A weather prediction-based intelligent agricultural resource scheduling method and scheduling system

Abstract

The application discloses a kind of wisdom agricultural resource scheduling method and scheduling system based on weather forecast, belong to wisdom agricultural field.Method includes: collecting farmland meteorological environment data;Variational mode decomposition is executed to barometric pressure data, and high-frequency disturbance term is extracted;The kurtosis factor of high-frequency disturbance term is calculated, and based on the judgment of continuous threshold value is identified extreme weather precursor;Trigger defense state scheduling, including based on hydrological model dynamic calculation reserved depth and carry out drainage pretreatment, and greenhouse side window is contracted to preset wind opening degree and is reinforced;After weather returns to normal, based on environmental stress data, calculate physiological deficiency amount by crop growth model, and generate water and fertilizer compensation scheme.System uses three-layer architecture, end node has the ability of independent operation based on local emergency strategy when communication is interrupted.The present application overcomes the hysteresis of traditional post-response by accurately identifying disaster precursor and executing preset scheduling and accurate compensation, and improves the risk resistance of agriculture.

Description

Technical Field

[0001] This application belongs to the field of smart agriculture, specifically involving a smart agricultural resource scheduling method and scheduling system based on weather forecasting. Background Technology

[0002] With the deep integration of IoT and big data technologies in agricultural production, smart agricultural resource scheduling systems have become a key infrastructure for ensuring stable crop yields and enhancing disaster prevention and mitigation capabilities. By integrating environmental sensing, automatic control, and data analysis technologies, modern agricultural management can achieve real-time monitoring and precise intervention of farmland microclimates, thereby optimizing resource allocation and reducing production risks.

[0003] Among them, automated scheduling technology based on meteorological monitoring aims to achieve intelligent management of production processes such as irrigation, drainage, and temperature control by collecting environmental parameters in real time and driving the actions of agricultural facilities. The basic principle of this technology is to use a sensor network deployed in the field to obtain the time series of meteorological elements and trigger the physical response of the actuators according to preset logical thresholds to maintain the ideal environmental conditions for crop growth.

[0004] However, existing technologies typically employ feedback control based on real-time thresholds, a reactive mechanism that exhibits significant lag when facing sudden extreme weather events. Due to the inherent time consumption in the physical execution processes of agricultural infrastructure such as drainage pumping stations and greenhouse curtains, delayed scheduling instructions often prevent protective measures from being deployed before disaster strikes, leading to irreversible physical damage to crops. Furthermore, traditional scheduling algorithms have limited ability to analyze non-stationary meteorological signals, making it difficult to effectively isolate and identify hidden weak fluctuations and precursory features from complex background noise. In addition, existing solutions largely rely on linear logic analysis, failing to capture nonlinear state changes during meteorological evolution, resulting in a lack of predictability in resource allocation under extreme conditions.

[0005] Therefore, a smart agricultural resource scheduling method and system based on weather forecasting is desired. Summary of the Invention

[0006] The purpose of this invention is to provide a smart agricultural resource scheduling method and system based on weather forecasting, which can effectively solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Firstly, a smart agricultural resource scheduling method based on weather forecasting includes the following steps: Collect multidimensional time-series data of farmland meteorological environment, wherein the multidimensional time-series data includes at least air pressure data; Variational mode decomposition is performed on the air pressure data, and high-frequency disturbance terms characterizing the precursory features of extreme weather are extracted through adaptive signal decomposition. The high-frequency disturbance terms include the fourth intrinsic mode component and the fifth intrinsic mode component. Identify precursor features of extreme weather, calculate kurtosis factors for the instantaneous amplitude sequences of the fourth and fifth intrinsic mode components respectively, and when any of the kurtosis factors exceeds the corresponding preset safety threshold and continues for a preset time, confirm the existence of extreme weather precursors and generate a warning sign. Based on the warning sign, defensive resource scheduling is triggered, and drainage pretreatment instructions and facility reinforcement instructions are generated and issued. The drainage pretreatment instructions are based on the dynamic calculation of the required storage space depth using a hydrological model, and the facility reinforcement instructions include retracting the greenhouse side windows to a preset windproof opening. In response to the return of meteorological data to normal, the system switches from a defensive state back to a production state, calculates the physiological deficit of crops based on environmental stress data during the defense period, and generates and executes a water and fertilizer compensation scheme based on the physiological deficit.

[0008] Preferably, multidimensional time-series data of farmland meteorological environment are collected, specifically including: The distributed weather stations deployed synchronously collect air pressure, wind speed and humidity data at a sampling frequency of 5Hz, and use a circular buffer to manage the collected raw data to construct a 60-second sliding time window. The sampled data undergoes physical extreme value verification and rate of change constraint verification, and automatically switches to the nearest backup weather station data source when a sensor failure is detected.

[0009] Preferably, variational mode decomposition processing is performed, specifically including: The number of decomposition modes K is set to 5, the penalty parameter α is set to 2000, a variational constraint model is constructed and iteratively solved by the alternating direction multiplier method to decompose the air pressure signal into 5 intrinsic mode components. A multi-core parallel computing architecture is used to accelerate the iterative solution process, so that the time taken for a single decomposition operation is controlled within 0.5 seconds; The first intrinsic mode component obtained from the decomposition is labeled as the low-frequency trend term, and the fourth and fifth intrinsic mode components are labeled as the high-frequency disturbance terms.

