Rice soil nutrient synergistic regulation method and system based on big data analysis

Abstract

The present application relates to the field of soil nutrient management, in particular to a rice soil nutrient synergistic regulation method and system based on big data analysis, first, data is collected through a multi-source sensing array, and a characteristic state space with time sequence memory is constructed. By introducing physical field theory, the buffer effect of soil on nutrients is modeled as a chemical potential energy field using the variable Hook's law to quantify system instability and solve the problem of excessive fertilization caused by hysteresis effect. In the control decision, the rolling time domain optimization is adopted, the potential energy burden is introduced into the genetic algorithm as a penalty term, and the search space constraint boundary is adaptively corrected through the prediction residual of the last period to prevent blind topdressing under model mismatch. Finally, a smooth instruction is generated through low-pass filtering to realize accurate and stable regulation of rice soil nutrients.

Description

Technical Field

[0001] This invention relates to the field of soil nutrient management, specifically to a method and system for the synergistic regulation of soil nutrients in rice based on big data analysis. Background Technology

[0002] In modern rice cultivation, using sensor arrays to monitor the soil environment in real time and automatically adjust the concentration of nutrients such as nitrogen, phosphorus, and potassium based on the monitoring data is an important means to achieve high crop yields and efficient resource utilization.

[0003] Existing technologies for solving the aforementioned fertilization control problems typically employ standard genetic algorithms or greedy optimization strategies based on error feedback as their core algorithms. The main logic of these algorithms is to quickly search for and generate fertilization formulas based on the deviation between the current sensor readings and the set target values, aiming to eliminate errors in the shortest possible time. However, in the actual application scenario of paddy fields, soil is a complex colloidal system that exhibits a significant buffering and hysteresis effect on changes in nutrient concentration.

[0004] Existing algorithms often overlook this physicochemical mechanism, treating the soil environment as a linear, immediate-response system. When the soil exhibits strong buffering capacity, leading to a lag in nutrient concentration increases, traditional algorithms may misinterpret this as insufficient fertilization, thus continuously and aggressively outputting high-intensity fertilization commands. This optimization approach, which lacks consideration of historical cumulative effects, easily leads to excessive fertilizer accumulation in the soil. This not only wastes resources and causes soil acidification and compaction, but also results in severe overshoot and oscillations in the system due to the delayed release of subsequent nutrients, making it unable to meet the needs of precision farming. Summary of the Invention

[0005] To address the problem that existing algorithms can easily lead to excessive fertilizer accumulation in the soil, this invention proposes a method for coordinated regulation of soil nutrients in rice paddies based on big data analysis. This method includes: constructing a time-series characteristic state space based on real-time environmental data collected by a multi-source sensor array; establishing a digital model characterizing the evolution of soil buffering capacity and chemical potential field using a variant of Hooke's law, wherein the model includes: calculating elastic deformation potential energy based on the deviation between the target nutrient concentration and the current actual nutrient concentration, and calculating a historical cumulative dissipation term based on the rate of regulation operations within a historical time window, using the sum of the elastic deformation potential energy and the historical cumulative dissipation term as the current chemical potential burden; obtaining regulation instructions using a multi-objective genetic algorithm, wherein the search space constraint boundary of the multi-objective genetic algorithm is adaptively adjusted in response to the prediction residual of the previous control cycle; and performing low-pass filtering and shaping on the regulation instructions to obtain smooth instructions, using the current operation's consumption of the soil's future buffering capacity as the initial boundary condition for the next control cycle.

[0006] The chemical potential burden is defined by equating the soil's buffering capacity for nutrient concentration to elastic deformation potential energy and the cumulative effect of historical regulatory operations to dissipation terms. This physical mapping mechanism fully considers the memory of the soil ecosystem and avoids over-fertilization or regulatory oscillations caused by neglecting the soil's buffering role. Simultaneously, a multi-objective genetic algorithm that adaptively adjusts the boundary based on the predictive residual, combined with low-pass filtering shaping, ensures the smoothness of regulatory commands and the robustness of the system, achieving precise, dynamic, and synergistic regulation of paddy soil nutrients that conforms to the soil's physicochemical properties.

[0007] Furthermore, the construction of the time-series characteristic state space includes: establishing a first-in-first-out circular buffer in the controller memory to collect and store the real-time environmental data at a set frequency; performing physical clamping processing on the real-time environmental data to remove non-physical jump noise, and performing Min-Max normalization processing.

