Artificial intelligence-based water and fertilizer integrated intelligent control system

Abstract

The application discloses an intelligent control system for water and fertilizer integration based on artificial intelligence and belongs to the technical field of intelligent control of agricultural irrigation, and comprises: an upper partition alternation irrigation decision layer and a lower pipe network execution correction layer; the decision layer generates irrigation area switching timing and water and fertilizer target quantity dynamically according to crop water stress prediction; the execution layer converts the target into pump valve parameters and corrects execution based on pressure, flow and concentration feedback; and post-irrigation crop state feedback is used to update the prediction model of the decision layer, thereby forming a double-layer closed-loop self-learning architecture. Through dynamic dry and wet stage division based on crop water stress prediction and water and fertilizer allocation consistency control based on pipe network hydraulic feedback, the application achieves accurate decision of irrigation timing and irrigation quantity of each partition and stable execution of water and fertilizer allocation, and solves the problems that irrigation parameters are difficult to be determined in real time according to dynamic data in the prior art and that control targets are difficult to be stably converted due to pipe network fluctuation.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for agricultural irrigation, and in particular to an intelligent control system for integrated water and fertilizer management based on artificial intelligence. Background Technology

[0002] With the increasing scarcity of agricultural water resources and the growing problem of non-point source pollution from chemical fertilizers, integrated water and fertilizer management technology has become an important means to improve water and fertilizer use efficiency and achieve sustainable agricultural development. The integrated water and fertilizer management intelligent control system collects real-time data on soil, crops, and pipeline conditions through sensors, dynamically generates irrigation and fertilization decisions based on intelligent algorithms, and automatically adjusts pump valve parameters according to pipeline hydraulic feedback to achieve precise water and fertilizer distribution to each irrigation zone. Especially in large-scale alternating irrigation scenarios, the system needs to dynamically switch irrigation zones according to the crop's water stress state, while suppressing fluctuations in pipeline pressure and concentration caused by valve opening and closing and pumping status changes, ensuring that the actual irrigation and fertilization amounts received by each zone are consistent with the decision objectives.

[0003] The existing integrated water and fertilizer alternating irrigation system uses a liquid supply pump to distribute water and fertilizer to each irrigation area through the main pipeline and branch pipelines, or uses an intermittent drainage structure to discharge fertilizer solution through each channel and open each branch pipeline in sequence to achieve zoned irrigation and fertilization.

[0004] For example, Chinese invention patent CN118235592B discloses a fertilization device for large-scale pasture planting, comprising: a main frame of the device; fixed shafts rotatably connected to the front and rear sides of the bottom of the main frame of the device, with wheels rotatably connected to the outside of the fixed shafts; auxiliary support columns hinged to the middle of the left and right sides of the bottom of the main frame of the device; a main liquid tank installed on the rear side of the top of the main frame of the device; a secondary liquid tank installed on the front side of the top of the main frame of the device; a liquid supply pump connected to the middle of the front wall of the main liquid tank; a main pipeline connected to the front of the liquid supply pump; quick-connect components connected to the left and right ends of the bottom of the main pipeline; and horizontal spray pipes and spray heads inserted into the left and right ends of the bottom of the main pipeline.

[0005] For example, Chinese invention patent CN120858726B discloses an array-type multi-channel intelligent precision fertilizer and its usage method, which includes: sequentially discharging equal amounts of fertilizer from a connected water pipe and opening the drip irrigation pipes sequentially to reduce the possibility of fertilizer settling in the drip irrigation pipes and clogging the nozzles.

[0006] However, in the process of implementing the inventive technical solution in the embodiments of this application, it was found that the above-mentioned technology has at least the following technical problems: In the process of alternating irrigation in different zones, existing integrated water and fertilizer irrigation systems usually use fixed rotation cycles or preset thresholds to control the irrigation sequence and amount of water in each zone. On the one hand, this cannot adapt to real-time changes in soil moisture conditions, crop water requirements and weather conditions, which can easily lead to excessive water stress or over-irrigation in some root zones.

[0007] On the other hand, when pressurized pipelines are used to distribute water and fertilizer and irrigation zones are frequently switched, the pipeline pressure, branch flow rate and fertilizer concentration are prone to fluctuations due to valve opening and closing, changes in pumping status and sudden changes in branch flow resistance. This makes it difficult to stably and accurately convert the alternating irrigation control targets generated by the system into the actual irrigation and fertilizer amounts obtained by each zone. Summary of the Invention

[0008] To address the technical problems of delayed decision-making and large execution deviations in existing water and fertilizer irrigation technologies, this invention provides an intelligent water and fertilizer integration control system based on artificial intelligence. The technical solution is as follows: An AI-based intelligent water and fertilizer integration control system is provided, which includes: The data synchronization acquisition and validity processing module is used to collect soil water potential, soil moisture content, crop water stress index, weather forecast data, main pipe pressure, branch pipe flow and fertilizer concentration in each zone, read the historical irrigation response records written back in the previous round, and generate the current status record, available historical irrigation response records and the first round prediction zone marker.

[0009] The zonal irrigation response prediction module is used to generate future water stress prediction results and post-irrigation response prediction results for different irrigation volumes based on current status records, available historical irrigation response records, first-round prediction zonal markers, and preset initial prediction model parameters.

[0010] The irrigation trigger determination and dry / wet phase division module is used to determine the remaining duration of the dry phase in each zone and the set of zones for which irrigation execution targets need to be generated in this round, based on the current status record, the future water stress prediction results under non-irrigation conditions, the preset allowable water stress upper limit, and the preset early warning margin.

[0011] The partition switching timing and water and fertilizer target generation module is used to generate the partition set of irrigation execution targets, the remaining duration of the dry phase and the post-irrigation response prediction results according to the needs of this round, and to generate the partition switching timing, target irrigation volume, initial duration of the wet phase and the set of executable partitions for this round.

[0012] The pump valve initial execution parameter conversion module is used to generate the current irrigation zone number, target branch flow rate, target fertilizer concentration, initial pump speed and initial proportional fertilizer injection valve opening based on the zone switching sequence, the set of executable zones in this round, the target irrigation volume, the target fertilizer application volume and the initial duration of the wet phase.

[0013] The pipeline feedback real-time correction execution module is used to perform feedback correction based on the target branch pipe flow rate and target fertilizer solution concentration, combined with the real-time collected main pipe pressure, branch pipe flow rate and fertilizer solution concentration, and outputs the corrected water pump speed, proportional fertilizer injection valve opening, actual irrigation volume, actual fertilizer application volume and remaining fertilizer application volume.

[0014] The zonal irrigation end determination and switching module is used to determine whether the current zonal irrigation has ended based on the actual irrigation volume, actual fertilization volume, remaining fertilization volume, and zonal switching sequence, and to generate the next execution zonal number, the actual irrigation volume and fertilization volume obtained in the current zonal, and the water and fertilizer deviation record.

[0015] The historical data update and cycle closure module is used to write the current irrigation data into the historical irrigation response record, generate training samples for the next round of model update, and trigger the next round of decision loop after the current round is completed.

[0016] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: The first step of this invention, based on artificial intelligence, is to provide an integrated water and fertilizer intelligent control system. This system collects multi-source data through a data synchronization and effectiveness processing module, generating initial prediction zone markers. A zone irrigation response prediction module then outputs future water stress predictions and post-irrigation response predictions. Subsequently, an irrigation trigger determination and dry / wet stage division module determines the remaining duration of the dry stage and the set of irrigation zones. A zone switching sequence and water / fertilizer target generation module generates the switching sequence and target amounts of water and fertilizer. A pump valve initial execution parameter conversion module converts the targets into pump valve execution parameters. A pipeline feedback real-time correction execution module corrects the execution parameters based on real-time hydraulic feedback. A zone irrigation end determination and switching module determines the zone switching and records deviations. A historical data update and cycle closed-loop module feeds back post-irrigation data to the next round of model training. This forms a complete closed loop from perception, decision-making, execution to self-learning optimization, enabling dynamic matching of irrigation sequences in each zone with the actual water requirements of the crop. Water and fertilizer allocation remains stable and accurate even under pipeline fluctuations, effectively solving the problems of lagging irrigation decisions and large execution deviations in existing technologies.