[0010] Preferably, the precursor features for identifying extreme weather include: Based on a sliding window containing 300 sampling points, the kurtosis factors of the 4th intrinsic mode component and the 5th intrinsic mode component are calculated respectively. The kurtosis factor is compared with the preset thresholds K4=8.5 and K5=7.2 respectively. When any kurtosis factor exceeds its corresponding threshold and exceeds the threshold continuously for 3 sampling periods, the existence of extreme weather precursors is confirmed. Risk levels are classified into general risk and high risk based on the degree to which the kurtosis value exceeds the threshold.

[0011] Preferably, the execution process of the wastewater pretreatment instruction includes: Drive all electric gates to open fully to clear the drainage channels and start the pumping station to force down the water level in the ditch; The hydrological model dynamically calculates the required depth of storage space by iteratively solving the water balance principle based on the historical maximum rainfall intensity, farmland soil permeability, and current soil moisture content. When the soil is saturated, the reserved depth is increased by a second preset depth.

[0012] Preferably, the execution process of the facility reinforcement order includes: A high-torque stepper motor is driven to retract the opening of the greenhouse side window to a preset windproof opening position, which is 15% of the maximum opening set after optimization based on fluid dynamics simulation; At the same time, the electromagnetic locking device is activated to fix the windproof reinforcement rod to the preset position.

[0013] Preferably, calculating the physiological deficit of the crop specifically includes: Extract the decrease in total light radiation, the cumulative increase in ambient humidity, and the integral value of temperature fluctuation during the defense period to construct an environmental stress feature vector; The environmental stress feature vector is input into a preset crop growth model. The reduction in photosynthetic products due to insufficient photosynthetically effective radiation and the increase in respiratory consumption caused by abnormal temperature and humidity during the defense period are calculated by integral calculation. The sum of the two is the physiological deficit.

[0014] Preferably, the crop growth model includes a rectangular hyperbola model representing the rate of photosynthesis and an exponential model representing the rate of respiration; The right-angle hyperbolic model is a function of the net photosynthetic rate with respect to the initial light energy utilization rate, light intensity, maximum photosynthetic rate, and dark respiration rate. The exponential model is a function of the respiration rate at actual temperature with respect to the respiration rate at 25°C, the temperature coefficient, and the actual temperature.

[0015] Preferably, a water and fertilizer compensation plan is generated and executed, specifically including: Based on the physiological deficit, the preset compensation coefficient table is consulted to determine the supplementary ratio of the three nutrients, nitrogen, phosphorus, and potassium. The additional ratio is converted into control parameters for the integrated water and fertilizer unit, including the target nutrient solution concentration and total irrigation volume; A proportional-integral-derivative (PID) control algorithm is adopted to dynamically adjust the pulse width of the solenoid valves in each nutrient solution branch based on the deviation between the real-time feedback conductivity value and the target concentration, so as to achieve precise mixing.

[0016] Secondly, a smart agricultural resource scheduling system based on weather forecasting includes: The central processing server is equipped with a high-performance floating-point arithmetic unit to run variational mode decomposition algorithms and extreme weather precursor identification models, and to generate scheduling instructions. The field control gateway is deployed in farmland areas and establishes a communication connection with the end execution nodes through a wireless mesh network for forwarding the scheduling instructions; The end execution node, including drainage gates, pumping stations, and greenhouse actuators, is used to receive and execute the scheduling instructions; The end execution node has a built-in local storage unit and an independent controller. Emergency policy rules are pre-written in the local storage unit. When a communication interruption with the field control gateway is detected, the end execution node automatically switches to independent operation mode, reads the last valid warning level received in the local storage unit, and performs basic defense actions according to the emergency policy rules.

[0017] In summary, this application includes at least one of the following beneficial technical effects: 1. This invention introduces a variational mode decomposition algorithm to adaptively decompose non-stationary air pressure signals into multiple intrinsic mode components, specifically extracting the 4th and 5th high-frequency disturbance terms associated with severe convective weather. Based on this, the impulse characteristics of the high-frequency disturbance terms are statistically quantified using a kurtosis factor, and confirmed through a jitter-reducing delay mechanism across multiple consecutive sampling periods. This effectively removes background noise and random interference, enabling the precise capture of weak precursory features before extreme weather events from complex meteorological backgrounds. Compared to traditional real-time threshold triggering modes, this predictive identification mechanism significantly improves the accuracy and lead time of warnings, providing valuable operational time for subsequent defensive measures such as drainage and reinforcement.

[0018] 2. Upon confirming precursors to extreme weather, this invention immediately triggers a defensive scheduling mechanism. For drainage, the system abandons fixed drainage strategies and instead uses a hydrological model to dynamically and iteratively calculate the optimal depth of reserved storage space, taking into account historical maximum rainfall intensity, current soil moisture content, and soil permeability, thus achieving pre-positioned capacity reduction in drainage channels. For greenhouse reinforcement, the side window opening is precisely reduced to 15% of its maximum opening, optimized through fluid dynamics simulation, ensuring basic air exchange while minimizing wind pressure load. This pre-positioned physical intervention based on precise calculations effectively overcomes the lag caused by the inherent time consumption of facility execution, elevating the disaster response mode from "passive response" to "active prevention," and greatly improving the disaster redundancy of agricultural infrastructure.