[0008] This invention effectively improves the quality and efficiency of data processing by establishing a first-in-first-out circular buffer and performing physical clamping and standardization. Compared with directly using raw data, this scheme can automatically eliminate non-physical jump noise caused by poor sensor contact or electrical interference, and map data of different dimensions to a unified interval, eliminating the impact of data scale differences on the convergence speed of subsequent modeling algorithms, and ensuring the purity and usability of the temporal feature state space.

[0009] Furthermore, the fitness function of the multi-objective genetic algorithm includes: using the chemical potential burden and its gradient with respect to time as a future trend penalty term.

[0010] Furthermore, the specific method for calculating the chemical potential burden is as follows:

[0011] ;

[0012] in This indicates the chemical potential burden; This indicates the target nutrient concentration; This indicates the current actual nutrient concentration; This indicates the set soil buffer coefficient; This represents the antagonistic damping factor, used to characterize the strength of interactions between different ions; Representing historical moments The rate of chemical reaction or the rate of fertilizer injection; Indicates the length of the time window for integration.

[0013] Compared to traditional independent nutrient regulation, this model explicitly quantifies the interaction strength between different ions, enabling it to accurately describe nonlinear relationships in multi-ion coexistence environments. The introduction of an integral term quantifies the long-term impact of fertilization history, allowing the system to consider past inputs when making decisions, thus avoiding resource waste from ineffective fertilization even after soil adsorption saturation.

[0014] Furthermore, the real-time environmental data includes the soil's N / P / K ion concentration, pH value, and ORP value.

[0015] In a second aspect, the present invention provides a rice soil nutrient synergistic regulation system based on big data analysis, comprising: a data acquisition and preprocessing unit connected to a multi-source sensor array, used to acquire real-time environmental data, perform physical clamping and standardization processing on the environmental data, and construct a time-series characteristic state space covering a set time window in memory through a circular buffer; a potential energy field modeling unit connected to the data acquisition and preprocessing unit, used to construct a nutrient chemical potential energy field model based on the time-series characteristic state space using a variant of Hooke's law, and calculate the chemical potential energy burden characterizing the instability of the soil system; a closed-loop optimization decision unit connected to the potential energy field modeling unit, used to perform rolling time-domain optimization, introduce the chemical potential energy burden as a penalty term into the fitness function to generate candidate fertilization formulas, and adaptively correct the search space constraint boundary of the optimization algorithm according to the prediction residual of the previous control cycle; and a control execution and feedback unit connected to the closed-loop optimization decision unit and the execution mechanism, used to perform low-pass filtering and shaping on the candidate fertilization formulas to generate control signals to drive the execution mechanism, and generate a resource reservation snapshot as the initial boundary condition for the next control cycle.

[0016] Furthermore, the potential field modeling unit is configured to calculate the chemical potential burden by:

[0017] ;

[0018] in This indicates the chemical potential burden; This indicates the target nutrient concentration; This indicates the current actual nutrient concentration; This indicates the set soil buffer coefficient; This represents the antagonistic damping factor, used to characterize the strength of interactions between different ions; Representing historical moments The rate of chemical reaction or the rate of fertilizer injection; Indicates the length of the time window for integration.

[0019] Furthermore, the closed-loop optimization decision unit is configured to: read the previous control cycle. Predicted residuals Determine the predicted residual Does it exceed the set convergence dead zone range? ; in response to Multiply the upper bound of the search space of the current optimization algorithm by the shrinkage factor. This will tighten the maximum allowable amount of fertilizer.

[0020] Compared to traditional optimization algorithms with fixed search boundaries, this system automatically reduces the maximum allowable fertilization amount through a shrinkage factor when it detects a significant deviation between the model prediction and the actual situation. This mechanism effectively prevents the system from blindly outputting high-intensity fertilization commands in the event of model mismatch or external disturbances, greatly improving the system's safety and self-stabilization capability under fault conditions.

[0021] Furthermore, the control execution and feedback unit is configured to calculate the current execution opening degree using the following formula. :

[0022] ;

[0023] in The theoretical pump valve opening output by the closed-loop optimization decision unit; The actual opening degree executed at the previous moment; The set filter coefficients.

