Water and fertilizer integrated multi-objective optimization intelligent regulation and control system based on artificial intelligence driving

Abstract

The invention relates to the technical field of automatic control, and further relates to a water and fertilizer integrated multi-objective optimization intelligent regulation and control system based on artificial intelligence driving, and the system comprises a water and fertilizer load analysis unit and a water and fertilizer combined regulation and control criterion generation unit. The water and fertilizer load analysis unit is used for acquiring meteorological data from a meteorological monitoring system; simultaneously receiving data from the soil sensor and the leaf surface sensor, acquiring growth state data of the crops, and evaluating water demand and fertilizer demand of the crops in unit area at each time; and the water and fertilizer combined regulation and control criterion generation unit is used for dynamically calculating the water application amount and the fertilization amount of the crops in unit area at each time according to the soil state and the root system absorption capacity of the crops by using an AI dynamic resource optimization engine. According to the invention, the utilization efficiency of moisture and nutrients is effectively improved, the waste of resources is reduced, and the intelligence, adaptability and robustness of regulation and control are enhanced.

Description

Technical Field

[0001] The present invention belongs to the field of automatic control technology, and specifically relates to an artificial intelligence-driven water-fertilizer integrated multi-objective optimization intelligent control system. Background Art

[0002] In modern agricultural production systems, crop water and nutrient management is a key factor influencing yield and resource utilization efficiency. Especially in the context of the rapid development of precision agriculture, traditional timed and quantitative irrigation and conventional proportional fertilization technologies have become unable to meet the multiple requirements of high yield, water conservation, environmental protection, and intelligent management. Consequently, a large number of researchers and agricultural engineering application units have conducted technical explorations focusing on water-fertilizer integration, automated irrigation control, and precise nutrient supply, and have made considerable progress. However, while these existing technologies have laid the foundation for agricultural informatization and automation, a series of unresolved issues remain in areas such as coordinated control, dynamic response capabilities, adaptive regulation, and closed-loop optimization, which restrict their universal adaptability and long-term stability under large-scale, multi-climate, and heterogeneous plot conditions.

[0003] Currently, widely used water and fertilizer regulation systems often employ static formulation strategies based on empirical models. For example, many regions establish empirical "crop-period-dosage" databases based on years of crop cultivation data, and then formulate irrigation and fertilization plans based on weather forecasts and manual field inspections. While this approach performs acceptably in normal seasons and on stable plots, it lacks real-time adaptability to factors such as abnormal climate (such as sudden high temperatures and high evapotranspiration days), complex terrain conditions (such as sloping land and varying soil bulk density distributions), and individual crop differences, making it difficult to ensure the timeliness of regulation strategies and their alignment with actual crop needs.

[0004] To overcome the limitations of static models, some studies have introduced real-time monitoring platforms based on sensor data, using parameters such as soil moisture, electrical conductivity, temperature and humidity to perform rule-triggered irrigation and fertilization control. This type of system generally uses a threshold mechanism. For example, when the soil moisture content falls below a preset value, valve control is automatically triggered to perform irrigation operations, or the appropriate fertilizer concentration is determined based on changes in EC values. Although this type of method introduces preliminary feedback control logic, its essence is still a "single factor control" or "preset threshold trigger" logic structure, and it has not formed a system-level dynamic collaborative optimization mechanism. More importantly, this method cannot identify the physiological feedback of crops, and only formulates operations based on soil or meteorological signals, ignoring the need for comprehensive regulation of crop absorption capacity, nutrient transport rate and growth stage characteristics. Summary of the Invention

[0005] The main object of the present invention is to provide an intelligent control system for multi-objective optimization of water and fertilizer integration driven by artificial intelligence. By constructing a water and fertilizer load analysis unit, a water and fertilizer joint control criterion generation unit, and a water and fertilizer feedback control unit, the system realizes real-time assessment, dynamic optimization calculation, and closed-loop correction control of the water and fertilizer requirements of crops. The system integrates an AI dynamic optimization engine for water application rate and fertilizer application rate, and incorporates a historical residual compensation mechanism, which can accurately output control quantities driven by multi-source environmental data and crop physiological parameters, ensuring that the water and fertilizer control not only responds to the actual needs of crops but also has long-term stability and environmental adaptability. The system effectively improves the utilization efficiency of water and nutrients, reduces resource waste, and enhances the intelligence, self-adaptability, and robustness of the control, and is applicable to intelligent agricultural production in multi-plot and multi-crop scenarios.

[0006] To solve the above problems, the technical solution of the present invention is realized as follows: An intelligent control system for multi-objective optimization of water and fertilizer integration driven by artificial intelligence, the system comprising: a water and fertilizer load analysis unit and a water and fertilizer joint control criterion generation unit; the water and fertilizer load analysis unit is used to obtain meteorological data from a meteorological monitoring system; at the same time, receive data from soil sensors and leaf surface sensors to obtain the growth state data of crops, and evaluate the water demand and fertilizer demand of crops per unit area at each time; the water and fertilizer joint control criterion generation unit is used to use an AI dynamic resource optimization engine to dynamically calculate the water application rate and fertilizer application rate of crops per unit area at each time according to the soil state and the root absorption capacity of crops; the AI dynamic resource optimization engine includes: a water application rate AI dynamic optimization engine and a fertilizer application rate AI dynamic optimization engine; the water application rate AI dynamic optimization engine is obtained by adding a water application rate AI dynamic optimization model and a water application rate AI historical residual accumulation part; the fertilizer application rate AI dynamic optimization engine is obtained by adding a fertilizer application rate AI dynamic optimization model and a fertilizer application rate AI historical residual accumulation part.

[0007] Further, the execution process of the irrigation water volume AI dynamic optimization engine includes: calling the residual sequence of the most recent complete growth cycle in the irrigation water volume AI historical residual accumulation part, setting a negative definite convergence domain based on the linear quadratic Lyapunov function, and writing it into the residual accumulation register and the safety threshold register; real-time collecting the soil state and the crop root absorption capacity, writing them together with the irrigation record of the previous time period into the state observation buffer, and forming a state vector after unified scaling; the irrigation water volume AI dynamic optimization model calls a linear matrix inequality solver to output the original irrigation water control amount that satisfies the Lyapunov descent constraint; at the same time, the adaptive gain modulator dynamically updates the matrix weight according to the convergence rate; the irrigation water volume AI historical residual accumulation part incorporates the latest error into the residual accumulation register, generates a compensation amount of the same dimension through a variable gain compensator, and buffers rapid disturbances; the original irrigation water control amount and the compensation amount are added dimension by dimension, and are limited within the execution threshold range by a saturation protector to obtain the target irrigation water volume, which is written into the execution instruction queue.

[0008] Further, the process of calling the residual sequence of the most recent complete growth cycle in the irrigation water volume AI historical residual accumulation part and setting a negative definite convergence domain based on the linear quadratic Lyapunov function includes: Step A1: Perform integrity verification on the residual sequence of the irrigation water volume AI historical residual accumulation part, linearly interpolate the missing segments in chronological order to obtain the verified sequence; Step A2: Write the verified sequence into the residual buffer stack, and use a double-weight exponential decay filter to suppress high-frequency noise to obtain a filtered sequence; Step A3: Detect the range of the filtered sequence. If it exceeds the threshold of the safety threshold register, trigger the abnormal backtracking logic and return to Step A1; Step B1: Dynamically generate a positive definite weighting matrix according to the variance distribution of the filtered sequence and write it into the matrix register; Step B2: Construct a linear quadratic Lyapunov function with the positive definite weighting matrix and the residual state vector in quadratic form, and verify the strict descent for all samples through a gradient iterator; Step B3: If the descent rate of any sample is non-negative, adaptively shrink the positive definite weighting matrix and return to Step B2 until all samples meet the conditions; Step C1: Define the boundary of the negative definite convergence domain with the maximum acceptable energy level of the linear quadratic Lyapunov function and write it into the negative definite convergence domain register; Step C2: Synchronize the boundary parameters to the safety locking logic and the irrigation water volume AI dynamic optimization engine to complete the setting of the negative definite convergence domain.

[0009] Further, the execution process of the fertilization amount AI dynamic optimization engine includes: retrieving the residual sequence of the most recent complete growth cycle from the fertilization amount AI historical residual accumulation part, writing it into the residual buffer stack, establishing the initial value of the sliding mode surface and recording the zero-bias baseline; collecting the soil state and the crop root absorption capacity, reading the fertilization record of the previous time period, standardizing them to the same scale and writing them into the state observation buffer area to form a state vector; the fertilization amount AI dynamic optimization model generates the original fertilization control amount based on the difference between the sliding mode surface and the state vector, and adopts the Dahn switching law; within the same cycle, the adaptive gain modulator updates the switching rate according to the approximation speed; the fertilization amount AI historical residual accumulation part superimposes the nutrient deviation of this cycle, and is converted into an equal-dimensional compensation amount through a variable gain compensator to suppress high-frequency disturbances; the original fertilization control amount and the compensation amount are added dimension by dimension, and are limited to the safety threshold by a saturation protector to obtain the fertilization amount, which is then written into the execution instruction queue.