[0017] In the second step, the present invention dynamically determines the remaining duration of the dry stage in each zone based on the current crop water stress index and future prediction results through the irrigation trigger determination and dry / wet stage division module. This ensures that the timing of irrigation switching matches the actual water demand of the crop, thereby reducing water waste caused by premature irrigation and crop growth inhibition caused by late irrigation. At the same time, by comparing the difference at the end of the prediction window with the early warning margin, the zone with the nearest water demand is included in the candidate waiting set in advance. This enables the system to respond to potential water stress demands in a timely manner after the current cycle is completed, effectively improving irrigation timeliness and water resource utilization efficiency.

[0018] The third step involves using a real-time correction module based on pipeline feedback to determine deviations in branch flow rate and fertilizer concentration over multiple consecutive control cycles. This avoids erroneous adjustments caused by instantaneous fluctuations. Furthermore, the module compensates for the delay in fertilizer concentration based on pipeline volume and corrects the opening of the proportional fertilizer injection valve. This ensures that water and fertilizer distribution can accurately track the target value even under pipeline transmission delays. Consequently, it reduces the impact of hydraulic fluctuations caused by changes in zone valve switching and pumping status on irrigation uniformity, thereby improving the consistency and control accuracy of actual irrigation and fertilizer application in each zone. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This application provides a structural diagram of an AI-based intelligent water and fertilizer integration control system. Figure 2 This is a diagram of the irrigation trigger determination and dry / wet stage division module provided in the embodiments of this application; Figure 3 The flowchart for real-time correction of pipeline feedback provided in this application embodiment; Figure 4 The flowchart for partition switching timing and water and fertilizer target generation provided in the embodiments of this application is shown. Detailed Implementation

[0021] The technical solution provided in this application will now be described with reference to the accompanying drawings.

[0022] This scheme comprises an upper-level zoned alternating irrigation decision layer and a lower-level pipeline network execution correction layer. The decision layer is responsible for generating irrigation district switching sequences and water and fertilizer target amounts based on crop water stress predictions; the execution layer is responsible for converting the targets into pump and valve parameters, and correcting the execution based on pressure, flow, and concentration feedback, and calculating the actual water and fertilizer delivery amount; post-irrigation crop status feedback is then used to update the decision layer prediction model.

[0023] like Figure 1 The diagram shown is a structural diagram of the AI-based integrated water and fertilizer intelligent control system provided in this application embodiment. It includes: a data synchronization acquisition and effectiveness processing module, a zoned irrigation response prediction module, an irrigation trigger determination and dry / wet stage division module, a zone switching sequence and water and fertilizer target generation module, a pump and valve initial execution parameter conversion module, a pipeline network feedback real-time correction execution module, a zoned irrigation end determination and switching module, and a historical data update and cycle closed-loop module. The specific functions of each module are as follows: The data synchronization acquisition and validity processing module is used to perform data synchronization acquisition and validity processing.

[0024] Each irrigation zone is numbered according to the pre-defined irrigation unit based on the distribution of plots, pipeline layout, and crop planting structure of the irrigation area. Each irrigation zone has a unique number and is associated with the soil water potential sensor, soil moisture sensor, canopy temperature sensor, and corresponding branch valves and branch flow meters within that zone.

[0025] First, the preset irrigation zone numbers are obtained. The current soil water potential, current soil moisture content, and current crop water stress index for each zone are collected. Soil water potential is collected by root zone soil water potential sensors, and soil moisture content is collected by soil moisture sensors. Based on the temperature difference between canopy temperature and air temperature, relative humidity, solar radiation intensity, and wind speed, the canopy-air temperature difference correction value is calculated using an empirical formula for the crop water stress index. This empirical formula is a regression model or empirical equation established based on the aforementioned variables. The calculated canopy-air temperature difference correction value is normalized to a preset numerical range, and the normalized value is used as the crop water stress index. Simultaneously, weather forecast data is obtained through a weather forecast interface. The duration of the weather forecast data is the future prediction duration, which is a pre-set time length based on the available weather forecast time and the time range required for crop water stress prediction. The data includes temperature, humidity, solar radiation, and wind speed at each forecast time within this duration. The acquired weather forecast data is arranged in chronological order to form a weather forecast sequence. In addition, the system reads the updated historical irrigation response records of each zone from the previous round of historical data updates and writes them back to the historical irrigation response database from the round closed-loop module. It also collects the main pipe pressure, the flow rate of each branch pipe, and the fertilizer solution concentration. The main pipe pressure is collected by the main pipe pressure sensor at the pump outlet, the branch pipe flow rate is collected by the flow meter of each branch pipe, and the fertilizer solution concentration is collected by the online concentration sensor at the fertilizer injection pipe outlet.

[0026] After data collection is complete, perform the following processing steps: The first step is to establish a correspondence between the current soil water potential, current soil moisture content, and current crop water stress index of each irrigation zone at the current sampling time and the meteorological forecast sequence within the future prediction period starting from the sampling time, using the irrigation zone number as an index, and generate the current status record of each zone. Each current status record includes the zone number, soil water potential, soil moisture content, crop water stress index and the corresponding meteorological forecast sequence.

[0027] The second step is to align the main pipe pressure, branch pipe flow rate, and fertilizer solution concentration at the same sampling time.

[0028] The third step involves extracting records for each irrigation zone from the historical irrigation response records of each zone. If records are extracted, they are cleaned by deleting records that lack actual irrigation volume, actual fertilizer application, post-irrigation soil water potential, post-irrigation soil moisture content, or post-irrigation crop water stress index. The remaining valid records after cleaning are arranged in chronological order as the available historical irrigation response records for that zone. If no valid records are found after cleaning, the zone number is added to the first-round prediction zone label set, and the available historical irrigation response records for that zone are set to empty.

[0029] If no records are retrieved for the partition, the partition number is directly added to the first-round prediction partition tag set, and the available historical irrigation response records for the partition are set to empty.

[0030] After the above processing, the current status record of each partition, the available historical irrigation response record of each partition, and the first round prediction partition mark are finally obtained for use by the subsequent partition irrigation response prediction module.

[0031] The zoned irrigation response prediction module is used to generate zoned irrigation response prediction results.

[0032] The module receives data from the synchronous acquisition and validity processing module, which generates current status records for each partition, available historical irrigation response records for each partition, first-round prediction partition markers, and preset initial prediction model parameters. These preset initial prediction model parameters are either pre-trained using historical general data or obtained through random initialization. This module also uses preset irrigation step size, fertilizer step size, and preset sample threshold. The irrigation step size and fertilizer step size are set according to the required prediction accuracy, and the preset sample threshold is a positive integer set based on model complexity and the minimum number of samples required for training convergence.

[0033] The maximum allowable irrigation amount is determined based on the zone area, root depth, field capacity, and current soil moisture content. The maximum allowable irrigation amount is the volume value obtained by multiplying the difference between field capacity and current soil moisture content by the root depth and then by the zone area. A candidate irrigation amount set is generated based on the maximum allowable irrigation amount and a preset irrigation amount step size, starting from zero and increasing by the irrigation amount step size until the maximum allowable irrigation amount is reached. Zero and the irrigation amount value corresponding to each increment step are used as elements in the candidate irrigation amount set. The maximum reference fertilizer amount is determined based on the target fertilizer-water ratio and the maximum allowable irrigation amount. A candidate fertilizer amount set is generated based on the maximum reference fertilizer amount and a preset fertilizer amount step size, starting from zero and increasing by the fertilizer amount step size until the maximum reference fertilizer amount is reached. Zero and the fertilizer amount value corresponding to each increment step are used as elements in the candidate fertilizer amount set.

