Agricultural intelligent irrigation method and system based on big data analysis

Abstract

The invention discloses an agricultural intelligent irrigation method and system based on big data analysis, and relates to the technical field of agricultural informatization and irrigation control, and the method comprises the following steps: collecting irrigation related data, preprocessing the irrigation related data, and obtaining a standardized irrigation feature vector; performing water demand prediction on the standardized irrigation feature vector by using a BP neural network to obtain a water demand estimated value; performing collaborative optimization calculation on the irrigation control parameters by adopting a group parameter collaborative regulation method based on the water demand estimated value to obtain an optimal irrigation strategy instruction set; and according to the optimal irrigation strategy instruction set, performing control processing on the irrigation execution parameters to obtain a corresponding irrigation execution control result. According to the invention, automation and intelligentization of the whole irrigation process from data perception, intelligent prediction and global optimization to accurate execution are realized, and scientificity of irrigation decision, utilization efficiency of water resources and self-adaptive capability of the system are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural informatization and irrigation control technology, and in particular to an intelligent agricultural irrigation method and system based on big data analysis. Background Technology

[0002] In recent years, water scarcity and inefficient water use have become global challenges hindering sustainable agricultural development. Traditional irrigation methods rely heavily on fixed irrigation regimes, human experience, or simple feedback control based on single-point sensors. These methods are ill-suited to the highly heterogeneous water requirements of crops across time and space, as well as the dynamic fluctuations in weather and soil conditions. This often leads to under- or over-irrigation, resulting in water waste, increased energy consumption, and potential negative environmental impacts.

[0003] Currently, Chinese invention patent CN111524024A discloses a smart water and fertilizer irrigation system and analysis method based on big data. This system includes a water and fertilizer irrigation data center, a water and fertilizer irrigation model upload center, a water and fertilizer irrigation operation model analysis center, a smart integrated water and fertilizer irrigation module, and a pipeline disinfection module. Building upon traditional water and fertilizer irrigation systems that address fertilization and watering issues, this system further analyzes and explores these issues, combining big data analytics, AI, and IoT technologies to upgrade and optimize the traditional technician-dependent water and fertilizer irrigation operation model. Through big data processing technology, it collects, analyzes, processes, and applies water and fertilizer irrigation data models to achieve smart agriculture. However, this technology lacks collaborative optimization and dynamic adjustment of irrigation control parameters, a mechanism for intelligent optimization using historical data, and an effective mechanism for dealing with extreme weather conditions (such as drought). Summary of the Invention

[0004] The technical problem addressed by this invention is that existing technologies struggle to achieve dynamic and coordinated optimization of irrigation control parameters, particularly in real-time adjustments to irrigation start time, duration, and flow distribution ratios. They are unable to adequately address the spatiotemporal heterogeneity of crop water requirements, weather changes, and soil conditions. Furthermore, existing technologies lack effective mechanisms for utilizing historical irrigation data for strategy optimization, and they cannot automatically adjust irrigation strategies in the face of extreme weather conditions such as drought, leading to water waste or insufficient crop water supply.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: Firstly, an intelligent agricultural irrigation method based on big data analysis, comprising the following steps: Step S1: Collect irrigation-related data, preprocess the irrigation-related data, and obtain standardized irrigation feature vectors; Step S2: Use a BP neural network to predict water demand from the standardized irrigation feature vector to obtain an estimated water demand. Step S3: Based on the estimated water demand, the irrigation control parameters are optimized by using a group parameter collaborative adjustment method to obtain the optimal irrigation strategy instruction set. Step S4: Control the irrigation execution parameters according to the optimal irrigation strategy instruction set to obtain the corresponding irrigation execution control results.

[0006] As a preferred embodiment of the intelligent agricultural irrigation method based on big data analysis described in this invention, step S1 includes: Step S101: Collect irrigation-related data through a sensor network; The sensor network includes soil moisture sensors, soil temperature sensors, light intensity sensors, air temperature sensors, air humidity sensors, wind speed sensors, wind direction sensors, rainfall sensors, and crop canopy temperature sensors. The irrigation-related data include soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature. Step S102: Perform data cleaning and outlier processing on the irrigation-related data to obtain cleaned data; Step S103: Using the crop water stress model, soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature in the cleaned data are calculated to obtain crop water stress index values. Step S104: Normalize the cleaned data and crop water stress index values ​​to obtain normalized data. Step S105: Extract water demand prediction features from the normalized data to obtain a standardized irrigation feature vector.

[0007] As a preferred embodiment of the intelligent agricultural irrigation method based on big data analysis described in this invention, step S105, which involves extracting water demand prediction features from the normalized data, specifically includes: Extract the current real-time values ​​of soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature from the normalized data; Calculate the average value of normalized data within a preset time window; The current real-time value and the average value are combined to obtain a standardized irrigation feature vector.

