Multi-source sensing intelligent irrigation system fused with multi-scale optimization

Abstract

The invention relates to the technical field of intelligent irrigation, in particular to a multi-source sensing intelligent irrigation system fused with multi-scale optimization. Comprising a supporting module, a photovoltaic power generation module, a light detection module, a water treatment module, a drip irrigation module, a temperature and humidity detection module, a control module and a prediction module. The supporting module is used for carrying out position supporting on equipment of the photovoltaic power generation module, the light detection module, the drip irrigation module, the water treatment module and the temperature and humidity detection module. According to the multi-source sensing intelligent irrigation system integrated with multi-scale optimization, rainwater is effectively collected, and after the rainwater is treated, the safety and quality of irrigation water are ensured, and the drip irrigation system is prevented from being blocked; and the photovoltaic power generation module provides green and sustainable power support for the whole irrigation system, ensures efficient and independent operation of the irrigation system, and can accurately estimate the water demand of crops based on the constructed evapotranspiration prediction model, thereby optimizing the irrigation frequency and water volume of the drip irrigation system.

Description

Technical Field

[0001] The present invention relates to the field of smart irrigation technology, and in particular to a multi-source sensing smart irrigation system integrating multi-scale optimization. Background Art

[0002] As the global population continues to rise and climate change intensifies, the contradiction between water supply and demand has become increasingly acute, especially in arid and semi-arid areas, where water scarcity poses a severe challenge to agricultural production. As the main consumer of freshwater resources in the world, improving its water use efficiency and reducing water waste are crucial to ensuring food security and the long-term sustainability of water resources.

[0003] In China, agricultural water use accounts for 63.0% of total water use, and in 2022 the total water use reached 599.82 billion cubic meters, of which agricultural water use accounted for 378.13 billion cubic meters

[0004] However, agricultural water waste is still serious, and the effective utilization coefficient of farmland irrigation water is only 0.572, far below the international advanced level (0.8-0.9). According to research by the Food and Agriculture Organization of the United Nations, crop evapotranspiration is an important indicator for measuring crop water consumption.

[0005] Currently, evapotranspiration (ET O ) can be determined by actual measurement and mathematical model. Although the actual measurement method can accurately measure ET O ,However, this method is cumbersome, costly and technically complex, and its practical application is limited. O The mathematical relationship between the ET and meteorological elements is estimated, which requires little work and is simple to calculate. Among them, the reference crop evapotranspiration calculation model (FAO 56P-M) proposed by the Food and Agriculture Organization of the United Nations (FAO) is the most widely used standard calculation method in the world, but its calculation process is complicated and requires a large amount of meteorological data, including solar radiation, temperature, sunshine time, wind speed and average relative humidity, etc. Therefore, its application is limited in areas where meteorological data is lacking [3]. To this end, domestic and foreign scholars have proposed many ET calculation methods based on less meteorological data. O Simplified calculation models, such as Hargreave-Samani, Jensen-Haise, Makkink, McCloud, Irmark Allen, etc. Although these models have good performance, their calculation accuracy varies significantly between different climate zones due to differences in model principles, geographical location of the study area, and hydrological and meteorological conditions. Therefore, it is more convenient to obtain more accurate and universal ET using less meteorological data. O Computational models still need to be further explored.

[0006] In recent years, with the development of machine learning, machine learning algorithms have been widely used in computing ET. O The application of machine learning in this area has made great progress. For example, Fang et al. used machine learning methods to construct a reference evapotranspiration prediction model, demonstrating its practical value in improving crop production efficiency and adaptability. Yu et al. proposed an extreme gradient boosting model (XGBoost) optimized by a particle swarm algorithm and applied it to the prediction of reference evapotranspiration in a solar greenhouse, achieving significant results, indicating that the model can accurately estimate reference evapotranspiration under different data modes.

[0007] Wu et al. used an extreme learning machine (ELM) optimized by a particle swarm algorithm to estimate the reference evapotranspiration in different climate zones in China. The experimental results showed that the model had good prediction results.

[0008] Petkovic et al. constructed a radial basis function network (RBFN) and optimized it using a particle swarm algorithm to predict reference evapotranspiration in Serbia during the period 1980–2010;

[0009] The results show that the model has good statistical properties and is helpful for the estimation of reference evapotranspiration.

