High-standard farmland intelligent irrigation system based on Internet of Things and data analysis

Abstract

The invention relates to the technical field of agricultural intelligent irrigation, and discloses a high-standard farmland intelligent irrigation system based on Internet of Things and data analysis, and the system comprises a soil moisture content sensing module which collects data through a multi-source sensor to construct a three-dimensional soil moisture content distribution model, and generates a soil moisture content characteristic spectrum; the soil moisture content prediction module generates water demand prediction data based on the soil moisture content characteristic spectrum, the meteorological data and the crop growth stage; the irrigation strategy module fuses terrain elevation and pipe network pressure parameters to generate an irrigation control map; the equipment state monitoring module collects water pump current waveform and other data to generate equipment health degree parameters; the pipe network optimization module optimizes pipe network topology and generates an adjusting instruction; the multi-source data fusion module generates fusion evaluation indexes by using an evidence theory, an entropy weight method and the like; and the intelligent execution module generates an execution control instruction accordingly. All the modules cooperate to achieve precise irrigation, and the utilization efficiency of water resources and the intelligent level of farmland management are improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural intelligent irrigation technology, specifically to a high-standard intelligent irrigation system for farmland based on the Internet of Things and data analysis. Background Technology

[0002] In the process of global agricultural modernization, the contradiction between water scarcity and low efficiency of farmland irrigation is becoming increasingly prominent. Traditional farmland irrigation methods rely on manual experience to judge soil moisture, which has problems such as delayed irrigation timing, unreasonable water allocation, and lack of monitoring of equipment operation status. This leads to serious water waste, uneven crop growth, and difficulty in meeting the needs of high-standard farmland for refined management.

[0003] From the perspective of soil moisture monitoring, existing technologies mostly use single-point sensors to collect data, which cannot comprehensively analyze the three-dimensional soil moisture distribution of farmland soil layers and accurately extract soil water-holding capacity characteristics, resulting in a lack of scientific basis for irrigation decisions. For example, traditional methods can only obtain the surface soil moisture content and cannot reflect the water migration patterns at different soil depths, which can easily lead to drought in deep soil or waterlogging in shallow soil, affecting crop root growth.

[0004] In terms of soil moisture forecasting, existing systems mostly process meteorological data based on simple statistical models, failing to effectively analyze the chaotic characteristics of meteorological elements and making it difficult to accurately identify dynamic trends in soil moisture evaporation. Simultaneously, the lack of refined water requirement analysis for crop growth stages prevents the scientific classification of water requirement levels, leading to a mismatch between irrigation volume and actual crop water needs, thus affecting crop yield and quality.

[0005] In the irrigation strategy formulation stage, traditional methods do not fully consider the coupling relationship between farmland topographic elevation data and pipeline pressure parameters, making it difficult to construct hydraulic balance equations to solve for the optimal irrigation duration and to achieve dynamic optimization of regional water allocation schemes. This makes uneven irrigation prone to occur in areas with complex terrain, with some areas receiving insufficient irrigation and others receiving excessive irrigation, thus reducing water resource utilization efficiency.

[0006] In terms of equipment status monitoring, current technologies rely mainly on manual inspections to assess the health of irrigation equipment such as water pumps. This makes it impossible to collect key operating parameters of the equipment in real time and provide early warnings of malfunctions. For example, potential problems such as worn motor bearings and aging insulation cannot be detected in time, leading to sudden equipment failures, affecting the normal operation of the irrigation system, and increasing maintenance and time costs.

[0007] In terms of pipeline network optimization, traditional irrigation pipeline network layouts lack dynamic adjustment mechanisms and cannot optimize the pipeline network topology based on equipment health parameters and historical irrigation data. This results in significant pipeline resistance loss, obvious fluctuations in end-point pressure, and affects the stability and uniformity of irrigation effects.

[0008] The application of multi-source data fusion technology in agricultural irrigation is still immature. Existing systems are unable to effectively integrate and weight heterogeneous data such as soil moisture, meteorological data, and equipment status, and cannot accurately determine the strength of the interaction between parameters, resulting in insufficient scientificity and reliability of irrigation decisions.

[0009] In the intelligent execution stage, traditional control methods lack a lag compensation mechanism for adjusting the opening degree of solenoid valves and pulse frequency, making it impossible to optimize the dynamic response of water flow. This can easily lead to irrigation delays or over-irrigation, affecting irrigation accuracy and efficiency. Summary of the Invention

[0010] The purpose of this invention is to provide a high-standard intelligent irrigation system for farmland based on the Internet of Things and data analysis, so as to solve the problems mentioned in the background art.

[0011] To achieve the above objectives, the present invention provides the following technical solution: a high-standard intelligent irrigation system for farmland based on the Internet of Things and data analysis, the system comprising:

[0012] The soil moisture perception module, based on data collected from multiple sources of sensors, performs layered analysis of farmland soil structure, constructs a three-dimensional soil moisture distribution model, extracts soil water holding capacity characteristics, and generates a soil moisture feature map.

[0013] The soil moisture forecasting module analyzes the chaotic characteristics of meteorological data sequences based on soil moisture feature maps, identifies the dynamic trend of soil moisture evaporation, classifies water demand levels according to crop growth stages, and generates water demand forecast data.

[0014] The irrigation strategy module, based on water demand prediction data, integrates farmland topographic elevation data and pipeline pressure parameters to construct a hydraulic balance equation to solve for the optimal irrigation duration, and uses a dynamic programming algorithm to generate regional water allocation schemes and generate irrigation control maps.

[0015] The equipment status monitoring module, based on the irrigation control map, collects the water pump operating current waveform in real time, separates electromagnetic interference noise and extracts the vibration characteristics of the motor windings, and generates equipment health parameters.

[0016] The pipeline optimization module reconstructs the irrigation pipeline topology based on equipment health parameters, applies graph theory algorithms to solve for the path with the minimum pipeline resistance loss, establishes a pressure compensation model based on historical irrigation data, and generates pipeline regulation commands.

[0017] The multi-source data fusion module, based on soil moisture feature maps, water demand prediction data, irrigation control maps, equipment health parameters, and pipeline regulation instructions, uses evidence theory to weight the heterogeneous data on credibility, determines the interaction strength between parameters through the entropy weight method, and generates fusion evaluation indicators.

[0018] The intelligent execution module, based on fusion evaluation indicators, matches the opening and closing sequence of irrigation valves, adjusts the opening degree and pulse frequency of solenoid valves by combining a hysteresis compensation mechanism, optimizes the dynamic response of water flow by using fuzzy PID control, and generates execution control commands.

[0019] Preferably, the soil moisture characteristic map specifically includes soil bulk density distribution, porosity gradient, and infiltration rate; the water demand prediction data includes crop root water absorption intensity, canopy transpiration coefficient, and critical water shortage threshold; the irrigation control map specifically refers to the priority ranking of zonal irrigation and the water allocation ratio; the equipment health parameters include motor bearing wear index and insulation aging degree; the pipeline network regulation command includes branch pipe pressure compensation value and valve response delay parameter; the integrated evaluation index includes soil improvement weight and equipment energy consumption coefficient; and the execution control command specifically includes pulse irrigation cycle parameter and flow rate adjustment step size.

[0020] Preferably, the soil moisture perception module includes a data acquisition submodule, a noise suppression submodule, and a feature extraction submodule;

[0021] The data acquisition submodule synchronously acquires the soil dielectric constant based on a capacitive sensor array, uses an adaptive sliding window algorithm to eliminate temperature drift error, calculates the volumetric water content through the frequency domain reflection principle, and generates the original soil moisture dataset.

[0022] The noise suppression submodule is based on the original soil moisture dataset. It uses an improved wavelet threshold denoising algorithm to separate the effective signal from high-frequency noise, identifies abnormal data points through kurtosis detection and performs interpolation repair, and generates a filtered data matrix.

[0023] The feature extraction submodule extracts soil moisture migration features based on the filtered data matrix using variational mode decomposition algorithm, constructs water-holding curve models for different soil layers using kernel density estimation method, and generates soil moisture feature maps.

[0024] Preferably, the soil moisture prediction module includes a meteorological analysis submodule, a demand modeling submodule, and a grade classification submodule;

[0025] The meteorological analysis submodule uses a grey prediction model to process wind speed and radiation intensity sequences, identifies chaotic attractor characteristics of meteorological elements through a phase space reconstruction algorithm, calculates the dynamic fluctuation of latent heat of evaporation, and generates an evaporation trend map.

[0026] The demand modeling submodule is based on the evaporation trend map, uses a long short-term memory network to build a crop evapotranspiration prediction model, and uses an attention mechanism to strengthen the characteristics of key growth stages to generate a daily water demand prediction curve.