[0034] For each irrigation zone, the first-round prediction zone marker is queried using the irrigation zone number, and processed according to the following cases: Scenario 1: The irrigation zone number was not marked by the first round of prediction zones, indicating that there are available historical irrigation response records for this zone.

[0035] At this point, the irrigation response prediction model for this partition is updated using the available historical irrigation response records for that partition, including pre-irrigation soil water potential, pre-irrigation soil moisture content, pre-irrigation crop water stress index, actual irrigation amount, actual fertilizer application amount, post-irrigation soil water potential, post-irrigation soil moisture content, post-irrigation crop water stress index, and corresponding meteorological data. This irrigation response prediction model uses any one of the following: a temporal neural network model, a gradient boosting tree model, or a random forest regression model. Its inputs are pre-irrigation soil water potential, pre-irrigation soil moisture content, pre-irrigation crop water stress index, future meteorological sequence, candidate irrigation amounts, and candidate fertilizer applications. The outputs are post-irrigation soil water potential, post-irrigation soil moisture content, and post-irrigation crop water stress index. During model training, irrigation amount is used to predict post-irrigation soil water potential and soil moisture content, while fertilizer application amount is used to predict fertilizer concentration, nutrient application per unit area, and concentration safety status. When soil conductivity or fertilizer conductivity is collected, it is used as a correction factor for the crop water stress index. When the number of valid historical samples in a certain partition is less than the preset sample threshold, a shared prediction model trained with historical data of the same crop type, soil type or pipeline area is invoked; after the number of valid historical samples in the partition reaches the preset sample threshold, the shared prediction model is fine-tuned based on the partition's own data to obtain the irrigation response prediction model for that partition.

[0036] After the model is updated, the weather forecast sequence in the current state record of the region, as well as each combination of candidate irrigation and candidate fertilizer amounts from the candidate irrigation amount set and candidate fertilizer amount set, are sequentially input into the updated irrigation response prediction model to obtain the post-irrigation response prediction results for the region under different irrigation and fertilizer amounts. Simultaneously, the model output when the candidate irrigation amount is set to zero is used as the future water stress prediction result for the region under non-irrigation conditions. This result includes soil water potential, soil moisture content, and crop water stress index at each prediction time within the future prediction period.

[0037] Scenario 2: The irrigation zone number is marked by the first round of prediction, indicating that there are no available historical irrigation response records for this zone. In this case, the initial irrigation response prediction model for this zone is directly constructed using the preset initial prediction model parameters. Each set of candidate irrigation and fertilizer amounts from the current state record of this zone, along with the candidate irrigation amount and candidate fertilizer amount sets, is combined and sequentially input into the initial irrigation response prediction model to obtain the post-irrigation response prediction results for this zone under different irrigation and fertilizer amount conditions. Simultaneously, the model output when the candidate irrigation amount is zero is used as the future water stress prediction result for this zone under non-irrigation conditions.

[0038] Thus, the prediction results of future water stress under non-irrigation conditions in each zone, as well as the prediction results of post-irrigation response in each zone under different irrigation and fertilization conditions, are obtained for use by the irrigation trigger determination and dry / wet stage division module and the zone switching sequence and water and fertilizer target generation module.

[0039] like Figure 2 The diagram shown is a module diagram of irrigation trigger determination and dry / wet stage division provided in the embodiment of this application. It shows the process by which the module determines the remaining duration of the dry stage and the irrigation trigger reference state of each zone according to three scenarios based on the current crop water stress index and future prediction results.

[0040] The irrigation trigger determination and dry / wet stage division module is used to receive the current status records of each zone generated by the data synchronization acquisition and validity processing module, the future water stress prediction results of each zone under non-irrigation conditions generated by the zone irrigation response prediction module, as well as the preset allowable water stress upper limit and preset warning margin. The allowable water stress upper limit is the critical value of the water stress index determined according to the crop type and the current growth stage, and the preset warning margin is a certain proportion based on the allowable water stress upper limit.

[0041] For each irrigation zone, firstly, the current crop water stress index, current soil water potential, and current soil moisture content are read from the current state record of that zone. Secondly, the crop water stress index at each prediction time is read sequentially from the future water stress prediction results under the non-irrigated condition of that zone. Then, the remaining dry phase duration and irrigation trigger reference state of that zone are determined according to the following three scenarios.

[0042] Scenario 1: The current crop water stress index is greater than or equal to the upper limit of allowable water stress. In this case, the remaining dry phase duration of the partition is recorded as zero, the current soil water potential and current soil moisture content are used as the reference state for triggering irrigation in this round for the partition, and the partition is added to the set of partitions that need to generate irrigation execution targets in this round.

[0043] Scenario 2: The current crop water stress index is less than the allowable water stress upper limit, and there is a predicted moment within the future prediction period when the predicted value will first reach the allowable water stress upper limit. In this case, the time interval between the current moment and the predicted moment when the allowable water stress upper limit is first reached is taken as the remaining dry phase duration of the partition, and the soil water potential and soil moisture content corresponding to the predicted moment are taken as the reference state for triggering irrigation in this round for the partition. The partition is then added to the set of partitions that need to generate irrigation execution targets in this round.

[0044] Scenario 3: The current crop water stress index is less than the allowable water stress upper limit, and the crop water stress index at all prediction times within the future prediction period does not reach the allowable water stress upper limit. In this case, this partition will not be added to the set of partitions that need to generate irrigation execution targets in this round, and no irrigation trigger reference state will be generated. The difference between the allowable water stress upper limit and the crop water stress index at the end of the prediction window will be further calculated: if the difference is less than the preset warning margin, which is a certain proportion based on the allowable water stress upper limit, then this partition will be included in the candidate waiting set, and will be recalculated first after all executable partitions in the current round have been irrigated; if the difference is greater than or equal to the preset warning margin, then this partition will not be processed in this round. The current state record of this partition will be retained, and will be used by the data synchronization acquisition and validity processing module to re-acquire data in the next round and re-enter the irrigation trigger determination and dry / wet stage division module for judgment.

[0045] After the above processing, the reference state for triggering irrigation in this round of each zone, the remaining time of the dry phase in each zone, and the set of zones for which irrigation execution targets need to be generated in this round are obtained, which are used by the zone switching timing and water and fertilizer target generation module.

[0046] like Figure 4 The diagram shown is a flowchart of the partition switching timing and water and fertilizer target generation provided in the embodiment of this application. It shows the process of generating partition switching timing by sorting the remaining dry stage time and crop water stress index, calculating the expected maintenance time and screening candidate irrigation amounts to determine the target irrigation amount, and determining the initial branch flow and the initial wet stage time.

[0047] The partition switching timing and water and fertilizer target generation module is used to generate the partition switching timing, the initial duration of the wet phase, and the target amount of water and fertilizer.

[0048] The system receives the set of zones for the current irrigation execution target generated by the irrigation trigger determination and dry / wet stage division module, the remaining duration of the dry stage in each zone, the reference status of the current irrigation trigger in each zone, and the post-irrigation response prediction results for each zone generated by the zone irrigation response prediction module. Simultaneously, it uses preset upper limits for field water holding capacity, minimum wet stage duration requirements, upper limits for fertilizer concentration, upper limits for fertilizer application per unit area, crop growth period fertilization plans, target fertilizer-to-water ratio, designed rated branch pipe flow, and the same preset value for the allowable water stress upper limit used in the irrigation trigger determination and dry / wet stage division module. The upper limit of field water holding capacity is the maximum soil moisture content determined based on soil type and root depth. The minimum wet stage duration requirement is the shortest duration of the wet stage determined based on crop root water absorption and irrigation system. The upper limit of fertilizer concentration is the maximum safe concentration determined based on fertilizer type and crop tolerance. The upper limit of fertilizer application per unit area is the maximum single application per unit area determined based on crop nutrient requirements, soil fertility, and fertilization plan. The crop growth period fertilization plan is the total fertilizer application and nutrient ratio preset for each stage of crop growth. The target fertilizer-water ratio is the ratio of the target fertilizer application to the target irrigation water amount determined based on crop nutrient requirements and irrigation plan.