[0010] Although scholars have achieved good results in predicting evapotranspiration, there is still a problem of low prediction accuracy caused by the instability and nonlinear coupling of evapotranspiration data.

[0011] To this end, a multi-source sensing intelligent irrigation system integrating multi-scale optimization was designed to provide another technical solution to the above technical problems. Summary of the invention

[0012] Based on this, it is necessary to provide a multi-source sensing intelligent irrigation system integrating multi-scale optimization to solve the technical problems raised in the above background technology.

[0013] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0014] A multi-source sensing smart irrigation system integrating multi-scale optimization, including a support module, a photovoltaic power generation module, a light detection module, a water treatment module, a drip irrigation module, a temperature and humidity detection module, a control module, and a prediction module;

[0015] The support module is used to provide position support for the photovoltaic power generation module, the light detection module, the drip irrigation module, the water treatment module and the temperature and humidity detection module;

[0016] The light detection module is used to detect the light intensity inside the support module, determine whether the light intensity inside the greenhouse meets the requirements, and transmit the detection signal to the control module;

[0017] The water treatment module is used to process the collected rainwater and transmit the processed rainwater to the drip irrigation module;

[0018] The drip irrigation module is used for drip irrigation of the land;

[0019] The temperature and humidity detection module is used to detect the humidity inside the support module and transmit the signal to the control module;

[0020] The control module is used to control the operation of the drip irrigation module after receiving signals from the light detection module and the temperature and humidity detection module;

[0021] The prediction module is used to predict the evapotranspiration of water through the evapotranspiration prediction model, and dynamically adjust the irrigation frequency and water volume of the drip irrigation module.

[0022] As a preferred implementation of the multi-source sensing intelligent irrigation system integrating multi-scale optimization provided by the present invention, the support module is constructed outside the land where drip irrigation is required to form a greenhouse.

[0023] As a preferred implementation of the multi-source sensing intelligent irrigation system integrating multi-scale optimization provided by the present invention, the inclination angle of the top of the greenhouse is 30° to 45°.

[0024] As a preferred implementation of the multi-source sensing intelligent irrigation system integrating multi-scale optimization provided by the present invention, a nano titanium dioxide (TiO2) coating is used on the inner side of the greenhouse.

[0025] As a preferred implementation of the multi-source sensing intelligent irrigation system integrating multi-scale optimization provided by the present invention, the light detection module is a light temperature sensor;

[0026] The humidity detection module is a temperature and humidity sensor.

[0027] As a preferred embodiment of the multi-source sensing intelligent irrigation system integrating multi-scale optimization provided by the present invention, the water treatment module at least includes a water storage tank, and the collected rainwater is stored in the water storage tank;

[0028] The inner side of the water storage tank includes a pretreatment module, a sedimentation module and a filtration module;

[0029] The pre-treatment module is used to remove large particles in the rainwater entering the water storage tank;

[0030] The sedimentation module is used to separate the sediment and clean water from the rainwater treated by the pretreatment module through sedimentation;

[0031] The filtering module is used to filter the clean water obtained by the sedimentation module again.

[0032] As a preferred implementation of the multi-source sensing intelligent irrigation system integrating multi-scale optimization provided by the present invention, the evapotranspiration prediction model is constructed as follows:

[0033] Construct a CNN-BiGRU-Attention neural network model through convolutional neural network, bidirectional gated recurrent unit and self-attention mechanism;

[0034] The CNN-BiGRU-Attention neural network model is optimized through the adaptive spiral flying sparrow search algorithm to form an evapotranspiration prediction model.

[0035] As a preferred implementation of the multi-source perception intelligent irrigation system integrating multi-scale optimization provided by the present invention, the evapotranspiration prediction model is formed by optimizing the CNN-BiGRU-Attention neural network model through an adaptive spiral flying sparrow search algorithm, and the steps are as follows:

[0036] 1) Extract features from measured data, normalize the data, and randomly divide the data into two groups, one of which is used as a training set for parameter learning of the neural network model, and the other is used as a test set to verify the generalization performance of the model;

[0037] 2) Set up the structure of the CNN-BiGRU-Attention neural network;

[0038] 3) Initialize the population through the adaptive spiral flying sparrow search algorithm;

[0039] 4) Sort individuals according to their fitness values ​​to ensure that individuals with the best fitness have priority in the population;

