Intelligent agricultural production system

Abstract

The application relates to the technical field of agricultural production, and discloses a smart agricultural production system, which comprises the following modules: an operation parameter generation module, which generates target application amount data and collects working condition parameter data; an actuator model library module, which stores model entries corresponding to different working condition intervals; a model selection module, which selects a target model entry; a control instruction generation module, which generates control instruction data; an execution control module, which drives and controls actuators according to the control instruction data; and a model updating module, which determines deviation data based on application amount feedback data. Through individualized identification of each actuator, an inverse mapping relationship between control quantity and application amount is established, and in the control instruction generation stage, the target application amount is converted and compensated according to the inverse mapping, so that the control instruction is matched with the actual output characteristics of the actuator, thereby reducing the deviation between the target application amount and the actual application amount and improving the application amount control precision and consistency.

Description

Technical Field

[0001] This invention relates to the field of agricultural production technology, specifically to a smart agricultural production system. Background Technology

[0002] In agricultural production, smart agricultural production systems are typically used for the information-based and automated management of production processes in fields or greenhouses. Their core function revolves around key production factors such as water and fertilizer, including operation planning, parameter generation, equipment control, and data acquisition. This enables unified scheduling and execution control of irrigation and fertilization operations. Furthermore, smart agricultural production systems for water and fertilizer operations generally require coordination with actuators such as valves and pumps, as well as sensors for flow, pressure, temperature, and fertilizer solution parameters. By generating target application rates and driving actuators to complete the application, a closed-loop control link is formed for the specific operation. Existing smart agricultural production systems typically employ control methods based on empirical parameters or fixed coefficients in water and fertilizer operations, directly converting the target application rate into the actuator's control quantity and issuing it for execution.

[0003] However, in current technology, the target application rate is converted into the actuator control quantity based on a fixed mapping relationship during water and fertilizer operations and then issued for execution. Due to individual differences in actuators and the fact that their characteristics change with operation, the fixed mapping relationship is prone to mismatch, resulting in a deviation between the target application rate and the actual application rate. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a smart agricultural production system that solves the problem of discrepancies between the target application rate and the actual application rate caused by the mismatch between the fixed mapping relationship and the individual differences and characteristic variations of the actuators.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a smart agricultural production system, comprising: The operation parameter generation module generates target application rate data for water and fertilizer operations and collects operating condition parameter data related to the water and fertilizer operations. The actuator model library module stores model entries corresponding to different operating conditions for each actuator related to water and fertilizer operations. The model entries include operating condition information, mapping data between control quantities and application quantities, and model version information. The model selection module selects target model entries based on the operating condition parameter data and the operating condition interval information; The control instruction generation module generates control instruction data based on the mapping data between the target application rate data and the target model entries; The execution control module drives and controls the actuator to perform water and fertilizer operations according to the control command data; The model update module determines the deviation data based on the application rate feedback data and updates the mapping data of the target model entry when the preset update conditions are met; the model update module meets the update range limit during the update process and reverts to the model entry corresponding to the model version information before the update when the preset rollback conditions are met. The model update module enters a frozen state after performing a rollback, prohibiting the updating of the mapping data within the frozen window, and setting the rolled-back model entries as the priority candidate model entries of the model selection module within the frozen window.

[0006] Preferably, the generation of the job parameters includes: Obtain the task object identifier and task time information, and determine the target application rate benchmark value based on the task object identifier and task time information; Collect the measured values ​​of the working condition parameters corresponding to the work object identifier, and perform time alignment processing on the measured values ​​of the working condition parameters to form the working condition parameter data; The target application rate data is determined based on the target application rate baseline value and the operating condition parameter data.

[0007] Preferably, the actuator model library module stores model entries including: Multiple operating condition intervals are set based on the value range of the operating condition parameter data, and corresponding operating condition interval information is generated; Model entries are established for the same actuator in different operating conditions, and the operating condition information of each model entry is associated with the mapping data between the control quantity and the applied quantity and stored accordingly. Set model version information for each model entry and establish the association between the model version information and the corresponding mapping data.

[0008] Preferably, the model selection module includes: Based on the operating condition parameter data, determine the operating condition range into which the current operating condition parameter falls, and determine the set of candidate model entries accordingly; When the current operating condition range changes, it is determined whether to switch the target model entry according to the anti-shake switching rules; The anti-shake switching rules include the working condition interval switching hysteresis threshold rule and the minimum dwell time rule; The target model entry is determined from the set of candidate model entries.