[0027] The classification submodule is based on the daily water demand prediction curve, uses the fuzzy C-means clustering algorithm to classify the irrigation urgency level, determines the irrigation start threshold of each region through the membership function, and generates water demand prediction data.

[0028] Preferably, the irrigation strategy module includes a terrain analysis submodule, a hydraulic calculation submodule, and a scheme generation submodule;

[0029] The terrain analysis submodule uses UAV lidar to scan the surface elevation of farmland, constructs a digital elevation model through the Delaunay triangulation algorithm, and calculates the terrain slope and runoff coefficient of each irrigation unit.

[0030] The hydraulic calculation submodule is based on the digital elevation model, establishes the flow conservation equation of the pipeline node, applies Newton's iteration method to solve the pipeline velocity and pressure distribution, and generates a set of hydraulic balance parameters.

[0031] The proposed scheme generation submodule uses a set of hydraulic balance parameters and a chaotic optimization algorithm to search for the optimal combination of irrigation durations. It also avoids local optima by employing a tabu search strategy to generate an irrigation control map.

[0032] Preferably, the device status monitoring module includes a signal acquisition submodule, a feature separation submodule, and a health assessment submodule;

[0033] The signal acquisition submodule captures the water pump current waveform based on the Rogowski coil sensor, and uses a synchronous compression transformation algorithm to enhance the time-frequency resolution of the signal and generate a high-precision current spectrum.

[0034] The feature separation submodule extracts the stator current feature components of the motor based on the current spectrum diagram and uses the singular value decomposition algorithm. It then separates the bearing fault frequency components through envelope demodulation technology to generate a fault feature vector.

[0035] The health assessment submodule is based on fault feature vectors and uses a deep belief network to construct an equipment degradation assessment model. It extracts hidden layer health indicators through a sparse autoencoder to generate equipment health parameters.

[0036] Preferably, the pipeline optimization module includes a topology reconstruction submodule, a loss calculation submodule, and a pressure compensation submodule;

[0037] The topology reconstruction submodule is based on the pipeline network GIS coordinate data. It uses a depth-first search algorithm to traverse the pipeline connection relationships and optimizes the pipeline network layout structure through a minimum spanning tree algorithm to generate a topology connection matrix.

[0038] The loss calculation submodule is based on the topology connection matrix, applies the Kolmogorov equation to calculate the friction loss of each pipe segment, solves the water hammer pressure fluctuation amplitude by the characteristic line method, and generates a hydraulic loss parameter set.

[0039] The pressure compensation submodule is based on a set of hydraulic loss parameters and uses a particle swarm optimization algorithm to dynamically adjust the speed of the variable frequency pump. It also compensates for pressure fluctuations at the end of the pipeline through a feedforward control model and generates pipeline regulation commands.

[0040] Preferably, the multi-source data fusion module includes an evidence synthesis submodule, a weight calculation submodule, and an indicator construction submodule;

[0041] The evidence synthesis submodule is based on the DS evidence theory framework. It integrates multi-sensor data using a conflict evidence correction algorithm, calculates the support of each proposition through a basic probability allocation function, and generates a fused confidence distribution.

[0042] The weight calculation submodule is based on information entropy theory, applies the improved coefficient of variation method to quantify the dispersion of parameters, and combines expert experience matrix to correct the index weight coefficients to generate a dynamic weight allocation table.

[0043] The indicator construction submodule is based on a dynamic weight allocation table, uses the TOPSIS method to calculate the closeness of each scheme, sorts the optimal decision schemes by the grey relational projection method, and generates fusion evaluation indicators.

[0044] Preferably, the intelligent execution module includes an instruction decomposition submodule, a hysteresis compensation submodule, and a parameter tuning submodule;

[0045] The instruction decomposition submodule is based on the irrigation control map, uses a time automaton model to analyze the valve opening and closing sequence, verifies logical conflicts through Petri nets, and generates a sequence of actionable actions.

[0046] The hysteresis compensation submodule is based on the dynamic response test data of the solenoid valve, constructs a second-order hysteresis transfer function model, applies a predictive control algorithm to calculate the advance trigger compensation amount, and generates a correction control signal.

[0047] The parameter tuning submodule optimizes the PID control parameters based on the modified control signal using a chaotic ant colony algorithm, verifies the convergence of the adjustment process using the Lyapunov stability criterion, and generates execution control commands.

[0048] Preferably, the grey prediction model adopts a metabolic data update mechanism, and adjusts the development coefficient and grey action amount through a background value optimization algorithm to construct a dynamic prediction equation that adapts to the characteristics of meteorological abrupt changes.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] The system, through a multi-source sensor array and a three-dimensional soil moisture distribution model constructed by the soil moisture perception module, can analyze the soil structure of farmland in layers, accurately extract features such as soil bulk density distribution, porosity gradient and infiltration rate, and generate soil moisture feature maps. This provides comprehensive and accurate soil moisture status data for irrigation decisions, overcomes the limitations of traditional single-point monitoring, and makes the judgment of irrigation timing and water volume more scientific and reasonable. It can effectively avoid deep drought or shallow water accumulation and optimize the crop root growth environment.

[0051] The soil moisture forecasting module, based on chaotic characteristic analysis and the classification of water requirements at different crop growth stages, can predict the dynamic trend of soil moisture evaporation by combining meteorological data sequences. It generates water requirement forecast data including crop root water absorption intensity, canopy transpiration coefficient, and critical water shortage threshold. This function enables irrigation systems to respond proactively to the water needs of crops at different growth stages, shifting from "passive irrigation" to "active forecasting." This significantly improves the timeliness and targeting of irrigation, reducing crop yield reduction caused by water shortages or over-irrigation, and enhancing crop yield and quality.

[0052] The irrigation strategy module integrates farmland topographic elevation data with pipeline pressure parameters, and generates irrigation control maps through hydraulic balance equations and dynamic programming algorithms. This enables the optimization of regional water allocation schemes and the precise calculation of irrigation duration. The module fully considers the impact of factors such as terrain slope and runoff coefficient on irrigation uniformity, ensuring that crops in different areas receive reasonable water distribution. It effectively solves the problem of uneven irrigation in areas with complex terrain, improves water resource utilization efficiency, and reduces irrigation costs.

[0053] The equipment status monitoring module dynamically generates equipment health parameters, such as the motor bearing wear index and insulation aging degree, by real-time acquisition of the water pump operating current waveform and extraction of motor winding vibration characteristics. This function enables real-time monitoring and fault early warning of irrigation equipment operation status, allowing maintenance personnel to promptly identify potential equipment problems and perform maintenance, reducing the probability of sudden equipment failures, extending equipment lifespan, ensuring stable operation of the irrigation system, and lowering manual inspection costs and downtime losses.

[0054] The pipeline optimization module optimizes the pipeline topology based on equipment health parameters and historical irrigation data, using graph theory algorithms and pressure compensation models to generate pipeline adjustment commands. This module effectively reduces pipeline resistance losses, compensates for end-point pressure fluctuations, improves the stability and efficiency of pipeline operation, ensures that irrigation water can be delivered evenly and stably to each irrigation unit, further enhances the uniformity and reliability of irrigation effects, and reduces irrigation deviations caused by unstable pipeline pressure.

[0055] The multi-source data fusion module employs evidence theory and entropy weighting to perform credibility weighting and influence strength analysis on heterogeneous data, generating fusion evaluation indicators. This technology achieves deep fusion of multi-source data such as soil moisture, water demand prediction, and equipment status, providing a comprehensive and reliable integrated evaluation basis for irrigation decisions. This improves the scientific nature and accuracy of decision-making, avoids decision biases caused by single data sources, and enables the irrigation system to be dynamically adjusted and optimized according to actual conditions.

[0056] The intelligent execution module, through a hysteresis compensation mechanism and a fuzzy PID control algorithm, achieves precise matching of the opening and closing sequence of irrigation valves and optimization of the dynamic response of water flow. This module effectively compensates for the response delay of solenoid valves, improves the accuracy and efficiency of irrigation control, ensures a high degree of matching between irrigation volume and predicted water demand, reduces irrigation delays and over-irrigation, and further enhances water resource utilization efficiency and the stability of irrigation effects. Attached Figure Description

[0057] Figure 1 This is a schematic diagram illustrating the working principle of the high-standard farmland intelligent irrigation system based on the Internet of Things and data analysis described in this invention.