[0010] Further, the process by which the fertilization amount AI dynamic optimization model generates the original fertilization control amount based on the difference between the sliding mode surface and the state vector includes: establishing a sliding mode surface register in the fertilization amount AI dynamic optimization model, and mapping the crop nutrient deviation and its time sum into a sliding mode surface vector; reading the state vector difference in real time, forming a multi-dimensional deviation mark through a difference comparator and writing it into a deviation latch; the output of the deviation latch is sent to a Dahn switching law generator to generate a symbol matrix; the Dahn switching law generator matches the symbol matrix dimension by dimension according to a pre-stored switching gain table to generate an initial switching pulse sequence; the adaptive gain modulator increases or decreases the modulation of the switching gain according to the deviation amplitude and the real-time threshold of the fertilization amount AI historical residual accumulation part, and outputs a gain modulation matrix; the initial switching pulse sequence is multiplied by the gain modulation matrix to obtain an open-loop control sequence, which is shaped by a rate limiter and a boundary layer filter in a two-stage manner to suppress high-frequency chattering.

[0011] Further, the system also includes: a water and fertilizer feedback regulation unit, which is used to measure the crop water content state index, combine it with the water application amount, calculate the dynamic corrected water application amount of the crop per unit area at each time, perform closed-loop feedback regulation of crop water application, and measure the crop nutrient sufficiency index, combine it with the fertilization amount, calculate the dynamic corrected fertilization amount of the crop per unit area at each time, and perform closed-loop feedback regulation of crop fertilization.

[0012] Further, the execution process of the water and fertilizer feedback control unit includes: reading the dynamically corrected water application rate and dynamically corrected fertilizer application rate stored in the previous operation cycle, writing them into the water feedback register and the nutrient feedback register, establishing a dual-channel error latch and clearing it; collecting the measured crop water content status index and the measured crop nutrient sufficiency index in real time, suppressing high-frequency noise through a dual-window index filter and writing them into the status observation buffer; calling the current water application rate and fertilizer application rate, forming a water content error vector and a nutrient error vector in the error latch respectively, and recording the change rate mark; performing dual-loop proportional-integral modulation, including: the water content error vector enters the water integral modulator to generate a water integral quantity; the nutrient error vector enters the nutrient integral modulator to generate a nutrient integral quantity; the two channels simultaneously perform proportional-integral superposition and output a first correction quantity; the first correction quantity is limited within the crop safety threshold through a limiter saturator to generate a second correction quantity and write it into the dynamic correction register; the second correction quantity in the dynamic correction register is added to the original water application rate and the original fertilizer application rate one by one to obtain the dynamically corrected water application rate and dynamically corrected fertilizer application rate of the crop per unit area at the current time, and at the same time push them to the execution instruction queue; after irrigation and fertilization are completed, re-collect the measured crop water content status index and the measured crop nutrient sufficiency index; if the absolute value of any error vector does not decrease for two consecutive cycles, increase the proportional-integral gain and return to the dual-loop proportional-integral modulation, otherwise maintain the current gain and enter the next regular cycle.

[0013] Further, the meteorological data includes: crop evapotranspiration , in millimeters per day; rainfall , in millimeters per day; the proportion of rainfall that is not absorbed by the soil and directly runs off, represented by the surface runoff rate ; the growth status data of the crop includes: soil water potential , in kPa; crop root suction , in kPa; the nitrogen demand of the crop per unit area , in g / m² / day; the phosphorus demand of the crop per unit area , in g / m² / day; the potassium demand of the crop per unit area in g / m² / day; soil adsorption ratio ; leaf chlorophyll concentration ; is time.

[0014] Further, the water demand of the crop per unit area evaluated at each time is: ; Among them, is the water demand of the crop per unit area at time ; the fertilizer demand of the crop evaluated at each time is: ; Among them, is the fertilizer requirement of crops per unit area per unit time; is the theoretical chlorophyll concentration of crops at the current growth stage at time.

[0015] The artificial intelligence-driven integrated water and fertilizer multi-objective optimization intelligent control system of the present invention has the following beneficial effects: By constructing a complete control system integrating data perception, collaborative decision-making, dynamic optimization and closed-loop feedback, the intelligent integration and real-time precise management of crop irrigation and fertilization are realized. Compared with the traditional single-channel, quantitative or threshold-triggered water and fertilizer control technologies, the present invention first introduces a dual-channel AI dynamic optimization engine architecture at the control layer, and couples the AI dynamic optimization model of water application amount and the AI dynamic optimization model of fertilization amount with the historical residual compensation mechanism respectively, effectively realizing the fine fitting and continuous correction of the dynamic demands of crop water and fertilizer. At the execution level, through the real-time monitoring of the crop water content status index and nutrient sufficiency index by the water and fertilizer feedback control unit, the closed-loop tracking and dynamic correction of the actual control results are realized, and the adaptive ability and stability of the control system are improved. In terms of the scheduling strategy, the system adopts a small time difference execution logic of giving priority to water application and delaying fertilization, significantly reducing the risk of rhizosphere loss. In terms of ensuring the system stability, the present invention adopts linear quadratic Lyapunov control and sliding mode control technologies to ensure that the system control solution maintains asymptotic convergence throughout the whole cycle, avoiding the damage to the crop root zone caused by long-term error drift. Generally speaking, the present invention can effectively improve the water and nutrient use efficiency, reduce the waste of agricultural resources, enhance the robustness of the system to environmental disturbances, be applicable to large-scale automatic deployment under multiple crop types, multiple meteorological regions and complex terrains, and has good application and popularization value. Description of the Drawings

[0016] Figure 1 is a schematic diagram of the system structure of the artificial intelligence-driven integrated water and fertilizer multi-objective optimization intelligent control system provided by an embodiment of the present invention; Figure 2 is a schematic diagram of the sliding mode control effect of the AI dynamic optimization engine for fertilization amount provided by an embodiment of the present invention; Figure 3 is a schematic diagram of the collaborative effect between the water and fertilizer load analysis unit and the AI dynamic resource optimization engine provided by an embodiment of the present invention; Figure 4 is a schematic diagram of the comparison of the control effect of the AI dynamic optimization engine for water application amount provided by an embodiment of the present invention. Detailed Embodiments

[0017] To enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solution in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Reference Figure 1 : An intelligent control system for integrated water and fertilizer multi-objective optimization driven by artificial intelligence. The system includes: a water and fertilizer load analysis unit and a water and fertilizer joint control criterion generation unit; the water and fertilizer load analysis unit is used to obtain meteorological data from a meteorological monitoring system; at the same time, receive data from soil sensors and leaf sensors to obtain the growth state data of crops, and evaluate the water demand and fertilizer demand of crops per unit area at each time; the water and fertilizer joint control criterion generation unit is used to use an AI dynamic resource optimization engine to dynamically calculate the water application amount and fertilizer application amount of crops per unit area at each time according to the soil state and the root absorption capacity of crops; the AI dynamic resource optimization engine includes: a water application amount AI dynamic optimization engine and a fertilizer application amount AI dynamic optimization engine; the water application amount AI dynamic optimization engine is obtained by adding a water application amount AI dynamic optimization model and a water application amount AI historical residual accumulation part; the fertilizer application amount AI dynamic optimization engine is obtained by adding a fertilizer application amount AI dynamic optimization model and a fertilizer application amount AI historical residual accumulation part.

[0019] Regarding the farmland as an open thermodynamic system that continuously evolves over time: the radiation, rainfall, and evapotranspiration input from the atmospheric boundary layer result in continuous exchange of water energy; the ion adsorption, desorption, and microbial mineralization in the rhizosphere environment determine the real-time transfer of nutrient potential energy. The water and fertilizer load analysis unit constructs a dynamic causal chain with the main line of "state - drive - response" through synchronous monitoring of these two energy paths. This causal chain first normalizes the scale of the external drive from the meteorological monitoring system, then aligns the time series with the water content state data output by the soil sensor and the photosynthetic pigment concentration data output by the leaf sensor, and then maps the three types of information into the same multi-dimensional phase space. The mapping process adopts an end-to-end time series embedding strategy, so that each sampling occupies a unique coordinate in the phase space, thus forming a continuous trajectory. The tangent vector of the trajectory in space describes the instantaneous change rate of water demand and fertilizer demand, and the normal vector reflects the vulnerability of the crop to external disturbances. It is precisely by using this "tangent - normal" coupling relationship that the water and fertilizer load analysis unit transforms the complex and variable physiological process into a smooth curve that can be immediately processed by a computer, and then obtains the water demand and fertilizer demand per unit area at each time point.

[0020] The water-fertilizer combined regulation criterion generation unit is located in the "decision-making layer" of the system. Its principle can be understood as finding a real-time updatable elastic balance point among the "time-varying demand curve - finite supply threshold - environmental constraints". The AI dynamic resource optimization engine is the computational core of this unit. It deploys both the AI dynamic optimization engine for irrigation water volume and the AI dynamic optimization engine for fertilization amount simultaneously. The two channels follow the same set of control philosophies in principle, but maintain differences in implementation details to avoid the explosion of solution dimensions caused by large-scale coupling. The AI dynamic optimization model for irrigation water volume inside the AI dynamic optimization engine for irrigation water volume regards "the dissipation of water in soil pores through both seepage and transpiration channels" as the main control object, and constructs an asymptotically stable domain with the help of an energy function. The accumulated historical residuals part of the irrigation water volume AI records the deviation between the model prediction and the real feedback, and injects these deviations as novel perturbations into the control law to achieve adaptive gain adjustment. This architecture of "model output - residual compensation" parallel superposition enables the AI dynamic optimization engine for irrigation water volume to quickly correct the width and interval of irrigation pulses through residual drive even when there are drastic mutations in external meteorological conditions, and maintain the steady decline of the water potential in the rhizosphere.