[0009] Preferably, the mapping data for the model entries includes lookup table mapping data; The model entries include minimum effective control parameters, response delay parameters, response inertia parameters, and effective range information of the mapping data; The effective range information is used to limit the range of values ​​for which the mapping relationship between the control quantity and the application quantity applies.

[0010] Preferably, the generation of the control command includes: Retrieve the control quantity corresponding to the target application rate data from the mapping data of the target model entry; When the target application rate data is located between adjacent retrieval points, the control quantity is determined according to a preset interpolation rule; When the target application rate data exceeds the effective range information of the mapping data, the control amount is determined according to the preset limit rule.

[0011] Preferably, the application rate feedback data is obtained through a flow measurement device; The model update module determines the deviation data based on the application rate feedback data and the target application rate data.

[0012] Preferably, the preset update condition is that the deviation data continuously exceeds a preset deviation threshold for a preset number of times.

[0013] Preferably, the update range limitation includes: The amount of data updated in a single instance shall not exceed a preset upper limit; Furthermore, updating the mapping data is prohibited when the data quality index of the application rate feedback data is lower than a preset quality threshold; The data quality indicators are determined based on the packet loss ratio, outlier ratio, and time alignment deviation, which are obtained by the model update module from the collected data.

[0014] Preferably, the preset rollback conditions include: Within the verification window after parameter updates, the deviation data did not meet the preset improvement criteria; The freeze window is a fixed duration window, and the freeze state is lifted after the freeze window expires. Within the frozen window, the model selection module prioritizes selecting the rolled-back model entry when the current operating condition range remains unchanged.

[0015] This invention provides a smart agricultural production system. It has the following beneficial effects: 1. This invention establishes an inverse mapping relationship between control quantity and application quantity by individually identifying each actuator, and performs control quantity conversion and compensation on the target application quantity based on the inverse mapping during the control command generation stage, so that the control command matches the actual output characteristics of the actuator, thereby reducing the deviation between the target application quantity and the actual application quantity and improving the control accuracy and consistency of the application quantity.

[0016] 2. This invention divides operating parameters such as pressure, temperature, and fertilizer concentration into intervals and establishes corresponding sub-model entries. During operation, it automatically selects or switches matching sub-models based on the current operating parameters, so that the system can still generate control commands using the corresponding mapping relationship when the operating conditions change, thereby improving the consistency and stability of application control under different operating conditions.

[0017] 3. In the process of online identification and mapping update, this invention sets update range limits and data quality gating, and triggers version rollback based on deviation criteria in the verification window. After rollback, it enters the freeze window and prioritizes the use of rollback model entries, thereby avoiding divergence of online self-learning under abnormal data or short-term disturbances and enhancing the reliability and availability of long-term operation. Attached Figure Description

[0018] Figure 1 This is an architectural diagram of a smart agricultural production system according to the present invention. Detailed Implementation

[0019] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see the appendix Figure 1 This invention provides a smart agricultural production system, comprising: The operation parameter generation module generates target application rate data for water and fertilizer operations and collects operation condition parameter data related to water and fertilizer operations. Furthermore, the generation of job parameters includes: Obtain the object identification and operation time information, and determine the target application rate baseline value based on the object identification and operation time information; Collect the measured values ​​of working parameters corresponding to the work object identifier, and perform time alignment processing on the measured values ​​of working parameters to form working parameter data; The target application rate data is determined based on the target application rate baseline value and operating condition parameter data.