[0040] 5) Update position through adaptive spiral flight sparrow search algorithm;

[0041] 6) After completing the position update, recalculate the fitness value of each individual, and record the new optimal and worst fitness values ​​and their corresponding positions;

[0042] 7) Check whether the current iteration has reached the preset maximum number of iterations or accuracy requirements. If the termination condition is met, output the optimal individual position and its corresponding minimum fitness value, and proceed to step 8). If the termination condition is not met, return to step 5) and continue iterative updating;

[0043] 8) Assign the optimal hyperparameters optimized by the adaptive spiral flying sparrow search algorithm to the CNN-BiGRU-Attention neural network model to initialize the network parameter configuration;

[0044] 9) Form a stable and accurate evapotranspiration prediction model.

[0045] As a preferred implementation of the multi-source sensing intelligent irrigation system integrating multi-scale optimization provided by the present invention, the position is updated by an adaptive spiral flying sparrow search algorithm, and the steps are as follows:

[0046] Update the position of the discoverer individual to find high fitness areas in the population;

[0047] Update the position of the follower individual;

[0048] Update the location of individuals who sense danger

[0049] It can be seen without a doubt that the above-mentioned technical solution of the present application can definitely solve the technical problem to be solved by the present application.

[0050] At the same time, through the above technical solutions, the present invention has at least the following beneficial effects:

[0051] 1. The multi-source sensing intelligent irrigation system integrated with multi-scale optimization provided by the present invention effectively collects rainwater and uses thermochromic materials to automatically adjust the transmittance, thereby reducing the light intensity in high temperature weather and reducing water evaporation loss; at the same time, rainwater is monitored in real time by online water quality sensors for indicators such as pH value and turbidity, and is filtered and processed to ensure the safety and quality of irrigation water and prevent clogging of the drip irrigation system; the photovoltaic power generation module provides green and sustainable power support for the entire irrigation system to ensure its efficient and independent operation, and based on the constructed evaporation prediction model, the system can accurately estimate the water demand of crops, thereby optimizing the irrigation frequency and water volume of the drip irrigation system; this comprehensive solution provides a sustainable water resource management solution for modern agriculture, has good application prospects and economic benefits, and can effectively respond to the increasingly severe water resource challenges.

[0052] 2. The invention further improves the prediction accuracy and stability through the hyperparameter configuration of the ASFSSA optimization model; through the accurate prediction of evapotranspiration, the intelligent drip irrigation system can dynamically adjust the irrigation frequency and water volume to ensure that the water supply of crops matches the actual demand, which not only reduces water evaporation and runoff losses, but also improves water utilization efficiency.

[0053] 3. The irrigation system of the present invention delivers water directly to the roots of plants, significantly reducing water evaporation and surface runoff, and can usually save 30% to 70% of water resources; at the same time, the system can flexibly adjust the water supply according to the needs of different growth stages of crops, further optimizing the utilization of water resources; in addition, the drip irrigation system provides a stable soil moisture environment, which is conducive to the healthy development of crop roots, promotes overall health and growth rate, and thus improves crop yield and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 It is a model diagram of the present invention;

[0056] Figure 2 The photovoltaic power generation principle diagram of the present invention;

[0057] Figure 3 It is the water purification principle diagram of the present invention;

[0058] Figure 4 The crop drip irrigation map of the present invention;

[0059] Figure 5 It is a model flow chart of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0061] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0062] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features and technical solutions in the embodiments may be combined with each other.

[0063] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0064] Reference Figure 1-Figure 5 , a multi-source sensing intelligent irrigation system integrating multi-scale optimization, including support module, photovoltaic power generation module, light detection module, water treatment module, drip irrigation module, temperature and humidity detection module, control module and prediction module;

[0065] The support module is used to provide position support for the photovoltaic power generation module, light detection module, drip irrigation module, water treatment module and temperature and humidity detection module, and can be built outside the land where drip irrigation is required to form a greenhouse;

[0066] The inclination angle of the top of the greenhouse is 30° to 45°, which can better collect rainwater through the inclination and guide the collected rainwater into the water treatment module;

