Integrated intelligent irrigation self-adaptive control system and method

Abstract

The invention provides an integrated intelligent irrigation self-adaptive control system and method, and relates to the technical field of intelligent irrigation. Comprising the steps of collecting historical irrigation data and real-time irrigation data; preprocessing the irrigation data to obtain preprocessed irrigation data; an irrigation decision model is constructed through BiLSTM-CNN-Attention, and the irrigation demand change trend in a short time in the future is predicted; according to the irrigation demand change trend in a short time in the future, the irrigation amount needed by plants is calculated, and a self-adaptive irrigation control strategy is generated; and evaluating the irrigation decision model by using the model evaluation index, and dynamically adjusting hyper-parameters of the irrigation decision model according to feedback. According to the method, the irrigation decision model is constructed, the change trend of the irrigation demand in a short time in the future is predicted, the irrigation amount required by plants is calculated, and the self-adaptive irrigation control strategy is generated, so that the overall utilization efficiency of water resources is improved, and the irrigation work is ensured to be efficiently and continuously propelled.

Description

Technical Field

[0001] The present invention relates to the field of intelligent irrigation technology, and in particular to an integrated intelligent irrigation adaptive control system and an integrated intelligent irrigation adaptive control method. Background Art

[0002] Traditional irrigation methods, which mostly rely on manual experience to determine the timing and quantity of irrigation, present numerous drawbacks. For one thing, it's difficult for humans to accurately determine key information such as the actual soil moisture content and crop water requirements. This often results in insufficient irrigation, impacting crop growth, or excessive irrigation, wasting water resources and potentially causing environmental problems such as secondary soil salinization. Furthermore, with the scale and modernization of agriculture, the number of large areas requiring irrigation, such as farmland, orchards, and gardens, continues to increase. Relying solely on manual irrigation management is becoming increasingly unrealistic and inefficient.

[0003] While some existing intelligent irrigation systems can achieve a certain degree of automated control, traditional adaptive irrigation control systems often lack comprehensive and accurate data support and rely heavily on empirical experience or simple statistical analysis, resulting in poorly accurate irrigation decisions. Furthermore, these systems lack effective performance evaluation and corresponding dynamic adjustment mechanisms, making it impossible to continuously optimize irrigation decision-making models based on actual conditions. These systems struggle to adapt to complex and changing environments and plant growth stages, leading to problems such as irrational irrigation and water waste. Summary of the Invention

[0004] The present invention provides an integrated intelligent irrigation adaptive control system and method, which is used to solve the defects of some existing intelligent irrigation systems in the prior art that they cannot change in real time according to actual conditions, resulting in unreasonable irrigation and waste of water resources.

[0005] In one aspect, the present invention provides an integrated intelligent irrigation adaptive control system, comprising: Data collection module, used to collect historical irrigation data and real-time irrigation data.

[0006] The data preprocessing module is used to preprocess historical irrigation data and real-time irrigation data to obtain preprocessed irrigation data.

[0007] The irrigation analysis module is used to build an irrigation decision-making model using BiLSTM-CNN-Attention to predict the changing trend of irrigation demand in the short term.

[0008] The irrigation execution module generates an adaptive irrigation control strategy based on the changing trend of irrigation demand in the short term and the amount of irrigation required by the plants.

[0009] The performance evaluation module uses model evaluation indicators to evaluate the irrigation decision model and dynamically adjusts the hyperparameters of the irrigation decision model based on feedback.

[0010] According to an integrated intelligent irrigation adaptive control system provided by the present invention, the historical irrigation data and the real-time irrigation data include: soil moisture data, irrigation parameter data, and irrigation parameter data.

[0011] Soil moisture data include: field water holding capacity, current soil moisture content, minimum soil moisture content allowed for crops, ratio of allowable soil moisture content to field water holding capacity, and effective soil irrigation depth.

[0012] Irrigation parameter data include: total soil volume in the irrigation area, irrigation water flow, irrigation water utilization efficiency, irrigation area, ratio of irrigation depth to crop water requirement, flow rate of a single sprinkler, number of sprinklers working simultaneously, and irrigation water loss coefficient.

[0013] Irrigation parameter data include: soil bulk density, crop coefficient, and reference evaporation and transpiration.

[0014] According to an integrated intelligent irrigation adaptive control system provided by the present invention, the data preprocessing module includes: The limiting filter unit is used to filter the historical irrigation data and the real-time irrigation data.