[0008] As a preferred embodiment of the intelligent agricultural irrigation method based on big data analysis described in this invention, the processing logic for predicting water demand using a BP neural network on a standardized irrigation feature vector includes: Standardized irrigation feature vector Each component Assign the corresponding nodes to the input layer of the neural network; For the The formulas for calculating the input and output values ​​of each hidden layer node are as follows: ; ; in, Indicates the first The input values ​​of each hidden layer node, This indicates the number of the input layer node. This indicates that the pre-training process determines the sequence of events from the input layer. The node to the first The connection weights of each hidden layer node Indicates the first The bias threshold determined by the pre-training of each hidden layer node. Indicates the first The output values ​​of each hidden layer node This represents a predefined activation function. This represents the total number of nodes in the input layer; The formulas for calculating the input values ​​and estimated water demand for the output layer nodes are as follows: ; ; in, This represents the input value of the output layer node. This indicates that the pre-training is determined, starting from the first... The connection weights from hidden layer nodes to output layer nodes. This represents the bias threshold determined during the pre-training of the output layer nodes. This indicates the estimated water demand. This represents the activation function of the output layer.

[0009] As a preferred embodiment of the intelligent agricultural irrigation method based on big data analysis described in this invention, the processing logic for collaboratively optimizing irrigation control parameters using a group parameter collaborative adjustment method based on the estimated water demand includes: Based on the estimated water demand, an objective function is constructed with the goal of minimizing the deviation between the actual total irrigation volume and the estimated water demand. The search space for determining irrigation duration and flow parameters is based on the estimated water demand. An initial irrigation strategy population is generated within the search space, and each individual irrigation strategy in the population is composed of irrigation control parameters. The irrigation control parameters include irrigation start time, irrigation duration, and irrigation flow distribution ratio; Based on the group parameter collaborative adjustment method, the group state distribution information and historical strategy memory information are introduced to iteratively update the parameter state of individual irrigation strategies until the preset termination condition is met and the optimal irrigation strategy instruction set is output. The population state distribution information includes distribution information entropy, which reflects the density of population individuals in the search space; The historical strategy memory information includes a historical strategy memory guidance vector obtained by fusing the strategic position vectors of elites from different historical generations. The parameter state is a position vector in the search space.

[0010] As a preferred embodiment of the intelligent agricultural irrigation method based on big data analysis described in this invention, the group parameter collaborative adjustment method includes a diverse sowing strategy, a historical strategy memory optimization operator, and a drought-responsive perturbation mechanism. The diverse seeding strategy is used to uniformly explore and initialize the search space, obtaining an initial irrigation strategy population, the calculation formula of which is: ; in, Describes the initial policy population The entropy of distributed information in the search space. This represents the total number of discrete sub-regions into which the solution space is divided. This indicates that an individual in the population falls into the first... Probability density estimation for each region This indicates the index number range.

[0011] As a preferred embodiment of the intelligent agricultural irrigation method based on big data analysis described in this invention, the historical strategy memory optimization operator performs fractional-order integral memory fusion calculation on the position vectors of elite strategies from different historical generations to obtain the historical strategy memory guidance vector, the calculation formula of which is: ; in, Indicates the first Historical strategy memory guidance vector. Represents the gamma function. Represent the memory decay factor and satisfy , Indicates the first in history Elite strategy position vectors in the archive. This indicates the current iteration number.

[0012] As a preferred embodiment of the agricultural intelligent irrigation method based on big data analysis described in this invention, the drought-responsive perturbation mechanism specifically includes: The drought response factor is calculated based on the estimated water demand and the preset crop water demand baseline. The calculation formula is as follows: ; in, Indicates the first Drought response factor at the next iteration Indicates the first The estimated water demand for the next iteration. This indicates the preset crop water requirement baseline value; Determine whether the drought response factor meets the preset triggering conditions. ; in, This indicates the preset drought trigger threshold; If the preset triggering condition is met, a disturbance update operation is performed; Then, a perturbation update operation is performed on the irrigation strategy individuals in the current irrigation strategy population; The perturbation update operation determines the perturbation amplitude based on the drought response factor, and uses the perturbation amplitude to weight the random perturbation vector to obtain the target perturbation vector. This target perturbation vector is then superimposed on the current position vector of the irrigation strategy individual to obtain the updated position vector of the irrigation strategy individual. The calculation formula is as follows: ; in, This represents the individual location vector of the updated irrigation strategy. Indicates the first During the nth iteration The location vector of each irrigation strategy individual This indicates the disturbance amplitude control parameter. Represents a random perturbation vector that follows a preset distribution; If the preset triggering conditions are not met, the current position vector of the current irrigation strategy individual remains unchanged, and this position vector is used as the initial state for the next iteration.

[0013] As a preferred embodiment of the intelligent agricultural irrigation method based on big data analysis described in this invention, the irrigation execution parameters are controlled and processed according to the optimal irrigation strategy instruction set to obtain the corresponding irrigation execution control result. The processing logic includes: Analyze the irrigation start time, irrigation duration, and irrigation flow allocation ratio in the optimal irrigation strategy instruction set; The water demand estimate is converted to a time scale over the duration of irrigation and dynamically weighted according to the irrigation flow allocation ratio to calculate the real-time execution flow control sequence. Using the irrigation start time as the time axis reference origin, the real-time execution flow control sequence is discretized and mapped on the time axis to establish the correspondence between each time step and the equipment flow output intensity; The irrigation task is triggered at the start time of irrigation, and the irrigation equipment is adjusted according to the real-time flow control sequence to obtain the corresponding irrigation execution control result; The irrigation execution control results include the flow output value corresponding to each time step, the irrigation start time, the irrigation end time, and the cumulative irrigation water volume during the irrigation process.

