CN-121997262-A - Multi-farmland cooperative irrigation method and system integrating meteorological conditions and AI large model

CN121997262ACN 121997262 ACN121997262 ACN 121997262ACN-121997262-A

Abstract

The invention belongs to the field of automatic irrigation control, and particularly relates to a multi-farmland collaborative irrigation method and system integrating meteorological conditions and an AI large model, wherein the method is characterized in that a growth period water demand model with rainfall prediction is constructed, rainfall intensity, duration and soil layering water content change are comprehensively considered, and standard water demand deviation of each farmland is accurately calculated; the BP neural network is optimized by adopting a particle swarm algorithm, the water content consistency probability is maximized, multidimensional parameters such as water flow diffusion mapping weight, farmland area, soil characteristics and the like are combined, the optimal opening value and opening time stamp of each digital water-saving gate opening are dynamically solved, an irrigation simulation control platform is introduced to perform closed loop verification and iterative optimization, the water content consistency probability and the irrigation deviation value are monitored in real time, the precise control of multi-farmland collaborative irrigation is realized, the water resource utilization efficiency and irrigation uniformity are effectively improved, and the water demand of each farmland is ensured to be accurately met.

Inventors

WANG YUNBO
XU LIANG
HUANG WU
CHEN Jin
LU JIAYUAN
LI MIN
QIN YUNZHEN
Chao Mengfan

Assignees

上海熊猫机械(集团)有限公司

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (10)