[0015] The missing data filling unit is used to supplement the missing data in the historical irrigation data and the real-time irrigation data using the K-nearest neighbor algorithm.

[0016] Normalization unit, used to standardize historical irrigation data and real-time irrigation data from different fields.

[0017] According to the present invention, an integrated intelligent irrigation adaptive control system is provided, which uses BiLSTM-CNN-Attention to build an irrigation decision model, including: The historical irrigation data in the preprocessed irrigation data is input into the BiLSTM-CNN-Attention model.

[0018] Set the initial weights and bias values ​​for each BiLSTM-CNN-Attention layer.

[0019] Use BiLSTM to capture time series features, use CNN to extract features and residual connection, use Attention to process the output and calculate the error.

[0020] Determine whether the number of iterations meets the maximum number of iterations. If so, terminate the training. Otherwise, perform backpropagation based on the calculated error, update the parameters of each layer, and recalculate the error.

[0021] The integrated intelligent irrigation adaptive control system provided by the application comprises the following steps: The time characteristics of the preprocessed irrigation data are obtained through the BiLSTM.

[0022] The data characteristics of the preprocessed irrigation data are obtained through three CNN layers.

[0023] The output of the first convolutional layer after the BiLSTM is added to the output of the third convolutional layer after batch normalization in a residual manner.

[0024] The data characteristics are processed in the Attention layer, and the final value of the model is output.

[0025] The final value of the model is compared with the actual data to determine the error value.

[0026] The process of model prediction of the integrated intelligent irrigation adaptive control system provided by the application is as follows: The real-time irrigation data are input into the trained BiLSTM-CNN-Attention to obtain a predicted output value.

[0027] The predicted output value is converted into a standardized value, and the standardization formula is as follows:

[0028] In the formula, xi is the standardized recovery value, yi is the predicted output value of the BiLSTM-CNN-Attention, s is the standard deviation of the input data, and x is the average value of the input data.

[0029] The recovery result is obtained, and the prediction of the irrigation demand change trend in the future short time is completed.

[0030] The integrated intelligent irrigation adaptive control system provided by the application comprises the following steps: The irrigation duration unit is used to determine the time length of each irrigation operation.

[0031] The irrigation amount unit is used to accurately calculate and set the specific water amount for plants in each irrigation.

[0032] The irrigation frequency unit is used to calculate the number of irrigation operations.

[0033] The integrated intelligent irrigation adaptive control system provided by the application comprises the following steps:

[0034] In the formula, t is the irrigation duration. fcis the field holding capacity, θa is the current soil moisture content, Vs is the total volume of soil in the irrigation area, ρb is the soil bulk density, d is the ratio of the soil moisture content allowed to be consumed to the field holding capacity, I is the irrigation water flow rate, and η is the irrigation water use efficiency.

[0035] According to the present invention, an integrated intelligent irrigation adaptive control system includes a model evaluation unit for evaluating performance indicators of an irrigation decision model.

[0036] A feedback analysis unit is used to perform analysis based on performance indicators.

[0037] The hyperparameter adjustment unit is used to dynamically adjust the hyperparameters of the irrigation strategy decision model based on the analysis results.

[0038] On the other hand, the present invention also provides an integrated intelligent irrigation adaptive control method, comprising: S1: Collect historical irrigation data and real-time irrigation data.

[0039] S2: Preprocess the historical irrigation data and the real-time irrigation data to obtain preprocessed irrigation data.

[0040] S3: Build an irrigation decision model using BiLSTM-CNN-Attention to predict the changing trend of irrigation demand in the short term.

[0041] S4: Based on the changing trend of irrigation demand in the short term in the future and the calculation of the irrigation amount required by the plants, an adaptive irrigation control strategy is generated.

[0042] S5: Use model evaluation indicators to evaluate the irrigation decision model and dynamically adjust the hyperparameters of the irrigation decision model based on feedback.