[0021] Specifically, the operation parameter generation module outputs the target application rate data before each water and fertilizer operation or within each control cycle, and simultaneously collects the operating condition parameter data corresponding to that water and fertilizer operation. The operating condition parameter data is generated by sensors associated with the operation object. By completing the generation and collection of the target application rate and operating condition parameters, the subsequent model selection and control command generation use the same cycle of input data, which facilitates maintaining the data correspondence. For example, when operating on valve area A, the module simultaneously provides the target application rate data of valve area A at the beginning of the cycle and collects the pressure and temperature measurement values ​​of the valve area. The operation parameter generation module obtains the operation object identifier and operation time information, and determines the target application rate benchmark value from the preset data source based on the operation object identifier and operation time information. The preset data source is the corresponding entry in the operation plan table, crop stage parameter table or historical operation record table. By introducing the operation object and operation time, the target application rate benchmark value is established to correspond with the specific operation scenario, which makes it easy for different valve zones and different plots to call the matching benchmark value at different operation time periods. For example, when the operation object is valve zone A of greenhouse zone 1 and the operation time is the first cycle in the morning, the module reads the benchmark value corresponding to the entry. The operation parameter generation module collects the measured values ​​of the operating parameters corresponding to the operation object identifier, and performs time alignment processing on the measured values ​​of the operating parameters to form operating parameter data. The time alignment processing includes aligning the measured values ​​at different sampling times to the same control cycle and generating a representative value for that control cycle. Through time alignment, the operating parameter data and the control cycle are established to establish a correspondence, which facilitates subsequent model selection based on the operating condition interval. For example, if the pressure sensor and the temperature sensor have different sampling frequencies, the module aligns them to the same cycle before outputting the operating parameter data. After obtaining the target application rate baseline value and operating condition parameter data, the operation parameter generation module determines the target application rate data according to the preset correction rules and applies upper and lower limit constraints to the target application rate data. By correcting the baseline value in combination with the current operating conditions, the target application rate data can reflect the current operating conditions, which facilitates the subsequent control command generation module to generate matching control commands. For example, when the current operating conditions of valve zone A change, the module corrects the baseline value and outputs the target application rate data for that cycle.

[0022] The actuator model library module stores model entries for various actuators related to water and fertilizer operations and corresponding to different operating conditions. The model entries include operating condition information, mapping data between control quantities and application quantities, and model version information. Furthermore, the actuator model library module stores model entries including: Multiple operating condition intervals are set based on the value range of the operating condition parameter data, and corresponding operating condition interval information is generated; Model entries are established for the same actuator in different operating conditions, and the operating condition information of each model entry is associated with the mapping data between the control quantity and the applied quantity and stored accordingly. Set model version information for each model entry and establish the association between the model version information and the corresponding mapping data.

[0023] Specifically, the actuator model library module receives operating condition parameter data and divides it into multiple continuous intervals based on the value range of the operating condition parameter in the preset configuration or historical statistics. Then, it generates an interval identifier, upper and lower bounds, and effective conditions for each interval, thereby forming operating condition interval information that can be indexed and called. That is, by discretizing the continuously changing operating condition parameters into operating condition intervals, the interval identifier can be used as a unified index to manage model entries, which facilitates quick location and matching of entries when the operating condition changes. For example, for the actuator of valve area A, the module divides it into several intervals based on the change range of the corresponding operating condition parameters of the valve area and generates corresponding interval identifiers and interval boundary information for each interval. The actuator model library module sets an actuator identifier for each actuator and uses the actuator identifier and the operating condition interval identifier as an index to create model entries. Each model entry stores the operating condition interval information corresponding to that interval and the lookup table mapping data between the control quantity and the applied quantity, and saves the two together within the same entry. By creating entries separately according to the operating condition interval and fixing the mapping relationship, the subsequent model selection module can select the matching entry when the operating condition interval changes. The control command generation module generates control commands that match the current operating condition. For example, for the same actuator in valve zone A, the module creates corresponding entries for interval one, interval two and interval three, and saves the lookup table mapping data of that interval in each entry. Simultaneously, a version identifier field is set in each model entry, and the currently effective lookup mapping data is bound to this version identifier. When the mapping data is updated, the module retains the mapping data before the update and the corresponding version identifier as historical versions, and binds and stores the updated mapping data with the new version identifier. Based on the establishment of the binding relationship between the version identifier and the mapping data, a clear recovery target can be provided for subsequent rollback operations, so that the system can restore to the mapping data corresponding to the version before the update when rollback is required. For example, when the mapping data of a certain interval entry in valve zone A is updated, the module retains the version before the update and generates a new version identifier, and stores the old and new versions and their respective mapping data separately for selection or rollback.

[0024] The model selection module selects target model entries based on operating condition parameter data and operating condition interval information. Furthermore, the model selection module includes: Based on the operating condition parameter data, determine the operating condition range into which the current operating condition parameters fall, and determine the set of candidate model items accordingly; When changes occur in the current operating condition range, determine whether to switch the target model entry according to the anti-shake switching rules; Anti-shake switching rules include operating condition range switching hysteresis threshold rules and minimum dwell time rules; The target model item is determined from the set of candidate model items.