[0014] Secondly, an intelligent agricultural irrigation system based on big data analysis includes: a data processing module, a water demand prediction module, a collaborative optimization module, and an irrigation execution module; The data processing module is used to collect irrigation-related data, preprocess the irrigation-related data, and obtain standardized irrigation feature vectors. The water demand prediction module is used to input the standardized irrigation feature vector into the BP neural network to predict water demand and obtain the water demand estimate. The collaborative optimization module is used to perform collaborative optimization calculations on irrigation control parameters based on the estimated water demand, using a group parameter collaborative adjustment method to obtain the optimal irrigation strategy instruction set. The irrigation execution module is used to control and process irrigation execution parameters according to the optimal irrigation strategy instruction set, and obtain the corresponding irrigation execution control results.

[0015] The beneficial effects of this invention are as follows: This application innovatively combines a population parameter collaborative adjustment method with a BP neural network, introducing a diverse sowing strategy. By controlling the information entropy of population distribution during the optimization process, the search diversity is improved, effectively avoiding the problem of premature convergence to a local optimum. Simultaneously, by introducing a drought response factor to dynamically adjust the perturbation amplitude, the irrigation strategy can be adjusted in a timely and reasonable manner according to changes in actual water demand. Furthermore, a historical strategy memory-guided vector construction method based on fractional integral memory fusion is proposed, further enhancing the method's guidance and adaptability, and solving the practical pain points of "data dimension deficiency," "optimization easily getting trapped in local optima," and "slow response to extreme environments" in agricultural irrigation. In the feature engineering stage, the system does not directly use the original data but introduces a crop water stress model to calculate index values ​​and concatenates real-time environmental parameters with historical average trends within a preset time window. This design allows the BP neural network to not only judge based on a single point in time but also simultaneously capture the instantaneous response and temporal evolution of crop water demand, greatly improving prediction accuracy. In terms of algorithm logic, this invention ensures uniform exploration of the initial population within the search space from the source by controlling the diversity of seeding strategies based on distributed information entropy, effectively avoiding the premature convergence problem common in traditional optimization algorithms. More ingeniously, the system introduces a memory fusion operator based on fractional integrals, using a gamma function and memory decay factor to nonlinearly weight and accumulate the elite strategies of each generation. This gives the optimization process a stronger global information guidance capability and adaptability to complex environments than traditional algorithms. Furthermore, this invention ingeniously designs a drought-responsive perturbation mechanism, dynamically triggering perturbation updates by calculating the deviation between the estimated water demand and the baseline value. When extreme weather causes the drought factor to exceed a threshold, the amplitude of the random perturbation vector is automatically increased, forcing the irrigation strategy to make leapfrog adjustments to cope with extreme demand. This dynamic adaptive capability is not possessed by conventional static optimization models. In terms of execution logic, through time scale transformation and dynamic weight allocation, the macroscopic optimal strategy instruction set is finely mapped into a sequence of device flow output intensity at each time step, realizing a complete intelligent closed loop from high-order big data analysis to precise execution by the underlying hardware. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of an intelligent agricultural irrigation method based on big data analysis, as provided in one embodiment of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Example, refer to Figure 1 This paper provides an intelligent agricultural irrigation method and system based on big data analysis, including the following steps: Step S1: Collect irrigation-related data, preprocess the irrigation-related data, and obtain standardized irrigation feature vectors; Step S2: Use a BP neural network to predict water demand from the standardized irrigation feature vector to obtain an estimated water demand. Step S3: Based on the estimated water demand, the irrigation control parameters are optimized by using a group parameter collaborative adjustment method to obtain the optimal irrigation strategy instruction set. Step S4: Control the irrigation execution parameters according to the optimal irrigation strategy instruction set to obtain the corresponding irrigation execution control results.

[0019] In specific implementation, step S1 includes: Step S101: Collect irrigation-related data through a sensor network; The sensor network includes soil moisture sensors, soil temperature sensors, light intensity sensors, air temperature sensors, air humidity sensors, wind speed sensors, wind direction sensors, rainfall sensors, and crop canopy temperature sensors; Irrigation-related data include soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature; Step S102: Perform data cleaning and outlier processing on the irrigation-related data to obtain cleaned data; Step S103: Using the crop water stress model, soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature in the cleaned data are calculated to obtain crop water stress index values. Step S104: Normalize the cleaned data and crop water stress index values ​​to obtain normalized data. Step S105: Extract water demand prediction features from the normalized data to obtain a standardized irrigation feature vector.

[0020] It should be noted that by utilizing a sensor network to collect irrigation-related data in real time, including soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature, multi-dimensional environmental information was effectively obtained, providing rich data support for irrigation decisions. Through data cleaning and outlier removal, noisy data was eliminated, ensuring the accuracy and reliability of the data and guaranteeing the precision of subsequent calculations and predictions.