[0043] This invention provides an integrated intelligent adaptive irrigation control system and method. Using BiLSTM-CNN-Attention to build an irrigation decision model, it predicts the changing trend of irrigation demand in the short term, calculates the required irrigation amount for plants, and generates an adaptive irrigation control strategy. The beneficial effects achieved are: Precision irrigation is achieved. The system widely collects historical irrigation data and real-time irrigation data through the data collection module, and comprehensively grasps the past irrigation conditions and the current real-time information such as plant growth environment, soil moisture, meteorological conditions, etc. After standardized processing by the data preprocessing module, it provides a high-quality data foundation for subsequent analysis. The irrigation analysis module uses the advanced BiLSTM-CNN-Attention model to deeply mine the data and accurately predict the changing trend of irrigation demand in the short term in the future. The irrigation execution module calculates the required irrigation amount of the plant and generates an adaptive strategy based on this. This series of processes ensures that the irrigation amount, irrigation duration and irrigation frequency can be accurately matched with the actual needs of the plant, avoiding insufficient or excessive irrigation, so that the plants are always in a suitable water environment, which helps them thrive, improve crop yield and quality, or maintain the good condition of landscape plants.

[0044] Precise irrigation control means that water resources will not be wasted unnecessarily. The system can supply water on demand, meeting the needs of plant growth while minimizing unnecessary water consumption. This is of great significance for areas with tight water resources. It is in line with the current sustainable development requirements for water conservation, improves the overall utilization efficiency of water resources, and ensures that irrigation work can be carried out stably in the long term without worrying about water shortages.

[0045] The automated and intelligent operation of the entire system replaces the traditional model of frequent manual monitoring and adjustment of irrigation operations. Each module performs its own functions and works together to systematically complete irrigation decision-making and execution without constant human oversight. This makes irrigation management easy and convenient, freeing up significant manpower and allowing personnel to focus on other more critical tasks, effectively reducing irrigation management operating costs.

[0046] The performance evaluation module evaluates the irrigation decision-making model through scientific model evaluation indicators and dynamically adjusts hyperparameters based on feedback, so that the model can continuously adapt to environmental changes, changes in plant growth stages and new data conditions, ensuring that the system always maintains good performance during long-term use, flexibly responds to various complex irrigation scenarios, and ensures that irrigation work continues to advance scientifically, reasonably and efficiently. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. 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 any creative work.

[0048] Figure 1This is a module diagram of an integrated intelligent irrigation adaptive control system provided by an embodiment of the present invention; Figure 2 This is a flow chart of an integrated intelligent irrigation adaptive control system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0050] The following combination Figure 1-Figure 2 An integrated intelligent irrigation adaptive control system and method of the present invention are described.

[0051] Figure 1 This is a module diagram based on an integrated intelligent irrigation adaptive control system provided by an embodiment of the present invention.

[0052] like Figure 1 As shown, an embodiment of the present invention provides an integrated intelligent irrigation adaptive control system and method, the system comprising: Data collection module, used to collect historical irrigation data and real-time irrigation data.

[0053] Soil moisture data include: field water holding capacity, current soil moisture content, minimum soil moisture content allowed for crops, ratio of allowable soil moisture content to field water holding capacity, and effective soil irrigation depth.

[0054] Field capacity is the maximum amount of capillary water that the soil can retain after all gravity-induced water has been completely drained. The field plot irrigation method involves fully irrigating selected field plots, covering them to prevent evaporation, and then taking soil samples in separate layers to determine their moisture content. Current soil moisture content is the actual amount of water currently contained in the soil. A drying method involves drying a soil sample in an oven to a constant weight, and calculating the current soil moisture content based on the difference in mass before and after drying. The minimum allowable soil moisture content for crops is the lowest moisture content the soil can reach during crop growth. This is determined through long-term field trials, monitoring of crop physiological indicators, and observation of crop growth performance. The ratio of permissible soil moisture content to field capacity is the ratio of soil moisture content that can be consumed relative to field capacity while ensuring normal crop growth. The effective irrigation depth is the soil depth range at which irrigation water can effectively penetrate and replenish the crop's root zone. Soil texture affects water infiltration rate and water holding capacity. Loose sandy soils infiltrate quickly but hold little water, while clay soils infiltrate slowly but can reach higher water contents at shallow depths. The depth and density of crop roots determine the effective range of irrigation water penetration. Shallow-rooted crops, for example, have relatively shallow effective irrigation depths, while deep-rooted crops require deeper effective irrigation depths. Irrigation parameter data includes: total soil volume in the irrigation area, irrigation water flow rate, irrigation water use efficiency, irrigated area, the ratio of irrigation depth to crop water requirements, flow rate of a single sprinkler, the number of sprinklers operating simultaneously, and the irrigation water loss coefficient.