[0058] Figure 2 A flowchart illustrating the workflow and data generation of the soil moisture forecasting module;

[0059] Figure 3 A flowchart for terrain analysis and scheme generation in the irrigation strategy module;

[0060] Figure 4 This is a flowchart of the pipeline optimization module's topology reconstruction and pressure compensation. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] Please see Figures 1-4 The present invention relates to a high-standard intelligent irrigation system for farmland based on the Internet of Things and data analysis, the specific implementation steps of which are as follows:

[0063] The soil moisture perception module collects farmland soil data through multi-source sensors, performs layered analysis of soil structure, constructs a three-dimensional soil moisture distribution model, extracts soil water holding capacity characteristics, and generates a soil moisture feature map that includes soil bulk density distribution, porosity gradient, and infiltration rate.

[0064] The soil moisture prediction module uses soil moisture feature maps to perform chaotic characteristic analysis on meteorological data sequences, identify the dynamic trend of soil moisture evaporation, classify water demand levels according to crop growth stages, and generate water demand prediction data that includes crop root water absorption intensity, canopy transpiration coefficient and critical water shortage threshold.

[0065] The irrigation strategy module is based on water demand prediction data, integrates farmland topographic elevation data and pipeline pressure parameters, constructs a hydraulic balance equation to solve for the optimal irrigation duration, and uses a dynamic programming algorithm to generate regional water allocation schemes, forming an irrigation control map that includes the priority ranking of regional irrigation and the proportion of water allocation.

[0066] The equipment status monitoring module is based on the irrigation control map, collects the water pump operating current waveform in real time, separates electromagnetic interference noise and extracts the vibration characteristics of the motor winding, and generates equipment health parameters including the motor bearing wear index and insulation aging degree.

[0067] Based on equipment health parameters, the pipeline optimization module reconstructs the topology of the irrigation pipeline network, applies graph theory algorithms to solve for the path with the minimum pipeline resistance loss, establishes a pressure compensation model by combining historical irrigation data, and generates pipeline regulation commands that include branch pipe pressure compensation values ​​and valve response delay parameters.

[0068] The multi-source data fusion module is based on soil moisture characteristic maps, water demand prediction data, irrigation control maps, equipment health parameters, and pipeline regulation instructions. It uses evidence theory to weight the heterogeneous data based on credibility, determines the interaction strength between parameters through the entropy weight method, and generates a fusion evaluation index that includes soil improvement weights and equipment energy consumption coefficients.

[0069] The intelligent execution module is based on fusion evaluation indicators, matches the opening and closing sequence of irrigation valves, adjusts the opening degree of solenoid valves and pulse frequency by combining hysteresis compensation mechanism, optimizes the dynamic response of water flow by fuzzy PID control, and generates execution control instructions that include pulse irrigation cycle parameters and flow adjustment step size.

[0070] The present invention will be further described below with reference to Examples 1 to 5:

[0071] Example 1:

[0072] The soil moisture perception module of the system includes a data acquisition submodule, a noise suppression submodule, and a feature extraction submodule. Each submodule works in collaboration with specific algorithms and hardware to achieve multi-level data processing and feature extraction of farmland soil moisture.

[0073] The data acquisition submodule synchronously acquires the soil dielectric constant based on a capacitive sensor array. The capacitive sensor array consists of multiple distributed sensor nodes, deployed at preset spatial intervals at different soil depths in the farmland (e.g., top 0-20cm, middle 20-40cm, deep 40-60cm). Each node contains a capacitive sensing unit and a miniature data acquisition circuit. When the sensor is in contact with the soil, the soil medium acts as the dielectric between the capacitor plates, and its dielectric constant changes with soil moisture content. The dielectric constant can be indirectly obtained by measuring the change in capacitance. To eliminate the influence of ambient temperature on sensor measurement accuracy, this submodule employs an adaptive sliding window algorithm to monitor sensor temperature data in real time and establish a temperature-dielectric constant error correction model. Specifically, the algorithm uses a sliding window to segment and statistically analyze continuous time-series temperature and dielectric constant data, calculates the deviation coefficient between temperature changes and dielectric constant measurements within each window, and then dynamically corrects the original dielectric constant data to eliminate temperature drift errors. After obtaining the corrected dielectric constant data, the submodule calculates the soil volumetric water content using the frequency domain reflection principle: by emitting electromagnetic waves of a specific frequency into the soil, the volumetric water content data of each soil layer is obtained by inversion based on the relationship between the propagation speed of electromagnetic waves in the soil and the dielectric constant (i.e., the empirical formula between the dielectric constant and the volumetric water content). Finally, the original soil moisture dataset is generated according to the time series and spatial location. This dataset contains multi-dimensional original data such as dielectric constant, temperature, and volumetric water content at different soil depths and different sampling times.

[0074] The noise suppression submodule uses the original soil moisture dataset as its processing target. First, it applies an improved wavelet thresholding denoising algorithm to separate the effective signal from high-frequency noise. The core steps of wavelet thresholding denoising include signal wavelet decomposition, threshold processing, and signal reconstruction: the original data sequence is decomposed into multiple layers of wavelets to obtain wavelet coefficients for different frequency components; for each layer of wavelet coefficients, an improved threshold function (such as a hybrid function combining soft and hard thresholding functions) is used to process them. Coefficients with absolute values ​​less than the threshold are set to zero to remove noise, while coefficients greater than the threshold are shrunk or retained to preserve the effective signal components. The improved algorithm enhances its ability to suppress non-stationary noise by adaptively adjusting the threshold parameters (e.g., dynamically calculating the threshold based on the noise standard deviation). After wavelet denoising, the submodule identifies outlier data points using kurtosis detection. Kurtosis, as a statistic measuring the thickness of the tail of a data distribution, is sensitive to outliers. When the kurtosis value of a data point exceeds a preset threshold, that point is determined to be an outlier. For identified outlier data points, linear or polynomial interpolation algorithms are used for repair. Specifically, based on the valid data points adjacent to the outlier, an estimated value for that point is fitted using an interpolation function and used to replace the original outlier, thus generating a filtered data matrix. This matrix eliminates high-frequency noise and outlier data, retaining the effective signal sequence reflecting changes in soil moisture.

[0075] The feature extraction submodule extracts soil moisture migration features and constructs water-holding curve models based on the filtered data matrix. First, the variational mode decomposition (VMD) algorithm is used to perform multi-scale decomposition of the volumetric water content time series in the data matrix. VMD is an adaptive signal decomposition method that separates different frequency components in a complex signal by decomposing the original signal into multiple intrinsic mode functions (IMFs) with different center frequencies. In soil moisture analysis, different frequency IMF components correspond to different physical processes; for example, high-frequency components may reflect rapid changes in moisture caused by short-term precipitation or irrigation, while low-frequency components may reflect long-term soil moisture migration trends. By analyzing the energy distribution and frequency characteristics of each IMF component, feature parameters that characterize the soil moisture migration pattern (such as the energy proportion of each component and its center frequency value) are extracted. Subsequently, the submodule constructs water-holding curve models for different soil layers using kernel density estimation. Kernel density estimation is a non-parametric estimation method used to estimate the probability density function of a random variable. For each soil depth, with volumetric water content as the independent variable and soil matric potential as the dependent variable, the kernel density estimation method is used to smooth the joint distribution of the two, resulting in a continuous water-holding curve. This curve describes the soil's ability to retain water at different moisture contents and is a key parameter characterizing soil water-holding capacity. Analysis of the water-holding curves for each soil layer allows for the extraction of characteristic parameters such as soil bulk density distribution, porosity gradient, and infiltration rate, ultimately generating a soil moisture characteristic map containing these parameters. This map is presented in the form of a spatial three-dimensional grid or a two-dimensional contour map, visually displaying the differences in soil moisture distribution and water-holding capacity in different areas and soil layers of farmland, providing basic data support for subsequent moisture prediction modules. The soil water-holding curve analysis uses a centrifuge method to determine the relationship between soil water suction and volumetric water content in different soil layers (0-20cm, 20-40cm, 40-60cm). Three parallel samples are set for each soil layer, and the average value is used to construct the basic data for the water-holding curve. The characteristic parameter extraction method uses the van Genuchten model formula as follows:

[0076] in This refers to the volumetric water content. Residual moisture content saturated moisture content, , , For fitting parameters, To determine the water suction, the water holding curve is fitted using the least squares method to optimize parameters. The parameter calculation logic is based on the fitted van Genuchten model, where the rate of change of soil water suction is obtained by differentiating the water content. This, combined with soil texture data (the proportions of sand, silt, and clay), is used to calculate the porosity gradient. This is then applied using Darcy's law.

[0077] in For the permeation rate, Unsaturated hydraulic conductivity Substitute the soil layer thickness into the water-holding curve. and The corresponding relationship is used to solve for the permeation rate.