[0067] Thermochromic materials (such as VO2-doped thermochromic microgels) are used inside the greenhouse to automatically adjust the light transmittance according to changes in ambient temperature, ensuring that when the temperature in the greenhouse rises, the light entering is automatically reduced, thereby reducing water evaporation; when the temperature drops, the material increases the light transmittance to ensure that the crops get enough light. It performs well in controlling temperature and reducing energy consumption;

[0068] In other embodiments, a nano-titanium dioxide (TiO2) coating may be used on the inside of the greenhouse, which can effectively reflect infrared rays while allowing visible light to pass through, thereby helping to control the temperature in the greenhouse, and also has excellent weather resistance and adhesion, maintaining its performance for a long time, thereby reducing energy consumption and improving energy saving effects;

[0069] Generate electricity through sunlight, so that the irrigation system can form a power reserve, which will not affect the crop lighting and can effectively utilize solar energy resources;

[0070] Photovoltaic power generation modules specifically include photovoltaic panels. By installing photovoltaic panels on the top of the greenhouse, photovoltaic panels use high-efficiency monocrystalline or polycrystalline silicon photovoltaic cell modules with a conversion efficiency of 15%-22%, providing sustainable power support for the entire irrigation system, ensuring efficient operation of the system, and daily power generation of 2-5 kWh. The battery system capacity is 5-10kWh to ensure normal operation of the equipment on cloudy days or at night. The inverter has a maximum power point tracking (MPPT) function to optimize the conversion of photovoltaic power.

[0071] The light detection module is used to detect the light intensity inside the supporting module, determine whether the light intensity inside the greenhouse meets the requirements, and transmit the detection signal to the control module. Specifically, the light detection module is a light temperature sensor.

[0072] The water treatment module is used to process the collected rainwater and transmit the processed rainwater to the drip irrigation module, so that the rainwater can be collected and utilized;

[0073] The water treatment module at least includes a water storage tank, through which the collected rainwater is stored;

[0074] The inside of the water storage tank includes a pretreatment module, a sedimentation module, and a filtration module;

[0075] The pre-treatment module is used to remove large particles in the rainwater entering the water storage tank;

[0076] The sedimentation module is used to separate the sediment and clean water from the rainwater treated by the pretreatment module through sedimentation;

[0077] The filtration module is used to filter the clean water obtained from the sedimentation module again to prevent impurities from entering the drip irrigation module, which in turn causes the drip irrigation module to be blocked. Specifically, sand filtration and activated carbon filtration can be used to effectively remove suspended matter and organic matter.

[0078] It also includes an online water quality sensor to detect the pH value of the water filtered by the filtration module to maintain the pH value between 6.0-7.5, as well as indicators such as turbidity (should be less than 5NTU) to ensure that the water quality meets irrigation standards and prevent clogging of the drip irrigation system.

[0079] The drip irrigation module is used to drip water the land, thereby reducing water evaporation and runoff losses, while improving water utilization efficiency;

[0080] The temperature and humidity detection module is used to detect the humidity inside the support module and transmit the signal to the control module. Specifically, the humidity detection module is a temperature and humidity sensor;

[0081] The control module is used to control the operation of the drip irrigation module after receiving the signals from the light detection module and the temperature and humidity detection module;

[0082] The prediction module is used to predict the evapotranspiration of water through the evapotranspiration prediction model, and then dynamically adjust the irrigation frequency and water volume of the drip irrigation module.

[0083] The evapotranspiration prediction model is constructed as follows:

[0084] 1. Adaptive Spiral Flying Sparrow Search Algorithm (ASFSSA)

[0085] Sparrow Search Algorithm (SSA) is an optimization algorithm based on the foraging behavior of sparrows. The algorithm simulates the various behaviors of sparrows in the process of finding food, which are mainly divided into three categories: discoverers, joiners and guards. The discoverer is responsible for finding food sources and guiding the group, the joiner obtains food by following the discoverer, and the guard sounds an alarm when a predator is found, prompting other sparrows to quickly hide. SSA simulates these natural behaviors and uses individual fitness values ​​to share information and update positions to solve the optimization problem. However, due to the randomness of population position generation, SSA may lead to slow convergence and easy to fall into local optimal solutions when the initial population quality is poor.