[0021] The principle of the AI dynamic optimization engine for fertilization amount starts from the perspective of sliding mode control. Its core is to set a slidable target hyperplane for the "concentration of ions that can be absorbed by crops". The AI dynamic optimization model for fertilization amount calculates the distance of the sliding mode surface in real time according to the difference between the root absorption ability of the crop and the soil buffer capacity, and then generates a bipolar pulse sequence with the reaching law switching, so that the system state slides towards the hyperplane with a high slope within a finite time. The accumulated historical residuals part of the fertilization amount AI continues to track the accumulation and distribution of nutrient deviations on this basis, and converts this information into an additional compensation vector to achieve the suppression of high-frequency oscillations of the concentration. In principle, this "two-stage" strategy ensures the rapid response of the model to step disturbances, and at the same time enables the residual channel to have long-term immunity to low-frequency drifts.

[0022] The AI dynamic resource optimization engine adopts an asynchronous parallel mechanism in overall scheduling: the AI dynamic optimization engine for irrigation water volume and the AI dynamic optimization engine for fertilization amount run in independent threads respectively. After each completes prediction, compensation, superposition and issuing instructions, the scheduling coordinator realizes serial queuing within a millisecond time slot. The priority arbitration rule is set inside the scheduling coordinator. When the system detects that the irrigation operation may change the ion migration path in the rhizosphere, it will temporarily increase the time slot occupancy ratio of the fertilization instruction to ensure that the water pulse has a precursor effect on the nutrient pulse, thus meeting the agronomic principle of "water first, then fertilizer, and fertilize with water".

[0023] The target water application rate and target fertilization rate output by the water-fertilizer combined regulation criterion generation unit enter the field execution layer and are distributed to the solenoid valve array and variable flow fertilization device through a high-speed field bus. The hardware in the execution layer follows the event-driven protocol and only starts to act when receiving a timestamped pulse packet, avoiding resource interlocks caused by multitasking preemption. After each hardware action ends, the execution layer immediately generates a feedback packet and sends it back to the water-fertilizer feedback regulation unit. The principle of the water-fertilizer feedback regulation unit essentially decomposes a large system into two independent integral loops: the water loop processes the water content state index, and the fertilizer loop processes the nutrient sufficiency index. The dual-loop proportional-integral modulator assigns different integral time constants to enhance the system's adaptability to different dynamic characteristics. The amplitude limiter and saturator are set after the two loops to shield the impact of abnormal peaks on the crop safety threshold. The corrected compensation amount is rewritten into the execution buffer and simultaneously written into the AI historical residual accumulation part of the water application rate and the AI historical residual accumulation part of the fertilization rate, forming a true closed loop of "measurement - decision - execution - feedback - compensation".

[0024] The system uses the multi-source fusion feature mapping of the water-fertilizer load analysis unit to transform the "nonlinear, strongly coupled, time-varying" attributes of crop requirements into a continuously traceable curve trajectory. Then, the water-fertilizer combined regulation criterion generation unit disassembles the resource allocation problem dimensionally through the "independent model prediction - residual compensation - asynchronous parallel scheduling" mechanism. Finally, the dual-loop proportional-integral modulation of the water-fertilizer feedback regulation unit realizes fast correction and asymptotic balance. Each sub-module in the three-layer architecture has an adaptive ability: the load analysis relies on continuous data streams to update the feature mapping; the dynamic optimization model reconstructs the stable domain or sliding mode surface based on real-time errors; the residual accumulation part uses the cumulative deviation as an adjustment trigger to offset the model assumption drift on a long-term scale. This global-local dual-scale adaptive framework ensures that the system can maintain real-time optimization of water and nutrient supply in the face of short-term meteorological impacts or long-term physiological changes, and also enables artificial intelligence to get rid of the controversy of pure "black box" optimization in agricultural scenarios, establishing an interpretable, verifiable, and traceable control link through residual compensation and explicit stability criteria.

[0025] Furthermore, the execution process of the irrigation water volume AI dynamic optimization engine includes: calling the residual sequence of the most recent complete growth cycle in the irrigation water volume AI historical residual accumulation part, setting a negative definite convergence domain based on the linear quadratic Lyapunov function, and writing it into the residual accumulation register and the safety threshold register; when calling the residual sequence of the most recent complete growth cycle in the irrigation water volume AI historical residual accumulation part, the system first retrieves the deviation trajectory stored in chronological order in the solid-state buffer and performs integrity verification on it. Any missing segment will be linearly interpolated based on adjacent observation points to ensure that the deviation trajectory covers all the water response characteristics of the crop from sowing to harvesting. Subsequently, the deviation trajectory enters the double-weight exponential decay filter, which compresses the residual energy within a continuous and smooth bandwidth by simultaneously attenuating high-frequency noise and low-frequency drift; this bandwidth directly determines the range of the weight matrix of the subsequent linear quadratic Lyapunov function. The system generates a symmetric positive definite weighted matrix in the temporary operation register area and constructs a linear quadratic Lyapunov function in combination with the filtered residual vector. Based on the energy level of this function, the descent rate is calculated point by point along the time axis. If a non-negative descent rate appears at any point, the system implements fine-grained adjustment on the weight matrix through the adaptive contraction algorithm until all sampling points are in a strictly decreasing state. After verification, the system uses the maximum acceptable value of the energy level as the boundary of the negative definite convergence domain, first writes the complete residual vector into the residual accumulation register for adaptive gain compensation in the next cycle; then writes the boundary of the negative definite convergence domain into the safety threshold register to provide a hard constraint for the real-time control law. Through this process, the intelligent control system for multi-objective optimization of integrated water and fertilizer ensures that the irrigation control loop can converge within the negative definite convergence domain under any external disturbance, and at the same time lays a stable foundation for subsequent residual compensation and gain self-tuning.

[0026] Soil conditions and crop root absorption capacity are collected in real time and written to the state observation buffer along with irrigation records from the previous period. After unified scaling, the state vector is formed. The system's tensiometers, conductivity probes, and variable-depth root pressure probes deployed within the field simultaneously output soil moisture trends, changes in ion exchange capacity, and rhizosphere osmotic potential with sub-second resolution. The acquisition thread synchronously writes these high-frequency streams of data into a high-speed circular buffer using a timestamp alignment mechanism. A parallel irrigation execution thread immediately writes the valve duty cycle, pulse width, and flow integral to a historical segment in the same physical address space after each valve pulse. The system aggregates these two streams of data at the on-chip bus level using dual-channel DMA to form a complete raw observation frame. Subsequently, a unified scaling pipeline is triggered within the FPGA logic. The first stage of the pipeline calls the discrete dynamic benchmark library to assign a corresponding real-time benchmark to each raw metric type. It then performs threshold clipping for sudden high-amplitude noise and sliding window drift removal for gradual trends across days. Finally, a parameter-free normalization based on quantile adaptation is used to convert all metrics to a unified dimensionless interval. The final stage of the pipeline uses a configurable concatenation table to sequentially push fields such as water gradient, absorption rate, and historical valve pulse width into a fixed-length array. This generates a "state vector" in the on-chip cache. This state vector is mapped to the input registers of the AI dynamic resource optimization engine via the AXI bus. This not only ensures that each optimization iteration uses data with the same physical time tag, but also ensures the numerical stability of model weights through scale alignment, regardless of dimensional differences. This ensures that the AI dynamic optimization engines for water and fertilizer application can converge to a unique solution even in high-concurrency scenarios, and deliver the judgment criteria to the execution layer within milliseconds.

[0027] Upon receiving the state vector, the AI dynamic optimization model for irrigation water application immediately locks the control law template matching the current crop stage and instantiates a linear matrix inequality solver in the on-chip cache. The solver uses the distributed interior point method to synchronously write the second-order approximation term of the objective function and the Lyapunov descent constraint into the reconfigurable logic array. Through an iterative process, it continuously adjusts the decision variables within an extremely small step size to keep the residual error monotonically decreasing in each iteration. To avoid the real-time loss caused by excessive iteration depth, the solver is divided into a main core and parallel co-cores at the hardware level: the main core is only responsible for searching the global feasible region, and the co-cores simultaneously perform local step tests on multiple candidate directions. Once it detects that the Lyapunov descent constraint is satisfied and the numerical condition number is within the safe range, it triggers an "early stop" signal and directly writes the solution vector corresponding to this direction as the original irrigation water application control quantity into the output register bank. The original irrigation water application control quantity undergoes a pipelined saturation protection and quantization encoding before being written out to ensure its plug-and-play compatibility with the timing pulse protocol of the solenoid valve drive stage. At the same time, the adaptive gain modulator resides in the bypass channel of the model calculation path, continuously monitors the descent slope of the Lyapunov function along the time axis, and subtracts the slope from the built-in target convergence rate of the system. The difference is mapped to the matrix weight adjustment coefficient after proportional leading and integral accumulation; this coefficient directly acts on the weight matrix elements of the linear matrix inequality in the next iteration to achieve fine-grained compression or amplification of the control law gain. When the external environment experiences severe disturbances resulting in a decrease in the convergence speed, the adaptive gain modulator will increase the weight of the key diagonal elements within a few milliseconds to quickly expand the Lyapunov descent margin; when the system enters a stable interval, the modulator will gradually decrease the gain to prevent root hypoxia caused by over-irrigation. The entire process forms a closed loop at the hardware level, maintaining the stable, fast, and energy-saving characteristics of the irrigation water application control without software interruption, thus providing a reliable reference trajectory for the subsequent residual error compensation stage.