[0049] First, the partition switching sequence for this round is generated. Each partition is selected from the set of partitions for which irrigation execution targets need to be generated in this round, and then sorted in ascending order of the remaining time for each dry stage. If several stages have the same remaining time, the current crop water stress index for each partition is read from the current status record generated by the data synchronization and validity processing module, and then sorted in descending order of the current crop water stress index. This yields the partition switching sequence for this round.

[0050] Secondly, the target irrigation amount and target fertilizer amount for each partition in the partition switching time sequence are determined. For each partition in the partition switching time sequence, the post-irrigation response prediction results generated by the partition irrigation response prediction module are queried, and candidate irrigation amounts that simultaneously meet the following conditions are selected: the post-irrigation crop water stress index is less than the upper limit of allowable water stress, the post-irrigation soil moisture content does not exceed the upper limit of field capacity, and the expected duration meets the minimum wet stage duration requirement. The expected duration is determined as follows: from the post-irrigation response prediction results of the partition, the crop water stress index at each prediction time after irrigation is read in chronological order. The time corresponding to the prediction time that is equal to or first exceeds the upper limit of allowable water stress is subtracted from the post-irrigation start time. The difference is the expected duration. The post-irrigation start time is the end time of irrigation for the partition, corresponding to the start time of the first prediction time after irrigation in the post-irrigation response prediction results. If the crop water stress index at all prediction times after irrigation is less than the upper limit of allowable water stress, then the difference between the end time of the prediction window and the post-irrigation start time is taken as the expected duration. Among the candidate irrigation amounts that meet the above conditions, the minimum irrigation amount is selected as the target irrigation amount for this zone. Then, the target irrigation amount is multiplied by the target fertilizer-water ratio to obtain the initial target fertilizer amount. After being corrected by the upper limit constraint of fertilizer concentration, the upper limit constraint of fertilizer amount per unit area, and the crop growth period fertilization plan constraint, the target fertilizer amount for this zone is determined.

[0051] Next, determine the initial wet phase duration for each partition in the partition switching sequence. For each partition in the partition switching sequence, the initial branch flow rate is determined as follows: If a valid record for that partition exists in the historical irrigation response records generated by the data synchronization acquisition and validity processing module, extract the branch flow rate record for the execution period from the most recent valid record. Sum the branch flow rates at each sampling time during that execution period and divide by the number of sampling times. The result is used as the initial branch flow rate for that partition. If a valid record for that partition does not exist in the historical irrigation response records generated by the data synchronization acquisition and validity processing module, directly use the design rated branch flow rate as the initial branch flow rate. The design rated branch flow rate is the rated branch flow rate value determined during the irrigation network design phase based on the partition area and irrigation requirements. In implementation, it can be obtained by multiplying the partition area by the design irrigation intensity and then dividing by the number of branches corresponding to that partition, or by directly using the branch design flow rate value from the network hydraulic calculation sheet. The above initial branch flow rate is corrected through network feedback in the subsequent network feedback real-time correction execution module. Divide the target irrigation volume of the partition by the initial branch flow rate to obtain the initial wet phase duration of the partition. In the above division operation, if the initial branch flow rate is zero, the initial wet phase duration of the partition cannot be calculated, and the partition will not be added to the set of executable partitions in this round.

[0052] Finally, the set of executable partitions for this round is generated. Partitions whose initial wet phase duration was successfully calculated in the partition switching sequence are added to the set of executable partitions for this round.

[0053] After the above processing, the following parameters are obtained: the zoning switching sequence, the target irrigation amount for each zoning zone, the target fertilizer amount for each zoning zone, the initial duration of the wet phase for each zoning zone, and the set of executable zoning zones for this round, which are then used by the pump valve initial execution parameter conversion module.

[0054] The pump and valve initial execution parameter conversion module converts the partition target into the pump and valve initial execution parameters.

[0055] The system receives the partition switching timing and the partition switching timing generated by the water and fertilizer target generation module, the set of executable partitions in this round, the target irrigation amount for each partition, the target fertilizer amount for each partition, and the initial duration of the wet phase for each partition. It also uses the preset pump speed range and proportional fertilizer valve opening range. The pump speed range is the interval from the lowest speed to the highest speed allowed by the pump frequency converter, and the proportional fertilizer valve opening range is the opening range corresponding to the fertilizer valve from fully closed to fully open.

[0056] The first step is to select the current execution partition. From the partition switching sequence, select the first irrigation partition that is simultaneously present in the set of executable partitions in this round, and use its number as the current execution irrigation partition number.

[0057] The second step is to obtain the target value of the currently executing partition. Based on the current executing irrigation partition number, extract the target irrigation amount of the partition from the target irrigation amount of each partition, extract the target fertilizer amount of the partition from the target fertilizer amount of each partition, and extract the initial wet phase duration of the partition from the initial wet phase duration of each partition.

[0058] The third step is to calculate the target branch flow rate and the target fertilizer concentration. If the target irrigation volume is greater than zero, the branch flow rate used in the partition switching sequence and the water and fertilizer target generation module to calculate the initial duration of the wet phase of the partition is taken as the target branch flow rate of the current partition, and the target fertilizer concentration is obtained by dividing the target fertilizer amount by the target irrigation volume. In the above division operation, the target fertilizer concentration cannot be calculated when the target irrigation volume is zero. At this time, the partition is marked as completed in this round, and the process moves to the next partition in the partition switching sequence. The process checks whether there is a next partition in the partition switching sequence: if there is a next partition, the process moves to that next partition and returns to the first step to reselect the current execution partition; if there is no next partition, it means that all executable partitions in this round have been processed, and the process jumps directly to the historical data update and round closure module to summarize the execution records of each partition in this round and start the decision for the next round.

[0059] The fourth step is to determine the initial pump speed. Based on the target branch flow rate and the pump characteristic curve, the initial pump speed is determined within the specified range. The pump characteristic curve is a curve or table showing the correspondence between the pump speed and the main outlet flow rate. The sum of the main outlet flow rate and the flow rates of each branch pipe must satisfy the pipe flow distribution relationship. If no pump characteristic curve is pre-stored, the median value of the pump speed range is selected as the default initial pump speed. The actual pump speed corresponding to the target branch flow rate is gradually approximated through measured feedback of the main outlet pressure and branch flow rate in the feedback correction loop of the real-time correction execution module.

[0060] Step 5: Determine the initial proportional fertilizer injection valve opening. Within the proportional fertilizer injection valve opening range, based on the valve characteristic curve or a preset opening-flow mapping relationship, select the opening corresponding to the target fertilizer concentration as the initial proportional fertilizer injection valve opening. The valve characteristic curve is a curve or table showing the correspondence between the proportional fertilizer injection valve opening and the outlet flow rate of the fertilizer injection pipe. The ratio of the outlet flow rate of the fertilizer injection pipe to the current branch pipe flow rate determines the fertilizer concentration. If neither the valve characteristic curve nor the opening-flow mapping relationship is pre-stored, the median value of the proportional fertilizer injection valve opening range is selected as the default initial proportional fertilizer injection valve opening. In the feedback correction loop of the real-time correction execution module, the actual opening corresponding to the target fertilizer concentration is gradually approximated by the measured feedback of the fertilizer concentration after delay compensation.

[0061] Step 6: Generate valve opening and closing commands. If a previous execution zone exists, first close the valves in the previous execution zone; then open the valves corresponding to the current irrigation zone number to generate the valve opening and closing commands for the current zone. Execution begins based on the initial pump speed and the initial proportional fertilizer valve opening.