[0086] The adaptive spiral flying sparrow search algorithm (ASFSSA) is a new algorithm that improves on the SSA. ASFSSA proposes four innovative strategies to improve the convergence accuracy of the algorithm and overcome the limitations of SSA. By introducing adaptive mechanisms and spiral search strategies, the algorithm enhances the ability to explore the solution space and makes the search process more efficient. Specifically, ASFSSA can dynamically adjust the search strategy according to the changes in individual fitness, thereby optimizing group behavior and improving overall performance. Through these improvements, ASFSSA shows stronger robustness and faster convergence speed when solving complex optimization problems.

[0087] 1.1 Random variable strategy based on chaotic mapping

[0088] The introduction of tent mapping strategy makes the population initialization more orderly and enhances the controllability of the algorithm. The formula is as follows:

[0089]

[0090] Where N represents the number of particles in the chaotic sequence.

[0091] Then the expression after Bernoulli transformation is:

[0092]

[0093] According to the characteristics of tent mapping, the process of generating chaotic sequences is as follows:

[0094] 1) Randomly generate an initial value of z0, z0∈(0,1); i=1.

[0095] 2) Use formula (2) to generate a z sequence through iteration, and i increases by 1.

[0096] 3) If the number of iterations reaches the maximum value, stop and store the generated z sequence.

[0097] 1.2 Adaptive Weighting Strategy

[0098] The introduction of adaptive weights can improve the quality of the discoverer's individual position, allowing other individuals to converge to the optimal position faster and speed up the convergence. The formula for adaptive weights is as follows:

[0099]

[0100] In the formula, w represents the property of nonlinear change between [0,1]. The formula for improving the discoverer position update is as follows:

[0101]

[0102] By introducing adaptive weights to dynamically adjust the position of sparrows, and using different guide modes and discoverers at different times, the algorithm's search flexibility is improved. As the number of iterations increases, the sparrows converge to the best position, and the larger the weight, the faster the individual moves, thus improving the convergence speed of the algorithm.

[0103] 1.3 Levy flight mechanism strategy

[0104] Levy flight obeys Levy distribution and covers the search space by randomly generating long-distance and short-distance mechanisms. The Levy flight mechanism is introduced to improve the performance of the proposed algorithm. The position update formula is as follows:

[0105] x i (t) = x i (t)+l⊕levy(λ) (5)

[0106] In the formula, x i (t) represents the position of the i-th sparrow in the t-th iteration; ⊕ represents the arithmetic symbol of point-to-point multiplication; l represents the step size control parameter, l = 0.01 (x i (t)-x p ); Levy(λ) represents the path that obeys the Levy distribution, 1<λ≤3.

[0107] Since the Levy distribution is relatively complex, the Mantegant algorithm is usually used to simulate it. The formula is as follows:

[0108]

[0109] In the formula, σv=1, Y=1.5.

[0110] The combination of the Levy flight mechanism and the adaptive weights balances the search method and improves the quality of each solution obtained, thus greatly enhancing the search capability of the algorithm.

[0111] 2.3.4 Variable spiral search strategy

[0112] The variable spiral position update strategy is introduced to make the follower position update more flexible, develop various search paths for position update, and balance the global search and local search of the algorithm. The formula of the variable spiral position update strategy is as follows:

[0113]

[0114] In the formula, the z parameter changes with the number of operations, and the size and amplitude of the spiral are dynamically adjusted according to the COS function. k represents the coefficient of change, k = 5; l represents a uniformly distributed random number, l∈[-1,1].

[0115] 2. CNN-BiGRU-Attention Neural Network Model

[0116] Convolutional Neural Network (CNN): CNN has good spatial feature extraction capabilities. The convolution layer is mainly used to extract the impact of input variables on the current evapotranspiration, and the pooling layer compresses the extracted features to reduce the complexity of the model. However, CNN cannot learn the temporal features in the evapotranspiration sequence, so the Bi-GRU model is further used to learn the temporal dependency in the load sequence.

[0117] Bidirectional Gated Recurrent Unit (Bi-GRU): Bi-GRU uses a bidirectional recurrent network structure to extract feature information from both the forward and backward directions of the time series. Compared with traditional unidirectional GRU or LSTM, Bi-GRU not only captures the historical information of the sequence in the forward pass, but also learns future contextual information in the backward pass. This bidirectional feature makes it more flexible in evapotranspiration prediction, better reflects the changes in crop water demand, and improves the model's ability to capture time-dependent features. The gating mechanism of Bi-GRU dynamically controls the retention and forgetting of memory by updating the gate and resetting the gate, thereby effectively reducing the gradient vanishing problem.