[0028] The AI historical residual accumulation part of the water application amount immediately calls the real-time error calculation result after each irrigation feedback closed-loop ends, writes the latest error into the residual accumulation register at the same physical address, and maintains a sliding window inside the register in chronological order, so that the cumulative curve continuously grows over time without being slowed down by outdated data. When the sliding window update triggers a threshold event, the variable gain compensator is synchronously called. The variable gain compensator obtains the gain baseline matching the current crop growth stage through table lookup, and dynamically adjusts the amplification factor in combination with the root mean square value of the residuals within the window, and directly applies this factor to the latest residual to generate a compensation amount with the same dimension as the original water application control amount. At the hardware level, the compensation amount and the original water application control amount are added dimension by dimension through a pipeline adder, and the generated instantaneous vector then enters the saturation protector immediately. The saturation protector sets the upper and lower limits based on the soil structure safety threshold, the crop root pressure tolerance threshold, and the maximum flow threshold of the irrigation equipment, and performs hard truncation on any component outside the range to ensure that the target water application amount neither causes soil erosion nor leads to insufficient water supply. The processed target water application amount is directly encoded as a protocol frame on the chip, written into the execution instruction queue, and compared with the previous queue element, and then the insertion position is determined by the hardware arbiter to ensure that all irrigation pulses reach the valve control node in chronological order. Through this implementation method of three-level coordination of residual accumulation, variable gain compensation, and saturation protection, when the system faces rapid disturbances such as sudden high wind evaporation or short-term post-rain humidification, it can not only quickly correct the control vector through the compensation amount, but also maintain the overall operation boundary through the saturation protector, achieving the unity of high precision and high security.

[0029] Further, when executing step A1, first, the residual sequence of the AI historical residual accumulation part of the water application amount is quickly mapped from non-volatile storage to the on-chip dynamic cache through the bus controller, and each data unit carries a sequentially increasing timestamp and a cyclic redundancy check bit; the integrity check logic compares the timestamp interval and the check bit value item by item in a pipeline manner on the chip. If a timestamp break is found, a marker placeholder is immediately inserted into the cache. If a redundant check mismatch is detected, the corresponding unit is marked as invalid. All markers are then sent to the missing segment reconstruction module. The reconstruction module assembles a list of continuous missing segments according to the markers and calls the linear interpolation operator; this operator extracts the nearest valid data from both ends of the missing segment as endpoints and generates an interpolation sequence at equal steps in the on-chip multiplier-accumulator array. The whole process is completed in parallel by hardware to ensure that the complemented data maintains a smooth transition with the original sequence and does not introduce artifacts beyond the range. After interpolation, the verification module calculates the cyclic redundancy check bit for the new data again and updates the timestamp, and finally writes the verified sequence back to the AI historical residual accumulation part of the water application amount, and at the same time triggers an interrupt to send a status synchronization signal to the adaptive gain modulator to ensure that subsequent control law iterations can run in a data domain without gaps, thereby stabilizing the timing consistency of the intelligent control system for multi-objective optimization of water and fertilizer integration.

[0030] When the system writes the verified sequence to the residual buffer stack, it first calls the on-chip memory controller to allocate a contiguous address space to ensure the physical contiguity of the residual data along the timeline. This design enables subsequent pipeline processing to complete the migration of the entire frame in a single burst read / write, avoiding wait cycles caused by cross-page access. After the write is complete, the residual buffer stack automatically updates the top-of-stack pointer and generates a new data arrival interrupt signal, triggering the dual-weight exponential decay filter to operate. The dual-weight exponential decay filter utilizes a two-stage multiply-add structure: the first stage uses a larger attenuation coefficient to rapidly attenuate transient high-amplitude peaks, while the second stage uses a smaller attenuation coefficient to smooth inter-cycle low-frequency drift. These two stages operate in parallel in hardware through a shared multiplier array, ensuring that each data sample completes the dual attenuation within a single clock cycle. The filter also incorporates dynamic coefficient adjustment logic to fine-tune the two-stage attenuation ratio in real time based on the variance of the most recent samples in the residual buffer stack, ensuring optimal bandwidth even when external weather conditions change rapidly. The processed data directly overwrites the same address in the residual buffer stack, maintaining the compact structure of the time series and outputting a read-only copy for subsequent linear quadratic stability determination. At this point, the filter sequence is synchronized across both the system memory and cache storage levels. This not only provides a high-frequency noise-free error driver for the AI dynamic optimization engine for water application, but also achieves millisecond-level latency through hardware-level parallelism, laying a solid foundation for the real-time and stability of the integrated multi-objective optimization and intelligent control system for water and fertilizer. The filtering sequence is immediately sent to the range detection module after being written back to the residual buffer stack. The range detection module consists of two parallel pipelines. One extracts the maximum value of the current window in a single-cycle pipelined manner at the cache reading end, and the other extracts the minimum value simultaneously. The two pipelines share the same address generator to ensure data alignment. The maximum and minimum values are subtracted by hardwiring to obtain the range result, and the result is sent to a hardware comparator connected to the safety threshold register in the next clock cycle. The safety threshold register is written with a set of fixed upper and lower limits by the agronomy expert database during the system power-on calibration phase, and can be finely tuned by the adaptive gain modulator according to the long-term statistical offset during operation. When the range detection value first exceeds the threshold in the safety threshold register, the comparator outputs an overload flag to the interrupt controller; the interrupt controller generates a high-priority exception signal, immediately freezes the residual buffer stack write pointer and locks the current filtering sequence version, and at the same time issues a backtracking instruction to the exception backtracking logic. The exception backtracking logic directly accesses the verified sequence backup area written in step A1 through DMA, reloads the original data with the fastest on-chip bus width to overwrite the filtering sequence, then clears the attenuation weight history of the residual buffer stack and resets the internal state of the dual-weight exponential decay filter to its initial gain configuration. After completing the data reload, the exception backtracking logic sends a reset pulse to the range detection module, cancels the freeze and releases the write pointer, and the system automatically jumps back to step A1 to re-execute the integrity check and linear interpolation process. Through the hardware closed-loop formed by the range detection module, comparator, safety threshold register, interrupt controller, and exception backtracking logic, the system can cut off the harmful data link within the microsecond level when abnormal spikes or continuous drifts occur in the residual sequence, and roll back to the stable data state, ensuring that the intelligent control system for multi-objective optimization of water and fertilizer integration can maintain the convergence reliability of the control law and the numerical safety of the hardware path under any extreme environmental disturbance; When performing step B1, the system first conducts a fast variance scan on the filtering sequence within the current sliding window. During the scan process, the on-chip vectorized multiply-accumulate array is utilized to obtain the variance values of each dimension within a single cycle. Then, the feature statistics logic is invoked to map these variance values into a set of diagonal weights, thereby forming an initial draft of the positive definite weighted matrix. Subsequently, a controlled perturbation is injected into each diagonal line through the diagonal compensation unit to ensure that all eigenvalues of the matrix are within a strictly positive interval, and the generated result is immediately written into a dedicated matrix register. When step B2 is initiated, the controller performs a quadratic form operation on the just-written positive definite weighted matrix and the residual state vector in the multi-port arithmetic logic unit to construct the linear quadratic Lyapunov function in real time. After the construction is completed, the gradient iterator immediately executes a batch gradient evaluation on the same data channel. It calculates the descent rate of the Lyapunov function along the time axis for all samples within the window in parallel. The descent rate vector is fed back to the controller in the next clock cycle. If the gradient iterator finds that the descent rate of any sample is non-negative, step B3 is triggered. The system invokes the adaptive shrinkage logic to proportionally compress all diagonal weights of the positive definite weighted matrix and synchronously updates the matrix register, and then re-enters step B2 for re-verification. This loop continuously runs at a nanosecond-level rhythm on the hardware pipeline until all the descent rates returned by the gradient iterator are negative, indicating that the linear quadratic Lyapunov function has achieved a strict descent within the current energy surface. At this point, the positive definite weighted matrix is locked. The final version in the matrix register will serve as the core weight for subsequent control law solution and is simultaneously written into the log buffer for long-term tracking by the adaptive gain modulator, providing a traceable stability proof for the intelligent control system of integrated water and fertilizer multi-objective optimization.

[0031] When the system enters step C1, the controller first calls the energy curve of the linear quadratic Lyapunov function locked by the previous round of gradient iterator, searches for its energy peak along the time axis and marks it as the maximum acceptable energy level; this level value is immediately written into the register mapping table, written into the negative definite convergence domain register through a hardwired path, and upper and lower edge double threshold marks are generated inside the register to ensure that any subsequent state evolution can be immediately captured by the hardware clipping logic once it crosses the boundary. Then the system enters step C2, and the boundary parameters in the negative definite convergence domain register are copied to the local register groups of the safety interlock logic and the water application AI dynamic optimization engine through the high-speed bus. After obtaining the boundary parameters, the safety interlock logic automatically refreshes the internal comparator threshold to perform real-time out-of-bounds detection on each upcoming frame of energy data. Once the detection result is non-zero, it triggers an emergency irrigation stop or a decreasing sequence to ensure that the field irrigation hardware does not exceed the crop tolerance range under extreme disturbances. After the water application AI dynamic optimization engine synchronizes to the same boundary parameters, it writes them into the energy constraint register of the control law solver, so that the linear matrix inequality solving process has the latest boundary reference during each step size estimation, thus avoiding the convergence domain diffusion caused by environmental mutations. The synchronization operation is completed through a hardware handshake protocol, and any clock drift during this period will be forced to be aligned by the timing corrector to ensure that the negative definite convergence domain boundary remains absolutely consistent between the sensing layer, the decision-making layer and the execution layer, so that the intelligent control system for multi-objective optimization of integrated water and fertilizer management can always lock the same stable attraction domain during long-term operation, realizing precise, asymptotic and verifiable dynamic control of irrigation resources.