[0024] Furthermore, the system adopts a distributed Internet of Things architecture, and the multi-source sensor array includes an ion-selective electrode for detecting the concentration of nitrogen, phosphorus and potassium ions in the soil, a glass electrode sensor for detecting pH value, and a platinum electrode for detecting redox potential; the actuator includes a solenoid valve and a peristaltic pump.

[0025] The use of ion-selective electrodes enables low-cost, online nutrient element detection; the use of peristaltic pumps allows for precise fluid delivery at the micro-level. The distributed architecture gives the system good scalability and deployment flexibility, enabling it to adapt to the field network needs of paddy fields of different sizes, reducing the cable costs and maintenance difficulties of large-scale deployments.

[0026] The technical effects of this invention are as follows:

[0027] This invention models the soil's nutrient buffering capacity as elastic deformation potential energy and the historical cumulative effect of fertilization as a dissipative term, thereby obtaining the chemical potential burden characterizing the instability of the soil system. By combining a multi-objective genetic algorithm that considers ion antagonism and adaptive constraints on prediction residuals, this invention solves the problem of neglecting soil physicochemical inertia and multi-ion coupling in traditional agricultural regulation, achieving precise, smooth, and memory-enabled intelligent fertilization. Attached Figure Description

[0028] Figure 1 This is a schematic flowchart illustrating the rice soil nutrient synergistic regulation method based on big data analysis, as described in an embodiment of the present invention.

[0029] Figure 2 This is a schematic illustration of a thermal map showing the spatial distribution of pH values ​​in paddy field soil obtained by a sensor array in an embodiment of the present invention.

[0030] Figure 3 This is a schematic diagram illustrating the nitrogen concentration gradient field and potential energy vector distribution in an embodiment of the present invention;

[0031] Figure 4 This is a schematic diagram illustrating the comparison of the evolution of the elastic potential energy term and the historical dissipation term over time in the chemical potential energy field model of this invention embodiment;

[0032] Figure 5 This is a schematic diagram illustrating the convergence of the N / P / K search boundary value based on feedback closed-loop logic with time steps in an embodiment of the present invention.

[0033] Figure 6 This is a schematic diagram illustrating the spatial distribution of fertilization points optimized based on the potential energy field in an embodiment of the present invention;

[0034] Figure 7 This is a schematic diagram illustrating the comparison of the effects of the low-pass filter shaping mechanism on the smoothing of control signals in an embodiment of the present invention;

[0035] Figure 8 This is a schematic diagram illustrating the cumulative consumption of different fertilizer elements over time in an embodiment of the present invention.

[0036] Figure 9 This is a schematic diagram illustrating a comparison between the proposed solution and the traditional algorithm in terms of cumulative fertilizer application and savings in this embodiment of the invention.

[0037] Figure 10 This is a schematic diagram illustrating the structural block of a rice soil nutrient synergistic regulation system based on big data analysis, according to an embodiment of the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0040] Example of a method for synergistic regulation of soil nutrients in rice based on big data analysis:

[0041] like Figure 1 As shown, the rice soil nutrient synergistic regulation method based on big data analysis of the present invention includes:

[0042] S1. Collect real-time environmental data based on a multi-source sensor array and construct a time-series feature state space.

[0043] To address the lack of historical reference in existing single-point sampling technologies, this embodiment establishes a data structure with temporal memory. The controller allocates a memory segment with a capacity of [missing information]. The first-in-first-out (FIFO) circular buffer is used, and the sampling frequency is set to once every 30 minutes, thus forming a time sliding window covering the past 24 hours.

[0044] The controller reads the current N / P / K ion concentration, pH value, and ORP value through a sensor array. Before storing the data in the buffer, the system performs physical clamping on the raw signal to remove non-physical jump noise.

[0045] Specifically, the system presets the physical effective range of pH value as follows: If the sensor's returned value exceeds this range, the system determines that the data is transient interference and keeps the previous sampled value unchanged, thereby ensuring the physical authenticity of the original data.

[0046] Subsequently, the system will place the buffer... Heterogeneous data from each period is processed using Min-Max normalization and mapped to... The interval is used to obtain a composite state space composed of the current state vector and the historical trend tensor, which provides a data foundation for subsequent potential field modeling.