[0021] By employing a crop water stress model to calculate the processed data, the water stress status of crops can be effectively predicted, providing a more accurate basis for irrigation demand forecasting and avoiding over- or under-irrigation. Data normalization resolves scale differences between different data sources, enabling the data to be uniformly used for feature extraction and neural network model training, further improving model stability and predictive accuracy. Extracting standardized irrigation feature vectors provides more precise parameters for water demand prediction, further optimizing irrigation schemes and improving agricultural irrigation efficiency and water resource utilization.

[0022] In specific implementation, step S105 involves extracting water demand prediction features from the normalized data, specifically including: Extract the current real-time values ​​of soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature from the normalized data; Calculate the average value of normalized data within a preset time window; The current real-time value and the average value are combined to obtain a standardized irrigation feature vector.

[0023] It should be noted that by combining the current real-time values ​​and average values ​​using vector concatenation, the extracted real-time values ​​(soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature) are arranged in a fixed order to form a real-time feature vector. The calculated average values ​​(average values ​​of the same set of parameters over a preset time window) are arranged in the same order to form a historical trend feature vector. These two vectors are then connected dimensionally to form a standardized irrigation feature vector. This provides the subsequent neural network with input features that have higher information density and stronger temporal correlation, significantly improving the neural network's ability to learn and predict the complex nonlinear relationship of crop water demand. This method not only inputs signals reflecting the instantaneous state of the environment into the neural network but also provides background information reflecting short-term trends, enabling the neural network to simultaneously capture the immediate response and cumulative effect of crop water demand. This effectively enhances the model's adaptability to field dynamics and the stability of prediction results. This feature construction strategy directly optimizes the quality of the neural network's input layer and is a crucial technical foundation for its ability to accurately model crop water demand patterns and support intelligent irrigation optimization decisions.

[0024] In specific implementation, the processing logic for predicting water demand using a BP neural network on standardized irrigation feature vectors includes: Standardized irrigation feature vector Each component Assign the corresponding nodes to the input layer of the neural network; For the The formulas for calculating the input and output values ​​of each hidden layer node are as follows: ; ; in, Indicates the first The input values ​​of each hidden layer node, This indicates the number of the input layer node. This indicates that the pre-training process determines the sequence of events from the input layer. The node to the first The connection weights of each hidden layer node Indicates the first The bias threshold determined by the pre-training of each hidden layer node. Indicates the first The output values ​​of each hidden layer node This represents a predefined activation function. This represents the total number of nodes in the input layer; The formulas for calculating the input values ​​and estimated water demand for the output layer nodes are as follows: ; ; in, This represents the input value of the output layer node. This indicates that the pre-training is determined, starting from the first... The connection weights from hidden layer nodes to output layer nodes. This represents the bias threshold determined during the pre-training of the output layer nodes. This indicates the estimated water demand. This represents the activation function of the output layer.

[0025] It should be noted that by utilizing a backpropagation (BP) neural network to predict water demand from standardized irrigation feature vectors, the accuracy and reliability of irrigation demand prediction are significantly improved. By assigning each component of the standardized irrigation feature vector to a corresponding node in the neural network's input layer and calculating using the input-output formulas of the hidden layer nodes, the complex nonlinear relationship between input features and water demand is effectively captured. The calculation of each hidden layer node is performed using pre-trained connection weights and bias thresholds, ensuring the model's training effectiveness and generalization ability.

[0026] Meanwhile, the input values ​​of the output layer nodes are combined with the calculation formula for the water demand estimate, incorporating information passed from the hidden layer. Pre-trained weights and biases further optimize the prediction results. This processing logic enables irrigation demand prediction not only to be based on historical data but also to fully utilize various environmental characteristics, ensuring the accuracy of irrigation decisions. This effectively improves water resource utilization efficiency, reduces waste, and provides more efficient and sustainable technical support for smart irrigation in practical applications.

[0027] In practice, the processing logic for collaboratively optimizing irrigation control parameters using a group parameter collaborative adjustment method based on the estimated water demand includes: Based on the estimated water demand, an objective function is constructed with the goal of minimizing the deviation between the actual total irrigation volume and the estimated water demand. The search space for determining irrigation duration and flow parameters is based on the estimated water demand. An initial irrigation strategy population is generated within the search space, and each individual irrigation strategy in the population is composed of irrigation control parameters. Irrigation control parameters include irrigation start time, irrigation duration, and irrigation flow distribution ratio; Based on the group parameter collaborative adjustment method, the group state distribution information and historical strategy memory information are introduced to iteratively update the parameter state of individual irrigation strategies until the preset termination condition is met and the optimal irrigation strategy instruction set is output. Population state distribution information includes distribution information entropy, which reflects the density of population individuals in the search space; Historical strategy memory information includes historical strategy memory guidance vectors obtained by fusing strategic position vectors of elites from different historical periods; The parameter state is the position vector in the search space.

[0028] It should be noted that this application innovatively combines a population parameter collaborative adjustment method with a BP neural network, enabling precise adjustments to the irrigation strategy based on the estimated water demand, thereby achieving effective optimization of irrigation control parameters. By constructing the objective function, the deviation between the actual irrigation volume and the estimated water demand is minimized, thus improving irrigation accuracy and water resource utilization. Simultaneously, the search space determined by the estimated water demand allows the irrigation strategy to search within a reasonable range, avoiding the occurrence of unreasonable parameters. The population parameter collaborative adjustment method, by introducing population state distribution information and historical strategy memory information, effectively avoids the influence of local optima, ensuring that the optimization process of the irrigation strategy has strong adaptability and stability, thereby improving the overall efficiency and sustainability of agricultural irrigation.