[0032] The fertilization rate AI dynamic optimization engine loads the residual sequence of the most recent complete growth cycle in batches from the historical residual accumulation part of the fertilization rate AI through a high-speed bus in the first clock segment of startup. After loading, it is immediately written into the residual buffer stack. The residual buffer stack adopts a single-chip ring storage structure to ensure the physical continuity of data on the time axis. Therefore, all residual values can be sequentially read by the downstream logic without page jitter. In the same cycle, the system calls the sliding mode initialization unit to calculate the difference between the top value of the residual buffer stack and the center value of the crop nutrient safety belt, and calibrates the initial value of the sliding mode surface with this difference. At the same time, the zero-offset baseline is recorded in the register table as the symmetric reference surface for subsequent deviation integration. In the subsequent perception stage, the multi-threaded DMA simultaneously pulls the soil state and crop root absorption capacity data streams, and reads the fertilization records of the previous time period at the field level. The hardware-scale alignment pipeline compresses all indicators into a unified dimensionless interval, and then concatenates them into a state vector in a fixed field order in the state observation buffer. The fertilization rate AI dynamic optimization model then starts the sliding mode control solution. It uses the difference between the sliding mode surface and the state vector as the driving quantity to enter the Dahn switching law generator. Inside the generator, parallel look-up table operations are triggered for the high-dimensional symbolic matrix. After the sign bits are matched with the preset gains dimension by dimension, the original fertilization control quantity is output. This output maintains an integer ratio architecture in all logical paths, facilitating direct parsing by the subsequent actuator. The adaptive gain modulator monitors the convergence speed of the sliding mode surface distance over time, dynamically compares the actual approach slope with the target approximation speed. If the actual speed is slow, the switching rate is increased in the same cycle; if it is fast, it is decreased slowly to avoid overshoot of the ion concentration accumulated in the rhizosphere. The historical residual accumulation part of the fertilization rate AI instantaneously accumulates the nutrient deviation at the end of the current feedback cycle, and then maps it to a compensation vector of the same dimension as the original fertilization control quantity through a variable gain compensator. This compensation vector is added to the original fertilization control quantity dimension by dimension in the hardware parallel adder, and then the overall vector is clipped by the upper and lower threshold values of the saturation protector. Any component that exceeds the crop salt tolerance threshold or the device maximum flow threshold will be immediately truncated. The clipped fertilization rate vector is encoded into an instruction frame compatible with the field bus and written into the execution instruction queue. The queue scheduler sorts according to the time stamp to ensure that the pulse packets arrive at the variable flow fertilization device in order. The entire process is centered around on-chip parallel logic, ensuring that residual drive, sliding mode solution, gain modulation, compensation superposition, and safety saturation are completed within one control cycle, enabling the fertilization rate AI dynamic optimization engine to maintain smooth transition and asymptotic stability of the rhizosphere nutrient concentration even in the event of sudden temperature drops or sharp changes in soil microbial activity.

[0033] Furthermore, in the AI dynamic optimization model for fertilization amount, the system first allocates an independent register cluster for the sliding mode surface register in the on-chip programmable logic. The bit width of this register cluster is strictly aligned with the dimension of the state vector to ensure that the crop nutrient deviation can be mapped into binary fixed-point numbers without loss. Subsequently, the model calls the integral accumulation unit integrated in the same logic area, synchronously sending the crop nutrient deviation obtained from real-time sampling and the deviation integral within the historical window into the accumulation unit. The continuous-time sum data output by the integral accumulation unit reaches the sliding mode surface register through a hardwired path, and the register automatically completes vector splicing to generate a sliding mode surface vector with a stable topological structure. This vector is written into the shared cache by the DMA engine in the next clock cycle for subsequent operation stages to call. The second parallel data channel of the model continuously reads the state vector at the current moment from the state observation buffer and calculates the state vector difference with the fertilization record in the previous period through a dedicated subtractor. This difference is pushed into the difference comparator. The difference comparator adopts parallel threshold determination logic, encoding the positive and negative polarities and amplitude intervals of each dimension into multi-dimensional deviation marks and writing these marks into the deviation latch in the same clock cycle. The deviation latch holds each dimension mark at a fixed data line position relying on the hardware grid, ensuring that the Darboux switching law generator can read it with zero delay in subsequent cycles. After receiving the multi-dimensional deviation marks from the deviation latch, the Darboux switching law generator looks up the corresponding switching coefficients in its internal look-up table unit, converts the positive and negative polarities into elements of the symbol matrix, and uses pipeline logic to complete all matrix assembly within one clock cycle, finally outputting the symbol matrix as the symbol carrier of the control law driving quantity. Through this implementation method of full-hardware parallelism and pipeline splicing, the system does not require any software interrupt or cache write-back during the whole process of generating the sliding mode surface vector, capturing the state vector difference, and constructing the symbol matrix, greatly reducing the round-trip bus delay, enabling the AI dynamic optimization model for fertilization amount to complete a full sliding mode control iteration in milliseconds, thus ensuring that the crop rhizosphere nutrient concentration can still maintain a smooth transition during rapid climate fluctuations or sudden soil ion adsorption changes.

[0034] The Dashi switching law generator pre-sets a switching gain table that exactly corresponds to the dimension in the on-chip read-only memory. After the symbol matrix enters the generator, each symbol bit is paired with the gain coefficient in the same column one by one through the row-first control logic, and the multiplication and accumulation are completed in a single-cycle pipeline, and an initial switching pulse sequence of one frame is output; this sequence is written into the register array in the form of a wide parallel bus to ensure that the pulse width and duty cycle are precisely aligned with the clock beats. The adaptive gain modulator continuously monitors the deviation amplitude and reads the real-time threshold from the fertilization amount AI historical residual accumulation part. After obtaining the gain modulation factor through the differential amplifier, it calls the incremental-decrement micro-step algorithm to update the corresponding entries in the switching gain table according to the dimension, and generates the gain modulation matrix; the gain modulation matrix and the initial switching pulse sequence complete the vector multiplication in the on-chip multi-way parallel multiplication unit to obtain the open-loop control sequence. The open-loop control sequence first enters the rate limiter, and the rate limiter compresses the front edge of any pulse according to the rising slope limit window to prevent mechanical shock caused by the valve control actuator being driven too fast; then the sequence flows into the boundary layer filter, and the filter uses the moving average kernel to weaken the spike energy in the switching frequency band, and the output pulse is compressed at both the amplitude and frequency ends, thereby suppressing high-frequency chattering. The shaped pulse is then written into the execution register, waiting for the scheduler to allocate it to the field-end variable flow fertilization device, ensuring that the switching law not only maintains the fast approaching characteristic of the sliding mode surface, but also does not introduce unnecessary oscillations due to the discretization effect.

[0035] The water and fertilizer feedback control unit is located in the core area of the closed-loop between the execution layer and the decision-making layer, taking over the on-site measurement values returned by the field sensing network after each irrigation and fertilization pulse. The system converts the real-time outputs of tensiometers, water potential probes, and leaf reflectance spectrometers into crop water content state indices through multiple synchronous acquisition channels, and at the same time converts the synchronous outputs of ion-selective electrodes and fluorometric colorimetric probes into crop nutrient sufficiency indices; after zero-holding processing and noise clipping at the hardware level for the two sets of indices, the difference is taken in the same arithmetic logic unit with the water application amount and fertilizer application amount executed in the previous control cycle to obtain the water content error vector and the nutrient error vector. The error vectors first enter a dual-window exponential filter to eliminate instantaneous spikes, and then are respectively injected into a water integration modulator and a fertilizer integration modulator. Inside the modulator, the short-period proportional correction and the long-period integral correction are superimposed into a single correction signal. The system then converts the correction signal into a dynamically corrected water application amount and a dynamically corrected fertilizer application amount through variable gain mapping; during the conversion process, the limit saturation logic performs two-way clipping on each component according to the crop safety threshold, soil mechanism threshold, and equipment physical threshold to avoid overshoot or under-supply. The clipped dynamically corrected water application amount and dynamically corrected fertilizer application amount are written into the execution instruction buffer, and the queue scheduler assigns the next available time slot to them, ensuring that the dynamically corrected water application amount is triggered first in the valve control sequence, and the dynamically corrected fertilizer application amount follows with the minimum delay after the water pulse ends, fully realizing the closed-loop feedback regulation of crop water application and the closed-loop feedback regulation of crop fertilization. After each round of regulation, the water and fertilizer feedback control unit writes the new error accumulation value back to the residual accumulation register, providing the latest marginal data for the historical residual accumulation part of the water application amount AI and the historical residual accumulation part of the fertilizer application amount AI. At the same time, it encapsulates the current gain configuration, threshold clipping record, and exponential filter state into a log frame and synchronizes it upward to the cloud model repository for long-term iterative update of the control law template in the decision-making layer. This entire process is completed within a millisecond-level hardware path, enabling the water and nutrient supply to remain dynamically corrected water application amount and dynamically corrected fertilizer application amount to be always closed in the face of dual disturbances of rapid climate fluctuations and soil buffer changes, thereby ensuring that the crop water content state index and the crop nutrient sufficiency index fall back to the target range along an asymptotically stable trajectory throughout the growth cycle.