[0118] Self-attention mechanism (Attention): The Attention mechanism enhances the model's ability to focus on important features by calculating the relevance and weight of input features. In evapotranspiration prediction, different time steps or input features have different effects on the results. The Attention mechanism can assign higher weights to important time steps and input features to ensure that the role of key features in the model is fully reflected. This can suppress irrelevant or low-impact variables, optimize the model's computing resource allocation, and make evapotranspiration prediction more accurate.

[0119] 3) CNN-BiGRU-Attention neural network model based on ASFSSA optimization

[0120] The proposed CNN-BiGRU-Attention neural network model combines the feature extraction capability of CNN, the time-dependent learning advantage of Bi-GRU, and the feature weighting capability of the Attention mechanism. In order to further improve the performance of the model, the ASFSSA algorithm is used to optimize its learning rate, the number of BiGRU neurons, the key-value dimension of the Attention mechanism, and the convolution kernel size to obtain the best combination in order to improve the prediction accuracy of the model. The specific steps are as follows:

[0121] 1) Data preprocessing. Extract features from measured data and normalize the data to improve the stability and convergence speed of training. Randomly divide the data into two groups: one group is used as a training set for parameter learning of the neural network model, and the other group is used as a test set to verify the generalization performance of the model.

[0122] 2) Initialize the neural network parameters. Set the structure of the CNN-BiGRU-Attention neural network, including the configuration of the input layer, BiGRU hidden layer, Attention mechanism, and convolutional layer. Define the network's initial weights, biases, and activation functions (such as ReLU), and set error evaluation indicators (such as mean square error MSE) to provide standards for subsequent training and optimization.

[0123] 3) ASFSSA population initialization. According to formula (2), the initial position of the sparrow population is randomly generated, and the population size and maximum number of iterations are set. The fitness value of each individual is calculated to measure the effect of the individual solution under the current network parameter configuration. The best fitness value and its position (i.e., the best individual) of the population, as well as the worst fitness value are recorded separately to guide the subsequent iterative updates of the population.

[0124] 4) Individual fitness sorting. Sort individuals by fitness value to ensure that individuals with the best fitness have priority in the population. This sorting step helps to select and retain the current optimal solution in subsequent iterations and lays the foundation for the allocation of discoverer, follower and scout roles.

[0125] 5) Position update. ASFSSA implements iterative updates of individuals through the following steps to improve the global exploration and local mining capabilities of the population:

[0126] Update the position of the discoverer: Use formula (3) and formula (4) to update the position of the discoverer individual so that it can find the high fitness area in the population.

[0127] Follower position update: Update the position of the follower individual according to formula (7), and balance the global search ability and local development ability of the population through the following strategy.

[0128] Danger-sensing individual position update: For individuals that sense danger, their positions are updated according to formula (4). The adaptive spiral flight strategy is used to improve their ability to escape the local optimum, ensuring that the population finds the global optimal solution in the complex solution space.

[0129] 6) Fitness value update After completing the position update, recalculate the fitness value of each individual, record the new best and worst fitness values ​​and their corresponding positions. This process ensures that the best position in the solution space can be captured after each iteration, thereby gradually approaching the global optimal solution.

[0130] 7) Termination condition judgment Check whether the current iteration has reached the preset maximum number of iterations or accuracy requirements. If the termination condition is met, the optimal individual position and its corresponding minimum fitness value are output; if the condition is not met, return to step 5 and continue iterative update.

[0131] 8) Assign the optimal hyperparameters (learning rate, number of neurons, key-value dimension, and convolution kernel size) after ASFSSA optimization to the CNN-BiGRU-Attention neural network model to initialize the network parameter configuration. This step ensures that the prediction effect of the network is improved under the optimized initial parameters.

[0132] Finally, a more stable and accurate evapotranspiration prediction model was formed. The overall structure of the model is as follows: Figure 5 shown.