[0062] After the above processing, the current irrigation zone number, target branch flow rate, target fertilizer concentration, initial pump speed, initial proportional fertilizer valve opening, and current zone valve opening and closing commands are obtained, which are then used by the pipeline network feedback real-time correction execution module.

[0063] like Figure 3 The diagram shown is a flowchart of the real-time correction process for pipeline feedback provided in this application embodiment. It illustrates the process by which the module collects the main pipe pressure, branch pipe flow rate, and fertilizer concentration in each control cycle, determines the correction of the pump speed and proportional fertilizer injection valve opening based on the deviation of the number of consecutive cycles (threshold number of cycles), and performs fertilizer concentration delay compensation and accumulation of actual irrigation and fertilizer application.

[0064] The pipeline feedback real-time correction execution module corrects execution parameters based on pipeline feedback.

[0065] The system receives the current irrigation zone number, target branch flow rate, target fertilizer concentration, target irrigation volume, target fertilizer application rate, initial pump speed, initial proportional fertilizer valve opening, and current zone valve opening / closing commands generated by the initial pump and valve execution parameter conversion module. Simultaneously, it uses preset parameters such as the pump speed range, proportional fertilizer valve opening range, control cycle, pressure limit, flow rate tolerance, concentration tolerance, pump speed adjustment step size, and proportional fertilizer valve opening adjustment step size, as well as a preset continuous cycle threshold. The control cycle is the time interval between system sampling and control, set according to sensor response speed and control accuracy requirements; the pressure limit is the network design safety pressure or the pump's maximum operating pressure; the flow rate tolerance and concentration tolerance are the allowable error ranges for branch flow rate and fertilizer concentration control, respectively, set according to control accuracy requirements; the pump speed adjustment step size and proportional fertilizer valve opening adjustment step size are the adjustment amounts for speed and opening during each feedback correction, set according to the actuator's adjustment accuracy and stability requirements; the continuous cycle threshold is an integer greater than 1, representing the number of cycles in which the deviation conditions are continuously met, used to avoid unnecessary adjustments caused by instantaneous fluctuations.

[0066] During execution, the main pipe pressure, current branch pipe flow rate, and fertilizer solution concentration are collected in real time. The main pipe pressure is collected by the main pipe pressure transmitter at the water pump outlet, the current branch pipe flow rate is collected by the electromagnetic flow meter of each branch pipe, and the fertilizer solution concentration is collected by the online conductivity sensor at the fertilizer injection pipe outlet.

[0067] First, initial settings are performed. The initial pump speed is used as the current pump speed before correction, and the initial proportional fertilizer valve opening is used as the current proportional fertilizer valve opening before correction. The valves corresponding to the current irrigation zone number are confirmed to be open according to the current zone valve opening / closing command. During the opening of the zone valves, the following correction steps are executed cyclically according to the control cycle. After each control cycle, the corrected pump speed and corrected proportional fertilizer valve opening are sent to the frequency converter and fertilizer valve controller. The next control cycle uses the corrected values ​​as the current values. The target fertilizer concentration is allowed to be updated by the zone irrigation end determination and switching module during the cycle. Each time the module returns from the zone irrigation end determination and switching module, the latest target fertilizer concentration value transmitted by the module is used.

[0068] Perform the following operations within each control cycle.

[0069] First, collect the main pipe pressure, current branch pipe flow rate, and fertilizer solution concentration for this control cycle.

[0070] Second, the current branch flow rate is compared and corrected with the target branch flow rate. If, within a consecutive control period of a threshold number of cycles, the current branch flow rate is consistently lower than the difference between the target branch flow rate and the allowable flow deviation, and the main pipe pressure is less than the upper pressure limit, the current pump speed is increased by one pump speed adjustment step to obtain the corrected pump speed. If, within a consecutive control period of a threshold number of cycles, the current branch flow rate is consistently lower than the difference between the target branch flow rate and the allowable flow deviation, but the main pipe pressure is greater than or equal to the upper pressure limit, the current pump speed remains unchanged, the corrected pump speed is equal to the current pump speed, and the irrigation volume continues to accumulate based on the current branch flow rate. The zonal irrigation end determination and switching module extends the execution time of the current zonal based on the actual accumulated irrigation volume. If, within a consecutive control period of a threshold number of cycles, the current branch flow rate is consistently higher than the sum of the target branch flow rate and the allowable flow deviation, the current pump speed is decreased by one pump speed adjustment step to obtain the corrected pump speed. When the current branch flow rate is within the allowable deviation range of the target branch flow rate plus or minus the flow rate, or when the deviation condition of the consecutive cycle number threshold number of control cycles has not been met, the current pump speed remains unchanged, and the corrected pump speed is equal to the current pump speed.

[0071] Third, the fertilizer concentration is compared and corrected after delay compensation. The concentration transmission delay is calculated based on the pipe volume from the fertilizer inlet to the current branch inlet and the current branch flow rate. The detected fertilizer concentration value is then compensated for the delay to obtain the compensated fertilizer concentration. Delay compensation is performed as follows: If the current branch flow rate is greater than zero, the pipe volume is divided by the current branch flow rate to obtain the transmission delay time; this transmission delay time is subtracted from the current time to obtain the delay matching time; from the fertilizer concentrations collected and recorded in previous control cycles, the fertilizer concentration recorded in the control cycle whose collection time is closest to this delay matching time is taken as the compensated fertilizer concentration for the current time. If this delay matching time is earlier than the start time of irrigation in this zone, the fertilizer concentration collected at the start time of irrigation in this zone is used as the compensated fertilizer concentration. If the current branch flow rate is zero, the delay compensation calculation is skipped, and the fertilizer concentration collected in this control cycle is directly used as the compensated fertilizer concentration. When the fertilizer concentration after delay compensation is lower than the target fertilizer concentration minus the allowable concentration deviation within a threshold number of consecutive control cycles, the current proportional fertilizer injection valve opening is increased by one proportional fertilizer injection valve opening adjustment step to obtain the corrected proportional fertilizer injection valve opening. When the fertilizer concentration after delay compensation is higher than the target fertilizer concentration plus the allowable concentration deviation within a threshold number of consecutive control cycles, the current proportional fertilizer injection valve opening is decreased by one proportional fertilizer injection valve opening adjustment step to obtain the corrected proportional fertilizer injection valve opening. When the fertilizer concentration after delay compensation is within the range of the target fertilizer concentration plus or minus the allowable concentration deviation, or when the deviation condition for the threshold number of consecutive control cycles has not yet been met, the current proportional fertilizer injection valve opening remains unchanged, and the corrected proportional fertilizer injection valve opening is equal to the current proportional fertilizer injection valve opening.

[0072] Fourth, boundary value constraints. If the corrected pump speed exceeds the upper limit of the pump speed range, the upper limit of the pump speed range is taken as the corrected pump speed; if the corrected pump speed is lower than the lower limit of the pump speed range, the lower limit of the pump speed range is taken as the corrected pump speed. If the corrected proportional fertilizer valve opening exceeds the upper limit of the proportional fertilizer valve opening range, the upper limit of the proportional fertilizer valve opening range is taken as the corrected proportional fertilizer valve opening; if the corrected proportional fertilizer valve opening is lower than the lower limit of the proportional fertilizer valve opening range, the lower limit of the proportional fertilizer valve opening range is taken as the corrected proportional fertilizer valve opening.

[0073] Fifth, accumulate the irrigation and fertilization amounts for this control cycle. The actual irrigation amount for the current zone is equal to the actual irrigation amount for the current zone in the previous cycle plus the current branch flow rate multiplied by the control cycle. The actual fertilization amount for the current zone is equal to the actual fertilization amount for the current zone in the previous cycle plus the current branch flow rate multiplied by the fertilizer concentration after delay compensation, and then multiplied by the control cycle. In the first cycle, both the actual irrigation and fertilization amounts for the current zone in the previous cycle are set to zero. The remaining fertilization amount for the current zone is equal to the difference between the target fertilization amount and the actual fertilization amount for the current zone.