[0036] The execution process of the water and fertilizer feedback control unit includes: reading the dynamically corrected water application rate and dynamically corrected fertilizer application rate stored in the previous operation cycle, writing them into the water feedback register and the fertilizer feedback register, establishing a dual-channel error latch and clearing it; collecting the measured crop water content status index and the measured crop nutrient sufficiency index in real time, suppressing high-frequency noise through a dual-window index filter and then writing them into the status observation buffer; calling the current water application rate and fertilizer application rate, forming a water content error vector and a nutrient error vector in the error latch respectively, and recording the change rate mark; performing dual-loop proportional-integral modulation, including: the water content error vector enters the water integral modulator to generate a water integral quantity; the nutrient error vector enters the fertilizer integral modulator to generate a fertilizer integral quantity; the two channels simultaneously perform proportional-integral superposition and output a first correction quantity; the first correction quantity is limited within the crop safety threshold through a limiter saturator to generate a second correction quantity and write it into the dynamic correction register; the second correction quantity in the dynamic correction register is added to the original water application rate and the original fertilizer application rate one by one to obtain the dynamically corrected water application rate and dynamically corrected fertilizer application rate of the crop per unit area at the current time, and at the same time push them to the execution instruction queue; after irrigation and fertilization are completed, re-collect the measured crop water content status index and the measured crop nutrient sufficiency index; if the absolute value of any error vector does not decrease for two consecutive cycles, increase the proportional-integral gain and return to the dual-loop proportional-integral modulation, otherwise maintain the current gain and enter the next regular cycle.

[0037] Further, the meteorological data includes: crop evapotranspiration , in units of millimeters per day; rainfall , in units of millimeters per day; the proportion of rainfall that is not absorbed by the soil and directly runs off, represented by the surface runoff rate ; the growth status data of the crop includes: soil water potential , in units of kPa; crop root suction , in units of kPa; the nitrogen demand of the crop per unit area , in units of g / m² / day; the phosphorus demand of the crop per unit area , in units of g / m² / day; the potassium demand of the crop per unit area in units of g / m² / day; soil adsorption ratio ; leaf chlorophyll concentration ; is time.

[0038] Further, the water demand of the crop per unit area evaluated at each time is: ; This formula comprehensively considers meteorological driving factors, soil water dynamic behavior, and crop root water absorption capacity, and can be divided into two parts in structure. The first part: Used to describe the actual loss of soil moisture per unit time, where represents the evapotranspiration of crops, which is the water lost due to evaporation and transpiration per unit area of crops in one day, with the unit of mm / day; represents the rainfall on the current day, also in the unit of mm / day. Subtracting the two gives the water gap faced by the crops. Since not all rainfall is absorbed by the soil and there is also surface runoff loss, the system uses to represent this ratio, thereby correcting the net input rainwater. The second part: reflects the coupling relationship between the dynamic process of soil water potential changing with time and the water absorption capacity of crop roots, where represents the rate of change of soil water potential per unit time, with the unit of kPa / day, represents the suction capacity of crop roots, also in the unit of kPa. If the soil water potential drops rapidly, it indicates intense water evaporation or migration and an increase in water demand intensity; while the stronger the root suction, the higher the ability of the crops to extract water from the soil. The two act jointly on the adjustment factor of water demand.

[0039] Among them, is the water demand of crops per unit area at time; the amount of fertilizer required by the crops evaluated at each time is: ; Among them, respectively represent the demand intensities of nitrogen, phosphorus, and potassium per unit area of crops at time , with the unit of . The sum of the three is the total theoretical nutrient demand; represents the soil adsorption ratio, which is used to describe the influence of the adsorption of ions in fertilization by the soil on the available nutrient content. If the soil adsorption capacity is high, more nutrients need to be applied to meet the actual absorption of crops; is the deviation correction factor between the actual chlorophyll concentration of crops and the theoretical concentration, is the measured value, is the theoretical value. The chlorophyll concentration of crops reflects their photosynthetic efficiency and nutrient status. When the measured value is lower than the theoretical value, it indicates that the crops currently have a nutrient deficiency and the amount of fertilizer required should be increased accordingly.

[0040] Furthermore, the cumulative part of the historical residual of the irrigation amount AI is: ; The cumulative part of the historical residual of the fertilization amount AI is: ; Among them, is Crop water requirement per unit area per unit time; is Crop fertilizer requirement per unit area per unit time; is Crop irrigation amount per unit area per unit time; is Crop fertilization amount per unit area per unit time; is Crop water content status index at a certain time; is Crop water content status index at a certain time; is Crop nutrient sufficiency index at a certain time; is Crop nutrient sufficiency index at a certain time.

[0041] In the AI-driven integrated water and fertilizer multi-objective optimization intelligent control system, the historical residual accumulation mechanism, as the core part of realizing intelligent closed-loop feedback control, forms an error feedback term by comparing the difference between the system predicted output and the actual crop response in each control cycle, which is used to dynamically adjust the irrigation amount and fertilization amount in the next cycle. This structure not only strengthens the system's self-learning and adaptation ability, but also constructs a time-series related strategy correction framework. Specifically, the system sets up historical residual accumulation models for the irrigation and fertilization processes respectively, which are used to record the control deviation in real time and quantify it as the feedback correction amount. The historical residual accumulation part of the irrigation amount is defined as follows: ; This formula has a clear structure, consisting of the product of three terms, each of which reflects a different feedback mechanism. The first term represents the difference between the theoretical water requirement in the previous cycle and the actual irrigation amount, reflecting the static deviation between the model prediction and the actual supply; if this value is positive, it means that the actual irrigation is insufficient, otherwise it is over-irrigation. The second term represents the relative change of the crop water content status index between two consecutive moments, measuring the dynamic impact of the irrigation behavior on the crop water content status; if the water content index decreases, this term is positive, indicating that the crop water condition deteriorates and the irrigation amount needs to be increased in the next round. The third term is the regulation ratio factor, which is used to standardize the weight of the regulation behavior and prevent abnormal extreme values from having too much interference on the model. Overall, this structure realizes the quantitative accumulation of historical errors, while maintaining numerical stability, dynamically responds to the changes of crop status, and thus realizes the automatic optimization of the irrigation strategy.

[0042] In terms of fertilization control, the system also sets up a corresponding historical residual accumulation module, and its structure is as follows: ; The structure of this formula is highly similar to the water application part, and its principle and mechanism of action are basically the same, but the variables of concern are transformed into crop nutrient responses. The first term represents the difference between the total amount of fertilizer required calculated by the system and the actual amount of fertilizer applied; the second term represents the change ratio of the crop nutrient sufficiency index, reflecting whether the fertilization behavior has improved the crop nutrient absorption status; the third term is used to adjust the overall influence weight to keep the residual amplitude within a reasonable range.

[0043] Furthermore, the dynamically corrected water application volume is: ; where is the photosynthetically active radiation intensity, with the unit of μmol / m² / s.

[0044] This formula is the third-layer correction mechanism in the entire intelligent water application strategy. Its principle is not based on the offline correction of historical residuals, nor does it solely rely on the previous water requirement estimation or soil condition assessment. Instead, it finely adjusts the water application volume in real-time based on the dynamic feedback of the current physiological state of the crop, forming a terminal closed-loop for regulation, enabling the actual response of the crop itself to directly affect the water input, thereby enhancing the system's adaptability to non-linear disturbances and complex growth states. First of all, in terms of structure, this formula is composed of the multiplication of three factors. The first term represents the degree of deficiency of the current water content state index of the crop. The water content state index is a standardized dimensionless variable, and its numerical range is usually between 0 and 1, reflecting the state of crop cell water saturation or leaf water tension. When , it means that the crop's water is completely sufficient and no water replenishment adjustment is required; when this value is much less than 1, it indicates that the crop is in a water stress state, and the value of this factor will increase, and the dynamic correction intensity will also increase. Therefore, this factor reflects the positive correlation between the current water shortage degree of the crop and the regulation intensity.

[0045] The second term is a water application - photosynthesis ratio adjustment factor. represents the water application volume per unit area in the current period, which is the regulation base value jointly calculated by the AI dynamic optimization engine and the historical residual accumulation part; and is the photosynthetically active radiation intensity, with the unit of , which measures the ability of crops to receive light, is one of the direct driving factors for crop transpiration and photosynthesis. This ratio actually establishes the relationship between the water application behavior and the light energy driving intensity: if the light is strong, the crop transpiration is vigorous and the water demand increases accordingly; if the light is weak, the water application should be reduced accordingly to avoid water excess leading to root hypoxia or diseases. Therefore, this factor essentially provides a physiological-environmental co-regulation mechanism to perform a standardized correction of the water application efficiency under the energy state.

[0046] The third item is the first-order derivative of the water content state index with respect to time, indicating the change trend of the crop water content state. When this derivative is positive, it means that the crop water content condition is improving. At this time, even if the current water is insufficient, the system can suspend the increase of water application; if this derivative is negative, it means that the water content state is deteriorating and rapid water replenishment is needed to prevent the expansion of physiological stress. Therefore, this factor represents the "rate adjustment channel" in the regulation, that is, not only considering the state value itself, but also considering its evolution speed, reflecting the system's response ability to the dynamic trend.

[0047] Multiplying the three items together forms the which is the water application compensation amount for dynamic emergency at this moment. In actual control, this compensation amount does not directly replace the main regulation value, but is superimposed on the original water application amount, that is, the system execution amount is: ; where is the output of the AI engine and the residual module, and is the correction increment driven by the crop state. The two jointly determine the actual water application amount.