[0078] During data acquisition, the deployment density and depth of the capacitive sensor array can be adjusted according to the farmland soil type and crop root distribution characteristics. For example, for sandy soil or shallow-rooted crops, the density of surface sensor nodes can be appropriately increased; for clay soil or deep-rooted crops, the number of deep sensors can be increased. The window length of the adaptive sliding window algorithm can be dynamically adjusted according to seasonal changes. For example, in seasons with drastic temperature changes (such as summer), the window length can be reduced to improve the real-time performance of temperature error correction; in seasons with relatively stable temperatures (such as winter), the window length can be increased to reduce computational load. The wavelet basis function type (such as the db series or sym series) in the improved wavelet threshold denoising algorithm can be selected according to data characteristics to achieve the best denoising effect. The mode number and penalty factor parameters of the variational mode decomposition algorithm can be determined through cross-validation to ensure that the decomposed IMF components can truly reflect the multi-scale characteristics of soil moisture migration. The kernel function type (such as Gaussian kernel or Epanechnikov kernel) and bandwidth parameter in kernel density estimation need to be optimized according to the data distribution characteristics to construct an accurate water-holding curve model. The selection criterion for the kernel function type is: when the data distribution is symmetrical and bell-shaped, the Gaussian kernel is preferred. The formula is as follows:

[0079] Where x is the input variable of the kernel function, it is suitable for farmland with relatively uniform soil moisture content distribution; when the data has a significant skewed distribution (such as large fluctuations in surface soil moisture content), the Epanechnikov kernel formula is used:

[0080]

[0081] in The indicator function is used to enhance robustness to outliers; the bandwidth parameter optimization method uses cross-validation to determine the bandwidth. The water-holding curve data were randomly divided into 5 groups, with 4 groups used as the training set and 1 group as the validation set. The mean square error between the kernel density estimate and the actual value in the validation set under different bandwidths is calculated using the following formula:

[0082] in This is an estimated value. This is the actual value. (For the number of validation set samples), select The minimum corresponding bandwidth is taken as the optimal value, and the bandwidth value is controlled between 0.02 and 0.15.

[0083] Through sequential processing by data acquisition, noise suppression, and feature extraction submodules, the soil moisture perception module achieves a complete process from raw data acquisition to noise suppression and feature parameter extraction. This module not only acquires real-time soil moisture data but also reveals the inherent laws of soil moisture distribution and migration through multi-level data processing and algorithm analysis, generating physically meaningful soil moisture feature maps. This provides scientific and reliable foundational data for subsequent decision-making in the entire intelligent irrigation system. The algorithm design and parameter optimization of each submodule are closely integrated with the actual needs of farmland, ensuring the module's applicability and stability under different soil conditions and environments, providing crucial perceptual support for precision irrigation of high-standard farmland.

[0084] Example 2:

[0085] The soil moisture forecasting module consists of a meteorological analysis submodule, a demand modeling submodule, and a classification submodule. Each submodule uses specific algorithms and data processing procedures to analyze soil moisture evaporation trends, predict crop water requirements, and classify irrigation demand levels.

[0086] The meteorological analysis submodule takes meteorological data sequences such as wind speed and radiation intensity as input and processes them using a grey prediction model. The grey prediction model is a forecasting method for uncertain systems with limited data and information, suitable for time series such as meteorological data that exhibit randomness and volatility. This module first preprocesses the raw meteorological data, transforming non-stationary sequences into approximately stationary sequences through an accumulation generation operation to reduce data randomness. To address potential abrupt changes in meteorological data (such as short-term strong winds and extreme radiation variations), the model employs a metabolic data update mechanism. Each time a new predicted value is generated, the oldest historical data is removed, and the latest data is added, forming a new sequence for remodeling, thus enhancing the model's adaptability to real-time data. Simultaneously, a background value optimization algorithm adjusts the model's development coefficient and grey action to optimize prediction accuracy. Specifically, the background value optimization algorithm modifies the traditional grey model's background value calculation formula by introducing weighted coefficients. The background value correction formula for the traditional grey model is as follows:

[0087] in, This is a cumulative sequence of the original meteorological data (wind speed / radiation intensity), where k represents the k-th time period or k-th data point in the grey model, and weighting coefficients are introduced. ( The value ranges from 0.4 to 0.6, adjusted according to the degree of fluctuation in meteorological data (the larger value is used when the fluctuation is large). The corrected background value is:

[0088] The dynamic prediction equation is constructed based on the corrected background value, and the dynamic prediction equation is established as follows:

[0089] in, For development coefficient, The gray action quantity is solved using the least squares method; This is a meteorological abrupt change correction term, applied when the rate of change in meteorological data is detected ( When it exceeds 0.3, ,otherwise This enables adaptation to mutation characteristics.

[0090] Background values ​​more closely reflect the changing trends of actual data, thus constructing dynamic prediction equations adapted to meteorological abrupt changes. After the model outputs predicted sequences of wind speed and radiation intensity, the submodule analyzes the chaotic characteristics of meteorological elements through a phase space reconstruction algorithm. Phase space reconstruction is a method to recover the dynamic characteristics of a system from a univariate time series. It maps a one-dimensional time series to a high-dimensional phase space using the delayed coordinate method, thereby revealing the characteristics of chaotic attractors hidden in the data (such as the dimension and shape of the attractors). Based on the phase space reconstruction results, the dynamic fluctuation of latent heat of evaporation is calculated—latent heat of evaporation is closely related to wind speed and radiation intensity. The predicted wind speed and radiation intensity are converted into changes in latent heat of evaporation using empirical formulas (such as a simplified form of the Penman-Monteith formula), thereby generating an evaporation trend map. This map uses time as the horizontal axis and the fluctuation of latent heat of evaporation as the vertical axis, intuitively displaying the strength trend of soil moisture evaporation at different times.

[0091] The demand modeling submodule uses an evaporation trend map and a Long Short-Term Memory (LSTM) network to build a crop evapotranspiration prediction model. LSTM is a special type of recurrent neural network that can effectively capture long-range dependencies in time series data, making it suitable for predicting complex time-series data such as crop evapotranspiration, which is influenced by multiple factors (e.g., meteorological conditions, crop growth stages). The module first collects historical crop evapotranspiration data (obtained through measurements using equipment such as weighing lysimeters), corresponding meteorological data (wind speed, radiation intensity, temperature, humidity, etc.), and crop growth stage characteristics (e.g., seedling emergence, jointing stage, booting stage, etc.). After normalizing the data, it is divided into training, validation, and test sets according to the time series. During model construction, the input layer receives a multi-dimensional feature vector containing meteorological factors such as latent heat of evaporation fluctuations, temperature, and humidity, as well as crop growth stage encodings. Through multi-layer LSTM neurons, long-term dependency features (e.g., periodic changes in crop evapotranspiration across different seasons) and short-term fluctuation features (e.g., a sudden drop in evapotranspiration after a rainfall event) are extracted from the time series. To enhance the model's focus on key crop growth stage characteristics (such as the sensitivity of the heading stage to water demand), the module introduces an attention mechanism. By calculating the attention weights of input features at each time step, the model prioritizes data relevant to key growth stages during prediction. The model outputs a predicted daily crop evapotranspiration value, which, after inverse normalization, generates a predicted daily water demand curve. This curve reflects the changes in crop water demand intensity on different days, providing a basis for subsequent irrigation level classification.