1. A multi-farmland collaborative irrigation method integrating meteorological conditions and an AI large model is characterized by comprising the following steps: acquiring the real-time water content of each farmland, and acquiring the standard water demand deviation amount of each farmland and the standard total water demand deviation amount of all farmlands by combining a preset growth period water demand model with rainfall prediction; Acquiring the area of each farmland, the distance between a corresponding digital water gate opening and a digital water gate total opening of each farmland, taking the preset water content consistency probability as a target, taking the standard water demand deviation of each farmland and the standard total water demand deviation of all farmlands as constraints, and combining a BP neural network optimized by a particle swarm algorithm and a preset water flow diffusion mapping weight to acquire the opening value and opening time stamp of the corresponding digital water gate opening, wherein the digital water gate opening corresponds to each farmland one by one; Responding to the opening value and the opening time stamp of each digital water gate opening, carrying out real-time irrigation simulation by combining a preset irrigation simulation control platform, monitoring the irrigation deviation value and the water content consistency probability corresponding to each farmland after irrigation simulation through a configured flowmeter and a configured camera, and feeding back the irrigation deviation value corresponding to each farmland after simulation to the irrigation simulation control platform for simulation until the water content consistency probability meets the preset threshold when the water content consistency probability does not meet the corresponding preset threshold.
2. The multi-farmland co-irrigation method for fusing meteorological conditions and large AI models according to claim 1, wherein the process of constructing and training the growth cycle water demand model with rainfall prediction comprises: obtaining standard growth water demand of each farmland target crop in each growth period, and carrying out time stamp alignment and pretreatment on precipitation parameters of corresponding farmland areas in each growth period and a rainfall soil layering water content change value after rainfall, wherein the precipitation parameters comprise rainfall intensity, rainfall duration and historical rainfall prediction accuracy confidence; and establishing a rainfall intensity-layered effective infiltration amount mapping function based on a water retention coefficient corresponding to the rainfall intensity-layered effective infiltration amount change value of the farmland soil after rainfall and the soil type, wherein the rainfall intensity-layered effective infiltration amount mapping function is obtained by acquiring the water content of each layer of soil after rainfall of each farmland under the conditions of the corresponding soil type under the conditions of the water content of each layer of soil before rainfall, the corresponding rainfall intensity and the duration time length.
3. The multi-farmland co-irrigation method of the fusion meteorological conditions and AI large model according to claim 2, wherein the construction and training process of the growth cycle water demand model with rainfall prediction further comprises: Constructing a rainfall prediction input sequence based on the obtained crop growth stage codes, rainfall intensity and rainfall duration of the corresponding growth period, historical rainfall prediction accuracy confidence level and rainfall intensity-layered effective infiltration amount mapping function; and constructing an output layer characteristic vector based on the rainfall-compensated farmland soil layering water content change value and the rainfall-compensated farmland water content.
4. The multi-farmland co-irrigation method of fusing meteorological conditions with AI large models of claim 3, wherein the process of constructing and training the growth cycle water demand model with rainfall prediction further comprises: Inputting the rainfall prediction input sequence and the output layer characteristic vector into a growth cycle water demand model with rainfall prediction constructed by a BP neural network, training by combining a preset training loss function, obtaining a training completed growth cycle water demand model with rainfall prediction, and outputting a predicted farmland soil layering water content change value after rainfall and a farmland water content after rainfall compensation; The preset training loss function is obtained by constructing a difference value between a predicted rainfall farmland soil layered water content change value and a truly measured rainfall farmland soil layered water content change value and a difference value between a predicted rainfall compensated farmland water content and a real-time measured rainfall compensated farmland water content; and subtracting the predicted rainfall compensated farmland water content from the standard growth water content of each farmland target crop in each growth period to obtain each farmland standard water demand deviation amount, wherein the predicted rainfall compensated farmland water content is obtained by combining the farmland water content before rainfall with the effective infiltration amount corresponding to rainfall.
5. The multi-farmland co-irrigation method of the fusion meteorological conditions and AI large model of claim 4, wherein the water flow diffusion mapping weight is used for representing the effective diffusion area and diffusion uniformity of irrigation water quantity in unit time in different farmlands under the conditions of soil water content, topography conditions and irrigation flow rate of each farmland, and is used for adjusting the opening value and opening time of a digital water gate dividing gate of different farmlands, and the construction process of the water flow diffusion mapping weight comprises the following steps: And obtaining the soil type, the average water content of the soil, the gradient of the farmland, the irrigation flow rate and the irrigation effective diffusion area and irrigation effective diffusion area ratio of different farmland in unit time, and carrying out pretreatment and alignment, wherein all the irrigation effective diffusion area ratios are constructed by the ratio of the effective diffusion area of irrigation in unit time to the total area of the corresponding farmland.
6. The multi-farmland co-irrigation method of fusing meteorological conditions with AI large models of claim 5, wherein the process of constructing the water flow diffusion mapping weights further comprises: taking the effective irrigation diffusion area after pretreatment as a dependent variable, taking irrigation flow rate as an independent variable, and carrying out fitting by combining a multiple regression algorithm with each farmland soil type, soil average water content and farmland gradient as covariates to obtain a diffusion fitting function; Based on a diffusion fitting function and an irrigation simulation control platform and a controlled variable method, with the aim of maximum uniformity of the irrigation effective diffusion area ratio, carrying out simulation adjustment on the irrigation flow velocity to obtain a regression coefficient corresponding to a farmland flow velocity sequence and a diffusion fitting function when the uniformity is maximum, and constructing to obtain the water flow diffusion mapping weight.
7. The multi-farm collaborative irrigation method according to claim 6, wherein obtaining the opening value and the opening time stamp of each digital water gate opening comprises: Acquiring the area of each farmland, the distance between a water gate separating gate corresponding to a plurality of bytes and a water gate total gate corresponding to each farmland, the soil type of each farmland, the average water content of soil, the gradient of the farmland, a mapping relation table of the opening degree and the flow velocity of the gate, the standard water demand deviation amount of each farmland and the standard total water demand deviation amount of all farmlands, and constructing an input sequence of a particle swarm algorithm; and constructing a constraint space by taking the product of the water content consistency probability and the consistency of the irrigation effective diffusion area ratio, the irrigation flow rate, the irrigation starting time stamp, the water flow diffusion mapping weight and the BP neural network construction function as fitness functions and taking the standard water demand deviation amount of each farmland, the standard total water demand deviation amount of all farmlands, the farmland area and the farmland gradient as fitness functions.
8. The multi-farm collaborative irrigation method according to claim 7, wherein the method for merging meteorological conditions and AI big models, wherein the method for obtaining the opening value and the opening time stamp of each digital water gate opening, further comprises: Inputting an input sequence, an fitness function and a constraint space of a particle swarm algorithm into the particle swarm algorithm, and carrying out iterative training by combining a preset training period to obtain irrigation flow rate and an irrigation starting time stamp which are output in each iteration; And generating coordinated irrigation control instructions based on the combination of the irrigation flow rate and the irrigation start time stamp of each farmland output by each iteration and a fuzzy control algorithm, and performing irrigation simulation based on the coordinated irrigation control instructions and an irrigation simulation control platform, so as to obtain the consistency of the real-time water demand deviation amount, the water content consistency probability and the irrigation effective diffusion area ratio of all farmlands.
9. The multi-farm collaborative irrigation method according to claim 8, wherein the method for merging meteorological conditions and AI big models, wherein the method for obtaining the opening value and the opening time stamp of each digital water gate opening, further comprises: When at least one of the real-time water demand deviation amount, the water content consistency probability and the consistency of the irrigation effective diffusion area ratio of all farmlands does not meet the corresponding threshold value, iterating until the corresponding threshold value is met, and obtaining the irrigation flow rate and the irrigation starting time stamp of each farmland; And when at least one of the iteration ends and the corresponding threshold value is not met, feeding the standard water demand deviation quantity of the corresponding farmland into the growth cycle water demand model with rainfall prediction for correction until the consistency of the real-time water demand deviation quantity, the water content consistency probability and the irrigation effective diffusion area ratio of all farmland meet the corresponding threshold value.
10. The multi-farmland cooperative irrigation system integrating the meteorological conditions and the large AI model is used for realizing the multi-farmland cooperative irrigation method integrating the meteorological conditions and the large AI model according to any one of claims 1 to 9, and is characterized by comprising a water demand prediction module, an optimization module and a simulation feedback module; The water demand prediction module is used for obtaining the standard water demand deviation amount of each farmland and the standard total water demand deviation amount of all farmlands by combining the real-time water content of each farmland with a preset growth period water demand model with rainfall prediction; The optimizing module is used for obtaining the area of each farmland, the distance between the corresponding digital water gate opening of each farmland and the total water gate opening of the digital water gate, taking the preset water content consistency probability as a target, taking the standard water demand deviation of each farmland and the standard total water demand deviation of all farmlands as constraints, and combining a BP neural network optimized by a particle swarm algorithm to calculate and obtain the opening value and the opening time stamp of the corresponding digital water gate opening of each digital water gate, wherein the digital water gate opening of each digital water gate corresponds to each farmland one by one; the simulation feedback module responds to the opening value and the opening time stamp of each digital sluice gate opening, performs real-time irrigation simulation by combining a preset irrigation simulation control platform, monitors the irrigation deviation value and the water content consistency probability corresponding to each farmland after irrigation simulation, and feeds back the irrigation deviation value corresponding to each farmland after simulation to the irrigation simulation control platform for simulation until the water content consistency probability meets the preset threshold when the water content consistency probability does not meet the corresponding preset threshold.