[0048] To illustrate the actual application logic of this mechanism, consider the following example: In a certain regulation period <http: / / www.example.com / , the crop monitoring system measures the crop water content state index <http: / / www.example.com / , which is lower than the target value of 0.70, indicating that the current crop water is lacking; the change trend of the water content index with time is <http: / / www.example.com / , indicating that its state is deteriorating slowly; the water application amount calculated by the current AI engine is <http: / / www.example.com / mm; the photosynthetically active radiation is <http: / / www.example.com / . Substitute into the formula for calculation: <http: / / www.example.com / <http: / / www.example.com / ; <http: / / www.example.com / Since the derivative is negative, it means that the state is deteriorating, but the result of this item is negative, indicating that the decrease in the water content state is not serious under strong light, so there is no need to significantly increase the water input. On the contrary, if the derivative drops significantly or the light is extremely strong, this item may become positive and have a significant supplement.

[0049] The greatest advantage of this feedback compensation mechanism lies in introducing the dynamic physiological response signal of the crop itself as the final control factor for watering behavior, realizing the closed-loop evolution from "environment - prediction - model" to "crop - response - regulation", which is a mathematical implementation form of the typical crop response-based intervention logic in intelligent agriculture. This formula has extremely high practicality in the regulation execution layer and can effectively reduce the problem of resource misallocation caused by non-model errors such as light fluctuations and water diffusion hysteresis. It is an indispensable dynamic correction mechanism in the precise regulation structure.

[0050] Furthermore, the dynamically corrected fertilization amount is: ; where is the plant nutrient conversion rate, with the unit of g / m² / day.

[0051] First of all, from a structural perspective, this formula consists of three factors. The first term is the deviation degree of the current nutrient sufficiency index of the crop, which is a static judgment of the crop's nutrient status. The nutrient sufficiency index is derived from integrated indicators such as the nitrogen concentration in crop tissues, chlorophyll concentration, stem color, and chlorophyll index, and ranges from 0 to 1, which is a standardized variable. When , it indicates that the crop has sufficient nutrients and no additional fertilization is required; when the index is low, this factor approaches 1, indicating nutrient deficiency and the need to increase the fertilization amount. This term directly constitutes the intensity benchmark of the dynamic correction amount.

[0052] The second term is the ratio factor of fertilization to nutrient conversion efficiency. represents the fertilization amount per unit area in the current period, which is a value obtained by superimposing historical residuals on the basis of the AI dynamic optimization engine and has considered the influence of the deviation between the theoretical demand and the previous cycle; while represents the plant nutrient conversion rate, with the unit of , which is the ability of the nutrients applied per unit area of the plant to be assimilated or converted into organic substances (such as proteins, enzymes, chlorophyll, etc.). It is a dynamic variable jointly determined by the crop growth model, metabolic rate, and current climate conditions, and is usually comprehensively estimated through photosynthesis rate, nitrogen use efficiency, nutrient redistribution efficiency, etc. The larger this ratio, the lower the efficiency of the effective conversion of the unit fertilization amount by the plant, and the system will increase the supplementation intensity in the correction; if the efficiency is high, this term will automatically decrease, so that the fertilization will not be excessive.

[0053] The third term is the first derivative of the nutrient sufficiency index with respect to time, representing the trend of the crop's nutrient status change. When it is positive, it indicates that the nutrient status is improving and immediate fertilization may not be required; when it is negative, it indicates that the nutrient status is deteriorating and the system needs to intervene quickly. Combining the first two terms, this derivative term constitutes the sensitive gating factor of the dynamic correction formula, enabling the system to perceive the "trend status" of the crop's nutrients. This mechanism allows the system to not only judge whether the current nutrients of the crop are sufficient, but further determine whether the nutrient status is on the "recovery" or "deterioration" path, thus deciding whether intervention is needed.

[0054] Integrating these three terms, what the system obtains is the current dynamic emergency fertilization compensation amount, and finally forms the corrected output of the total fertilization amount with the previous fertilization plan: ; where is the combined result of AI optimization and historical correction, which is the static regulation output, while is the dynamic intervention result driven by the crop's physiology. This mechanism significantly improves the system's reaction speed and physiological matching accuracy, and is the key technical path for the fertilization process to shift from "environment-centered" to "crop-centered".

[0055] Figure 2It is a schematic diagram of the sliding mode control effect of the fertilization amount AI dynamic optimization engine. This diagram details the application effect of the reaching law in fertilization amount control, especially highlighting the important contribution of the adaptive gain modulator to control performance. The abscissa represents the 30-day control period, and the ordinate represents the deviation of crop nutrient sufficiency. Negative values indicate nutrient deficiency, positive values indicate nutrient excess, and a zero value represents the ideal nutrient balance state. The dotted line in the figure represents the position of the sliding mode surface, set at zero deviation, which is the ideal control target of system design. The long dashed line represents the control effect of the reaching law without an adaptive gain modulator. This curve shows that starting from an initial deviation of -0.4, although the system can gradually approach the sliding mode surface, the convergence speed is relatively slow, and there is an obvious slowdown when approaching the sliding mode surface. Eventually, it still does not reach the zero deviation state at the 30th day. The solid line represents the control curve of the reaching law equipped with an adaptive gain modulator, showing significantly improved control performance. This curve not only has a faster initial convergence speed, but more importantly, it basically reaches near the sliding mode surface around the 15th day and remains stable in the subsequent time. The rectangular area marked by the thin dotted line in the figure represents the action range of the boundary layer filter, within the ±0.05 deviation interval near the sliding mode surface. Within this interval, the boundary layer filter effectively suppresses the high-frequency chattering phenomenon, making the control signal smoother and avoiding the common chattering problem in traditional sliding mode control. By comparing the two curves, it can be clearly seen that the adaptive gain modulator dynamically adjusts the switching gain according to the deviation amplitude and the real-time threshold of the cumulative part of the fertilization amount AI historical residuals, significantly improving the system's response speed and control accuracy, verifying the technical advantages of the method of converting the variable gain compensator into an equal-dimensional compensation amount.

[0056] Figure 3It is a schematic diagram of the collaborative effect between the water and fertilizer load analysis unit and the AI dynamic resource optimization engine. This diagram comprehensively compares the resource utilization efficiency performance of different control strategies during the complete growth period. The horizontal axis represents the 180-day complete growth period, and the vertical axis represents the resource utilization efficiency index. The higher the value, the better the utilization effect of water and fertilizer resources. The long dashed line in the figure represents the effect curve of the traditional water and fertilizer management method. This curve shows a relatively gentle upward trend throughout the growth period, gradually increasing from the initial 0.2 to 0.26. The overall improvement is limited, and the rising rate remains basically constant, lacking the ability to adaptively adjust to the changing demands of different growth stages. The medium dashed line represents the control effect of only using the water and fertilizer load analysis unit, showing significantly better performance than the traditional method. The resource utilization efficiency increases from 0.3 to 0.185, reflecting the technical advantages of load analysis based on meteorological data, soil sensor data, and leaf surface sensor data. The thick solid line represents the final effect of the collaborative action between the water and fertilizer load analysis unit and the AI dynamic resource optimization engine, demonstrating the most excellent control performance. This curve not only has the fastest initial rising speed, but more importantly, it reaches an efficient level above 1.0 after the 90th day and finally stabilizes around 1.4, achieving nearly a 5-fold performance improvement compared to the traditional method. The solid dots in the figure mark the time nodes where the historical residual accumulation part plays a key role, which are located at the 60th day, 90th day, 120th day, and 150th day respectively. These nodes correspond to the key transition periods of different growth stages of the crop. The rectangular area marked by the thin dotted line frame represents the main action range of the variable gain compensator, which is between the 60th and 120th days. Within this range, the compensator effectively improves the adaptive adjustment ability of the system by accumulating the deviation between the predicted state and the measured state. The legend clearly distinguishes the three different control strategies and the key points of historical residual accumulation, verifying the significant technical advantages of the described AI dynamic resource optimization engine in water and fertilizer collaborative control.

[0057] Figure 4It is a schematic diagram for comparing the control effects of the AI dynamic optimization engine for water application amount. This diagram clearly shows the significant advantages of the AI dynamic optimization engine for water application amount in the present invention compared with the traditional control method in terms of controlling the error of crop water content state. The abscissa represents the time axis, in days, covering a complete 60-day observation period; the ordinate represents the numerical change of the crop water content state error, reflecting the deviation degree between the actual water content state and the ideal water content state. The dotted line in the figure represents the control curve of the traditional control method. It can be observed that this curve shows obvious fluctuation characteristics throughout the time period, and the error value always remains between 0.26 and 0.29, and there are repeated oscillations, indicating that it is difficult for the traditional method to achieve precise control of the water content state and the system stability is poor. In contrast, the control curve of the AI dynamic optimization engine for water application amount represented by the solid line shows completely different characteristics. Starting from the initial error value of 0.30, this curve shows an obvious downward trend and basically stabilizes in the low error range below 0.08 after the 30th day. The dash-dotted line in the figure marks the boundary of the negative definite convergence domain set based on the linear quadratic Lyapunov function, located at the error value of 0.12. It can be clearly seen from the figure that the control curve of the AI dynamic optimization engine for water application amount successfully enters this convergence domain around the 25th day and stably remains within the domain in the subsequent time, fully verifying the effectiveness of the method for setting the negative definite convergence domain of the Lyapunov function. This phenomenon indicates that by invoking the residual sequence of the most recent complete growth cycle in the historical residual accumulation part of the AI for water application amount, combined with the synergistic effect of the double-weight exponential decay filter and the gradient iterator, the system can achieve asymptotically stable control and effectively suppress the common control oscillation problem in the traditional method.