[0092] The grading submodule uses the fuzzy C-means clustering algorithm (FCM) to classify irrigation urgency levels based on the daily water demand prediction curve. FCM is a soft clustering algorithm that allows samples to belong to multiple clusters with different membership degrees, making it suitable for classification problems with fuzzy boundaries, such as water demand levels. The module first sets the number of clusters (e.g., dividing into three levels: "low urgency," "medium urgency," and "high urgency"), randomly initializes the cluster centers, and then calculates the membership degree of each sample point to each cluster center. By iteratively updating the membership matrix and cluster centers, the objective function (the weighted sum of distances from sample points to each cluster center) is minimized, ultimately classifying the predicted daily water demand into different urgency levels. For each irrigation area (e.g., different areas divided by farmland plots), the irrigation initiation threshold for that area is determined by a membership function—the membership function defines the degree to which a sample belongs to a certain level. When the membership degree of a region's predicted daily water demand to the "high urgency" level exceeds a preset threshold (e.g., 0.7), it is determined that irrigation needs to be initiated immediately in that region. Simultaneously, the module incorporates soil water-holding capacity parameters (such as infiltration rate and porosity gradient) from farmland soil moisture characteristic maps to correct the clustering results: for sandy soil areas with poor water-holding capacity, the irrigation initiation threshold is appropriately lowered to avoid crop water shortage due to insufficient soil water retention; for clay soil areas with strong water-holding capacity, the irrigation initiation threshold is appropriately raised to prevent over-irrigation and water waste. The final generated water demand prediction data includes parameters such as crop root water absorption intensity, canopy transpiration coefficient, and critical water shortage threshold for each irrigated area at different time periods, providing a scientific basis for the irrigation strategy module to formulate regional water allocation plans. (mm / d) is calculated using the Feddes model, and the formula is as follows:

[0093] in, This represents the crop's potential evapotranspiration (mm / d, calculated using the Penman-Monteith formula). Root depth (cm, determined according to the crop growth stage, such as the jointing stage of wheat) cm), The water stress coefficient is the coefficient of soil water suction. (Water absorption threshold) , (Stress threshold) , hour Canopy evapotranspiration coefficient Referencing the FAO crop coefficient table and adjusting accordingly based on local crop varieties, such as the tasseling period of maize. Grouting period The system is calibrated monthly through on-site observation (weighing lysimeter), with a correction value not exceeding ±0.1; critical water shortage threshold. Based on crop wilting coefficient (Determined through indoor experiments, such as corn) (cm³ / cm³) (Safety margin), when the soil moisture content is lower than Irrigation is triggered at certain times.

[0094] The correction parameter is selected from the soil water holding capacity parameter, specifically the infiltration rate. (mm / h) and porosity gradient (% / cm) is used as a correction factor; the initial cluster centers are first calculated (based on the fuzzy C-means algorithm to obtain the water demand cluster centers of each irrigation area). Then, the correction factor is calculated based on the permeation rate and porosity gradient. The formula is as follows:

[0095] in, This represents the maximum permeability rate across all regions. The maximum porosity gradient is obtained for all regions; finally, the corrected cluster centers are obtained. ,when Time to take (Avoid over-correction) to achieve a match between clustering results and soil water-holding capacity.

[0096] In the meteorological analysis submodule, the metabolic cycle of the grey prediction model can be set according to the update frequency of meteorological data (e.g., updated daily) to ensure that the model reflects the latest meteorological changes in a timely manner. The delay time and embedding dimension parameters of the phase space reconstruction algorithm are determined using the autocorrelation function method and the GP algorithm to ensure that the reconstructed phase space accurately reflects the dynamic characteristics of the system. In the demand modeling submodule, the number of layers, neurons, and weight calculation method of the attention mechanism in the LSTM network can be optimized using the grid search method to improve the model's generalization ability. In the hierarchical classification submodule, the termination iteration condition (e.g., the change in the membership matrix is ​​less than 1e-5) and the shape parameters of the membership function of the fuzzy C-means clustering algorithm can be adjusted based on historical irrigation data to adapt to the water requirements of different crops.

[0097] Through the collaborative operation of the meteorological analysis submodule, demand modeling submodule, and irrigation level classification submodule, the soil moisture prediction module realizes a complete process from meteorological data processing to crop water requirement prediction and irrigation level classification. This module can not only predict the dynamic trend of soil moisture evaporation, but also accurately classify the irrigation urgency level of each region by combining crop growth stages and soil characteristics, generating water requirement prediction data with spatiotemporal specificity. The algorithm design of each submodule is closely integrated with the actual farmland scenario, fully considering the uncertainty of meteorological factors, the dynamic nature of crop water requirements, and the differences in soil conditions, ensuring the scientific validity and practicality of the prediction results, and providing crucial predictive support for the accurate decision-making of high-standard farmland intelligent irrigation systems.

[0098] Example 3:

[0099] The irrigation strategy module includes a terrain analysis submodule, a hydraulic calculation submodule, and a scheme generation submodule. Through the collaboration of hardware devices and algorithm models, each submodule realizes digital modeling of farmland terrain, calculation and analysis of pipeline hydraulic characteristics, and optimized generation of irrigation strategies.

[0100] The terrain analysis submodule acquires farmland surface elevation data based on UAV LiDAR technology. The UAV, equipped with a LiDAR sensor, performs low-altitude scanning of the farmland along a preset flight path. The LiDAR emits laser pulses and receives reflected signals from the ground surface, calculating the distance between the sensor and the target. Combined with the UAV's positioning system (such as GPS / RTK) and attitude sensors (such as an inertial measurement unit, IMU), high-density point cloud data is generated, with a point cloud density reaching tens to hundreds of points per square meter, ensuring accurate capture of surface details. After acquiring the point cloud data, the submodule uses the Delaunay triangulation algorithm to construct a digital elevation model (DEM). Delaunay triangulation is a method that constructs a non-overlapping triangular mesh from a discrete set of points. It maximizes the minimum angle property, ensuring that the generated triangular mesh is as regular as possible, thus accurately reflecting the surface undulations. During the triangulation process, noise points significantly higher or lower than the ground level (such as vegetation tops and field obstacles) are first removed. A moving surface fitting algorithm is then used to filter the ground point cloud, separating ground points from non-ground points. Next, triangulation is performed on the ground points to generate an irregular triangular network (TIN). An interpolation algorithm is then used to convert the TIN into a regular grid DEM. The grid resolution is set according to the farmland area and irrigation precision requirements (e.g., 0.5m × 0.5m to 2m × 2m). Based on the DEM, the submodule further calculates the terrain slope and runoff coefficient for each irrigation unit: the terrain slope is calculated using the elevation difference of the grid units, reflecting the degree of surface inclination; the runoff coefficient is estimated using empirical formulas (such as the CN method) based on parameters such as soil type and vegetation cover, characterizing the proportion of precipitation or irrigation water converted into surface runoff. These parameters provide the basic terrain data for the hydraulic calculation submodule.

[0101] The hydraulic calculation submodule establishes flow conservation equations for pipeline network nodes based on a digital elevation model. The pipeline system consists of water sources (such as pumps), main pipes, branch pipes, capillary pipes, and irrigation equipment (such as sprinklers and drippers). The flow at each node (pipeline junction, outlet) follows a continuity equation, meaning the flow into a node equals the sum of the flow out of the node and the node's outflow. The node outflow refers to the flow directly supplied to the irrigation unit (such as a plot of land or planting row) at the pipeline node, including sprinkler or dripper outflow, calculated using the following formula:

[0102] in The number of irrigation devices connected to the node (e.g., a node connects to 20 sprinklers). Design flow rate for a single irrigation device (such as a sprinkler head) The outflow rate at each node must meet the water demand of the irrigation unit and not exceed the maximum water delivery capacity of the pipeline. The submodule first performs topological modeling of the pipeline network, abstracting pipelines as directed edges and nodes as vertices to construct a pipeline network topology diagram. Then, based on the DEM, the elevation of each node is determined, and the geometric parameters of the pipeline (such as length, diameter, and burial depth) are calculated. Combined with the regional water distribution scheme in the irrigation control map, the target flow rate for each irrigation unit is set. Based on this, the Newton-Westbach iteration method is applied to solve for the pipeline velocity and pressure distribution. The Newton-Westbach iteration method is a numerical solution method for nonlinear equations, which iteratively updates variable values ​​to gradually approximate the solution to the equations. For each pipe segment, the friction loss is calculated according to the Darcy-Weisbach formula, and the local head loss is calculated according to the local resistance coefficient, establishing a nonlinear equation system with node pressure and pipe segment flow rate as variables. Through iterative solving, the velocity, pressure, and head loss distribution of each pipe segment are obtained, generating a hydraulic balance parameter set containing parameters such as node pressure, pipe segment flow rate, and head loss. This parameter set reflects the hydraulic operation status of the pipeline network under a specific water distribution scheme, providing a basis for the scheme generation submodule to optimize irrigation duration.