[0047] like Figure 2 As shown, this diagram illustrates a spatial distribution thermogram of soil pH values ​​generated by a sensor array during the system's initialization phase. The horizontal and vertical axes represent the length and width of the paddy field (in meters), and the color bars on the right represent pH readings (range 5.76-7.20). Significant differences in pH exist between different areas of the paddy field. This embodiment, by aligning these spatially heterogeneous data, provides the fundamental physical boundaries for subsequently constructing a precise potential energy field.

[0048] S2. A variant of Hooke's law is used to determine a digital model that can characterize the soil buffering capacity and the evolution of the chemical potential field.

[0049] In this embodiment, to address the soil overexploitation problem caused by greedy optimization, the soil environment is digitally defined as a nutrient chemical potential energy field by introducing field theory. Soil, as a complex colloidal system, has a buffering effect on nutrient changes, but existing technologies often neglect this hysteresis effect. Therefore, this embodiment introduces a variant of Hooke's law, analogizing the soil's resistance to nutrient concentration deviations from equilibrium to elastic potential energy, thereby quantifying the system's instability. Specifically, the system calculates the current moment using the following formula. Chemical potential burden :

[0050] ;

[0051] in This indicates the set target nutrient concentration balance point. For example, the standard value for nitrogen concentration is set to... ; This indicates the actual nutrient concentration detected by the sensor at the current moment; This represents the soil buffering coefficient, which is physically mapped to the soil's cation exchange capacity (CEC). In this embodiment, for clay paddy fields, [the specific value is missing]. Set as ; This represents the antagonistic damping factor, used to characterize the strength of interactions between different ions, such as the inhibition of magnesium ions by potassium ions. In this embodiment, it can be set to... It should be noted that since this antagonistic damping factor is calculated based on a fixed concentration of each ion, the corresponding unit is concentration. ; Representing historical moments The rate of chemical reaction or the rate of fertilizer injection; The time window length for integration is set to [value] in this embodiment. Hour.

[0052] The first term of the above formula This characterizes the elastic deformation potential energy. With the actual concentration... With target value An increase in deviation leads to a quadratic increase in potential energy, indicating that the system is in a high-energy unsteady state and urgently needs to return to equilibrium. The second term of the formula... This is the historical cumulative dissipation term. This integral term forces the system to remember the past. Historical regulation within a given time period. If high-intensity fertilization operations were performed within that historical period, i.e. The integral term is relatively large, and even if the current concentration has not yet reached the target, the total potential energy will be significantly increased. .

[0053] The above mechanism manifests itself in the following control strategy: when it is detected that the historical input is too large, the algorithm will determine that the current state is in a high potential energy burden state, thereby suppressing further aggressive fertilization instructions, effectively preventing excessive fertilizer accumulation and subsequent soil acidification caused by delayed soil response.

[0054] like Figure 3 As shown, the distribution of nitrogen (N) concentration in the paddy field at a certain moment and its corresponding gradient field are displayed. The gradient vector direction of the potential energy field is also shown, indicating the direction of the potential driving force for nutrient diffusion or fertilization optimization.

[0055] Based on this physical field, the system calculates the potential energy burden, such as... Figure 4 As shown in the figure, this diagram provides a detailed analysis of the dynamic evolution logic of the two core elements: the elastic potential energy term and the historical dissipation term. The elastic potential energy term curve exhibits high-frequency and dramatic fluctuations, reflecting the real-time deviation between the current nutrient concentration and the target value. In contrast, the historical dissipation term curve changes relatively smoothly and exhibits lag, reflecting the system's accumulated memory of past time windows.

[0056] Combination Figure 4 It can be seen that even when the instantaneous deviation shown by the elastic potential energy term curve is small at certain times (such as around June 3, 2024), the historical memory term shown by the historical dissipation term curve is still at a high level. At this time, the total potential energy calculated by the system is still large, thus effectively suppressing the aggressive fertilization command and preventing excessive regulation.

[0057] S3. Use feedback closed-loop logic to perform rolling time-domain optimization to adaptively correct the search parameter constraint boundary.

[0058] When the controller executes the genetic algorithm for optimization, it will use the results calculated in step S2. and its gradient with respect to time A fitness function is introduced as a future trend penalty term. The system predicts whether the soil potential energy will diverge in the future if the currently generated candidate solution, i.e., the fertilization formula, is implemented. For example, if a certain formula leads to a predicted... Exceeding the safety threshold If so, then the solution will be given a very large penalty weight.