[0019] 3. This invention overcomes the limitations of traditional techniques that rely solely on soil moisture for simple replenishment. After a disaster, it constructs an environmental stress feature vector composed of decreased solar radiation, accumulated humidity, and temperature fluctuations. Combined with a built-in rectangular hyperbolic photosynthetic model and exponential respiration model, it precisely quantifies the physiological deficit caused to crops by extreme weather. Furthermore, based on this deficit, it automatically generates a water and fertilizer compensation plan including the proportions of nitrogen, phosphorus, and potassium supplementation, and uses a proportional-integrated-differential control algorithm to drive the integrated water and fertilizer unit for precise mixing and application. This complete technical chain, from "environmental stress quantification" to "physiological deficit compensation," ensures the scientific and targeted nature of post-disaster recovery, effectively compensates for the negative impact of crop growth stagnation during the disaster response period, and provides strong technical support for achieving ultimate crop yield stability. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the intelligent agricultural resource scheduling method based on weather forecasting of the present invention. Figure 2 This is a schematic diagram of the core principle framework of the present invention based on variational mode decomposition and precursor feature recognition; Figure 3 This is a flowchart of the main stages of the bidirectional switching logic between production and defense states in this invention. Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow between the central processing server, the field control gateway, and the end execution nodes in this invention; Figure 5 This is a flowchart of the main stages of the recovery and compensation mechanism based on crop physiological deficit prediction in this invention. Detailed Implementation

[0021] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of specific embodiments based on the present invention is provided in conjunction with the accompanying drawings and preferred embodiments.

[0022] Reference Figures 1 to 5 As shown, in a smart agricultural resource scheduling method based on weather forecasting, a perception and decision-making system with predictive identification capabilities is constructed to achieve early response to extreme weather events and pre-positioned scheduling of physical facilities. Specifically, the method follows these steps: First, regarding step S1: collecting multi-dimensional time-series data of farmland meteorological environment, the aim is to acquire high-quality real-time data on air pressure, wind speed, and humidity through a sensor network deployed in the farmland, and to construct a standardized rolling time series to ensure the data's continuity, reliability, and real-time performance. The specific process includes the following: Step S101: Deploy distributed integrated weather stations and configure sensors. Multiple integrated weather stations are set up at preset locations in the farmland. Each weather station is a distributed sensing unit, which integrates a piezoresistive barometric pressure sensor, an ultrasonic time-of-flight wind speed sensor, and a capacitive thin-film humidity sensor. The barometric pressure sensor has micropascal resolution and can capture weak pulsations of atmospheric pressure.

[0023] The location of weather stations is determined based on farmland topography, crop distribution, and wireless communication coverage to ensure the representativeness of data collection and the reliability of network transmission.

[0024] Step S102: Set the sampling frequency and collect data synchronously. Based on the monitoring needs of sudden severe convective weather, the sampling frequency is set to 5Hz. All meteorological stations are equipped with GPS timing modules to collect air pressure, wind speed and humidity data synchronously according to a unified time step to ensure that the multi-element data are strictly aligned in time.

[0025] Step S103: Construct a rolling time series vector, using a circular buffer to manage the collected raw data. The length of the circular buffer is set to 300 sampling points, corresponding to a 60-second sliding time window.

[0026] The circular buffer is implemented in the form of a circular queue. Whenever new sampled data arrives, it is stored at the tail of the queue and the oldest sampled point at the head of the queue is automatically removed, thereby maintaining a sliding observation window that reflects the current atmospheric physical state in real time.

[0027] Step S104: Perform data quality monitoring. The processor performs logical consistency verification on the uploaded original messages, including physical extreme value checks and rate of change constraint checks.

[0028] The physical extreme value check is set with a normal air pressure range of 850 hPa to 1080 hPa, a humidity range of 0% to 100% RH, and a wind speed range of 0 m / s to 60 m / s.

[0029] The rate of change constraint check limits the rate of change between adjacent sampling points. That is, the rate of change per second is obtained by dividing the difference between adjacent sampling points by the sampling interval of 0.2 seconds, and then multiplying by 60 to obtain the rate of change per minute. The rate of change of air pressure does not exceed 2 hPa per minute, the rate of change of humidity does not exceed 10% RH per minute, and the rate of change of wind speed does not exceed 5 m / s per minute.

[0030] All data exceeding the above thresholds were marked as potential anomalies.

[0031] Step S105: Abnormal data processing and backup data source switching. When the sampled values ​​of three consecutive sampling points exceed the physical extreme value range, or the rate of change of three consecutive sampling points exceeds the rate of change constraint, it is determined that the current sensor has a hardware failure or is subject to severe environmental interference. At this time, the abnormal jump value is immediately removed, and the system automatically switches to the backup data source of the nearby weather station.

[0032] The system is pre-configured with redundant weather stations that operate simultaneously with the main weather station. Data from the backup stations is synchronized to the central processor in real time via a hot backup link. During the switchover, the central processor selects the backup weather station that is geographically closest to the faulty station and has normal data quality as the new data source based on a preset proximity priority list. Since the backup station and the main station use the same sampling frequency and time synchronization mechanism, the switchover process can achieve seamless data flow connection.

[0033] Through steps S101 to S105, high-frequency and high-precision data collection of farmland meteorological environment is completed, and the accuracy and integrity of the data are ensured through real-time quality monitoring.

[0034] The preprocessed multidimensional time series data will be used as input to the variational mode decomposition step S2, which will be used to further extract the precursor features of extreme weather hidden in meteorological signals.

[0035] Secondly, regarding step S2: performing variational mode decomposition (VMD) processing on the meteorological signal, the high-quality air pressure time series obtained in step S1 is input into the VMD algorithm. Adaptive signal decomposition extracts modal components of different frequency bands, thereby revealing the precursory features of extreme weather hidden in non-stationary meteorological data. The specific execution process is as follows: Step S201: Set the decomposition parameters and initialize the algorithm. The processor reads the original air pressure sequence output in step S1, determines the number of decomposition modes K to be 5, and sets the penalty parameter α to 2000.

