to the AquaCrop-OSPy module, and the current irrigation strategy is reset within the AquaCrop-OSPy module. This cycle repeats until the algorithm converges and the optimal irrigation strategy is output. The iteration period is primarily limited by the runtime of the AquaCrop simulation. To standardize the iteration time period and ensure synchronization, the iteration time step between the AquaCrop simulation and the Q-learning agent is set to 10 seconds, i.e., the time increment of each iteration is Δt = 10 seconds.

[0083] The collaborative simulation platform constructed through this construction can fully utilize the extensive data structure resources provided by Python. Executing variable definitions and calculations for the Q-learning algorithm in the Python working environment can effectively overcome the limitations of the preset control logic of the AquaCrop software.

[0084] Example 2

[0085] In order to implement the technical solution in the first embodiment and achieve the corresponding functions and technical effects, this embodiment also provides a smart irrigation strategy formulation system, including:

[0086] The current state acquisition module is used to obtain the current agricultural state of the target planting area; the current agricultural state includes meteorological data, soil data, crop parameters and field management data.

[0087] The current irrigation strategy determination module is used to determine the current irrigation strategy according to the current agricultural status based on a preset irrigation decision model and a Q-learning agent.

[0088] An experience pool construction module is used to perform irrigation simulation on the target planting area based on the current irrigation strategy to obtain the corresponding irrigation reward value and the next agricultural state; the irrigation reward value is determined based on the crop yield, crop water requirement and annual economic cost after irrigation; the current agricultural state, the current irrigation strategy, the irrigation reward value and the next agricultural state constitute an experience quadruple; multiple experience quadruples constitute an experience pool.

[0089] An optimal Q-value table determination module is used to randomly sample experience quadruples from the experience pool as training samples to train the Q-learning agent to obtain an optimal irrigation Q-value table; the optimal irrigation Q-value table is used to determine the corresponding optimal irrigation strategy according to any agricultural state of the target planting area.

[0090] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0091] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for formulating an intelligent irrigation strategy, characterized in that the method Including: Obtain the current agricultural status of the target planting area; the current agricultural status includes meteorological data, soil data, crop parameters, and field management data; Based on a preset irrigation decision model and a Q-learning agent, determine the current irrigation strategy according to the current agricultural status; Perform irrigation simulation on the target planting area based on the current irrigation strategy to obtain the corresponding irrigation reward value and the next agricultural status; The irrigation reward value is determined based on the crop yield after irrigation, the crop water requirement, and the annual economic cost; the current agricultural status, the current irrigation strategy, the irrigation reward value, and the next agricultural status form an experience quadruple; multiple such experience quadruples form an experience pool; Randomly sample experience quadruples from the experience pool as training samples to train the Q-learning agent to obtain an optimal irrigation Q-value table; the optimal irrigation Q-value table is used to determine the corresponding optimal irrigation strategy according to any agricultural status of the target planting area.

2. The method for formulating an intelligent irrigation strategy according to claim 1, wherein The preset irrigation decision model includes six irrigation strategies; The first irrigation strategy: rainfed irrigation; The second irrigation strategy: trigger irrigation when the soil water content in the crop root zone is below a preset threshold; The third irrigation strategy: implement irrigation at preset day intervals; The fourth irrigation strategy: a predefined irrigation plan; The fifth irrigation strategy: daily irrigation to fill all soil layer gaps and maintain the soil humidity at a preset humidity value; The sixth irrigation strategy: daily irrigation to a preset depth of the soil.

3. The method for formulating an intelligent irrigation strategy according to claim 1, wherein The calculation formula for the irrigation reward value is: Reward=-μ1QV-μ2GY-μ3CT; Where, Reward is the irrigation reward value, μ1, μ2, and μ3 are all weight coefficients, QV is the crop water requirement, GY is the crop yield, and CT is the annual economic cost.

4. The method for formulating an intelligent irrigation strategy according to claim 1, characterized in that The training update formula during the training of the Q-learning agent is as follows: Among them, Q(s t , a t ) represents the Q-value under the current agricultural state s t and action a t , where the action refers to the irrigation strategy; Let \(\alpha\) denote the learning rate; \(\beta\) denotes the discount factor, which is used to measure the present value of future rewards; Denote the next agricultural state s t+1 The maximum value of the Q-values for all possible actions, representing the expected value of the optimal future action, r t+1 Denote the next irrigation reward value.

5. The method for formulating an intelligent irrigation strategy according to claim 1, wherein The method further includes: Use AquaCrop software to build a preset irrigation decision model and collect the current agricultural status of the target planting area; Use Python software to build a Q-learning agent and an AquaCrop-OSPy module; Receive initialization data from the Q-learning agent in the Python software through the AquaCrop-OSPy module, and then combine the preset irrigation decision model built in the AquaCrop software and the current agricultural status to determine the current irrigation strategy; In the AquaCrop-OSPy module, perform irrigation simulation according to the current irrigation strategy, and send the simulation results to the Q-learning agent in the Python software for one iteration, output the irrigation reward value and the next agricultural status to the AquaCrop-OSPy module, and then reset the current irrigation strategy in the AquaCrop-OSPy module.

6. The method for formulating an intelligent irrigation strategy according to claim 1, wherein The meteorological data includes precipitation, temperature, and sunshine duration; the crop parameters include growth period parameters and yield per unit area data.

7. An intelligent irrigation strategy formulation system, characterized in that, The system includes: The current status acquisition module is used to acquire the current agricultural status of the target planting area; the current agricultural status includes meteorological data, soil data, crop parameters, and field management data; The current irrigation strategy determination module is used to determine the current irrigation strategy based on a preset irrigation decision model and a Q-learning agent according to the current agricultural status; The experience pool construction module is used to perform irrigation simulation on the target planting area based on the current irrigation strategy to obtain the corresponding irrigation reward value and the next agricultural status; the irrigation reward value is determined based on the crop yield, crop water requirement, and annual economic cost after irrigation; the current agricultural status, the current irrigation strategy, the irrigation reward value, and the next agricultural status form an experience quadruple; Multiple said experience quadruples form an experience pool; The optimal Q-value table determination module is used to randomly sample experience quadruples from the experience pool as training samples to train the Q-learning agent to obtain an optimal irrigation Q-value table; the optimal irrigation Q-value table is used to determine the corresponding optimal irrigation strategy according to any agricultural status of the target planting area.

Images