[0058] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent control system for multi-objective optimization of integrated water and fertilizer driven by artificial intelligence, characterized in that, The system includes: a water and fertilizer load analysis unit and a water and fertilizer combined regulation criterion generation unit; the water and fertilizer load analysis unit is used to obtain meteorological data from a meteorological monitoring system; at the same time, it receives data from soil sensors and leaf surface sensors to obtain the growth status data of crops, and evaluates the water demand and fertilizer demand of crops per unit area at each time; the water and fertilizer combined regulation criterion generation unit is used to use an AI dynamic resource optimization engine to dynamically calculate the water application amount and fertilizer application amount of crops per unit area at each time according to the soil status and the root absorption capacity of crops; the AI dynamic resource optimization engine includes: a water application amount AI dynamic optimization engine and a fertilizer application amount AI dynamic optimization engine; the water application amount AI dynamic optimization engine is obtained by adding a water application amount AI dynamic optimization model and a water application amount AI historical residual accumulation part; the fertilizer application amount AI dynamic optimization engine is obtained by adding a fertilizer application amount AI dynamic optimization model and a fertilizer application amount AI historical residual accumulation part.

2. The intelligent control system for integrated water and fertilizer multi-objective optimization driven by artificial intelligence according to claim 1, characterized in that The execution process of the water application amount AI dynamic optimization engine includes: calling the residual sequence of the most recent complete growth cycle in the water application amount AI historical residual accumulation part, setting a negative definite convergence domain based on the linear quadratic Lyapunov function, and writing it into the residual accumulation register and the safety threshold register; real-time collecting the soil status and the root absorption capacity of crops, writing them together with the irrigation record of the previous time period into the state observation buffer, and forming a state vector after unified scaling; the water application amount AI dynamic optimization model calls a linear matrix inequality solver to output the original water application control amount that satisfies the Lyapunov descent constraint; at the same time, the adaptive gain modulator dynamically updates the matrix weight according to the convergence rate; the water application amount AI historical residual accumulation part incorporates the latest error into the residual accumulation register, generates a compensation amount of the same dimension through a variable gain compensator, and buffers rapid disturbances; the original water application control amount and the compensation amount are added dimension by dimension, and are limited within the execution threshold range by a saturation protector to obtain the target water application amount, which is written into the execution instruction queue.

3. The intelligent control system for integrated water and fertilizer multi-objective optimization driven by artificial intelligence according to claim 2, characterized in that, The process of calling the residual sequence of the most recent complete growth cycle in the water application amount AI historical residual accumulation part and setting a negative definite convergence domain based on the linear quadratic Lyapunov function includes: Step A1: Perform integrity verification on the residual sequence of the water application amount AI historical residual accumulation part, linearly interpolate the missing segments in chronological order to obtain the verified sequence; Step A2: Write the verified sequence into the residual buffer stack, and use a double-weight exponential decay filter to suppress high-frequency noise to obtain a filtered sequence; Step A3: Detect the range of the filtered sequence. If it exceeds the threshold of the safety threshold register, trigger the abnormal backtracking logic and return to Step A1; Step B1: Dynamically generate a positive definite weighted matrix according to the variance distribution of the filtered sequence and write it into the matrix register; Step B2: Construct a linear quadratic Lyapunov function with the positive definite weighted matrix and the residual state vector in quadratic form, and verify the strict descent for all samples through a gradient iterator; Step B3: If the descent rate of any sample is non-negative, adaptively shrink the positive definite weighted matrix and return to Step B2 until all samples meet the conditions; Step C1: Define the boundary of the negative definite convergence domain with the maximum acceptable energy level of the linear quadratic Lyapunov function and write it into the negative definite convergence domain register; Step C2: Synchronize the boundary parameters to the safety interlock logic and the water application rate AI dynamic optimization engine to complete the setting of the negative definite convergence domain.

4. The intelligent control system for integrated water and fertilizer multi-objective optimization driven by artificial intelligence according to claim 3, characterized in that, The execution process of the fertilizer application rate AI dynamic optimization engine includes: retrieving the residual sequence of the most recent complete growth cycle from the fertilizer application rate AI historical residual accumulation part, writing it into the residual buffer stack, establishing the initial value of the sliding mode surface and recording the zero-offset baseline; collecting the soil state and the crop root absorption capacity, and reading the fertilizer application record of the previous time period, standardizing them uniformly and writing them into the state observation buffer area to form a state vector; the fertilizer application rate AI dynamic optimization model generates the original fertilizer application control quantity based on the difference between the sliding mode surface and the state vector, using the reaching law switching rule; within the same cycle, the adaptive gain modulator updates the switching rate according to the approaching speed; the fertilizer application rate AI historical residual accumulation part superimposes the nutrient deviation of this cycle, and is converted into an equal-dimensional compensation quantity through a variable gain compensator to suppress high-frequency disturbances; the original fertilizer application control quantity and the compensation quantity are added dimension by dimension, and are limited within the safety threshold by a saturation protector to obtain the fertilizer application rate, which is written into the execution instruction queue.

5. The intelligent control system for integrated water and fertilizer multi-objective optimization based on artificial intelligence drive according to claim 4, wherein, The process of the fertilizer application rate AI dynamic optimization model generating the original fertilizer application control quantity based on the difference between the sliding mode surface and the state vector includes: establishing a sliding mode surface register in the fertilizer application rate AI dynamic optimization model, mapping the crop nutrient deviation and its time sum into a sliding mode surface vector; reading the state vector difference in real time, forming a multi-dimensional deviation flag through a difference comparator and writing it into the deviation latch; the output of the deviation latch is sent to the reaching law switching rule generator to generate a sign matrix; the reaching law switching rule generator matches the sign matrix dimension by dimension according to the pre-stored switching gain table to generate an initial switching pulse sequence; the adaptive gain modulator increases or decreases the switching gain according to the deviation amplitude and the real-time threshold of the fertilizer application rate AI historical residual accumulation part, and outputs a gain modulation matrix; the initial switching pulse sequence is multiplied by the gain modulation matrix to obtain an open-loop control sequence, which is shaped by a rate limiter and a boundary layer filter in a two-stage manner to suppress high-frequency chattering; 6. The intelligent control system for integrated water and fertilizer multi-objective optimization driven by artificial intelligence according to claim 5, characterized in that, The system further includes: a water and fertilizer feedback control unit, which is used to measure the crop water content state index, combine it with the water application rate, calculate the dynamic corrected water application rate of the crop per unit area at each time, perform closed-loop feedback regulation of crop water application, and measure the crop nutrient sufficiency index, combine it with the fertilizer application rate, calculate the dynamic corrected fertilizer application rate of the crop per unit area at each time, and perform closed-loop feedback regulation of crop fertilizer application.

7. The intelligent control system for integrated water and fertilizer multi-objective optimization driven by artificial intelligence according to claim 6, wherein The execution process of the water and fertilizer feedback control unit includes: reading the dynamically corrected water application rate and dynamically corrected fertilizer application rate stored in the previous operation cycle, writing them into the water feedback register and the nutrient feedback register, establishing a dual-channel error latch and clearing it; collecting the measured crop water content status index and the measured crop nutrient sufficiency index in real time, and writing them into the status observation buffer after suppressing high-frequency noise through a dual-window index filter; calling the current water application rate and fertilizer application rate, forming a water content error vector and a nutrient error vector in the error latch respectively, and recording the change rate mark; performing dual-loop proportional-integral modulation, including: the water content error vector enters the water integral modulator to generate a water integral quantity; the nutrient error vector enters the nutrient integral modulator to generate a nutrient integral quantity; the two channels simultaneously execute proportional-integral superposition and output a first correction quantity; the first correction quantity is limited within the crop safety threshold through a limiter saturator to generate a second correction quantity and write it into the dynamic correction register; the second correction quantity in the dynamic correction register is added to the original water application rate and the original fertilizer application rate one by one to obtain the dynamically corrected water application rate and dynamically corrected fertilizer application rate of the crop per unit area at the current time, and at the same time push them to the execution instruction queue; after irrigation and fertilization are completed, re-collect the measured crop water content status index and the measured crop nutrient sufficiency index; if the absolute value of any error vector does not decrease for two consecutive cycles, increase the proportional-integral gain and return to the dual-loop proportional-integral modulation, otherwise maintain the current gain and enter the next regular cycle.

8. The intelligent control system for integrated water and fertilizer multi-objective optimization driven by artificial intelligence according to claim 7, wherein Meteorological data includes: crop evapotranspiration , in millimeters per day; rainfall , in millimeters per day; the proportion of rainfall that runs off directly without being absorbed by the soil, represented by the surface runoff rate ; Crop growth status data includes: soil water potential , in kPa; crop root suction , in kPa; nitrogen demand of crops per unit area , in g / m² / day; phosphorus demand of crops per unit area , in g / m² / day; potassium demand of crops per unit area , in g / m² / day; soil adsorption ratio ; leaf chlorophyll concentration ; is time.

9. The intelligent control system for integrated water and fertilizer multi-objective optimization driven by artificial intelligence according to claim 8, characterized in that, The water requirement of the crop per unit area evaluated at each time is: ； Among them, is the water requirement of crops per unit area of time; the fertilizer requirement of crops evaluated at each time is: ； Among them, is the fertilizer requirement of crops per unit area per unit time; is the theoretical chlorophyll concentration of crops at the current growth stage at unit time.

Images