[0025] Specifically, the model selection module receives the operating condition parameter data output by the operation parameter generation module and reads the configured operating condition interval information from the actuator model library module. It compares the current operating condition parameter with the boundaries of each operating condition interval to determine the operating condition interval identifier corresponding to the current operating condition parameter. After obtaining the operating condition interval identifier, the model selection module uses the actuator identifier and the operating condition interval identifier as search keys to retrieve the model entry corresponding to the interval from the actuator model library module, thereby obtaining a set of candidate model entries. This transforms the continuously changing operating condition parameter into a searchable interval identifier and limits the candidate range to the set of entries corresponding to the current operating condition interval, which facilitates the subsequent stable output of the target model entry. For example, for the actuator in valve zone A, when the current operating condition parameter is determined to fall into interval two, the module retrieves the model entry corresponding to interval two to form a set of candidate model entries. In each control cycle, the model selection module updates the current operating condition interval identifier and compares it with the operating condition interval identifier recorded in the previous control cycle. When the two are consistent, the model selection module keeps the target model entry output in the previous control cycle unchanged and directly continues to output the target model entry to the control command generation module. When the two are inconsistent, the model selection module enters the switching determination process to decide whether to allow the target model entry to be switched from the corresponding entry in the previous operating condition interval to the entry in the current candidate model entry set. By establishing the switching determination on the continuous update of the control cycle, it can ensure that the output of the model entry and the change of operating condition parameters form a clear correspondence, avoiding unnecessary entry switching when the operating condition is stable. For example, when the operating condition interval of valve zone A changes from interval two to interval three, the module triggers the switching determination process, while when the interval remains interval two, it continues to output the target model entry of interval two. The anti-jitter switching rules executed in the switching determination process include the operating condition interval switching hysteresis threshold rule and the minimum dwell time rule. The hysteresis threshold rule is used to require that the operating condition parameters meet the entry conditions before switching to the adjacent interval when the operating condition parameters are close to the interval boundary, and to meet the exit conditions before returning to the original interval, thereby reducing repeated jumps near the boundary. The minimum dwell time rule is used to start the dwell timer after the target model item is switched, and no switching is allowed before the dwell time reaches the preset value. By superimposing the above two types of rules when the operating condition interval changes, the model item switching can be made smoother, avoiding frequent switching of the control link caused by short-term disturbances. For example, when the operating condition parameters of valve area A fluctuate near the boundary between interval two and interval three, the module will not frequently switch back and forth items under the constraints of hysteresis threshold and dwell time. When the anti-shake switching rule allows switching, the model selection module selects a model entry from the current candidate model entry set that matches the current operating condition interval identifier as the target model entry, and outputs the target model entry to the control command generation module. When the anti-shake switching rule does not allow switching, the model selection module maintains the target model entry of the previous control cycle as the current target model entry output. Thus, by directly using the determination result of the target model entry as the input of the control command generation module, it can be ensured that the subsequent control command generation is always based on the determined and continuous model entry, forming a continuous link from operating condition parameters to target model entry and then to control command.

[0026] Furthermore, the mapping data for model entries includes lookup table mapping data; Model entries include minimum effective control parameters, response delay parameters, response inertia parameters, and effective range information for the mapped data; Among them, the effective range information is used to limit the range of values ​​for which the mapping relationship between the control quantity and the application quantity applies.

[0027] Specifically, after the model selection module determines the target model item, the control instruction generation module reads the lookup mapping data from the target model item. The lookup mapping data stores the correspondence between multiple sets of control quantities and application quantities in the form of a table. Each set includes the control quantity value and its corresponding application quantity value. By using the lookup mapping data, the corresponding control quantity can be directly retrieved from the table based on the target application quantity, so that the generation of control instructions has a clear data basis. The target model entry also stores the minimum effective control quantity parameter, response delay parameter, and response inertia parameter. The control command generation module calls the above parameters when generating control commands: the minimum effective control quantity parameter is used to limit the lower limit of the control quantity, the response delay parameter is used to configure the effective time or duration of the control command, and the response inertia parameter is used to constrain the change process of the control quantity. Thus, by fixing the actuator response-related parameters in the model entry, the control command generation is consistent with the actual response characteristics of the actuator. The target model entry further stores the effective range information of the mapping data. The effective range information limits the range of control quantity values ​​and application quantity values ​​applicable to the mapping relationship. Before determining the control quantity, the control instruction generation module first performs boundary judgment. When the target application quantity or candidate control quantity exceeds the range limited by the effective range information, it is restricted according to the boundary value. That is, by introducing the effective range information, the mapping relationship is avoided outside the coverage of the mapping data, thus ensuring the applicability of the mapping relationship.