[0036] The selection of the number of modes K is based on the analysis of the frequency band characteristics of meteorological signals. Specifically, the first mode is used to characterize the low-frequency trend term, the fourth and fifth modes are used to capture high-frequency disturbances related to sudden strong convection, and the remaining modes are intermediate transition frequency bands.

[0037] The penalty parameter α is used to balance the signal reconstruction accuracy and modal bandwidth limitation in the alternating direction multiplier method (ADMM) iteration. A value of 2000 can effectively suppress modal aliasing and ensure decomposition stability.

[0038] Step S202: Construct a variational constraint model and solve iteratively. The processor takes the original signal f(t) as input and constructs the following variational constraint optimization problem:

[0039] in, This represents the k-th intrinsic mode component. Its corresponding center frequency, For the Dirac function, Represents convolution operation. The imaginary unit, This indicates taking the partial derivative with respect to time. The square of the L2 norm represents the signal energy.

[0040] The goal of the variational constraint model described above is to find a set of modal components that minimizes the sum of the analytic signal bandwidths of each component, while ensuring that the sum of all components is exactly equal to the original signal.

[0041] The solution process employs the Alternating Direction Multiplier Method (ADMM). By introducing a penalty parameter α to construct an augmented Lagrangian function, the constrained optimization problem is transformed into an unconstrained problem for iterative solution. The algorithm alternately updates each mode in the frequency domain. and its center frequency The algorithm continues until convergence, and the numerical computation library embedded in the processor can be directly called to implement the algorithm.

[0042] Step S203: Separate and extract key modal components. After iterative convergence, the algorithm outputs five intrinsic modal components IMF1 to IMF5 and their center frequencies.

[0043] The system labels IMF1 as a low-frequency trend term to monitor the macroscopic evolution trend of atmospheric pressure; it labels IMF4 and IMF5 as high-frequency disturbance terms to reflect the microscopic pressure oscillation characteristics generated by atmospheric stratification before heavy precipitation or storms, and serves as the direct input for subsequent precursor identification; the remaining intermediate modes are used as background noise and are not involved in emergency decision-making, but can be used to assist in the analysis of the transitional state of meteorological evolution.

[0044] Step S204: Parallel computing ensures real-time performance. To meet the requirement of second-level response to sudden weather events, the processor adopts a multi-core parallel computing architecture and uses the Single Instruction Multiple Data (SIMD) instruction set to accelerate matrix operations and frequency domain transformations in ADMM iteration.

[0045] In actual deployment, the time taken for a complete decomposition operation containing 5 modalities is controlled within 0.5 seconds, and the total processing latency from the end of data acquisition to the output of feature vectors does not exceed 2 seconds, ensuring that the system has sufficient time margin to execute subsequent defensive state scheduling.

[0046] Step S205: Generate frequency domain feature vectors. After decomposition, encapsulate the instantaneous amplitude sequence, center frequency, and energy distribution parameters of each mode into frequency domain feature vectors and pass them to the extreme weather precursor identification module in step S3.

[0047] The frequency domain feature vector contains the original amplitude information of IMF4 and IMF5, and records their energy proportion, that is, the proportion of the instantaneous energy of the mode to the total signal energy, as well as the frequency drift trend, that is, the rate of change of the center frequency over time, providing fine input data for kurtosis factor calculation.

[0048] Through the above steps, the non-stationary air pressure signal is adaptively decomposed into frequency band components with clear physical meaning. The system successfully isolates high-frequency disturbance components closely related to severe weather, laying a reliable data foundation for subsequent precursor identification based on kurtosis factors.

[0049] Next, regarding step S3: identifying precursory features of extreme weather, statistical analysis is performed on the high-frequency mode components obtained from step S2. The kurtosis factor is calculated to quantitatively characterize the pulse characteristics of the signal, and a jitter reduction mechanism is used to eliminate noise interference, thereby accurately identifying precursory signals of extreme weather before a disaster occurs. The specific execution process is as follows: Step S301: Real-time extraction of high-frequency modal components. The processor directly extracts the instantaneous amplitude sequences of the fourth intrinsic modal component IMF4 and the fifth intrinsic modal component IMF5 from the frequency domain feature vector output in step S205. These two components mainly reflect the micro-pressure oscillations generated by the atmospheric stratification before heavy precipitation or strong winds, and are the input data source for subsequent kurtosis factor calculation.

[0050] The extraction operation is completed within each data update cycle to ensure that the analysis results are synchronized with the latest weather conditions.

[0051] Step S302: Calculate the kurtosis factor for each component. For the amplitude sequences of IMF4 and IMF5, calculate their kurtosis value K according to the fourth-order cumulant formula. The kurtosis factor measures the sharpness or tail thickness of the signal amplitude distribution relative to the normal distribution. The calculation formula is as follows:

[0052] Where x is the amplitude sequence of the current modal component, Let be the mean of the sequence. Standard deviation, This indicates the calculation of mathematical expectation.

[0053] When atmospheric stratification becomes unstable, the originally stable high-frequency disturbances will evolve into pulsed, violent oscillations, and the amplitude distribution will deviate significantly from the normal distribution, resulting in a rapid increase in kurtosis values.

[0054] During the calculation, the processor adopts a sliding window method. The window length is consistent with the rolling time series window in step S103, that is, the window contains 300 sampling points corresponding to 60 seconds of data. As new sampling data arrives, the window slides forward, always calculating the kurtosis value based on the latest 300 sampling points, ensuring the real-time performance and statistical reliability of the kurtosis estimation.

[0055] Step S303: Set a safety threshold and perform a preliminary judgment. Two independent kurtosis safety limits are preset, corresponding to the trigger thresholds of IMF4 and IMF5 respectively.