[0055] The total soil volume of the irrigation area is the volume of the entire soil area requiring irrigation. First, measure the planar surface area of ​​the irrigation area, then determine the soil depth involved in irrigation. Multiply the two to obtain the total volume. Irrigation water flow rate is the amount of water flowing through the irrigation system per unit time. This is measured using an electromagnetic flowmeter. Installed on the main or branch pipes of the irrigation system, it monitors water flow in real time. Irrigation water use efficiency measures the proportion of water actually used by crops during the irrigation process as a percentage of the total irrigation water volume. The irrigated area is the area where irrigation is being carried out, determined by measuring length and width. The ratio of irrigation depth to crop water requirement provides a quantitative relationship between irrigation depth and overall crop water requirement. The irrigation water loss coefficient measures the proportion of water lost during the irrigation process due to various factors, such as seepage and evaporation.

[0056] Irrigation parameter data include: soil bulk density, crop coefficient, and reference evaporation and transpiration.

[0057] Soil bulk density is the oven-dry mass of undisturbed soil per unit volume. The crop coefficient reflects the difference in water requirements of different crops at different growth stages relative to a reference crop. This is derived through extensive field trials, where water requirements of different crops at different growth stages are observed in different regions and years, and then compared with reference crops. Reference evapotranspiration (ET) is an estimate of the ET of a standard reference crop based on factors such as meteorological conditions.

[0058] The data preprocessing module is used to preprocess the irrigation data to obtain preprocessed irrigation data. The data preprocessing module includes a limiting filter unit, a missing data filling unit and a normalization unit.

[0059] The limiting filter unit is used to filter irrigation data. During IoT data collection, sensors are often subject to various random interferences, resulting in significant fluctuations in the collected data, which in turn affects the accuracy and reliability of the data. To ensure data accuracy, the collected data must be properly filtered. By comparing the data values ​​at adjacent time points and setting a maximum allowable error, the data is filtered and corrected. When the difference between the data collected at a certain moment and the data at the previous moment exceeds the preset maximum allowable error, the system will regard the data as an outlier and delete it. Conversely, if the difference is within the allowable range, it is considered valid data and retained. In this way, abnormal data caused by random interference is effectively eliminated, thereby improving the overall accuracy and reliability of IoT data collection.

[0060] The missing data filling unit is used to fill in missing data using the K-nearest neighbor algorithm. Due to potential factors such as sensor failure, network interruption, or other unforeseen reasons, data may be missing. In order to ensure the integrity of the data, the K-nearest neighbor algorithm is used to fill in missing data. KNN has the advantages of being simple, easy to understand and implement in filling in missing data in the agricultural Internet of Things. Euclidean distance is selected as the distance measurement method to measure the similarity between different sensor nodes. For each data point, the Euclidean distance between it and other complete data points is calculated, and its definition is:

[0061] Where x and y are the feature vectors of the two sensor nodes. Calculate the distances to other nodes and select the k closest nodes as nearest neighbors. Calculate the average humidity value of the k nearest neighbors to fill in the missing value for that node. Add the filled data to the dataset and update the data table.

[0062] The normalization unit is used to standardize historical and real-time irrigation data from different fields. Data normalization helps eliminate bias caused by data of different scales or ranges, which helps to speed up the convergence of the model and improve model performance.

[0063]

[0064] In the formula, x represents the normalized value of the data collected by the sensor, x represents the value actually measured by the sensor, min represents the minimum value in the sample, and max represents the maximum value.

[0065] The irrigation analysis module is used to determine the current irrigation demand and predict the changing trend of irrigation demand in the short term through the irrigation decision model constructed by the RBF neural network algorithm.

[0066] The BiLSTM-CNN-Attention model is a hybrid deep learning architecture that combines a bidirectional long short-term memory (BiLSTM) network, a convolutional neural network (CNN), and an attention mechanism. It is specifically designed for processing complex time series data. BiLSTM processes sequential data bidirectionally to capture long-term dependencies. CNN uses convolution operations to extract local features and spatial patterns. Residual connections fuse shallow and deep features to mitigate the vanishing gradient problem. The attention mechanism dynamically assigns weights to focus on key time points and environmental factors. In irrigation decision-making systems, this model can simultaneously learn from multiple sources of information, including historical irrigation patterns, soil moisture changes, and meteorological data, to accurately predict short-term trends in irrigation demand. It then combines crop water demand patterns with environmental parameters to calculate the optimal irrigation amount and ultimately generate an adaptive irrigation control strategy. Its advantages lie in end-to-end learning, automatic feature extraction, and high-precision prediction. It is particularly suitable for processing irrigation data with complex spatiotemporal dependencies, providing intelligent and efficient decision support for agricultural water resources management.