[0028] The control instruction generation module generates control instruction data based on the mapping data between the target application rate data and the target model entries; Furthermore, the generation of control commands includes: Retrieve the control quantity corresponding to the target application rate data from the mapping data of the target model entries; When the target application rate data is located between adjacent retrieval points, the control amount is determined according to the preset interpolation rules; When the target application rate data exceeds the effective range information of the mapped data, the control amount is determined according to the preset limit rule.

[0029] Specifically, after receiving the target model entry output by the model selection module and the target application rate data output by the operation parameter generation module, the control command generation module first reads the lookup table mapping data in the target model entry and locates the retrieval point corresponding to the target application rate data in the lookup table mapping data. In specific implementation, the control command generation module compares the target application rate data with the application rate value in the lookup table mapping data to determine the control quantity corresponding to the application rate value that is equal to the target application rate data, or to determine the set of adjacent retrieval points corresponding to the application rate interval into which the target application rate data falls. By searching in the lookup table mapping data, the target application rate can be converted into a control quantity that can be issued, providing a basis for subsequent interpolation or limiting. For example, when the target application rate data of valve zone A in the current period is a certain value, the module finds the corresponding application rate point and the control quantity corresponding to that point in the lookup table mapping data. When the target application rate data is located between adjacent retrieval points, the control command generation module determines the target control quantity between the control quantities corresponding to the adjacent retrieval points according to the preset interpolation rule. In specific implementation, the preset interpolation rule can be a linear interpolation rule. The control command generation module calculates the control quantity value between the two control quantity retrieval points based on the positional relationship between the target application rate data and the two adjacent application rate retrieval points. By determining the control quantity through interpolation, continuous control quantity output can be obtained between discrete points of the lookup table mapping data, making the generation of control commands smoother and avoiding control quantity jumps due to the sparse lookup table points. For example, when the target application rate data of valve zone A falls between two application rate retrieval points, the module obtains the control quantity between the control quantities of the two retrieval points according to the linear interpolation rule and uses it to generate control commands. When the target application amount data exceeds the application amount value range limited by the effective range information in the target model entry, the control command generation module determines the control amount according to the preset limiting rules. Specifically, the control command generation module first reads the lower limit and upper limit of the application amount in the effective range information. When the target application amount data is less than the lower limit, the target application amount is processed according to the lower limit and the corresponding control amount is selected. When the target application amount data is greater than the upper limit, the target application amount is processed according to the upper limit and the corresponding control amount is selected, thus forming the limited control amount. Through the limiting process, it can be ensured that the output control amount falls within the applicable range of the mapping relationship, avoiding the generation of unreliable control commands outside the coverage range of the mapping data.

[0030] The execution control module drives and controls the actuators to perform water and fertilizer operations based on control command data; Specifically, the execution control module receives control command data and converts it into drive control signals for the actuator. This drives the actuator to operate according to the control command data and complete the water and fertilizer operation. The execution control module controls the actuator's start / stop and operating status based on the control quantity values ​​and durations in the control command data, and terminates the drive at the end of the control cycle. By directly applying the control command data to the actuator, the corresponding control quantities for the target model items are executed. For example, when the control command data for valve zone A instructs the valve to operate at a preset opening degree for one control cycle, the execution control module outputs the corresponding drive signal to control the valve to open and close at the end of the cycle.

[0031] The model update module determines the deviation data based on the application rate feedback data and updates the mapping data of the target model item when the preset update conditions are met; the model update module meets the update range limit during the update process and reverts to the model item corresponding to the model version information before the update when the preset rollback conditions are met. The model update module enters a frozen state after performing a rollback. Within the frozen window, updating the mapping data is prohibited, and the rolled-back model entries are set as the priority candidate model entries for the model selection module within the frozen window.

[0032] Specifically, the model update module acquires application rate feedback data at the end of each control cycle and compares it with the target application rate data of the same cycle to obtain deviation data. When the preset update conditions are met, the model update module updates the mapping data of the current target model item and applies an update range limit to the single update during the update process so that the mapping data changes according to the controlled range. When the preset rollback conditions are met, the model update module rolls back the target model item to the model item corresponding to the model version information before the update. After the rollback, the model update module enters a frozen state and starts the freeze window timer. During the freeze window, updating the mapping data is prohibited, and the rolled-back model item is set as the priority candidate model item for the model selection module in the freeze window. This allows subsequent control to generate control instructions based on the rolled-back version first within the freeze window, thus forming a closed loop of version management that links update, rollback, and freeze.