[0056] The trigger thresholds for IMF4 and IMF5 are based on historical meteorological data and disaster records. The IMF4 and IMF5 kurtosis values ​​within 10 minutes before multiple heavy rainfall and strong wind events in the past three years were selected, and the 95th percentile was taken as the baseline threshold. After on-site calibration, the final threshold K4 for IMF4 was determined to be 8.5, and the threshold K5 for IMF5 was determined to be 7.2.

[0057] When the kurtosis value of any component exceeds its corresponding threshold, the system marks the component as a potential precursor and records the specific value exceeding the threshold, proceeding to the next stability confirmation step; if neither exceeds the limit, the current meteorological state is determined to be normal, and monitoring continues.

[0058] Step S304: Introduce a jitter-reducing delay mechanism to verify the stability of potential precursor signals in order to avoid false triggering caused by local gusts or instantaneous noise from the sensor.

[0059] The kurtosis value of the same component is required to exceed the threshold for at least 3 consecutive sampling periods. Each sampling period corresponds to 0.2 seconds at a sampling frequency of 5Hz in step S102, that is, the kurtosis value is always higher than the threshold for 0.6 consecutive seconds.

[0060] Only when this continuous condition is met can the existence of extreme weather precursors be confirmed. The introduction of shake-reduction delay effectively filters out brief random disturbances and improves the accuracy of early warnings.

[0061] Step S305: Generate risk level and trigger defensive state transition. After confirming that the precursor signal is stable, the system classifies the current risk level according to the degree to which the kurtosis value exceeds the corresponding threshold.

[0062] Specifically, a kurtosis value within 50% of the threshold is defined as general risk, while a kurtosis value above 50% is defined as high risk. For example, for IMF4, a kurtosis value between 8.5 and 12.75 is considered general risk, while a value greater than 12.75 is considered high risk; for IMF5, a kurtosis value between 7.2 and 10.8 is considered general risk, while a value greater than 10.8 is considered high risk.

[0063] The system generates an extreme weather warning sign and transmits the sign along with the current risk level to the defensive resource scheduling module in step S4. At the same time, it records the IMF4 and IMF5 amplitude, kurtosis value and corresponding timestamp at the trigger time for subsequent log analysis and algorithm optimization.

[0064] Through the above steps, the system achieves quantitative identification of precursors to extreme weather such as heavy precipitation and strong winds based on kurtosis analysis of high-frequency modal components, and ensures the accuracy of the judgment through a jitter-reducing mechanism.

[0065] Validated early warning features will drive subsequent resource pre-allocation and scheduling, gaining valuable early warning time for agricultural production.

[0066] Next, regarding step S4: defensive resource scheduling instructions, the system immediately activates pre-positioned protective measures upon confirming precursors to extreme weather. This is achieved by issuing instructions for drainage pretreatment and facility reinforcement, thus enabling the pre-deployment of physical facilities. The specific execution process is as follows: Step S401: System state switching and instruction generation. After receiving the extreme weather warning sign generated in step S305, the central controller immediately switches the system operating state from production state to defense state, and simultaneously generates drainage pretreatment instructions and facility reinforcement instructions. Both instructions are assigned high priority and are ready to be sent to the corresponding execution terminals via wireless network.

[0067] Step S402: Drainage pretreatment command execution. The central controller sends a high-priority control signal to the execution terminal of the farmland drainage canal through an industrial-grade wireless transmission protocol such as LoRa or 4G. The execution terminal consists of a programmable logic controller (PLC) and an electric actuator. Upon receiving the signal, it immediately drives all electric gates to fully open to ensure unobstructed drainage channels. At the same time, it starts the pumping station deployed in the low-lying area. The submersible pumps in the pumping station start working to forcibly lower the water level in the ditch.

[0068] The pump station's operating logic is based on a preset safety warning line. The system dynamically calculates the required depth of storage space by using a hydrological model, based on the water content data fed back by the current soil moisture sensor, combined with the farmland soil permeability and the historical maximum rainfall intensity.

[0069] The hydrological model is based on the principle of water balance. First, it determines the design rainfall per unit time based on the historical maximum rainfall intensity, and then calculates the total runoff based on the farmland catchment area. Next, it estimates the soil infiltration rate based on the soil permeability and corrects the actual infiltration capacity using the current soil moisture content. Finally, it subtracts the soil infiltration from the total runoff to obtain the amount of water that needs to be stored by the ditch, and divides it by the area of ​​the ditch water area to obtain the reserved space depth.

[0070] The above calculation process can be expressed as a functional relationship between the reserved depth H and the soil permeability P, the historical maximum rainfall intensity I, and the current soil moisture content θ. The system solves this iteratively through a built-in numerical model.

[0071] When the soil moisture sensor indicates that the soil is saturated, the system automatically increases the reserved depth to a second preset depth, for example, by 30% on the basis of the original calculated value, so as to reserve sufficient storage capacity for the upcoming surface runoff.

[0072] Step S403: Facility reinforcement command executed. After receiving the warning signal, the greenhouse controller drives the high-torque stepper motor to retract the opening of the side window to the preset windproof opening position.

[0073] The preset windproof opening value is determined through fluid dynamics simulation optimization. For example, it is set to 15% of the maximum opening. This can minimize the wind pressure load inside the greenhouse while ensuring basic air exchange and preventing sudden changes in air pressure. At the same time, the controller activates the electromagnetic locking device to fix the windproof reinforcement rod to the preset position to prevent extreme winds and heavy rainfall from overturning the greenhouse film.