[0029] The objective function is constructed primarily to minimize the deviation between the actual total irrigation volume and the estimated water demand. Specifically, the irrigation duration and flow allocation ratio are first mapped to a sequence of equipment flow output intensity at discrete time steps. If the equipment flow output intensity at a certain time step exceeds a preset maximum flow threshold, the actual flow at that time step is calculated based on the maximum flow threshold, i.e., a flow limiting process is applied. Finally, the actual flow at each time step is summed to obtain the actual total irrigation volume.

[0030] The core of the objective function is to optimize the difference between the actual irrigation volume (after mapping and limiting) and the estimated water demand. The estimated water demand is derived from the irrigation feature vector using a backpropagation neural network. The objective function uses the difference between the actual irrigation volume and the estimated water demand as the optimization objective, employing a deviation minimization method to optimize the irrigation strategy, ensuring that the total irrigation volume is as close as possible to the actual needs of the crop. During the optimization process, the objective function can be expressed as minimizing the absolute difference between the actual irrigation volume and the estimated water demand, ensuring efficient use of irrigation resources.

[0031] When constructing the search space for irrigation duration and flow rate parameters, a reasonable range for the relevant parameters is determined based on the estimated water demand. The estimated water demand is a key basis for irrigation strategies, reflecting the actual water requirements of crops. Based on the estimated water demand, the irrigation duration and flow rate allocation ratio can be matched with the estimated water demand through a predetermined functional form, thereby limiting the range of the search space. The search space for irrigation duration and flow rate ratio defines the feasible solution region in the optimization process, avoiding unreasonable calculations caused by excessively large or small parameter values. By reasonably limiting the range of these parameters, the search process can be ensured to proceed within an effective range, improving computational efficiency and optimization accuracy.

[0032] This invention employs a group parameter collaborative adjustment method to iteratively update the parameter states of individual irrigation strategies. The core of this process is to guide the optimization process by introducing group state distribution information and historical strategy memory information within the search space. The parameter states of each individual irrigation strategy (such as irrigation start time, irrigation duration, and flow allocation ratio) are represented as a position vector. In each iteration, individuals in the group update their parameters based on the current optimal solution and historical strategy information. By calculating the entropy of the group state distribution information, individuals can be guided to conduct uniform exploration or concentrated search within the search space, avoiding getting trapped in local optima. Historical strategy memory information, through the memorization of guiding vectors, helps individuals update their parameters in a more optimal direction. When the parameters of all individual strategies meet the preset termination conditions, the optimal irrigation strategy instruction set is output, completing the collaborative optimization of irrigation control parameters.

[0033] In practice, the methods for coordinated adjustment of population parameters include diverse seeding strategies, historical strategy memory optimization operators, and drought-responsive perturbation mechanisms. The diversity seeding strategy is used to uniformly explore and initialize the search space, resulting in an initial irrigation strategy population. Its calculation formula is as follows: ; in, Describes the initial policy population The entropy of distributed information in the search space. This represents the total number of discrete sub-regions into which the solution space is divided. This indicates that an individual in the population falls into the first... Probability density estimation for each region This indicates the index number range.

[0034] It should be noted that this application innovatively introduces a diversity seeding strategy. During the optimization process, the diversity of the search process is enhanced by controlling the distribution entropy of the population, effectively preventing premature convergence to a local optimum. Through this strategy, the search for irrigation control parameters is not only more comprehensive but also improves the stability and adaptability of the algorithm, making the optimization results more reliable and ultimately providing a more accurate and reasonable solution for intelligent irrigation.

[0035] In practice, the historical strategy memory optimization operator performs fractional-order integral memory fusion calculation on the position vectors of elite strategies from different historical generations to obtain the historical strategy memory guidance vector, the calculation formula of which is as follows: ; in, Indicates the first Historical strategy memory guidance vector. Represents the gamma function. Represent the memory decay factor and satisfy , Indicates the first in history Elite strategy position vectors in the archive. This indicates the current iteration number.

[0036] It should be noted that this application innovatively proposes constructing a historical strategy memory guidance vector through fractional-order integral memory fusion. This technique greatly enhances the adaptability of swarm optimization algorithms in complex environments. By employing fractional-order memory, the rate of memory decay can be controlled more precisely, making the influence of historical strategy information more flexible in different iterations. Compared to traditional simple memory strategies, this method can maintain strong global information over a longer search period, thereby improving the optimization efficiency of irrigation strategies and the accuracy of the final decision. This historical strategy guidance not only effectively improves the accuracy of parameter state updates but also avoids the limitations of a single strategy by fusing elite strategy information from different generations, thus providing more intelligent and refined optimization strategies for smart irrigation systems. This innovative design is particularly suitable for the ever-changing agricultural environment, enhancing the adaptability and flexibility of irrigation decisions.