[0059] The specific correction logic is as follows:

[0060] First, the system reads the previous control cycle ( The predicted residuals This refers to the deviation between the predicted nutrient increase at the previous moment and the current actual detected value. Furthermore, in this embodiment, the convergence dead zone range can be set to... If at this time If the actual lift is far less than expected, it indicates strong soil adsorption or severe leaching. The system will automatically correct the search space constraints of the current optimization algorithm. Specifically, the system will multiply the upper bound of the search space by a shrinkage factor. This physically tightens the maximum allowable amount of fertilizer.

[0061] Ultimately, through the aforementioned closed-loop feedback based on residuals, the system can avoid blindly adding fertilizer when the model is mismatched, thus preventing the system from falling into a state of instruction saturation or overshoot.

[0062] like Figure 5 The diagram illustrates the evolution of the upper bounds for the search of N, P, and K fertilizers over time steps. In the initial stage (time steps 0-50), the search boundaries are wide due to the unknown initial soil state. As the closed-loop feedback progresses, the system identifies the soil's adsorption characteristics and gradually tightens the boundaries using a shrinkage factor. Finally, after time step 150, all three curves converge to a steady-state region close to 0, indicating that the system has found the optimal control trajectory, avoiding random searches within a large, ineffective space.

[0063] Based on the above optimizations, the system plans the final execution actions. For example... Figure 6 As shown in the figure, the triangles mark the optimal discrete fertilization points calculated by the algorithm. (Comparison) Figure 3 As can be seen from the concentration field, these points are intelligently distributed in regions with lower concentration gradients.

[0064] S4. Obtain smooth control instructions based on low-pass filter shaping mechanism and update resource reservation snapshot.

[0065] To avoid mechanical wear caused by frequent start-stop cycles of the actuator, the optimized output is also shaped in this embodiment. The controller uses a first-order low-pass filter to convert the discrete formula output by the algorithm into a continuous control signal. Let the theoretical pump valve opening output by the algorithm be... The actual opening degree at the previous moment was The current execution openness The calculation is as follows:

[0066] ;

[0067] In this embodiment, the filter coefficient is set to This parameter setting makes the change curve of the control command smoother, avoiding the impact of abrupt changes on the soil microenvironment.

[0068] Finally, the system generates a resource reservation snapshot, writing the current operation's consumption of the soil's future buffering capacity—that is, the new added value of the integral term in step S2—into the database. This snapshot data will be used as the basis for the next control cycle. The initial boundary conditions are used to form a complete logical closed loop in the time dimension.

[0069] like Figure 7 As shown, in order to protect the actuator, the instruction is shaped in this embodiment. The figure shows a comparison between the optimized output signal and the final control signal: the output before filtering represents the original instruction directly generated by the algorithm, which presents high-frequency pulse fluctuations. If executed directly, it will cause the pump valve to start and stop frequently; while the control signal after filtering represents the final instruction after filtering, which removes high-frequency noise, retains low-frequency trends, and ensures the smoothness of the execution action.

[0070] In one embodiment, when the system detects that the current nutrient concentration is low but the historical integral term value is high, the total potential energy is calculated according to the above method. The concentration will remain high, allowing the controller to output a more conservative fertilization command. Compared to traditional greedy algorithms, this invention effectively avoids over-fertilization due to misjudgment in scenarios where sudden rainfall causes a temporary decrease in concentration. Furthermore, through the residual feedback mechanism in step S3, when the model prediction fails due to soil physicochemical property drift, compaction leading to a decrease in permeability, the system can use a shrinkage factor... The system automatically adjusts the search boundary; this adaptive adjustment mechanism enables the system to control the steady-state error within a certain range during long-term operation. Within the dead zone, the system's adaptability to nonlinear environmental changes is significantly enhanced.

[0071] like Figure 8 and Figure 9 As shown, where Figure 8 The cumulative consumption curves of N, P, and K fertilizers during one week of operation in this embodiment are shown. The curves rise in a step-like manner and gradually slow down, indicating that after reaching a steady state, the system can maintain low-consumption operation for a long time without exhibiting an exponential growth out-of-control phenomenon.