[0103] The solution generation submodule, based on the hydraulic balance parameter set, employs a chaotic optimization algorithm to search for the optimal combination of irrigation durations. The chaotic optimization algorithm is a global optimization method based on the ergodicity, randomness, and regularity of chaotic systems. By mapping optimization variables to a chaotic variable space, it utilizes the ergodic characteristics of chaotic variables to search within the solution space, avoiding getting trapped in local optima. The module first defines the optimization objective: to minimize the total irrigation duration or energy consumption while meeting the water demand of each irrigation unit. The optimization variables are the starting irrigation time and duration for each irrigation area, with constraints including that the pipeline pressure does not exceed the equipment's rated pressure and that the irrigation duration for each area is not less than the minimum irrigation cycle. The optimization variables are initialized chaotically (e.g., using Logistic mapping to generate chaotic sequences) and scaled to ensure they fall into the feasible solution space. During the iteration process, the hydraulic state of the pipeline (e.g., maximum pressure, minimum flow rate) under each chaotic variable's irrigation duration combination is calculated, the objective function value is evaluated, and the best solutions are retained. New solutions are generated through chaotic perturbation, gradually approaching the global optimum. To further improve search efficiency, the module introduces a tabu search strategy: a tabu list is established to record recently searched solutions, prohibiting repeated searches for the same or similar solutions. A disregard criterion is also set: if the objective function value of a tabu solution is better than the current optimal solution, the tabu is lifted and the solution is accepted. By combining chaotic optimization algorithms with the tabu search strategy, an irrigation control map is generated, including irrigation priority ranking, water allocation ratios, and optimal irrigation durations for each region. This map is presented in tabular or graphical form, clearly showing the irrigation sequence, start time, duration, and corresponding water volume for each irrigation region, providing operational instructions for the intelligent execution module.

[0104] In the terrain analysis submodule, the flight altitude and speed of the UAV LiDAR need to be adjusted according to the farmland area and terrain complexity. For example, a higher flight altitude can be used in flat areas to improve scanning efficiency, while a lower flight altitude should be used in complex areas to ensure point cloud density. The Delaunay triangulation algorithm can be implemented using open-source libraries (such as CGAL) or custom-developed code to ensure the efficiency and accuracy of mesh generation. In the hydraulic calculation submodule, the initial value setting of the Newton iteration method can be based on the design parameters of the pipeline network (such as design flow rate and design pressure), and the iteration termination condition is that the difference between the variables in two adjacent iterations is less than a preset threshold (such as 1e-4). In the scheme generation submodule, the iteration number of the chaotic optimization algorithm and the tabu length of the tabu search can be determined through trial and error to balance optimization accuracy and computation time.

[0105] Through the sequential execution of the terrain analysis, hydraulic calculation, and scheme generation submodules, the irrigation strategy module realizes a complete process from farmland terrain digitization to pipeline hydraulic calculation and then to irrigation scheme optimization. This module fully considers the impact of farmland terrain differences on irrigation flow, and through precise hydraulic modeling and intelligent optimization algorithms, generates a scientifically sound irrigation control map to ensure the uniform distribution and efficient utilization of water resources within the pipeline network. The technical approach of each submodule is closely integrated with the actual farmland irrigation engineering, utilizing the efficient modeling advantages of UAV remote sensing while solving the optimization challenges of complex pipeline systems through numerical calculations and intelligent algorithms, providing core support for the precise irrigation strategy formulation of high-standard farmland intelligent irrigation systems.

[0106] Example 4:

[0107] The equipment status monitoring module consists of a signal acquisition submodule, a feature separation submodule, and a health assessment submodule. Each submodule combines sensor technology with signal processing algorithms to achieve real-time monitoring and health assessment of the irrigation equipment's operating status.

[0108] The signal acquisition submodule captures the current waveform of the water pump motor based on a Rogowski coil sensor. A Rogowski coil is a hollow toroidal coil that can be uniformly and tightly wound around the conductor being measured. It measures alternating current through electromagnetic induction, offering advantages such as wide bandwidth response and no magnetic saturation. The sensor is installed at the water pump motor's input terminal, tightly wound around the three-phase cable, sensing current changes in real time during motor operation and outputting an analog voltage signal. To improve signal acquisition accuracy, the submodule employs a synchronous compression transform algorithm to process the original signal. Synchronous compression transform is a time-frequency analysis technique that resamples the time-frequency distribution after continuous wavelet transform, concentrating energy of the same frequency onto a single frequency point, thereby enhancing the signal's time-frequency resolution and clearly revealing the transient processes and frequency components in the current waveform. The specific processing flow is as follows: First, the analog voltage signal is converted into a digital signal sequence by analog-to-digital conversion (ADC) at a fixed sampling frequency (e.g., 10kHz); then, the digital signal is subjected to synchronous compressed wavelet transform to generate a high-precision current spectrum. This spectrum uses time as the horizontal axis and frequency as the vertical axis, and uses color or grayscale values ​​to characterize the energy intensity of each frequency component, making it easy to intuitively observe the time-frequency characteristics of the motor current.

[0109] The feature separation submodule takes the current spectrum as input and applies the Singular Value Decomposition (SVD) algorithm to extract the stator current feature components of the motor. Singular Value Decomposition is a matrix factorization technique that decomposes any matrix into a product of singular vectors and singular values, commonly used in signal denoising and feature extraction. For the time-frequency matrix corresponding to the current spectrum, SVD yields a sequence of singular values ​​arranged from high to low energy, along with corresponding left and right singular vectors. Low-order singular values ​​correspond to the main components of the signal (such as the fundamental frequency current during normal motor operation), while high-order singular values ​​correspond to noise or fault feature components. By retaining low-order singular values ​​and reconstructing the matrix, noise reduction of the current signal is achieved, extracting the stator current feature components reflecting the normal operating state of the motor. Subsequently, the submodule employs envelope demodulation technology to separate the bearing fault frequency components. The basic principle of envelope demodulation is to extract the high-frequency modulation components from the signal using a high-pass filter, then perform a Hilbert transform to obtain the envelope signal, and finally identify the fault feature frequencies through spectral analysis. For water pump motors, bearing failures (such as inner ring failure, outer ring failure, or rolling element failure) will cause periodic modulation of the current signal. The characteristic frequency of this failure can be calculated based on the bearing's geometric parameters and the motor speed (e.g., the frequency of an inner ring failure). ,in This refers to the motor speed. For the number of rolling elements, The diameter of the rolling element, The bearing pitch circle diameter, (Contact angle). By comparing the spectrum of the envelope signal with the theoretical fault characteristic frequencies, the frequency components corresponding to the bearing fault are extracted, and a fault feature vector containing the energy values ​​of each fault frequency is generated.

[0110] The health assessment submodule uses a Deep Belief Network (DBN) to construct an equipment degradation assessment model based on fault feature vectors. A DBN is a deep neural network composed of multiple Restricted Boltzmann Machines (RBMs) stacked together, possessing powerful nonlinear feature learning capabilities and automatically extracting high-level abstract features from complex data. The module first normalizes the fault feature vectors and inputs them into the DBN's input layer. Through unsupervised pre-training with multiple RBMs, it extracts the hidden representations of fault features layer by layer. Then, it optimizes the network parameters through supervised fine-tuning (such as backpropagation) to maximize the mapping between the network output and equipment health status labels (such as "normal," "minor wear," and "severe aging"). To avoid overfitting, a sparse autoencoder is introduced to constrain the hidden layers. By adding a sparsity penalty term to the loss function, the network is forced to activate only a small number of neurons during training, thereby extracting more representative health indicators. After model training, the input is a real-time fault feature vector, and the output is equipment health parameters, including the motor bearing wear index (a dimensionless parameter reflecting the degree of bearing wear; a larger value indicates more severe wear) and insulation aging degree (characterized by the deviation of stator current harmonic components from theoretical values; a larger deviation indicates more significant insulation aging). These parameters are presented in numerical or percentage form, reflecting the real-time health status of the pump motor, including the motor bearing wear index. Vibration signal analysis was performed, and vibration acceleration signals (sampling frequency 10kHz) at the motor bearing were collected. The spectrum was obtained using Fast Fourier Transform (FFT), and the formula for extracting the bearing outer ring fault frequency is as follows: (

[0111] in The motor speed (r / min) is given. The bearing pitch circle diameter (mm) is given. The diameter of the rolling element is in mm. (Number of rolling elements); Calculate Peak energy at The formula is as follows:

[0112] in, The wear index formula is as follows, representing the spectral amplitude:

[0113] in, For new bearings Peak energy at that point The value range is 0-5. The insulation was judged to be severely worn; the degree of aging was determined to be... The stator current signal of the motor is acquired, and the 50Hz fundamental component and harmonic components (3rd, 5th, and 7th harmonics) are obtained through wavelet transform decomposition. The total harmonic distortion rate is then calculated.

[0114] ( in, For fundamental current, , , (Harmonic currents); Insulation aging degree , The total harmonic distortion of the new motor (typically) ), The insulation was determined to be significantly aged.