[0074] After the above cyclic processing, the corrected water pump speed, corrected proportional fertilizer valve opening, actual irrigation volume of the current zone, actual fertilizer application volume of the current zone, and remaining fertilizer application volume of the current zone are output in each control cycle for use by the zone irrigation end determination and switching module.

[0075] The partition irrigation end determination and switching module is used to determine whether the current partition has ended and switch to the next partition.

[0076] The system receives feedback from the pipeline network and real-time correction execution module, which generates the actual irrigation volume, actual fertilization volume, and remaining fertilization volume for the current zone. It also receives the current irrigation zone number and zone switching sequence generated by the pump and valve initial execution parameter conversion module, along with the zone switching sequence generated by the water and fertilizer target generation module. Simultaneously, it uses preset allowable deviations for irrigation volume and fertilization volume, which are the control accuracy requirements for irrigation volume and fertilization volume, respectively, set according to the requirements for irrigation uniformity and fertilization accuracy.

[0077] First, compare the actual irrigation amount of the current zone with the target irrigation amount of the current zone, and compare the actual fertilization amount of the current zone with the target fertilization amount of the current zone. Then, determine whether the current zone has ended based on the following five scenarios.

[0078] Scenario 1: Both irrigation and fertilization meet the standards. If the absolute value of the difference between the actual irrigation amount and the target irrigation amount in the current zone is less than or equal to the allowable deviation of irrigation amount, and the absolute value of the difference between the actual fertilization amount and the target fertilization amount in the current zone is less than or equal to the allowable deviation of fertilization amount, then the irrigation in this zone is deemed to have met the standards. In actual execution, irrigation and fertilization usually meet the standards sequentially. If one of them meets the standard first while the other has not, Scenario 2 or Scenario 3 continues execution. When both of them meet the above conditions, the following operations are performed: Close the valve in the current zone, record the actual irrigation amount and actual fertilization amount in the current zone into the actual irrigation and fertilization amount record of the current zone, and generate the current zone end command. Find the next irrigation zone number after the currently executing irrigation zone number from the zone switching sequence, and use it as the next executing irrigation zone number. If the currently executing irrigation zone number is the last one in the zone switching sequence, then the next executing irrigation zone number is empty.

[0079] Scenario 2: If the actual irrigation amount in the current zone is less than the difference between the target irrigation amount and the allowable deviation of the irrigation amount, the irrigation amount has not met the standard; if the actual fertilization amount in the current zone is greater than or equal to the difference between the target fertilization amount and the allowable deviation of the fertilization amount, the fertilization amount has met the standard.

[0080] When the irrigation amount is below the target but the fertilization amount is above the target, if the actual irrigation amount in the current zone is less than the difference between the target irrigation amount and the allowable deviation of the irrigation amount, and the actual fertilization amount in the current zone is greater than or equal to the difference between the target fertilization amount and the allowable deviation of the fertilization amount, then the target fertilizer concentration will be updated to zero, and the feedback to the pipeline network real-time correction execution module will continue to accumulate the irrigation amount.

[0081] Scenario 3: Irrigation volume meets the target, but fertilization volume does not. If the absolute value of the difference between the actual irrigation volume and the target irrigation volume of the current zone is less than or equal to the allowable deviation of irrigation volume, but the actual fertilization volume of the current zone is less than the difference between the target fertilization volume and the allowable deviation of fertilization volume, then the remaining available irrigation volume is obtained by adding the target irrigation volume and the allowable deviation of irrigation volume to the actual irrigation volume of the current zone. If the remaining available irrigation volume is greater than zero, then the target fertilizer concentration is recalculated using the remaining fertilization volume and the remaining available irrigation volume of the current zone; if the recalculated target fertilizer concentration exceeds the preset upper limit of fertilizer concentration, then the upper limit of fertilizer concentration is applied, and the uncompensated fertilization volume is recorded as a fertilization deficiency deviation. Then, the feedback from the pipeline network is returned to the real-time correction execution module to correct the opening of the proportional fertilizer injection valve. If the remaining available irrigation water is zero, no additional fertilization action will be performed. The difference between the target fertilization amount and the actual fertilization amount of the current zone will be recorded as the fertilization deviation of the current zone. The valve of the current zone will be closed, the end command of the current zone will be generated, and the next irrigation zone number will be determined in accordance with the method of Case 1.

[0082] Scenario 4: Excessive Irrigation. If the actual irrigation amount in the current zone is greater than the sum of the target irrigation amount and the allowable deviation of irrigation amount, the valve of the current zone is closed, the difference between the actual irrigation amount and the target irrigation amount of the current zone is recorded as the excessive irrigation deviation of the zone, an end command for the current zone is generated, and the next irrigation zone number is determined in the same way as in Scenario 1.

[0083] Scenario 5: Over-fertilization. If the actual fertilization amount in the current zone is greater than the sum of the target fertilization amount and the allowable deviation, the target fertilizer concentration is updated to zero. The difference between the actual fertilization amount and the target fertilization amount in the current zone is recorded as the over-fertilization deviation for that zone. Irrigation continues with clean water until the cumulative actual irrigation amount in the current zone reaches the allowable range of the target irrigation amount. If the irrigation amount subsequently meets the standard (i.e., the absolute value of the difference between the actual irrigation amount and the target irrigation amount in the current zone is less than or equal to the allowable deviation), the valve is closed and the deviation is recorded.

[0084] Secondly, summarize the water and fertilizer deviation records for this partition during this irrigation. Extract the under-fertilization deviation caused by scenario three (fertilization failure at the end of the topdressing), the over-irrigation deviation triggered by scenario four, and the over-fertilization deviation recorded by scenario five. If any of the above deviations is missing, the corresponding item is not extracted. Combine all the extracted deviations into the current partition's water and fertilizer deviation record and output it along with the current partition's end command.

[0085] After the above processing, the current partition end command, the next irrigation partition number to be executed, the actual irrigation and fertilizer application records of the current partition, and the water and fertilizer deviation records of the current partition are output for use by the historical data update and cycle closure module.

[0086] Historical data update and cycle closure module: Updates historical irrigation response data and proceeds to the next round of decision-making.

[0087] The system receives records of the actual irrigation and fertilization amounts obtained in the current zone, the water and fertilizer deviation records for the current zone, and the next irrigation zone number generated by the zone irrigation end determination and switching module. It also receives target irrigation amounts, target fertilization amounts, and zone switching sequences for each zone generated by the zone switching sequence and water and fertilizer target generation module, as well as the current irrigation zone number generated by the pump valve initial execution parameter conversion module. Furthermore, it collects the post-irrigation soil water potential, post-irrigation soil moisture content, and post-irrigation crop water stress index for the current zone. The post-irrigation soil water potential is collected by the root zone soil water potential sensor for that zone after a delay following valve closure; the post-irrigation soil moisture content is collected by the soil moisture sensor for that zone after a delay following valve closure; and the post-irrigation crop water stress index is calculated from the canopy temperature sensor and meteorological data after a delay following valve closure. It also acquires meteorological data during the execution period, as well as records of main pipe pressure, branch pipe flow, and fertilizer solution concentration during the execution period. The meteorological data during the execution period comes from the records of the meteorological forecast interface of the data synchronization acquisition and validity processing module corresponding to the execution period. The records of main pipe pressure, branch pipe flow, and fertilizer solution concentration during the execution period come from the execution process records of the data synchronization acquisition and validity processing module and the pipeline feedback real-time correction execution module.