[0067] The crop irrigation quantity prediction method based on the BiLSTM-CNN-Attention model first preprocesses the collected irrigation data and divides it into training and test sets. Next, the BiLSTM-CNN-Attention model is constructed. In this paper's data processing, the data is first passed to the BiLSTM layer and then to the CNN layer. This choice is based on the consideration that applying the CNN layer first may result in loss of sequential information in the data. Although the BiLSTM layer can be used later, this order may not fully utilize the BiLSTM's advantages in sequence encoding, which may reduce the model's predictive performance.

[0068] This model uses BiLSTM to capture the temporal features of irrigation data. It further explores the internal characteristics of irrigation volume through a CNN architecture consisting of three one-dimensional convolutional layers and a pooling layer. The convolution layer efficiently extracts local features of the irrigation data, while the pooling layer simplifies these features through max pooling, highlighting more core information. To address potential sample shift, batch normalization is added after the first two convolutions. To address gradient issues, a residual link is introduced between the two convolutions. The outputs of the convolutional layers are then fed into an attention layer, which assigns different weights to the temporal features, strengthening the temporal correlation with the irrigation data. This mechanism helps highlight key time nodes while reducing interference from irrelevant information. A fully connected layer finally completes the irrigation volume prediction. To improve the model's robustness, the dropout technique is introduced to further enhance its versatility. For parameter optimization, the model uses the Adam optimizer and a mean squared error loss function for training. After training, the model is evaluated on a test set, and its prediction results are analyzed in depth to refine its performance.

[0069] The irrigation decision model constructed by BiLSTM-CNN-Attention includes: The preprocessed irrigation data is fed into the BiLSTM-CNN-Attention model.

[0070] Set the initial weights and bias values ​​for each BiLSTM-CNN-Attention layer.

[0071] The temporal characteristics of irrigation data are captured by BiLSTM, generating input for the CNN layer and producing corresponding output.

[0072] The data passes through three CNN layers (including convolution and pooling) to extract data features and generate output.

[0073] The residuals of the first convolutional layer output after BiLSTM and the third convolutional layer output after batch normalization are accumulated.

[0074] The output of CNN is processed in the Attention layer and the corresponding output is generated.

[0075] Get the output from the Attention layer and generate the final value of the model.

[0076] Compare the model output with the actual data to determine the error value.

[0077] Determine whether the end condition of the prediction process is met. If so, terminate the training. Otherwise, perform backpropagation based on the calculated error, update the parameters of each layer, and then return to step 4 and continue training.

[0078] The model prediction process is: The real-time irrigation data is input into the trained BiLSTM-CNN-Attention to obtain the predicted output value.

[0079] Convert the predicted output value into a standardized value. The standardization formula is as follows:

[0080] Where x i is the normalized recovery value, y i is the predicted output value of BiLSTM-CNN-Attention, s is the standard deviation of the input data, and x is the mean of the input data.

[0081] Get the restored results and complete the prediction process.

[0082] The irrigation execution module identifies the peak and trough periods of water demand based on the changing trend of irrigation demand in the short term. It also calculates the amount of irrigation required by plants and generates an adaptive irrigation control strategy by analyzing the peak, valley, and slope of the irrigation demand trend curve.

[0083] Based on the water requirements of different crops at different growth stages and current environmental conditions, key parameters such as the optimal irrigation duration, irrigation amount, and irrigation frequency are calculated.

[0084] The irrigation duration unit is used to determine the duration of each irrigation operation to ensure that the plants can get the appropriate amount of water. The calculation formula for irrigation duration is:

[0085] Where t represents the irrigation duration. fc θ is the field capacity, that is, the maximum capillary suspended water content that the soil can hold, representing the water content of the soil in an ideal moist state. a is the current soil moisture content, which can be obtained through the soil moisture sensor. s is the total volume of soil in the irrigated area, which needs to be calculated based on parameters such as the area of ​​the irrigated area and the soil depth, for example, V s =A×h (A is the irrigation area, h is the soil depth). b =soil bulk density, reflecting soil properties such as compaction. Different soil types have corresponding common bulk density ranges. d is the allowable soil moisture content as a percentage of field capacity, generally ranging from 0.3 to 0.7, depending on factors such as the crop's sensitivity to water. For example, d is typically lower for water-sensitive crops. I is the irrigation water flow rate, which depends on the specifications of the irrigation equipment and operating pressure. η is the irrigation water use efficiency, generally ranging from 0.7 to 0.95, taking into account water losses due to leakage, evaporation, and other factors during the irrigation process.