[0115] In the signal acquisition submodule, the number of turns and cross-sectional area of ​​the Rogowski coil sensor need to be selected based on the motor's rated current to ensure that the sensor's output signal is within the linear range of the analog-to-digital converter. The wavelet basis function type (e.g., Morlet wavelet, Mexican Hat wavelet) and scaling parameters of the synchronous compression transform algorithm can be adjusted according to the frequency range of the current signal to optimize time-frequency resolution. In the feature separation submodule, the order of singular value decomposition can be determined by the cumulative energy contribution rate (e.g., retaining low-order singular values ​​with a cumulative energy contribution rate of 95%), and the cutoff frequency of the high-pass filter in envelope demodulation needs to be higher than the motor's fundamental frequency (e.g., 50Hz) to effectively separate the modulated signal. In the health assessment submodule, the number of layers in the deep belief network, the number of RBM neurons, and the sparsity parameters of the sparse autoencoder can be optimized using cross-validation to improve the model's generalization ability.

[0116] Through the collaborative work of the signal acquisition submodule, feature separation submodule, and health assessment submodule, the equipment status monitoring module realizes a complete process from current signal acquisition to fault feature extraction and health status assessment. Utilizing the non-invasive measurement advantages of Rogowski coil sensors, combined with advanced time-frequency analysis and deep learning algorithms, this module can monitor the operating status of water pump motors in real time and with high accuracy, proactively detecting potential faults such as bearing wear and insulation aging, and generating quantified equipment health parameters. These parameters provide crucial information for the pipeline optimization module to adjust pipeline operation strategies and for the intelligent execution module to formulate maintenance plans, ensuring that irrigation equipment is always in a safe and efficient operating state and guaranteeing the continuous and stable operation of the high-standard farmland intelligent irrigation system. The technical solutions of each submodule are closely integrated with the monitoring needs of industrial equipment, ensuring the reliability of signal acquisition while achieving in-depth analysis of equipment health status through multi-level signal processing and model analysis, providing strong support for the intelligent operation and maintenance of farmland irrigation equipment.

[0117] Example 5:

[0118] The pipeline optimization module, multi-source data fusion module, and intelligent execution module achieve global optimization and precise control of the irrigation system through algorithm linkage and data interaction.

[0119] The pipeline network optimization module comprises three sub-modules: topology reconstruction, loss calculation, and pressure compensation. The topology reconstruction sub-module uses pipeline network GIS coordinate data (including pipeline start and end coordinates and node elevations) and a depth-first search (DFS) algorithm to traverse pipeline connections, starting from the water source node and recursively visiting adjacent nodes until all reachable nodes have been traversed, constructing a pipeline network node connection table. Subsequently, the minimum spanning tree (MST) algorithm optimizes the pipeline network layout structure—using parameters such as pipeline length and material as edge weights—to generate a minimum spanning tree covering all irrigation units, eliminating redundant connections and reducing the total pipeline length, generating a topology connection matrix (matrix elements...). Represents a node With nodes The connection status between them is 1 if there is a connection, and 0 otherwise.

[0120] The loss calculation submodule calculates the friction loss along each pipe segment based on the topology connection matrix and applying the Kolmogorov equation. The formula for friction loss is:

[0121] in, Friction loss along the path (unit: meters). This is the friction coefficient. The length of the pipe section is in meters. The inner diameter of the pipe (unit: meters). The average flow velocity of the fluid inside the pipe (unit: m / s). This is the acceleration due to gravity (unit: m / s²). It is determined by the Modigliani diagram or using the Cole formula. The value, combined with the flow velocity output by the hydraulic calculation submodule. The calculations are complete. Simultaneously, the amplitude of water hammer pressure fluctuations is solved using the method of characteristics, considering transient flow effects caused by valve opening / closing or pump start / stop. This generates a set of hydraulic loss parameters including friction loss and water hammer pressure. Parameters required for the method of characteristics include the pipe inner diameter. (m, determined according to the pipeline design drawings, such as the main pipe) m), pipe length (m, such as the main pipe) m), water flow velocity (m / s, obtained from hydraulic calculations, such as) m / s), bulk elastic modulus of water ( Pa), elastic modulus of pipe material (such as steel pipes) Pa), pipe wall thickness (m, such as) (m), first calculate the water hammer wave velocity. :

[0122] in, The density of water, kg / m³; then the characteristic line equation is used as follows: : ; :

[0123] in, For water hammer pressure head, The initial pressure head, For gravitational acceleration, combined with boundary conditions (such as valve closing time) s), to solve for the pressure fluctuation amplitude of each section of the pipeline at different times.

[0124] The pressure compensation submodule dynamically adjusts the variable frequency pump speed based on a set of hydraulic loss parameters using a particle swarm optimization (PSO) algorithm. The PSO algorithm simulates bird foraging behavior, treating each speed adjustment scheme as a "particle," and searches for the optimal solution by iteratively updating the particle positions (speed values) with the objective function of minimizing pressure fluctuations in the terminal pipeline. Combined with a feedforward control model to compensate for terminal pressure fluctuations—based on the pipeline topology and historical pressure data, a transfer function between the input (pump speed) and output (terminal pressure) is established, the speed adjustment amount is calculated in advance to offset the expected pressure fluctuations, and a pipeline regulation command including branch pressure compensation values ​​and valve response delay parameters is generated.

[0125] The multi-source data fusion module comprises an evidence synthesis submodule, a weight calculation submodule, and an index construction submodule. The evidence synthesis submodule, based on the DS evidence theory framework, integrates heterogeneous data from multiple sources, including soil moisture and water demand prediction, using a conflict evidence correction algorithm. It assigns support to each proposition (e.g., "Current soil moisture is suitable for irrigation") using the Basic Probability Assignment Function (BPA), generating a fusion confidence distribution (a set of confidence values ​​for each proposition). The weight calculation submodule, based on information entropy theory, applies an improved coefficient of variation method to quantify parameter dispersion—calculating the ratio of the standard deviation to the mean of each parameter, eliminating the influence of dimensions to generate initial weights, and then combining this with an expert experience matrix (a subjective rating of the importance of each parameter) to correct the weight coefficients, generating a dynamic weight allocation table (recording the final weight values ​​of each parameter). The index construction submodule, based on the dynamic weight allocation table, uses the TOPSIS method to calculate the closeness of each irrigation scheme to the ideal solution, and ranks the schemes using the grey relational projection method, generating a fusion evaluation index that includes soil improvement weights and equipment energy consumption coefficients.

[0126] The intelligent execution module consists of an instruction decomposition submodule, a lag compensation submodule, and a parameter tuning submodule. The instruction decomposition submodule, based on the irrigation control map, uses a time automata model to analyze the valve opening and closing sequence—generating a time sequence table of valve opening and closing by defining time nodes and state transition conditions. Petri nets are used to verify logical conflicts (such as multiple valves opening simultaneously causing pipeline pressure to exceed limits). By adding storage locations and transition adjustment logic, a conflict-free sequence of action actions is generated (e.g., valve A opens at time t1, and valve B closes at time t2).

[0127] The hysteresis compensation submodule constructs a second-order hysteresis transfer function model based on the dynamic response test data of the solenoid valve:

[0128] in, For the solenoid valve transfer function, Represents complex frequency. This is the gain coefficient. , The time constant (unit: seconds). The delay time (in seconds) is used. Predictive control algorithms are applied to calculate the advance trigger compensation – based on the model's prediction of the time delay before the solenoid valve opening reaches the target value, the timing of the control signal transmission is adjusted in advance to generate a corrective control signal (e.g., advancing the opening command). (Sent in seconds).

[0129] The parameter tuning submodule optimizes the PID control parameters (proportional coefficient) based on the corrected control signal using a chaotic ant colony algorithm. Integral coefficient Differential coefficients The chaotic ant colony algorithm generates the initial ant colony position through a logistic mapping and searches for the optimal solution in the parameter space by combining the positive feedback mechanism of ant foraging with the ergodicity of chaotic variables. The convergence of the regulation process is verified using the Lyapunov stability criterion—a Lyapunov function is constructed. ,like Positive definite and its derivative If the value is negative, the system is stable and will eventually generate an execution control command containing pulse irrigation cycle parameters and flow adjustment step size.

[0130] The above modules reduce hydraulic losses through pipeline network structure optimization, improve decision-making accuracy through multi-source data fusion, and compensate for system lag with intelligent control algorithms, forming a closed-loop optimization link from data perception to execution control, thereby achieving coordinated and efficient operation of the irrigation system.