[0088] First, the irrigation data for this partition is written into the historical irrigation response record for that partition. The written data includes: pre-irrigation status, taken from the current status record generated by the data synchronization acquisition and validity processing module, which contains the corresponding soil water potential, soil moisture content, and crop water stress index for that partition; target irrigation amount and target fertilizer amount, taken from the target irrigation amount and target fertilizer amount for each partition generated by the partition switching sequence and water and fertilizer target generation module; actual irrigation amount and actual fertilizer amount, taken from the actual irrigation and fertilizer amount records for the current partition; water and fertilizer deviation, taken from the water and fertilizer deviation record for the current partition; post-irrigation status, including the post-irrigation soil water potential, post-irrigation soil moisture content, and post-irrigation crop water stress index for the current partition; meteorological data during the execution period; and records of main pipe pressure, branch pipe flow, and fertilizer concentration during the execution period.

[0089] Secondly, the training samples for updating the zonal irrigation response prediction model are generated for the next round of zonal irrigation response prediction module. The post-irrigation response prediction result is selected from the candidate irrigation amount and candidate fertilizer amount combinations generated in the zonal irrigation response prediction module, and the combination closest to the actual irrigation amount and actual fertilizer amount is selected. The closest combination is determined by the absolute value of the difference between the candidate irrigation amount and the actual irrigation amount plus the absolute value of the difference between the candidate fertilizer amount and the actual fertilizer amount, and the combination with the smallest distance is selected. If there are multiple combinations with equal distances, the combination with the smallest candidate irrigation amount is selected. The post-irrigation response prediction result corresponding to the selected combination is compared with the current zonal post-irrigation soil water potential, current zonal post-irrigation soil moisture content, and current zonal post-irrigation crop water stress index to calculate the deviation. The complete record containing the zonal pre-irrigation soil water potential, pre-irrigation soil moisture content, pre-irrigation crop water stress index, actual irrigation amount, actual fertilizer amount, post-irrigation soil water potential, post-irrigation soil moisture content, post-irrigation crop water stress index, and corresponding meteorological data is used as the training sample for the next round of zonal irrigation response prediction module to update the zonal irrigation response prediction model.

[0090] Finally, determine whether to continue this round of irrigation. Search the partition switching sequence for the next irrigation partition number after the currently executing partition number, and iteratively check if this number belongs to the set of executable partitions for this round. If it does, return to the pump / valve initial execution parameter conversion module to continue irrigation for that partition. If it does not, continue searching the next irrigation partition number in the partition switching sequence until a partition belonging to the set of executable partitions for this round is found. If no partition belonging to the set of executable partitions for this round is found after traversing the partition switching sequence, it indicates that all executable partitions for this round have completed irrigation. At this point, summarize the records of all executed partitions as the execution records for each partition in this round, return to the data synchronization acquisition and validity processing module to re-acquire data, and generate the next round of alternating irrigation decisions.

[0091] After the above processing, the updated historical irrigation response data for each partition will be written back to the historical irrigation response database for the next round of data synchronization acquisition and validity processing module to read and use.

Claims

1. An intelligent control system for integrated water and fertilizer management based on artificial intelligence, characterized in that: include: The data synchronization acquisition and validity processing module is used to collect soil water potential, soil moisture content, crop water stress index, weather forecast data, main pipe pressure, branch pipe flow and fertilizer concentration in each zone, read the historical irrigation response records written back in the previous round, and generate current status records, available historical irrigation response records and first round prediction zone markers. The zonal irrigation response prediction module is used to generate future water stress prediction results and post-irrigation response prediction results under different combinations of irrigation volume and fertilizer application based on the current status record, available historical irrigation response record, first-round prediction zonal markers, and preset initial prediction model parameters. The irrigation trigger determination and dry / wet phase division module is used to determine the remaining duration of the dry phase in each zone and the set of zones for which irrigation execution targets need to be generated in this round, based on the current status record, the future water stress prediction results under non-irrigation conditions, the preset allowable water stress upper limit, and the preset early warning margin. The partition switching sequence and water and fertilizer target generation module is used to generate the partition set of irrigation execution targets, the remaining duration of the dry phase and the post-irrigation response prediction results according to the needs of this round, and generate the partition switching sequence, the target irrigation amount of each partition, the target fertilizer amount of each partition, the initial duration of the wet phase and the set of executable partitions for this round. The pump valve initial execution parameter conversion module is used to generate the current irrigation zone number, target branch flow rate, target fertilizer concentration, initial pump speed and initial proportional fertilizer injection valve opening based on the zone switching sequence, the set of executable zones in this round, the target irrigation volume, the target fertilizer application volume and the initial duration of the wet stage; The pipeline feedback real-time correction execution module is used to perform feedback correction based on the target branch pipe flow rate and target fertilizer concentration, combined with the real-time collected main pipe pressure, branch pipe flow rate and fertilizer concentration, and output the corrected water pump speed, proportional fertilizer valve opening, actual irrigation volume, actual fertilizer application volume and remaining fertilizer application volume. The zonal irrigation end determination and switching module is used to determine whether the current zonal irrigation has ended based on the actual irrigation volume, actual fertilization volume, remaining fertilization volume and zonal switching sequence, and to generate the next execution zonal number, the actual irrigation volume and fertilization volume obtained by the current zonal, and the water and fertilizer deviation record. The historical data update and cycle closure module is used to write the current irrigation data into the historical irrigation response record, generate training samples for the next round of model update, and trigger the next round of decision loop after the current round is completed.

2. The intelligent water and fertilizer integration control system based on artificial intelligence as described in claim 1, characterized in that, The data synchronization acquisition and validity processing module generates current status records, available historical irrigation response records, and first-round prediction partition markers, including: Extract the records for each region from the historical irrigation response records of each region, delete the records that are missing any one of the following: actual irrigation amount, actual fertilizer amount, post-irrigation soil water potential, post-irrigation soil moisture content, or post-irrigation crop water stress index, and arrange the remaining valid records in chronological order as the available historical irrigation response records for that region. If no valid record is found or no record is retrieved, the partition number is added to the first-round prediction partition marker, and the available historical irrigation response records are set to empty.

3. The intelligent water and fertilizer integration control system based on artificial intelligence as described in claim 1, characterized in that: The generation of future water stress prediction results and post-irrigation response prediction results for different irrigation volumes includes: The maximum allowable irrigation volume is determined based on the area of ​​the zone, root depth, field capacity, and current soil moisture content. The maximum allowable irrigation volume is the volume value obtained by multiplying the difference between field capacity and current soil moisture content by the root depth and then by the area of ​​the zone. If the volume value is less than or equal to zero, the maximum allowable irrigation volume is set to zero. Starting from zero, the water volume increments by a preset step size until the maximum allowable water volume is reached. Each increment is then used as an element of the candidate water volume set. Preset target fertilizer-to-water ratio; The maximum reference fertilizer application rate is determined based on the target fertilizer-water ratio and the maximum allowable irrigation volume. Starting from zero, the application rate is increased in preset increments until the maximum reference fertilizer application rate is reached. Each increment is used as an element of the candidate fertilizer application rate set. For each irrigation zone, query the first round of prediction zone label. If it is not labeled, update the irrigation response prediction model using available historical irrigation response records as training samples. If it is labeled, construct the initial irrigation response prediction model using preset initial prediction model parameters.

4. The intelligent water and fertilizer integration control system based on artificial intelligence as described in claim 1, characterized in that: The determination of the remaining time of each partition's dry phase and the set of partitions for which irrigation execution targets need to be generated in this round includes: For each irrigation zone, the current crop water stress index is read from the current status record, and the crop water stress index at each prediction time is read in chronological order from the future water stress prediction results under non-irrigation conditions. When the current crop water stress index is greater than or equal to the upper limit of allowable water stress, the remaining duration of the dry stage is recorded as zero. If the current crop water stress index is less than the upper limit of allowable water stress, and there is a forecast time within the future forecast period when the forecast value first reaches the upper limit of allowable water stress, the time interval from the current time to that forecast time is taken as the remaining dry period. Calculate the difference between the upper limit of allowable water stress and the crop water stress index at the end of the prediction window; If the difference is less than the preset warning margin, the partition will be included in the candidate waiting set and recalculated first after all executable partitions in the current round have been irrigated.