[0133] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the present invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A multi-source sensing smart irrigation system integrating multi-scale optimization, characterized by: It includes a support module, a photovoltaic power generation module, a light detection module, a water treatment module, a drip irrigation module, a temperature and humidity detection module, a control module and a prediction module; The support module is used to provide position support for the photovoltaic power generation module, the light detection module, the drip irrigation module, the water treatment module and the temperature and humidity detection module; The light detection module is used to detect the light intensity inside the support module, determine whether the light intensity inside the greenhouse meets the requirements, and transmit the detection signal to the control module; The water treatment module is used to process the collected rainwater and transmit the processed rainwater to the drip irrigation module; The drip irrigation module is used for drip irrigation of the land; The temperature and humidity detection module is used to detect the humidity inside the support module and transmit the signal to the control module; The control module is used to control the operation of the drip irrigation module after receiving signals from the light detection module and the temperature and humidity detection module; The prediction module is used to predict the evapotranspiration of water through the evapotranspiration prediction model, and dynamically adjust the irrigation frequency and water volume of the drip irrigation module.

2. The multi-source sensing intelligent irrigation system integrating multi-scale optimization according to claim 1 is characterized in that: The support module is constructed outside the land where drip irrigation is required to form a greenhouse.

3. The multi-source sensing intelligent irrigation system integrating multi-scale optimization according to claim 2 is characterized in that: The inclination angle of the top of the greenhouse is 30° to 45°.

4. The multi-source sensing intelligent irrigation system integrating multi-scale optimization according to claim 2 is characterized in that: The inner side of the greenhouse is coated with nano titanium dioxide (TiO2).

5. The multi-source sensing intelligent irrigation system integrating multi-scale optimization according to claim 1 is characterized in that: The light detection module is a light temperature sensor; The humidity detection module is a temperature and humidity sensor.

6. The multi-source sensing intelligent irrigation system integrating multi-scale optimization according to claim 1 is characterized in that: The water treatment module at least includes a water storage tank, through which the collected rainwater is stored; The inner side of the water storage tank includes a pretreatment module, a sedimentation module and a filtration module; The pre-treatment module is used to remove large particles in the rainwater entering the water storage tank; The sedimentation module is used to separate the sediment and clean water from the rainwater treated by the pretreatment module through sedimentation; The filtering module is used to filter the clean water obtained by the sedimentation module again.

7. The multi-source sensing intelligent irrigation system integrating multi-scale optimization according to claim 1 is characterized in that: The evapotranspiration prediction model is constructed as follows: Construct a CNN-BiGRU-Attention neural network model through convolutional neural network, bidirectional gated recurrent unit and self-attention mechanism; The CNN-BiGRU-Attention neural network model is optimized through the adaptive spiral flying sparrow search algorithm to form an evapotranspiration prediction model.

8. The multi-source sensing intelligent irrigation system integrating multi-scale optimization according to claim 7 is characterized in that: The CNN-BiGRU-Attention neural network model is optimized by the adaptive spiral flying sparrow search algorithm to form an evapotranspiration prediction model. The steps are as follows: 1) Extract features from measured data, normalize the data, and randomly divide the data into two groups, one of which is used as a training set for parameter learning of the neural network model, and the other is used as a test set to verify the generalization performance of the model; 2) Set up the structure of the CNN-BiGRU-Attention neural network; 3) Initialize the population through the adaptive spiral flying sparrow search algorithm; 4) Sort individuals according to their fitness values ​​to ensure that individuals with the best fitness have priority in the population; 5) Update position through adaptive spiral flight sparrow search algorithm; 6) After completing the position update, recalculate the fitness value of each individual, and record the new optimal and worst fitness values ​​and their corresponding positions; 7) Check whether the current iteration has reached the preset maximum number of iterations or accuracy requirements. If the termination condition is met, output the optimal individual position and its corresponding minimum fitness value, and proceed to step 8). If the termination condition is not met, return to step 5) and continue iterative updating; 8) Assign the optimal hyperparameters after optimization of the adaptive spiral flying sparrow search algorithm to the CNN-BiGRU-Attention neural network model to initialize the network parameter configuration; 9) Form a stable and accurate evapotranspiration prediction model.

9. The multi-source sensing intelligent irrigation system integrating multi-scale optimization according to claim 7 is characterized in that: The position is updated by the adaptive spiral flight sparrow search algorithm. The steps are as follows: Update the position of the discoverer individual to find high fitness areas in the population; Update the position of the follower individual; For individuals who sense danger, the location is updated.

Images