[0037] Considering that the optimization process of irrigation strategies is based on iterative algebra t, a discrete summation method is used for numerical calculation. Specifically, the system performs a weighted summation of the elite strategy position vectors archived from generation 1 to generation t, utilizing a kernel function. This approach achieves nonlinear weighted fusion of long-term and recent memories. This discretization ensures the algorithm's executability in computer systems, enabling more precise control over memory decay rates and thus improving the global search efficiency and adaptability of irrigation strategies in complex environments.

[0038] In practice, the drought response disturbance mechanism specifically includes: The drought response factor is calculated based on the estimated water demand and the preset crop water demand baseline. The calculation formula is as follows: ; in, Indicates the first Drought response factor at the next iteration Indicates the first The estimated water demand for the next iteration. This indicates the preset crop water requirement baseline value; Determine whether the drought response factor meets the preset triggering conditions. ; in, This indicates the preset drought trigger threshold; If the preset triggering condition is met, a disturbance update operation is performed; Then, a perturbation update operation is performed on the irrigation strategy individuals in the current irrigation strategy population; The perturbation update operation determines the perturbation amplitude based on the drought response factor, and uses the perturbation amplitude to weight the random perturbation vector to obtain the target perturbation vector. This target perturbation vector is then superimposed on the current position vector of the irrigation strategy individual to obtain the updated position vector of the irrigation strategy individual. The calculation formula is as follows: ; in, This represents the individual location vector of the updated irrigation strategy. Indicates the first During the nth iteration The location vector of each irrigation strategy individual This indicates the disturbance amplitude control parameter. Represents a random perturbation vector that follows a preset distribution; If the preset triggering conditions are not met, the current position vector of the current irrigation strategy individual remains unchanged, and this position vector is used as the initial state for the next iteration.

[0039] It should be noted that this application innovatively introduces a drought response factor to dynamically adjust the perturbation amplitude, enabling irrigation strategies to be adjusted promptly and reasonably according to changes in actual water demand. Traditional optimization algorithms often employ fixed perturbations or overly simplistic adjustment rules, while this invention determines the intensity of the perturbation based on the deviation between water demand and a baseline value, thus improving the system's adaptability and accuracy. This method not only enhances the flexibility of irrigation strategies but also avoids the negative impact of excessive perturbation on irrigation effectiveness, thereby improving water resource utilization efficiency and crop growth conditions. This innovative design is particularly suitable for variable agricultural irrigation environments, providing a more intelligent solution for precision agriculture irrigation.

[0040] The drought-responsive perturbation mechanism calculates the drought response factor based on the estimated water demand and the crop's baseline water demand. The drought response factor reflects the deviation between the actual water demand and the crop's baseline water requirement; when this deviation reaches a certain threshold, it indicates that the irrigation strategy needs adjustment. By calculating the drought response factor, it is possible to determine whether the current irrigation strategy needs to be updated with a perturbation.

[0041] Specifically, the calculation of the disturbance amplitude is directly related to the magnitude of the drought response factor. A larger drought response factor indicates a greater gap between actual and projected water demand, thus requiring a correspondingly larger disturbance amplitude to prompt greater adjustments to irrigation strategies. Conversely, a smaller drought response factor results in a smaller disturbance amplitude, avoiding unnecessary over-adjustment. In this way, the disturbance amplitude can be flexibly adjusted according to actual changes in current irrigation demand, ensuring timely adaptation of irrigation strategies.

[0042] During the perturbation update operation, the perturbation amplitude is weighted onto the random perturbation vector to obtain the target perturbation vector. This target perturbation vector, superimposed on the current position vector, updates the state of the irrigation strategy individual. The random perturbation vector increases the diversity and randomness of the search, while the weighting of the perturbation amplitude controls the degree of perturbation, ensuring that the update amplitude conforms to the adjustment range required by the current irrigation strategy. Through this dynamic adjustment, the system can adapt to constantly changing environmental conditions, allowing the irrigation strategy to flexibly respond to environmental changes during the optimization process and dynamically adjust the parameters of the irrigation strategy individual based on this factor.

[0043] In practice, the irrigation execution parameters are controlled according to the optimal irrigation strategy instruction set to obtain the corresponding irrigation execution control results. The processing logic includes: Analyze the irrigation start time, irrigation duration, and irrigation flow allocation ratio in the optimal irrigation strategy instruction set; The water demand estimate is converted to a time scale over the duration of irrigation and dynamically weighted according to the irrigation flow allocation ratio to calculate the real-time execution flow control sequence. Using the irrigation start time as the time axis reference origin, the real-time execution flow control sequence is discretized and mapped on the time axis to establish the correspondence between each time step and the equipment flow output intensity; The irrigation task is triggered at the start time of irrigation, and the irrigation equipment is adjusted according to the real-time flow control sequence to obtain the corresponding irrigation execution control result; The irrigation execution control results include the corresponding flow output value at each time step, irrigation start time, irrigation end time, and cumulative irrigation water volume during the irrigation process.

[0044] It should be noted that the specific implementation of dynamic weight allocation is as follows: the target irrigation volume for each time step of the irrigation duration can be determined by dividing the total estimated water demand by the irrigation duration. Combined with the irrigation flow allocation ratio for each time period, the flow is dynamically allocated. The flow control amount for each time period is the product of the target irrigation volume and the flow allocation ratio. In this way, the flow allocation can be dynamically adjusted according to the actual water demand in each time period, ensuring more precise water supply during the irrigation process.