5. The intelligent water and fertilizer integration control system based on artificial intelligence as described in claim 1, characterized in that: The generation of the partition switching sequence, target irrigation volume, initial duration of the wet phase, and set of executable partitions for this round includes: Take out each partition from the set of partitions that need to generate irrigation execution targets in this round, and sort them in ascending order of the remaining time of the dry stage; if the remaining time of several stages is the same, read the current crop water stress index from the current status record of each partition, and sort them in descending order of the current crop water stress index to obtain the partition switching sequence. For each partition, candidate irrigation amounts are selected based on whether the post-irrigation crop water stress index is less than the upper limit of allowable water stress, the post-irrigation soil moisture content does not exceed the upper limit of field capacity, and the expected duration meets the minimum wet stage duration requirement. The expected duration is obtained by reading the crop water stress index at each predicted time after irrigation in chronological order from the post-irrigation response prediction results, and subtracting the post-irrigation start time from the time corresponding to the predicted time when it is equal to or first exceeds the upper limit of allowable water stress. The post-irrigation start time is the irrigation end time for that partition. If the crop water stress index at all predicted times after irrigation is less than the upper limit of allowable water stress, then the difference between the end of the prediction window and the beginning of the post-irrigation time will be used as the expected duration. Select the minimum irrigation amount as the target irrigation amount; If the available historical irrigation response records contain a record for this partition, then extract the branch flow record during the execution period from the most recent record, sum the branch flow at each sampling time and divide by the number of samplings, and use the result as the initial branch flow for this partition. If the available historical irrigation response records do not contain a record for that partition, then the design rated branch flow rate is used as the initial branch flow rate.

6. The intelligent water and fertilizer integration control system based on artificial intelligence as described in claim 1, characterized in that, The pump valve initial execution parameter conversion module generates the current irrigation zone number, target branch flow rate, target fertilizer concentration, initial pump speed, and initial proportional fertilizer injection valve opening, including: If the target irrigation volume is zero, then the partition is marked as completed for this round, and the next partition in the partition switching sequence is selected to reselect the current execution partition; If the pump characteristic curve is not pre-stored, the median value of the pump speed range will be selected as the default initial pump speed. If the characteristic curve of the fertilizer injection valve is not pre-stored, the median value of the proportional fertilizer injection valve opening range is selected as the default initial proportional fertilizer injection valve opening.

7. The intelligent water and fertilizer integration control system based on artificial intelligence as described in claim 1, characterized in that: The output of the corrected pump speed, proportional fertilizer valve opening, actual irrigation volume, actual fertilizer application volume, and remaining fertilizer application volume includes: When the current branch flow rate is lower than the target branch flow rate minus the allowable deviation within the control cycle that continuously reaches the preset quantity threshold, and the main pressure is less than the pressure limit, the current pump speed will be increased by one pump speed adjustment step. When the current branch flow rate is lower than the difference between the target branch flow rate and the allowable flow deviation within the same number of consecutive control cycles, but the main pressure is greater than or equal to the pressure limit, the current pump speed remains unchanged, and the irrigation volume continues to be accumulated according to the current branch flow rate. The zonal irrigation end determination and switching module extends the execution time of the current zonal based on the actual accumulated irrigation volume. When the current branch flow rate is higher than the sum of the target branch flow rate and the allowable flow deviation within a certain number of consecutive control cycles, the current pump speed is reduced by one pump speed adjustment step to obtain the corrected pump speed. If the current branch flow rate is greater than zero, the pipeline volume is divided by the current branch flow rate to obtain the transmission delay time. The current time is subtracted from the transmission delay time to obtain the delay matching time. The value closest to the matching time is taken from the fertilizer concentration recorded in each previous control cycle as the fertilizer concentration after delay compensation. If the current branch flow rate is zero, the fertilizer solution concentration collected in this control cycle will be used as the fertilizer solution concentration after delay compensation. When the fertilizer concentration after delay compensation is lower than the target fertilizer concentration minus the allowable deviation within the control cycle that continuously reaches the preset quantity threshold, the current proportional fertilizer injection valve opening is increased by one proportional fertilizer injection valve opening adjustment step. When the fertilizer solution concentration after delay compensation is higher than the target fertilizer solution concentration plus the allowable concentration deviation within the control cycle that continuously reaches the preset quantity threshold, the current proportional fertilizer injection valve opening is reduced by one proportional fertilizer injection valve opening adjustment step.

8. The intelligent water and fertilizer integration control system based on artificial intelligence as described in claim 1, characterized in that: The process of determining whether the current partition has ended, generating the next execution partition number, the actual irrigation and fertilization amounts obtained in the current partition, and the water and fertilizer deviation record includes: When both irrigation and fertilization amounts meet the standards, close the valve and generate an end command. If the irrigation amount is below the target but the fertilizer amount is above the target, update the target fertilizer concentration to zero and continue to accumulate the irrigation amount; When the irrigation amount has reached the standard but the fertilizer amount has not, the remaining available irrigation amount is obtained by adding the target irrigation amount of the current zone to the allowable deviation of the irrigation amount, and then subtracting the actual irrigation amount of the current zone. If the result is greater than zero, the target fertilizer concentration is recalculated using the remaining fertilizer amount and the remaining available irrigation amount for topdressing. If the result is equal to zero, the fertilizer deviation is recorded and the process ends. If the amount of irrigation exceeds the limit, record the excess irrigation deviation and end the process. If the amount of fertilizer applied is excessive, update the target fertilizer concentration to zero and irrigate with clean water until the irrigation volume meets the standard.

9. The intelligent water and fertilizer integration control system based on artificial intelligence as described in claim 1, characterized in that: The generation of training samples for the next round of model updates, after the completion of this round, triggers the next round of decision-making loop, including: The post-irrigation response prediction results are selected from the candidate irrigation amount and candidate fertilizer amount combinations that are closest to the actual irrigation amount and actual fertilizer amount. The closest combination is selected by using the absolute value of the difference between the candidate irrigation amount and the actual irrigation amount plus the absolute value of the difference between the candidate fertilizer amount and the actual fertilizer amount as the distance metric. If no partition belonging to the set of executable partitions for this round is found after traversing the partition switching sequence, then all records of the executed partitions are summarized, and the data is returned to the data synchronization acquisition and validity processing module to re-acquire data and generate the next round of alternating irrigation decisions.

10. The intelligent water and fertilizer integration control system based on artificial intelligence as described in claim 1, characterized in that, The historical data update and cycle closure module writes the current irrigation data into the historical irrigation response record, generates training samples for the next round of model updates, and triggers the next round of decision-making loop after the current round is completed, including: The irrigation data will be written into the historical irrigation response record. The data to be written includes soil water potential before irrigation, soil moisture content before irrigation, crop water stress index before irrigation, target irrigation amount, target fertilizer amount, actual irrigation amount, actual fertilizer amount, water and fertilizer deviation, soil water potential after irrigation, soil moisture content after irrigation, crop water stress index after irrigation, as well as meteorological data and pipeline records during the implementation period. When generating training samples for the next round of model updates, the post-irrigation response prediction results corresponding to the combination of candidate irrigation amount and candidate fertilizer amount that is closest to the actual irrigation amount and actual fertilizer amount are selected. The closest combination is selected by using the absolute value of the difference between the candidate irrigation amount and the actual irrigation amount plus the absolute value of the difference between the candidate fertilizer amount and the actual fertilizer amount as a distance metric. If there are multiple combinations with equal distances, the combination with the smallest candidate irrigation amount is selected.

Images