[0131] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0132] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A high-standard intelligent irrigation system for farmland based on the Internet of Things and data analysis, characterized in that: The system includes: The soil moisture perception module, based on data collected from multiple sources of sensors, performs layered analysis of farmland soil structure, constructs a three-dimensional soil moisture distribution model, extracts soil water holding capacity characteristics, and generates a soil moisture feature map. The soil moisture forecasting module analyzes the chaotic characteristics of meteorological data sequences based on soil moisture feature maps, identifies the dynamic trend of soil moisture evaporation, classifies water demand levels according to crop growth stages, and generates water demand forecast data. The irrigation strategy module, based on water demand prediction data, integrates farmland topographic elevation data and pipeline pressure parameters to construct a hydraulic balance equation to solve for the optimal irrigation duration, and uses a dynamic programming algorithm to generate regional water allocation schemes and generate irrigation control maps. The equipment status monitoring module, based on the irrigation control map, collects the water pump operating current waveform in real time, separates electromagnetic interference noise and extracts the vibration characteristics of the motor windings, and generates equipment health parameters. The pipeline optimization module reconstructs the irrigation pipeline topology based on equipment health parameters, applies graph theory algorithms to solve for the path with the minimum pipeline resistance loss, establishes a pressure compensation model based on historical irrigation data, and generates pipeline regulation commands. The multi-source data fusion module, based on soil moisture feature maps, water demand prediction data, irrigation control maps, equipment health parameters, and pipeline regulation instructions, uses evidence theory to weight the heterogeneous data on credibility, determines the interaction strength between parameters through the entropy weight method, and generates fusion evaluation indicators. The intelligent execution module, based on fusion evaluation indicators, matches the opening and closing sequence of irrigation valves, adjusts the opening degree and pulse frequency of solenoid valves by combining a hysteresis compensation mechanism, optimizes the dynamic response of water flow by using fuzzy PID control, and generates execution control commands.

2. The high-standard farmland intelligent irrigation system based on the Internet of Things and data analysis according to claim 1, characterized in that: The soil moisture characteristic map specifically includes soil bulk density distribution, porosity gradient, and infiltration rate; the water demand prediction data includes crop root water absorption intensity, canopy transpiration coefficient, and critical water shortage threshold; the irrigation control map specifically refers to the priority ranking of zonal irrigation and the water allocation ratio; the equipment health parameters include motor bearing wear index and insulation aging degree; the pipeline network regulation instructions include branch pipe pressure compensation value and valve response delay parameters; the integrated evaluation index includes soil improvement weight and equipment energy consumption coefficient; and the execution control instructions specifically include pulse irrigation cycle parameters and flow rate adjustment step size.

3. The high-standard farmland intelligent irrigation system based on the Internet of Things and data analysis according to claim 1, characterized in that: The soil moisture perception module includes a data acquisition submodule, a noise suppression submodule, and a feature extraction submodule; The data acquisition submodule synchronously acquires the soil dielectric constant based on a capacitive sensor array, uses an adaptive sliding window algorithm to eliminate temperature drift error, calculates the volumetric water content through the frequency domain reflection principle, and generates the original soil moisture dataset. The noise suppression submodule is based on the original soil moisture dataset. It uses an improved wavelet threshold denoising algorithm to separate the effective signal from high-frequency noise, identifies abnormal data points through kurtosis detection and performs interpolation repair, and generates a filtered data matrix. The feature extraction submodule extracts soil moisture migration features based on the filtered data matrix using variational mode decomposition algorithm, constructs water-holding curve models for different soil layers using kernel density estimation method, and generates soil moisture feature maps.

4. The high-standard farmland intelligent irrigation system based on the Internet of Things and data analysis according to claim 1, characterized in that: The soil moisture forecasting module includes a meteorological analysis submodule, a demand modeling submodule, and a grade classification submodule; The meteorological analysis submodule uses a grey prediction model to process wind speed and radiation intensity sequences, identifies chaotic attractor characteristics of meteorological elements through a phase space reconstruction algorithm, calculates the dynamic fluctuation of latent heat of evaporation, and generates an evaporation trend map. The demand modeling submodule is based on the evaporation trend map, uses a long short-term memory network to build a crop evapotranspiration prediction model, and uses an attention mechanism to strengthen the characteristics of key growth stages to generate a daily water demand prediction curve. The classification submodule is based on the daily water demand prediction curve, uses the fuzzy C-means clustering algorithm to classify the irrigation urgency level, determines the irrigation start threshold of each region through the membership function, and generates water demand prediction data.

5. The high-standard farmland intelligent irrigation system based on the Internet of Things and data analysis according to claim 1, characterized in that: The irrigation strategy module includes a terrain analysis submodule, a hydraulic calculation submodule, and a scheme generation submodule. The terrain analysis submodule uses UAV lidar to scan the surface elevation of farmland, constructs a digital elevation model through the Delaunay triangulation algorithm, and calculates the terrain slope and runoff coefficient of each irrigation unit. The hydraulic calculation submodule is based on the digital elevation model, establishes the flow conservation equation of the pipeline node, applies Newton's iteration method to solve the pipeline velocity and pressure distribution, and generates a set of hydraulic balance parameters. The proposed scheme generation submodule uses a set of hydraulic balance parameters and a chaotic optimization algorithm to search for the optimal combination of irrigation durations. It also avoids local optima by employing a tabu search strategy to generate an irrigation control map.

6. The high-standard farmland intelligent irrigation system based on the Internet of Things and data analysis according to claim 1, characterized in that: The equipment status monitoring module includes a signal acquisition submodule, a feature separation submodule, and a health assessment submodule; The signal acquisition submodule captures the water pump current waveform based on the Rogowski coil sensor, and uses a synchronous compression transformation algorithm to enhance the time-frequency resolution of the signal and generate a high-precision current spectrum. The feature separation submodule extracts the stator current feature components of the motor based on the current spectrum diagram and uses the singular value decomposition algorithm. It then separates the bearing fault frequency components through envelope demodulation technology to generate a fault feature vector. The health assessment submodule is based on fault feature vectors and uses a deep belief network to construct an equipment degradation assessment model. It extracts hidden layer health indicators through a sparse autoencoder to generate equipment health parameters.

7. The high-standard farmland intelligent irrigation system based on the Internet of Things and data analysis according to claim 1, characterized in that: The pipeline optimization module includes a topology reconstruction submodule, a loss calculation submodule, and a pressure compensation submodule. The topology reconstruction submodule is based on the pipeline network GIS coordinate data. It uses a depth-first search algorithm to traverse the pipeline connection relationships and optimizes the pipeline network layout structure through a minimum spanning tree algorithm to generate a topology connection matrix. The loss calculation submodule is based on the topology connection matrix, applies the Kolmogorov equation to calculate the friction loss of each pipe segment, solves the water hammer pressure fluctuation amplitude by the characteristic line method, and generates a hydraulic loss parameter set. The pressure compensation submodule is based on a set of hydraulic loss parameters and uses a particle swarm optimization algorithm to dynamically adjust the speed of the variable frequency pump. It also compensates for pressure fluctuations at the end of the pipeline through a feedforward control model and generates pipeline regulation commands.

8. The high-standard farmland intelligent irrigation system based on the Internet of Things and data analysis according to claim 1, characterized in that: The multi-source data fusion module includes an evidence synthesis submodule, a weight calculation submodule, and an indicator construction submodule; The evidence synthesis submodule is based on the DS evidence theory framework. It integrates multi-sensor data using a conflict evidence correction algorithm, calculates the support of each proposition through a basic probability allocation function, and generates a fused confidence distribution. The weight calculation submodule is based on information entropy theory, applies the improved coefficient of variation method to quantify the dispersion of parameters, and combines expert experience matrix to correct the index weight coefficients to generate a dynamic weight allocation table. The indicator construction submodule is based on a dynamic weight allocation table, uses the TOPSIS method to calculate the closeness of each scheme, sorts the optimal decision schemes by the grey relational projection method, and generates fusion evaluation indicators.

9. The high-standard farmland intelligent irrigation system based on the Internet of Things and data analysis according to claim 1, characterized in that: The intelligent execution module includes an instruction decomposition submodule, a hysteresis compensation submodule, and a parameter tuning submodule; The instruction decomposition submodule is based on the irrigation control map, uses a time automaton model to analyze the valve opening and closing sequence, verifies logical conflicts through Petri nets, and generates a sequence of actionable actions. The hysteresis compensation submodule is based on the dynamic response test data of the solenoid valve, constructs a second-order hysteresis transfer function model, applies a predictive control algorithm to calculate the advance trigger compensation amount, and generates a correction control signal. The parameter tuning submodule optimizes the PID control parameters based on the modified control signal using a chaotic ant colony algorithm, verifies the convergence of the adjustment process using the Lyapunov stability criterion, and generates execution control commands.

10. The high-standard farmland intelligent irrigation system based on the Internet of Things and data analysis according to claim 4, characterized in that: The grey prediction model adopts a metabolic data update mechanism and adjusts the development coefficient and grey action through a background value optimization algorithm to construct a dynamic prediction equation that adapts to the characteristics of meteorological abrupt changes.

Images