[0045] The flow control value for each time step is discretized and mapped using the irrigation start time as the origin of the time axis. Through a timer control system, the flow value is mapped to each time point according to a preset flow control sequence, generating corresponding irrigation equipment control signals. This discretization process allows the irrigation equipment to adjust its flow output on time and as needed, ensuring that the irrigation volume for each period matches the predetermined target.

[0046] Irrigation tasks are triggered based on the irrigation start time. When the system detects that the preset irrigation start time has arrived, the control system automatically triggers the execution of the irrigation task. At this time, the control system gradually adjusts the working state of the irrigation equipment according to the real-time execution flow control sequence. This process regulates the water pump through controllers such as frequency converters, adjusting the pump's power output according to the predetermined flow and time sequence. Through timer interrupt signals, the control system ensures that the water pump operates according to the precisely set flow control sequence, thereby guaranteeing the accurate operation of the irrigation equipment and the rational distribution of water supply.

[0047] This invention employs a precise irrigation control method that dynamically adjusts irrigation water volume based on estimated water demand and flow allocation ratios. This ensures that water supply meets crop needs, avoiding over- or under-irrigation and improving water resource utilization efficiency. By transforming time scales and dynamically weighting the data, the irrigation flow distribution is optimized. Furthermore, the real-time flow control sequence is discretized and mapped to ensure precise adjustment of irrigation equipment, reducing uncertainty and enhancing execution stability. The automatic triggering mechanism for irrigation tasks allows the system to adjust automatically according to actual needs, reducing human intervention and further enhancing the adaptability and flexibility of the intelligent irrigation system. This provides agriculture with an efficient, water-saving intelligent irrigation solution, promoting the development of sustainable agriculture.

[0048] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0049] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. An intelligent agricultural irrigation method based on big data analysis, characterized in that, Includes the following steps: Step S1: Collect irrigation-related data, preprocess the irrigation-related data, and obtain standardized irrigation feature vectors; Step S2: Use a BP neural network to predict water demand from the standardized irrigation feature vector to obtain an estimated water demand. Step S3: Based on the estimated water demand, the irrigation control parameters are optimized by using a group parameter collaborative adjustment method to obtain the optimal irrigation strategy instruction set. Step S4: Control the irrigation execution parameters according to the optimal irrigation strategy instruction set to obtain the corresponding irrigation execution control results.

2. The agricultural intelligent irrigation method based on big data analysis as described in claim 1, characterized in that, Step S1 includes: Step S101: Collect irrigation-related data through a sensor network; The sensor network includes soil moisture sensors, soil temperature sensors, light intensity sensors, air temperature sensors, air humidity sensors, wind speed sensors, wind direction sensors, rainfall sensors, and crop canopy temperature sensors. The irrigation-related data include soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature. Step S102: Perform data cleaning and outlier processing on the irrigation-related data to obtain cleaned data; Step S103: Using the crop water stress model, soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature in the cleaned data are calculated to obtain crop water stress index values. Step S104: Normalize the cleaned data and crop water stress index values ​​to obtain normalized data. Step S105: Extract water demand prediction features from the normalized data to obtain a standardized irrigation feature vector.

3. The agricultural intelligent irrigation method based on big data analysis as described in claim 2, characterized in that, In step S105, water demand prediction features are extracted from the normalized data, specifically including: Extract the current real-time values ​​of soil moisture, soil temperature, light intensity, air temperature, air humidity, wind speed, wind direction, rainfall, and crop canopy temperature from the normalized data; Calculate the average value of normalized data within a preset time window; The current real-time value and the average value are combined to obtain a standardized irrigation feature vector.

4. The agricultural intelligent irrigation method based on big data analysis as described in claim 1, characterized in that, The processing logic for predicting water demand using a BP neural network on a standardized irrigation feature vector includes: Standardized irrigation feature vector Each component Assign the corresponding nodes to the input layer of the neural network; For the The formulas for calculating the input and output values ​​of each hidden layer node are as follows: ； ； in, Indicates the first The input values ​​of each hidden layer node, This indicates the number of the input layer node. This indicates that the pre-training process determines the sequence of events from the input layer. The node to the first The connection weights of each hidden layer node Indicates the first The bias threshold determined by the pre-training of each hidden layer node. Indicates the first The output values ​​of each hidden layer node This represents a predefined activation function. This represents the total number of nodes in the input layer; The formulas for calculating the input values ​​and estimated water demand for the output layer nodes are as follows: ； ； in, This represents the input value of the output layer node. This indicates that the pre-training is determined, starting from the first... The connection weights from hidden layer nodes to output layer nodes. This represents the bias threshold determined during the pre-training of the output layer nodes. This indicates the estimated water demand. This represents the activation function of the output layer.