[0074] Step S404: Energy optimization strategy execution. While starting the high-power pumping station, the central controller temporarily cuts off the power supply to non-essential landscape lighting, regular irrigation branches, and auxiliary ventilation equipment through the power monitoring unit and intelligent relays to ensure that power resources are prioritized for the stable operation of drainage and reinforcement facilities.

[0075] The system monitors water level changes in real time. Once the water level drops below the safe level or the warning is lifted, the controller gradually restores power supply to non-essential loads based on the current power load.

[0076] Through the above steps, the system quickly enters a defensive state after the early signs of a disaster are confirmed, completes the pre-set capacity reduction of the drainage system and the windproof reinforcement of greenhouse facilities, and dynamically optimizes energy distribution.

[0077] The aforementioned pre-positioned scheduling measures shorten preparation time for responding to upcoming extreme weather, effectively overcome the lag in traditional post-event responses, and lay the infrastructure foundation for the subsequent recovery of agricultural production.

[0078] Finally, regarding step S5: recovery compensation and state switching logic, the system automatically returns to the production state after the disaster warning is lifted, and performs precise nutrient compensation based on the potential losses to crops caused by environmental stress during the defense period. The specific execution process is as follows: Step S501: Determining and switching between normal meteorological conditions. The central processing unit continuously receives the real-time meteorological data updated in step S1 and determines the normal range of the three indicators: air pressure, wind speed, and humidity.

[0079] The normal range is preset to the normal range specified in step S104 and the rate of change is lower than the threshold. Specifically, the air pressure is between 850hPa and 1080hPa and the change per minute does not exceed 2hPa, the humidity is between 0% and 100%RH and the change per minute does not exceed 10%RH, and the wind speed is between 0m / s and 60m / s and the change per minute does not exceed 5m / s.

[0080] The system requires all monitoring data to remain stable within the normal range for 5 consecutive minutes before the disaster process can be confirmed to be over. After confirmation, the central processing unit automatically switches the system's operating status from defensive mode back to production mode and records the duration of the defensive mode and the extreme values ​​and cumulative amounts of various environmental parameters during the period.

[0081] Step S502: Environmental stress data extraction during the defense phase. The system reads key environmental parameters recorded during the defense phase from the historical database. This includes three data items: (1) The decrease in total solar radiation is equal to the difference between the actual accumulated photosynthetically active radiation during the defense period and the historical average total radiation during the same period, wherein the historical average total radiation during the same period is obtained by averaging the radiation data of the same day in the past three years.

[0082] (2) The cumulative increase in environmental humidity is equal to the difference between the actual cumulative humidity during the defense period and the cumulative value of the baseline humidity suitable for crop growth. The baseline humidity is set to 60% relative humidity according to the crop variety.

[0083] (3) Temperature fluctuation integral value is obtained by calculating the time integral of the instantaneous temperature deviation from the upper and lower limits of the suitable temperature range during the defense period. The suitable temperature range is set to 20℃ to 28℃ according to the crop variety.

[0084] The above three data points constitute the environmental stress feature vector, which serves as the input for subsequent loss assessment.

[0085] Step S503: Crop growth model call and physiological deficit calculation.

[0086] First, the system calls the built-in crop growth model, which is built based on existing crop physiological mechanisms and presets the response curve of photosynthetic rate to light and the response function of respiration rate to temperature and humidity for specific crop varieties.

[0087] The rate of photosynthesis is represented by a rectangular hyperbola model, which has the following form:

[0088] in, Indicates net photosynthetic rate, Indicates the initial light energy utilization rate. Indicates light intensity. Indicates the maximum photosynthetic rate. This indicates the rate of dark respiration.

[0089] The respiratory rate is represented by an exponential model, which is as follows:

[0090] in, This represents the respiration rate at actual temperature. This indicates the respiratory rate at 25°C. The temperature coefficient (i.e., the factor by which the respiration rate increases for every 10°C increase in temperature), T is the actual temperature (unit: °C), and the effect of humidity is introduced through a correction factor.

[0091] The crop growth model takes the environmental stress feature vector output in step S502 as input and calculates the reduction of photosynthetic products due to insufficient photosynthetically active radiation and the increase in respiration consumption caused by high temperature and high humidity during the defense period through integral calculation.

[0092] Specifically, the defense period is divided into several time steps. Within each step, the current photosynthetic product production and respiration consumption are calculated based on actual environmental parameters and compared with the simulated values ​​under suitable conditions. The total deficit is then accumulated, and the sum of the two is the physiological deficit of the crop, expressed in grams of dry matter or as a percentage of expected yield.

[0093] The model parameters were calibrated using field trial data before system deployment to ensure that the calculation results matched the actual response of local crops.

[0094] Step S504: Generation of water and fertilizer compensation scheme.

[0095] Based on the physiological deficit calculated in step S503, and combined with the current crop growth period's nutrient requirements, the supplementary ratios of nitrogen, phosphorus, and potassium are determined using a preset compensation coefficient table.

[0096] The compensation coefficient table is established based on the correlation test between deficit and yield loss. The table stores the nitrogen, phosphorus and potassium compensation ratios corresponding to different deficit ranges. For example, when the deficit is converted to 5% of the expected yield loss, the compensation coefficient is set to increase nitrogen by 15%, phosphorus by 10% and potassium by 12%.

[0097] The system converts the compensation ratio into control parameters for the integrated water and fertilizer unit, namely the target nutrient solution concentration and the total irrigation volume. The target nutrient solution concentration is calculated based on the compensation ratio and the base concentration, while the total irrigation volume is determined based on the supplementary water volume calculated from the deficit and the crop's water requirement pattern.