Description

Multi-farmland cooperative irrigation method and system integrating meteorological conditions and AI large model Technical Field The invention belongs to the field of automatic irrigation control, and particularly relates to a multi-farmland collaborative irrigation method and system integrating meteorological conditions and an AI large model. Background The multi-farmland accurate irrigation system is characterized by multiple water sources, multiple gates connected in parallel, uneven pipe network impedance distribution, complex control link coupling, dynamic migration of crops in need of water and the like along with large-scale deployment of the digital sluice gate, soil humidity sensing and weather prediction module in the irrigation system through an Internet of things interface; in a weak pipe network or high-permeability irrigation scene, a central control system is usually required to bear the tasks of water distribution, pressure/flow stabilization, irrigation time sequence coordination, multi-gate coordination and the like, irrigation deviation caused by the interaction of a controller and the dynamic state of the pipe network and multiple links of soil, crops and climate easily occurs, and a dominant deviation mode can migrate along with the topology of the pipe network, equivalent impedance of the gates, growth stages of crops and changes of meteorological conditions. In the prior art, one common scheme is to carry out irrigation planning and regulation based on a static model, for example, a soil moisture balance model or an equivalent hydraulic model is constructed, and an irrigation strategy is determined by combining methods such as water demand prediction, timing and quantitative control, regional rotation irrigation, empirical adjustment based on historical data and the like; the method has better feasibility when the climate is stable, the soil is homogeneous and the crop growth stage is known, but in the actual scene of parallel connection of multiple farmlands, various crop types, obvious soil space variation and frequent fluctuation of meteorological conditions, the model establishment and parameter calibration cost is high, the method is sensitive to real-time dynamic soil moisture, rainfall uncertainty, sensing noise and gate response lag, the result of the irrigation decision lag or the setting result deviates from the actual water requirement easily, and the accurate and balanced irrigation effect is difficult to maintain when the crop water requirement changes rapidly. The other scheme adopts a data driving or rule reasoning model to judge and decide the farmland state, such as feature extraction and irrigation judgment are carried out on soil humidity, weather forecast and other data by using fuzzy logic or a simple neural network, and the opening and closing of a gate or the adjustment of time length are triggered according to the feature extraction and the irrigation judgment, the method reduces the dependence on accurate physical modeling to a certain extent, but the conventional model still has insufficient suitability in an accurate irrigation scene, the averaging or zoning aggregation operation adopted for reducing the computational complexity possibly weakens the fine granularity identification on the specific water demand of a single farmland, the high-fidelity characterization is difficult to realize on the irrigation deviation of multiple farmland coupling, the modeling time sequence trend is difficult to model, the internal decision is mostly black box output, the contribution positioning basis for the collaborative water quantity allocation of multiple gates is difficult to be provided while the integral irrigation deviation is identified, the problems that the pressure adjustment and the water quantity allocation still tend to be uniformly set or experience rules, the hydraulic interference among parallel gates, the insufficient irrigation or excessive waste and the like are easy to occur. Therefore, the technical problem to be solved in the prior art is how to realize high-fidelity extraction of water demand deviation characteristics of soil-climate-object coupling under a complex irrigation system environment with multiple farmlands connected in parallel, and accurately position deviation contribution sources while identifying an overall irrigation unbalanced mode, so that targeted pressure regulation and water quantity dynamic collaborative distribution are performed. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a multi-farmland collaborative irrigation method and system integrating meteorological conditions and an AI large model, and the method is characterized in that a growth period water demand model with rainfall prediction is constructed, rainfall intensity, duration and soil layering water content change are comprehensively considered, and the standard water demand deviation of each farmland is accurately calculate