5. The agricultural intelligent irrigation method based on big data analysis as described in claim 1, characterized in that, Based on the estimated water demand, the processing logic for collaboratively optimizing irrigation control parameters using a group parameter collaborative adjustment method includes: Based on the estimated water demand, an objective function is constructed with the goal of minimizing the deviation between the actual total irrigation volume and the estimated water demand. The search space for determining irrigation duration and flow parameters is based on the estimated water demand. An initial irrigation strategy population is generated within the search space, and each individual irrigation strategy in the population is composed of irrigation control parameters. The irrigation control parameters include irrigation start time, irrigation duration, and irrigation flow distribution ratio; Based on the group parameter collaborative adjustment method, the group state distribution information and historical strategy memory information are introduced to iteratively update the parameter state of individual irrigation strategies until the preset termination condition is met and the optimal irrigation strategy instruction set is output. The population state distribution information includes distribution information entropy, which reflects the density of population individuals in the search space; The historical strategy memory information includes a historical strategy memory guidance vector obtained by fusing the strategic position vectors of elites from different historical generations. The parameter state is a position vector in the search space.

6. The agricultural intelligent irrigation method based on big data analysis as described in claim 5, characterized in that, The group parameter collaborative adjustment method includes a diverse seeding strategy, a historical strategy memory optimization operator, and a drought-responsive perturbation mechanism. The diverse seeding strategy is used to uniformly explore and initialize the search space, obtaining an initial irrigation strategy population, the calculation formula of which is: ； in, Describes the initial policy population The entropy of distributed information in the search space. This represents the total number of discrete sub-regions into which the solution space is divided. This indicates that an individual in the population falls into the first... Probability density estimation for each region This indicates the index number range.

7. The agricultural intelligent irrigation method based on big data analysis as described in claim 6, characterized in that, The historical strategy memory optimization operator performs fractional-order integral memory fusion calculation on the position vectors of elite strategies from different historical generations to obtain the historical strategy memory guidance vector, the calculation formula of which is: ； in, Indicates the first Historical strategy memory guidance vector. Represents the gamma function. Represent the memory decay factor and satisfy , Indicates the first in history Elite strategy position vectors in the archive. This indicates the current iteration number.

8. The agricultural intelligent irrigation method based on big data analysis as described in claim 6, characterized in that, The drought-responsive disturbance mechanism specifically includes: The drought response factor is calculated based on the estimated water demand and the preset crop water demand baseline. The calculation formula is as follows: ； in, Indicates the first Drought response factor at the next iteration Indicates the first The estimated water demand for the next iteration. This indicates the preset crop water requirement baseline value; Determine whether the drought response factor meets the preset triggering conditions. ; in, This indicates the preset drought trigger threshold; If the preset triggering condition is met, a disturbance update operation is performed; Then, a perturbation update operation is performed on the irrigation strategy individuals in the current irrigation strategy population; The perturbation update operation determines the perturbation amplitude based on the drought response factor, and uses the perturbation amplitude to weight the random perturbation vector to obtain the target perturbation vector. This target perturbation vector is then superimposed on the current position vector of the irrigation strategy individual to obtain the updated position vector of the irrigation strategy individual. The calculation formula is as follows: ； in, This represents the individual location vector of the updated irrigation strategy. Indicates the first During the nth iteration The location vector of each irrigation strategy individual This indicates the disturbance amplitude control parameter. Represents a random perturbation vector that follows a preset distribution; If the preset triggering conditions are not met, the current position vector of the current irrigation strategy individual remains unchanged, and this position vector is used as the initial state for the next iteration.

9. The agricultural intelligent irrigation method based on big data analysis as described in claim 1, characterized in that, According to the optimal irrigation strategy instruction set, the irrigation execution parameters are controlled and processed to obtain the corresponding irrigation execution control results. The processing logic includes: Analyze the irrigation start time, irrigation duration, and irrigation flow allocation ratio in the optimal irrigation strategy instruction set; The water demand estimate is converted to a time scale over the duration of irrigation and dynamically weighted according to the irrigation flow allocation ratio to calculate the real-time execution flow control sequence. Using the irrigation start time as the time axis reference origin, the real-time execution flow control sequence is discretized and mapped on the time axis to establish the correspondence between each time step and the equipment flow output intensity; The irrigation task is triggered at the start time of irrigation, and the irrigation equipment is adjusted according to the real-time flow control sequence to obtain the corresponding irrigation execution control result; The irrigation execution control results include the flow output value corresponding to each time step, the irrigation start time, the irrigation end time, and the cumulative irrigation water volume during the irrigation process.

10. An agricultural intelligent irrigation system based on big data analysis, applied in an agricultural intelligent irrigation method based on big data analysis as described in any one of claims 1-9, characterized in that, include: The system comprises a data processing module, a water demand prediction module, a collaborative optimization module, and an irrigation execution module. The data processing module is used to collect irrigation-related data, preprocess the irrigation-related data, and obtain standardized irrigation feature vectors. The water demand prediction module is used to input the standardized irrigation feature vector into the BP neural network to predict water demand and obtain the water demand estimate. The collaborative optimization module is used to perform collaborative optimization calculations on irrigation control parameters based on the estimated water demand, using a group parameter collaborative adjustment method to obtain the optimal irrigation strategy instruction set. The irrigation execution module is used to control and process irrigation execution parameters according to the optimal irrigation strategy instruction set, and obtain the corresponding irrigation execution control results.

Images