[0033] Furthermore, the application rate feedback data is obtained through a flow measurement device; The model update module determines the deviation data based on the application rate feedback data and the target application rate data.

[0034] Specifically, the flow measurement device is installed on the water and fertilizer delivery pipeline corresponding to the actuator to output flow measurement values ​​during water and fertilizer operations. The model update module reads the flow measurement values ​​within each control cycle or at the end of the control cycle and forms the application rate feedback data corresponding to that control cycle based on the flow measurement values. By directly collecting application rate-related data by the flow measurement device, objective measurement basis can be provided for subsequent deviation calculations, enabling the model update module to obtain feedback input corresponding to the actual water and fertilizer application process. For example, in the water and fertilizer operation in valve zone A, the flow measurement device continuously outputs flow measurement values ​​within the control cycle, and the model update module summarizes the measurement values ​​of that cycle at the end of the cycle to form the application rate feedback data for valve zone A. After obtaining the application rate feedback data, the model update module reads the target application rate data corresponding to the same control cycle and compares the application rate feedback data with the target application rate data to determine the deviation data for that control cycle. The deviation data can be represented as the difference between the two or the absolute value of the difference, and is stored in association with the corresponding work object identifier and control cycle identifier for subsequent update condition determination, mapping data update, and rollback determination. By determining the deviation data, the deviation between the actual application and the target application can be quantified, providing a unified input for subsequent update amplitude limits, verification window determination, and other processing. For example, when the application rate feedback data of valve zone A in that cycle is greater than the target application rate data, the model update module generates the corresponding deviation data and records it in the records of valve zone A and that control cycle for subsequent update and rollback processes.

[0035] Furthermore, the preset update condition is that the deviation data continuously exceeds the preset deviation threshold for a preset number of times.

[0036] Specifically, after obtaining the corresponding deviation data in each control cycle, the model update module compares the deviation data with a preset deviation threshold and maintains a continuous over-threshold count. When the deviation data in a certain control cycle exceeds the preset deviation threshold, the continuous over-threshold count is incremented by one. When the deviation data in a certain control cycle does not exceed the preset deviation threshold, the continuous over-threshold count is cleared or reset according to preset rules. When the continuous over-threshold count reaches a preset number, the model update module determines that the preset update conditions are met and triggers the update of the mapping data of the target model item. By adopting the update triggering method based on the number of consecutive over-thresholds, frequent updates caused by single occasional deviations can be avoided, making the update triggering more stable and more in line with continuous deviation scenarios. For example, when the deviation data in valve zone A exceeds the preset deviation threshold in multiple consecutive control cycles, the continuous over-threshold count is gradually accumulated. When it is accumulated to a preset number, the model update module triggers the update of the mapping data of the target model item corresponding to the current operating condition interval of valve zone A.

[0037] Furthermore, the update range limitations include: The amount of data updated in a single mapping operation shall not exceed a preset limit; Furthermore, updating the mapping data is prohibited when the data quality index of the application rate feedback data is lower than the preset quality threshold; Among them, the data quality indicators are determined based on the packet loss ratio, outlier ratio, and time alignment deviation. The packet loss ratio, outlier ratio, and time alignment deviation are obtained by the model update module based on the collected data.