[0098] Step S505: PID control of the water and fertilizer unit to perform compensation.

[0099] The central controller sends compensation commands to the integrated water and fertilizer unit. The unit's built-in proportional-integral-derivative control algorithm dynamically adjusts the pulse width of the solenoid valves in each nutrient solution branch according to the deviation between the target concentration and the real-time feedback conductivity (EC) value, so as to achieve precise mixing of nitrogen, phosphorus, and potassium.

[0100] The proportional gain, integral time constant, and derivative time constant of the PID algorithm are pre-tuned based on the pipeline flow characteristics and the mixing tank volume. The tuning method adopts the Ziegler-Nichols empirical formula or the field trial and error method. At the same time, the opening time of the irrigation main valve is adjusted to ensure that the compensating nutrient solution is applied to the crop root zone according to the set amount.

[0101] Step S506: Compensation effect tracking and closed-loop correction.

[0102] In subsequent irrigation cycles, the system continues to monitor the compensation effect through soil sensors and crop physiological sensors. Soil sensors monitor changes in root zone nutrient concentration, while crop physiological sensors monitor indicators such as chlorophyll fluorescence or leaf nitrogen content.

[0103] If monitoring data from multiple consecutive irrigation cycles shows that the recovery rate of crop growth indicators is lower than the expected model output, the system will automatically fine-tune the compensation coefficient for subsequent cycles until the crop growth returns to its normal trajectory.

[0104] The fine-tuning rule is: if the actual recovery speed is more than 10% slower than expected, the proportion of each element will be increased by 5% in the next compensation.

[0105] Through the above steps, the system achieves a smooth transition from a defensive state to a productive state after a disaster, and provides precise water and fertilizer compensation based on the physiological deficit caused by environmental stress, effectively offsetting the negative impact of crop growth stagnation during the defense period, and providing technical support for stable final yield.

[0106] At the system architecture level, the scheduling system adopts a three-layer structure design, including a central processing server, a field control gateway, and end execution nodes.

[0107] The central processing server is deployed in the control center and is equipped with a high-performance floating-point arithmetic unit to run variational mode decomposition algorithms and extreme weather precursor identification models, and undertake all signal processing and decision calculation tasks.

[0108] The field control gateway is deployed in farmland areas and establishes communication connections with each end execution node through an industrial-grade wireless mesh network. This network supports dynamic routing and multi-path transmission. When a node fails due to severe weather interference, the network protocol can automatically replan the communication path to bypass the faulty node, ensuring that control commands can still be delivered to the target node with predetermined reliability. The gateway device adopts a waterproof and dustproof design and has a built-in 4G backup communication module, which automatically switches to the backup link when the main network fails.

[0109] The end execution node integrates a dual redundancy mechanism of manual intervention and automatic control. Each node is equipped with a local storage unit and an independent controller. The local storage unit is pre-written with the last received valid meteorological parameters and preset emergency strategy rules.

[0110] Under normal operating conditions, the node receives instructions from the gateway to perform actions such as drainage and reinforcement. When wireless communication is interrupted for a long time due to extreme weather, the node detects the communication timeout through its built-in watchdog timer and automatically switches to independent operation mode.

[0111] In this mode, the node reads the last valid meteorological parameters stored locally and performs basic defense actions according to the preset emergency strategy, such as automatically opening drainage gates or tightening greenhouse side windows according to the stored warning level; the emergency strategy rules are stored in the form of conditional statements, such as opening all gates if the warning level is high risk and the current water level is higher than the threshold.

[0112] The dispatch system also integrates a long-term meteorological trend prediction module. This module is built on a long short-term memory neural network. It takes the time series of air pressure, wind speed and humidity over a preset historical period as input, extracts the long-term dependencies of meteorological evolution through deep learning, and outputs the macro-risk assessment level for the preset forecast period in the future.

[0113] The long-term weather trend prediction module is trained using historical meteorological data and corresponding disaster records during the offline phase. After training, it is deployed to the central server as a supplement to the short-term precursor identification in step S3.

[0114] When the long-term weather trend forecasting module predicts a high risk of continuous rain or strong winds in the next few hours, it can increase water and fertilizer reserves in advance or adjust the resource reserve plan under defensive conditions to achieve a combination of long-term and short-term weather warnings and resource allocation.

[0115] The aforementioned system architecture, through the organic combination of layered deployment, redundant communication, and local intelligence, ensures that the entire chain from data collection and early warning identification to defense execution remains reliable in extreme environments, providing agricultural production with all-weather disaster prevention and mitigation capabilities.

[0116] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects.

[0117] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment includes only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A smart agricultural resource scheduling method based on weather forecasting, characterized in that, Includes the following steps: Collect multidimensional time-series data of farmland meteorological environment, wherein the multidimensional time-series data includes at least air pressure data; Variational mode decomposition is performed on the air pressure data, and high-frequency disturbance terms characterizing the precursory features of extreme weather are extracted through adaptive signal decomposition. The high-frequency disturbance terms include the fourth intrinsic mode component and the fifth intrinsic mode component. Identify precursor features of extreme weather, calculate kurtosis factors for the instantaneous amplitude sequences of the fourth and fifth intrinsic mode components respectively, and when any of the kurtosis factors exceeds the corresponding preset safety threshold and continues for a preset time, confirm the existence of extreme weather precursors and generate a warning sign. Based on the warning sign, defensive resource scheduling is triggered, and drainage pretreatment instructions and facility reinforcement instructions are generated and issued. The drainage pretreatment instructions are based on the dynamic calculation of the required storage space depth using a hydrological model, and the facility reinforcement instructions include retracting the greenhouse side windows to a preset windproof opening. In response to the return of meteorological data to normal, the system switches from a defensive state back to a production state, calculates the physiological deficit of crops based on environmental stress data during the defense period, and generates and executes a water and fertilizer compensation scheme based on the physiological deficit.