[0086] The irrigation amount unit is designed to accurately calculate and set the specific amount of water given to plants during each irrigation to meet the water needs of plant growth. The irrigation amount calculation formula is:

[0087] Where Q is the irrigation amount, A is the irrigation area, and ρ b is the soil bulk density, V s is the total volume of soil in the irrigated area, θ fc θ is the field capacity, that is, the maximum capillary suspended water content that the soil can hold, representing the water content of the soil in an ideal moist state. a is the current soil moisture content, which can be obtained through the soil moisture sensor.

[0088] The irrigation frequency unit is used to plan the number of irrigation operations within a certain period of time so that plants can be replenished with water in a timely manner.

[0089]

[0090] Where f represents the irrigation frequency. Tc is the crop water requirement. a is the current soil moisture content, θ fc is the field capacity. w The minimum soil moisture content allowed for crops. Below this value, crop growth will be significantly affected. Different crops have different critical values. b is the soil bulk density. h is the effective irrigation depth of the soil, that is, the depth of the soil layer where water can effectively affect the crop root system.

[0091] The adaptive irrigation strategy achieves zoned irrigation through a wireless solenoid valve network, generating differentiated irrigation plans based on the soil texture and crop type of different plots.

[0092] For irrigation equipment, choose devices with intelligent control features, including sprinklers and drip irrigation pipes that can be remotely controlled to adjust the on / off switch, water flow, and irrigation duration. These devices should be well connected to sensors and the control system. Soil moisture sensors and weather stations should be strategically distributed throughout the planting area. Soil moisture sensors provide real-time feedback on soil moisture content, while weather stations continuously provide the latest weather information, which is promptly transmitted to the control system.

[0093] During the irrigation process, if the soil moisture reaches or approaches the upper limit of the suitable range, or if meteorological data indicates that rainfall or other conditions that can replenish water are imminent, the control system will promptly instruct the irrigation equipment to stop working, avoiding excessive irrigation that causes waste of water resources and possible adverse effects on plant growth, such as root hypoxia.

[0094] A feedback mechanism is implemented to continuously monitor soil moisture, plant growth status, and other indicators after each irrigation operation. The feedback is used to further optimize the parameters of the irrigation control strategy. In a specific embodiment, adjustments are made to the upper and lower soil moisture thresholds, the irrigation duration calculation coefficient, and other factors to ensure that the entire irrigation control strategy is more aligned with the actual irrigation needs of the plants, ensuring healthy plant growth while achieving efficient water resource utilization.

[0095] Taking into account the stability and fault tolerance of the system, backup power supplies and emergency plans to deal with abnormal situations such as sensor failure and communication interruption are set up to ensure that even if some emergencies occur, reasonable irrigation operations can be maintained as much as possible to avoid causing serious damage to plant growth.

[0096] The performance evaluation module uses model evaluation indicators to evaluate the irrigation decision model and dynamically adjusts the hyperparameters of the irrigation decision model based on feedback.

[0097] MAE (mean absolute error), MSE (mean square error), RMSE (root mean square error), and R2 (goodness of fit) are used as evaluation indicators for the model. The smaller the MAE, MSE, and RMSE, the better the model, and the larger the R2, the better the model. The MAE calculation formula is:

[0098] Where, is the predicted value, and yi is the actual value.

[0099] The MSE calculation formula is expressed as:

[0100] Where, is the predicted value, and yi is the actual value.

[0101] The RMSE calculation formula is expressed as:

[0102] R 2 The calculation formula is expressed as:

[0103] Where, is the predicted value, y i is a real value, is the average value.