[0038] Specifically, in one embodiment, when the preset update conditions are met and updates are allowed, the model update module first forms candidate update amounts for the mapped data based on the deviation data, and then applies an update magnitude limit to the single update amount, ensuring that the single update amount for the mapped data does not exceed a preset upper limit. Let the value of a certain mapped data item before the update be... The updated candidate value is The preset upper limit is Then the updated value will be retrieved. Determine using the following formula: ; in, Indicates will Limited to the range Within this framework, the mapping data can be gradually corrected over multiple control cycles, avoiding abrupt changes in the mapping relationship caused by a single update. For example, if valve zone A obtains a large number of candidate updates in a single update of the corresponding entry in a certain working condition range, the model update module will trim it to no more than the preset upper limit before writing back the mapping data of that entry. Before updating the mapping data, the model update module calculates the data quality index of the application rate feedback data and compares it with a preset quality threshold. If the data quality index is lower than the preset quality threshold, updating the mapping data is prohibited, and the current mapping data remains unchanged. For ease of implementation, the data quality index... Determined based on packet loss ratio, outlier ratio, and time alignment deviation, for example, by determining the minimum value of the three quality components: ; and with As a condition for prohibiting updates, among which To preset the quality threshold, The quality component is obtained from the packet loss ratio. The mass component is obtained from the proportion of outliers. For the mass component derived from time alignment deviation, introducing a mass gating mechanism before updating can prevent updating the mapping data when the feedback data is unreliable, thereby enhancing the stability of the online update process. For example, when there is significant packet loss and increased alignment deviation in the feedback data of valve zone A during a certain period, the model update module calculates... If the data is below the threshold, skip this mapping data update; The model update module calculates the packet loss rate, outlier rate, and time alignment deviation based on the collected data within a statistical window, and determines data quality indicators accordingly. Let the expected number of data points within the statistical window be... The actual number of data points received is The packet loss ratio Statistics can be calculated using the following formula: ; Let the number of data points identified as outliers in the statistics window be . The proportion of outliers Statistics can be calculated using the following formula: ; Time alignment deviation Used to characterize the degree of deviation between the acquisition timestamp and the control cycle time reference, for example, let the first... The collection time for each receiving data point is The alignment time reference for its control cycle is Then, the following formula can be used for statistics: ; In obtaining 、 and Then, the model update module constructs accordingly. , , And further obtain data quality indicators Through the above statistics, the model update module can quantify the reliability of feedback data using three types of information: packet loss, anomalies, and time alignment. This provides a basis for whether to allow the updating of mapping data. For example, within a certain statistical window of threshold zone A, if the number of received points is lower than the expected number and the alignment deviation increases, a higher packet loss ratio and a larger time alignment deviation will be obtained, thereby reducing the data quality index and triggering a prohibition on updates.

[0039] Furthermore, the preset rollback conditions include: Within the verification window after parameter updates, the deviation data did not meet the preset improvement criteria; The freeze window is a fixed-duration window, and the freeze will be lifted after the freeze window expires. Within the frozen window, the model selection module prioritizes selecting the rolled-back model entries when the current operating condition range remains unchanged.

[0040] Specifically, after completing the mapping data update, the model update module starts the verification window and records the corresponding deviation data within the verification window according to the control cycle. When the verification window ends, the model update module judges the deviation data within the verification window based on preset improvement criteria. If the preset improvement criteria are not met, a rollback operation is triggered, restoring the target model entry to the model entry corresponding to the model version information before the update. For ease of implementation, the preset improvement criteria can be set to the average absolute value of the deviation within the verification window not being greater than a preset threshold. The judgment of not meeting the improvement criteria can be expressed as follows: ; in, This indicates the number of control cycles contained in the verification window. Indicates the first Deviation data for each control cycle This indicates the preset threshold corresponding to the improvement criterion. By introducing a verification window and improvement criteria, the update effect can be constrained after the update, avoiding the continued use of the mapping data after the update. For example, if the deviation in the verification window of valve zone A continues to be large and exceeds the threshold after a mapping data update, the model update module will trigger a rollback and restore the entry to the version before the update. The model update module enters a frozen state after performing a rollback and starts a freeze window timer. The freeze window is a fixed-duration window during which updates to the mapping data are prohibited. After the freeze window expires, the frozen state is lifted, and the module returns to an updatable state. For ease of implementation, the frozen state can be calculated from the rollback time and maintained for a fixed duration, for example, it can be represented as: ; in, Indicates the time when the rollback occurred or the time when the rollback was completed. Indicates the fixed duration of the frozen window. This indicates the frozen state. By using a fixed-duration freeze window, a stable operating range can be provided after rollback, avoiding repeated updates in a short period of time that cause fluctuations in the model version. For example, after valve zone A is rolled back, it enters a frozen state. During the duration of the freeze window, the mapping data will not be updated. Only after the freeze window ends will it be allowed to re-enter the update judgment process. When the model update module is in a frozen state, the model selection module still determines the current operating condition range based on the operating condition parameter data. When the current operating condition range is consistent with the previous control cycle, the model selection module outputs the rolled-back model entries as priority candidate model entries within the frozen window, so that the control command generation module can continue to generate control commands. When the operating condition range changes, the model selection module determines the target model entry from the set of candidate model entries corresponding to the new operating condition range. For ease of implementation, the freeze priority selection rule can be expressed as follows: ; in, Indicates the current operating condition range identifier. This indicates the operating condition interval identifier of the previous control cycle. This represents the target model entry currently being output. This indicates the model entries after rollback. Through the freeze window priority selection mechanism, the validated rollback version can be prioritized when the operating conditions are stable after rollback, ensuring consistent control. For example, if valve zone A's operating range remains unchanged within the freeze window, the model selection module continuously prioritizes outputting the rolled-back model entries. When the operating range changes, the corresponding entries are selected and output according to the new range.