2. The smart agricultural resource scheduling method based on weather forecasting according to claim 1, characterized in that, Collect multidimensional time-series data of farmland meteorological environment, specifically including: The distributed weather stations deployed synchronously collect air pressure, wind speed and humidity data at a sampling frequency of 5Hz, and use a circular buffer to manage the collected raw data to construct a 60-second sliding time window. The sampled data undergoes physical extreme value verification and rate of change constraint verification, and automatically switches to the nearest backup weather station data source when a sensor failure is detected.

3. The smart agricultural resource scheduling method based on weather forecasting according to claim 1, characterized in that, Performing variational mode decomposition processing specifically includes: The number of decomposition modes K is set to 5, the penalty parameter α is set to 2000, a variational constraint model is constructed and iteratively solved by the alternating direction multiplier method to decompose the air pressure signal into 5 intrinsic mode components. A multi-core parallel computing architecture is used to accelerate the iterative solution process, so that the time taken for a single decomposition operation is controlled within 0.5 seconds; The first intrinsic mode component obtained from the decomposition is labeled as the low-frequency trend term, and the fourth and fifth intrinsic mode components are labeled as the high-frequency disturbance terms.

4. The smart agricultural resource scheduling method based on weather forecasting according to claim 1, characterized in that, Identifying precursory features of extreme weather includes: Based on a sliding window containing 300 sampling points, the kurtosis factors of the 4th intrinsic mode component and the 5th intrinsic mode component are calculated respectively. The kurtosis factor is compared with the preset thresholds K4=8.5 and K5=7.2 respectively. When any kurtosis factor exceeds its corresponding threshold and exceeds the threshold continuously for 3 sampling periods, the existence of extreme weather precursors is confirmed. Risk levels are classified into general risk and high risk based on the degree to which the kurtosis value exceeds the threshold.

5. The smart agricultural resource scheduling method based on weather forecasting according to claim 1, characterized in that, The execution process of the drainage pretreatment instruction includes: Drive all electric gates to open fully to clear the drainage channels and start the pumping station to force down the water level in the ditch; The hydrological model dynamically calculates the required depth of storage space by iteratively solving the water balance principle based on the historical maximum rainfall intensity, farmland soil permeability, and current soil moisture content. When the soil is saturated, the reserved depth is increased by a second preset depth.

6. The smart agricultural resource scheduling method based on weather forecasting according to claim 1, characterized in that, The execution process of a facility reinforcement order includes: A high-torque stepper motor is driven to retract the opening of the greenhouse side window to a preset windproof opening position, which is 15% of the maximum opening set after optimization based on fluid dynamics simulation; At the same time, the electromagnetic locking device is activated to fix the windproof reinforcement rod to the preset position.

7. The smart agricultural resource scheduling method based on weather forecasting according to claim 1, characterized in that, Calculating the physiological deficit of crops specifically includes: Extract the decrease in total light radiation, the cumulative increase in ambient humidity, and the integral value of temperature fluctuation during the defense period to construct an environmental stress feature vector; The environmental stress feature vector is input into a preset crop growth model. The reduction in photosynthetic products due to insufficient photosynthetically effective radiation and the increase in respiratory consumption caused by abnormal temperature and humidity during the defense period are calculated by integral calculation. The sum of the two is the physiological deficit.

8. The smart agricultural resource scheduling method based on weather forecasting according to claim 7, characterized in that, Crop growth models include the rectangular hyperbola model representing the rate of photosynthesis and the exponential model representing the rate of respiration; The right-angle hyperbolic model is a function of the net photosynthetic rate with respect to the initial light energy utilization rate, light intensity, maximum photosynthetic rate, and dark respiration rate. The exponential model is a function of the respiration rate at actual temperature with respect to the respiration rate at 25°C, the temperature coefficient, and the actual temperature.

9. The smart agricultural resource scheduling method based on weather forecasting according to claim 1, characterized in that, Generate and execute a water and fertilizer compensation plan, specifically including: Based on the physiological deficit, the preset compensation coefficient table is consulted to determine the supplementary ratio of the three nutrients, nitrogen, phosphorus, and potassium. The additional ratio is converted into control parameters for the integrated water and fertilizer unit, including the target nutrient solution concentration and total irrigation volume; A proportional-integral-derivative (PID) control algorithm is adopted to dynamically adjust the pulse width of the solenoid valves in each nutrient solution branch based on the deviation between the real-time feedback conductivity value and the target concentration, so as to achieve precise mixing.

10. A smart agricultural resource scheduling system based on weather forecasting, used to execute the smart agricultural resource scheduling method based on weather forecasting as described in any one of claims 1 to 9, characterized in that, include: The central processing server is equipped with a high-performance floating-point arithmetic unit to run variational mode decomposition algorithms and extreme weather precursor identification models, and to generate scheduling instructions. The field control gateway is deployed in farmland areas and establishes a communication connection with the end execution nodes through a wireless mesh network for forwarding the scheduling instructions; The end execution node, including drainage gates, pumping stations, and greenhouse actuators, is used to receive and execute the scheduling instructions; The end execution node has a built-in local storage unit and an independent controller. Emergency policy rules are pre-written in the local storage unit. When a communication interruption with the field control gateway is detected, the end execution node automatically switches to independent operation mode, reads the last valid warning level received in the local storage unit, and performs basic defense actions according to the emergency policy rules.

Images