[0104] Based on feedback from the evaluation, if the model's accuracy is insufficient, it may be due to an improperly set learning rate, which prevents the model from fully learning the patterns in the data. Appropriately adjusting the learning rate will allow the model to learn better in subsequent runs or retraining, thereby improving the accuracy of decisions such as irrigation volume prediction. If the model is unstable, it may be necessary to adjust some hyperparameters related to data normalization to enable the model to more appropriately process data of varying magnitudes, thereby enhancing its ability to produce stable irrigation decisions under different circumstances.

[0105] Adjustments are made continuously based on feedback from each evaluation. The environment is constantly changing, and plant growth is also dynamic. The hyperparameters of the irrigation decision-making model need to be continuously optimized to ensure that it consistently adapts to the actual situation and outputs high-quality irrigation decisions, ultimately achieving precise, efficient, and adaptive irrigation control.

[0106] In summary, this embodiment provides an integrated intelligent adaptive irrigation control system and method. Using BiLSTM-CNN-Attention, an irrigation decision model is constructed to predict the changing trend of irrigation demand in the short term, calculate the required irrigation amount for plants, and generate an adaptive irrigation control strategy. The following beneficial effects are achieved: Precision irrigation is achieved. The system widely collects historical irrigation data and real-time irrigation data through the data collection module, and comprehensively grasps the past irrigation conditions and the current real-time information such as plant growth environment, soil moisture, meteorological conditions, etc. After standardized processing by the data preprocessing module, it provides a high-quality data foundation for subsequent analysis. The irrigation analysis module uses the advanced BiLSTM-CNN-Attention model to deeply mine the data and accurately predict the changing trend of irrigation demand in the short term in the future. The irrigation execution module calculates the required irrigation amount of the plant and generates an adaptive strategy based on this. This series of processes ensures that the irrigation amount, irrigation duration and irrigation frequency can be accurately matched with the actual needs of the plant, avoiding insufficient or excessive irrigation, so that the plants are always in a suitable water environment, which helps them thrive, improve crop yield and quality, or maintain the good condition of landscape plants.

[0107] Based on the same general inventive concept, the present invention also protects an integrated intelligent irrigation adaptive control method. The integrated intelligent irrigation adaptive control method provided by the present invention is described below. The integrated intelligent irrigation adaptive control method described below and the integrated intelligent irrigation adaptive control system described above can be referenced to each other.

[0108] Figure 2 It is a structural diagram of an integrated intelligent irrigation adaptive control method provided by an embodiment of the present invention.

[0109] like Figure 2As shown, an integrated intelligent irrigation adaptive control method includes: S1: Collect historical irrigation data and real-time irrigation data.

[0110] S2: Preprocess the historical irrigation data and the real-time irrigation data to obtain preprocessed irrigation data.

[0111] S3: Build an irrigation decision model using BiLSTM-CNN-Attention to predict the changing trend of irrigation demand in the short term.

[0112] S4: Based on the changing trend of irrigation demand in the short term in the future and the calculation of the irrigation amount required by the plants, an adaptive irrigation control strategy is generated.

[0113] S5: Use model evaluation indicators to evaluate the irrigation decision model and dynamically adjust the hyperparameters of the irrigation decision model based on feedback.

[0114] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods of each embodiment or certain portions of the embodiments.

[0115] Finally, it should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will appreciate that modifications may be made to the technical solutions described in the aforementioned embodiments, or that some of the technical features may be replaced with equivalents. Such modifications or replacements do not deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An integrated intelligent irrigation adaptive control system, characterized in that: include: Data collection module, used to collect historical irrigation data and real-time irrigation data; A data preprocessing module, configured to preprocess the historical irrigation data and the real-time irrigation data to obtain preprocessed irrigation data; The irrigation analysis module is used to build an irrigation decision model using BiLSTM-CNN-Attention to predict the trend of irrigation demand changes in the short term. The irrigation execution module calculates the irrigation amount required by the plants based on the changing trend of the irrigation demand in the short term and generates an adaptive irrigation control strategy; The performance evaluation module uses model evaluation indicators to evaluate the irrigation decision model and dynamically adjusts the hyperparameters of the irrigation decision model based on feedback.

2. The integrated intelligent irrigation adaptive control system according to claim 1, characterized in that: Historical irrigation data and real-time irrigation data include: soil moisture data, irrigation parameter data, and irrigation parameter data; Soil moisture data include: field water holding capacity, current soil moisture content, minimum soil moisture content allowed for crops, ratio of soil moisture content allowed to be consumed to field water holding capacity, and effective soil irrigation depth; Irrigation parameter data include: total soil volume in the irrigation area, irrigation water flow, irrigation water use efficiency, irrigation area, ratio of irrigation depth to crop water requirement, flow rate of a single sprinkler, number of sprinklers working simultaneously, and irrigation water loss coefficient; Irrigation parameter data include: soil bulk density, crop coefficient, and reference evaporation and transpiration.