[0072] and Figure 9 The improved algorithm in this embodiment is further demonstrated to compare its performance with that of the traditional algorithm in the prior art that does not incorporate a potential energy field: in the early stages of the experiment, the difference between the two is not significant; however, as time goes on, the traditional algorithm, which ignores the hysteresis effect of the soil, frequently performs overcompensation, leading to a rapid increase in its cumulative fertilizer application. In contrast, the method in this embodiment maintains a lower cumulative amount through the suppression effect of historical memory terms.

[0073] Example of a rice-soil nutrient synergistic regulation system based on big data analysis:

[0074] On the other hand, this invention also provides a rice soil nutrient synergistic regulation system based on big data analysis. This system is built on a distributed Internet of Things (IoT) architecture, such as... Figure 10As shown, it includes an embedded main controller, a multi-source sensor array deployed in the paddy field tillage layer, and an actuator.

[0075] The embedded main controller (exemplarily, an STM32H7 series microcontroller based on an ARM Cortex-M7 core) serves as the core processing platform and is electrically connected to the multi-source sensor array and the actuator, respectively.

[0076] The system specifically includes the following interconnected and collaborative units in terms of logical function:

[0077] A data acquisition and preprocessing unit, connected to the multi-source sensor array, is used to acquire real-time environmental data and construct a time-series characteristic state space. The multi-source sensor array specifically includes an ion-selective electrode for detecting soil nitrogen, phosphorus, and potassium ion concentrations, a glass electrode sensor for detecting pH values, and a platinum electrode for detecting redox potentials. The data acquisition and preprocessing unit maintains a first-in-first-out (FIFO) circular buffer in memory, acquiring data at a set frequency (e.g., every 30 minutes) to form a time sliding window covering a set time window (e.g., 24 hours). This unit is also configured to perform physical clamping processing on the raw signal (e.g., limiting the pH value to between 3.0 and 9.0) to eliminate non-physical jump noise, and to perform Min-Max normalization processing on the heterogeneous data in the buffer, constructing a composite state space composed of the current state vector and the historical trend tensor.

[0078] The potential energy field modeling unit, connected to the data acquisition and preprocessing unit, is used to construct a digital model characterizing soil buffering capacity and the evolution of the chemical potential energy field using a variant of Hooke's law. This unit introduces field theory, digitally defining the soil environment as a nutrient chemical potential energy field, and analogizing the soil's resistance to nutrient concentration deviations from equilibrium to elastic potential energy. Based on the preset target nutrient concentration equilibrium point, the current actual nutrient concentration, the soil buffer coefficient, the antagonistic damping factor, and the historical chemical reaction rates, this unit calculates the chemical potential energy burden at the current moment. By introducing a historical cumulative dissipation term, the system is forced to remember the regulatory operations within the historical time window. When excessive historical input is detected, aggressive fertilization commands are suppressed to prevent soil acidification and fertilizer accumulation.

[0079] A closed-loop optimization decision unit, connected to the potential energy field modeling unit, is used to perform rolling time-domain optimization using feedback closed-loop logic to adaptively correct the search parameter constraint boundary. This unit employs an improved multi-objective genetic algorithm, introducing the chemical potential energy burden and its gradient with respect to time as a future trend penalty term into the fitness function, assigning penalty weights to candidate solutions that lead to potential energy divergence. This unit also includes closed-loop correction logic, used to read the prediction residual from the previous control cycle. When the prediction residual exceeds the set convergence dead zone, a shrinkage factor is used to automatically correct the upper bound of the search space of the current optimization algorithm, preventing blindly adding fertilizer in case of model mismatch.

[0080] The control execution and feedback unit, connected to the closed-loop optimization decision unit and the actuator, is used to generate smooth control commands and update resource reservation snapshots based on a low-pass filter shaping mechanism. The actuator consists of a solenoid valve and a peristaltic pump. This unit uses a first-order low-pass filter to convert the discrete formula output by the closed-loop optimization decision unit into a continuous control signal to avoid mechanical wear caused by frequent start-stop of the actuator. Simultaneously, this unit generates a resource reservation snapshot, writing the current operation's consumption of the soil's future buffering capacity into a database as the initial boundary condition for the next control cycle, thus forming a complete logical closed loop in the time dimension.