[0041] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart agricultural production system, characterized in that, include: The operation parameter generation module generates target application rate data for water and fertilizer operations and collects operating condition parameter data related to the water and fertilizer operations. The actuator model library module stores model entries corresponding to different operating conditions for each actuator related to water and fertilizer operations. The model entries include operating condition information, mapping data between control quantities and application quantities, and model version information. The model selection module selects target model entries based on the operating condition parameter data and the operating condition interval information; The control instruction generation module generates control instruction data based on the mapping data between the target application rate data and the target model entries; The execution control module drives and controls the actuator to perform water and fertilizer operations according to the control command data; The model update module determines the deviation data based on the application rate feedback data and updates the mapping data of the target model entry when the preset update conditions are met; the model update module meets the update range limit during the update process and reverts to the model entry corresponding to the model version information before the update when the preset rollback conditions are met. The model update module enters a frozen state after performing a rollback, prohibiting the updating of the mapping data within the frozen window, and setting the rolled-back model entries as the priority candidate model entries of the model selection module within the frozen window.

2. The intelligent agricultural production system according to claim 1, characterized in that, The generation of the job parameters includes: Obtain the task object identifier and task time information, and determine the target application rate benchmark value based on the task object identifier and task time information; Collect the measured values ​​of the working condition parameters corresponding to the work object identifier, and perform time alignment processing on the measured values ​​of the working condition parameters to form the working condition parameter data; The target application rate data is determined based on the target application rate baseline value and the operating condition parameter data.

3. The intelligent agricultural production system according to claim 1, characterized in that, The actuator model library module stores model entries including: Multiple operating condition intervals are set based on the value range of the operating condition parameter data, and corresponding operating condition interval information is generated; Model entries are established for the same actuator in different operating conditions, and the operating condition information of each model entry is associated with the mapping data between the control quantity and the applied quantity and stored accordingly. Set model version information for each model entry and establish the association between the model version information and the corresponding mapping data.

4. The intelligent agricultural production system according to claim 1, characterized in that, The model selection module includes: Based on the operating condition parameter data, determine the operating condition range into which the current operating condition parameter falls, and determine the set of candidate model entries accordingly; When the current operating condition range changes, it is determined whether to switch the target model entry according to the anti-shake switching rules; The anti-shake switching rules include the working condition interval switching hysteresis threshold rule and the minimum dwell time rule; The target model entry is determined from the set of candidate model entries.

5. The intelligent agricultural production system according to claim 1, characterized in that, The mapping data for the model entries includes lookup table mapping data; The model entries include minimum effective control parameters, response delay parameters, response inertia parameters, and effective range information of the mapping data; The effective range information is used to limit the range of values ​​for which the mapping relationship between the control quantity and the application quantity applies.

6. The intelligent agricultural production system according to claim 1, characterized in that, The generation of the control commands includes: Retrieve the control quantity corresponding to the target application rate data from the mapping data of the target model entry; When the target application rate data is located between adjacent retrieval points, the control quantity is determined according to a preset interpolation rule; When the target application rate data exceeds the effective range information of the mapping data, the control amount is determined according to the preset limit rule.

7. The intelligent agricultural production system according to claim 1, characterized in that, The application rate feedback data is obtained through a flow measurement device; The model update module determines the deviation data based on the application rate feedback data and the target application rate data.

8. The intelligent agricultural production system according to claim 1, characterized in that, The preset update condition is that the deviation data continuously exceeds the preset deviation threshold for a preset number of times.

9. A smart agricultural production system according to claim 1, characterized in that, The update range limitation includes: The amount of data updated in a single instance shall not exceed a preset upper limit; Furthermore, updating the mapping data is prohibited when the data quality index of the application rate feedback data is lower than a preset quality threshold; The data quality indicators are determined based on the packet loss ratio, outlier ratio, and time alignment deviation, which are obtained by the model update module from the collected data.

10. A smart agricultural production system according to claim 1, characterized in that, The preset rollback conditions include: Within the verification window after parameter updates, the deviation data did not meet the preset improvement criteria; The freeze window is a fixed duration window, and the freeze state is lifted after the freeze window expires. Within the frozen window, the model selection module prioritizes selecting the rolled-back model entry when the current operating condition range remains unchanged.

Images