3. The integrated intelligent irrigation adaptive control system according to claim 1, characterized in that: The data preprocessing module includes: A limiting filter unit, configured to filter the historical irrigation data and the real-time irrigation data; A missing data filling unit is used to fill in missing data in historical irrigation data and real-time irrigation data using a K-nearest neighbor algorithm; Normalization unit, used to standardize historical irrigation data and real-time irrigation data from different fields.

4. The integrated intelligent irrigation adaptive control system according to claim 1, characterized in that: The irrigation decision model constructed by BiLSTM-CNN-Attention includes: Inputting historical irrigation data in the preprocessed irrigation data into a BiLSTM-CNN-Attention model; Set the initial weights and bias values ​​for each layer of BiLSTM-CNN-Attention; Use BiLSTM to capture time series features, use CNN to extract features and residual connection, use Attention to process the output and calculate the error; Determine whether the number of iterations meets the maximum number of iterations. If so, terminate the training. Otherwise, perform backpropagation based on the calculated error, update the parameters of each layer, and recalculate the error.

5. The integrated intelligent irrigation adaptive control system according to claim 4, characterized in that: The specific steps of error calculation are: Obtaining the temporal characteristics of the pre-processed irrigation data through BiLSTM; Obtaining data features of the preprocessed irrigation data through three CNN layers; The residuals of the output of the first convolutional layer after BiLSTM and the output of the third convolutional layer after batch normalization are accumulated; Process the data features in the Attention layer and output the final model value; The final value of the model is compared with the actual data to determine the error value.

6. The integrated intelligent irrigation adaptive control system according to claim 1, characterized in that: The model prediction process is: Input the real-time irrigation data into the trained BiLSTM-CNN-Attention to obtain a predicted output value; The predicted output value is converted into a standardized value, and the standardization formula is: ； Where xi is the normalized recovery value, yi is the predicted output value of BiLSTM-CNN-Attention, s is the standard deviation of the input data, and x is the mean value of the input data; Obtain the results after recovery and complete the forecast of the changing trend of irrigation demand in the short term in the future.

7. The integrated intelligent irrigation adaptive control system according to claim 1, characterized in that: The irrigation execution module includes: Irrigation duration unit, used to determine the duration of each irrigation operation; Irrigation volume unit, used to accurately calculate and set the specific amount of water given to plants each time of irrigation; Irrigation frequency unit, used to calculate the number of times irrigation operations are performed.

8. The integrated intelligent irrigation adaptive control system according to claim 7, characterized in that: The calculation formula for irrigation duration is: Where, t is the irrigation duration; θ fc is the field capacity, θ a is the current soil moisture content, V s is the total volume of soil in the irrigated area, ρ b is the soil bulk density, d is the ratio of the soil water content allowed to be consumed to the field water holding capacity, I is the irrigation water flow rate, and η is the irrigation water use efficiency.

9. The integrated intelligent irrigation adaptive control system according to claim 1, characterized in that: The performance evaluation modules include: Model evaluation unit, used to evaluate the performance indicators of irrigation decision-making models; a feedback analysis unit, configured to perform analysis based on the performance indicators; The hyperparameter adjustment unit is used to dynamically adjust the hyperparameters of the irrigation strategy decision model based on the analysis results.

10. An integrated intelligent irrigation adaptive control method, applied to an integrated intelligent irrigation adaptive control system according to any one of claims 1 to 9, characterized in that: The adaptive control method comprises: S1: Collect historical irrigation data and real-time irrigation data; S2: preprocessing the historical irrigation data and the real-time irrigation data to obtain preprocessed irrigation data; S3: Build an irrigation decision model using BiLSTM-CNN-Attention to predict the irrigation demand trend in the short term. S4: generating an adaptive irrigation control strategy based on the irrigation demand change trend in the short future period and calculating the irrigation amount required by the plants; S5: Evaluate the irrigation decision model using a model evaluation index, and dynamically adjust hyperparameters of the irrigation decision model based on feedback.

Images