[0081] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

Claims

1. A method for synergistic regulation of soil nutrients in rice based on big data analysis, characterized in that, The method includes: constructing a time-series characteristic state space based on real-time environmental data collected by a multi-source sensor array; and establishing a digital model characterizing the soil buffering capacity and chemical potential field evolution using a variant of Hooke's law, wherein the model includes: The elastic deformation potential energy is calculated based on the deviation between the target nutrient concentration and the current actual nutrient concentration, and the historical cumulative dissipation term is calculated based on the control operation rate within the historical time window. The sum of the elastic deformation potential energy and the historical cumulative dissipation term is used as the chemical potential energy burden at the current moment. A multi-objective genetic algorithm is used to obtain control instructions, and the search space constraint boundary of the multi-objective genetic algorithm is adaptively adjusted in response to the prediction residual of the previous control cycle. The control command is subjected to low-pass filtering and shaping to obtain a smooth command, and the consumption of the soil's future buffering capacity by the current operation is used as the initial boundary condition for the next control cycle. The fitness function of the multi-objective genetic algorithm includes: using the chemical potential burden and its gradient with respect to time as a future trend penalty term; The specific method for calculating the chemical potential burden is as follows: ;in This indicates the chemical potential burden; This indicates the target nutrient concentration; This indicates the current actual nutrient concentration; This indicates the set soil buffer coefficient; This represents the antagonistic damping factor, used to characterize the strength of interactions between different ions; Representing historical moments The rate of chemical reaction or the rate of fertilizer injection; Indicates the length of the time window for integration; Characterizing elastic deformation potential energy; This is a historical cumulative dissipation term; The prediction residual is the deviation between the predicted nutrient increase at the previous time step and the current actual measured value.

2. The method for synergistic regulation of rice soil nutrients based on big data analysis according to claim 1, characterized in that, The construction of the temporally sequenced feature state space includes: A first-in-first-out circular buffer is established in the controller memory to collect and store the real-time environmental data at a set frequency; Physical clamping is performed on the real-time environmental data to remove non-physical jump noise, and Min-Max normalization is performed.

3. The method for synergistic regulation of rice soil nutrients based on big data analysis according to claim 1, characterized in that, The real-time environmental data includes the soil's N / P / K ion concentrations, pH value, and ORP value.

4. A rice soil nutrient synergistic regulation system for implementing the rice soil nutrient synergistic regulation method based on big data analysis as described in any one of claims 1-3, characterized in that, The system includes: a data acquisition and preprocessing unit connected to a multi-source sensor array, used to acquire real-time environmental data, perform physical clamping and standardization on the environmental data, and construct a time-series characteristic state space covering a set time window in memory through a circular buffer; a potential energy field modeling unit connected to the data acquisition and preprocessing unit, used to construct a nutrient chemical potential energy field model based on the time-series characteristic state space using a variant of Hooke's law, and calculate the chemical potential energy burden characterizing the instability of the soil system; a closed-loop optimization decision unit connected to the potential energy field modeling unit, used to perform rolling time-domain optimization, introducing the chemical potential energy burden and its gradient with respect to time as a future trend penalty term into the fitness function to generate candidate fertilization formulas, and adaptively correcting the search space constraint boundary of the optimization algorithm according to the prediction residual of the previous control cycle; and a control execution and feedback unit connected to the closed-loop optimization decision unit and the execution mechanism, used to perform low-pass filtering and shaping on the candidate fertilization formulas to generate control signals to drive the execution mechanism, and generate a resource reservation snapshot as the initial boundary condition for the next control cycle.

5. The rice soil nutrient synergistic regulation system according to claim 4, characterized in that, The closed-loop optimization decision unit is configured as follows: Read the previous control cycle Predicted residuals ; Determine the predicted residual Does it exceed the set convergence dead zone range? ; In response to Multiply the upper bound of the search space of the current optimization algorithm by the shrinkage factor. This will tighten the maximum allowable amount of fertilizer.

6. The rice soil nutrient synergistic regulation system according to claim 4, characterized in that, The control execution and feedback unit is configured to calculate the current execution level using the following formula. : ； in The theoretical pump valve opening output by the closed-loop optimization decision unit; The actual opening degree executed at the previous moment; The set filter coefficients.

7. The rice soil nutrient synergistic regulation system according to claim 4, characterized in that, The system adopts a distributed Internet of Things architecture. The multi-source sensor array includes an ion-selective electrode for detecting the concentration of nitrogen, phosphorus, and potassium ions in the soil, a glass electrode sensor for detecting pH value, and a platinum electrode for detecting redox potential. The actuator includes a solenoid valve and a peristaltic pump.

Images