A Data- and Mechanism-Driven Irrigation Forecasting Method and System for Rice-Shrimp Farming Areas
The irrigation forecasting method for shrimp-rice farming areas, driven by both data and mechanisms, solves the problem of coupled forecasting of water consumption and seepage characteristics between paddy fields and shrimp ditches, as well as water quality and water exchange in shrimp-rice farming areas. It enables precise scheduling of irrigation water and water quality control, improves forecast accuracy and robustness, and is suitable for digital twin irrigation area platforms.
Patent Information
- Application Number
- CN202611419416.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-14
- Publication Date
- 2026-11-17
AI Technical Summary
Existing irrigation forecasting methods cannot adapt to the water consumption and lateral seepage characteristics of rice paddy-shrimp ditch zones in rice-shrimp farming areas. They lack coupled forecasting of water quality and water exchange, have low utilization rates of irrigation district management experience data, and lack coupling between mechanistic models and data models, resulting in forecast accuracy that is difficult to meet actual needs.
An irrigation forecasting method for shrimp-rice farming areas based on both data and mechanism is adopted. By acquiring multi-source data, dividing the paddy field area into the shrimp ditch area, constructing a mechanism forecasting model and a data-driven model, and combining water balance and water quality control, the method can accurately forecast irrigation time, irrigation quota and water exchange volume.
It enables precise forecasting of irrigation water in rice-shrimp farming areas, meeting the dual requirements of water demand for rice and water quality control for shrimp, improving forecast accuracy and robustness, and is suitable for automated operation of digital twin irrigation district platforms.
Smart Images

Figure CN123117467A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of agricultural water and soil engineering and smart irrigation technology, specifically involving an irrigation forecasting method and system for shrimp-rice farming areas based on both data and mechanism-driven approaches. Background Technology
[0002] The integrated rice-crayfish farming model organically combines rice cultivation with the farming of red swamp crayfish (Procambarus clarkii), resulting in significant comprehensive benefits. It has been widely promoted in areas such as the Jianghan Plain and is an important industry for ensuring food security and increasing farmers' income. Compared to ordinary rice paddies, rice-crayfish fields often include circular crayfish farming ditches. Water management in these fields must simultaneously consider the water requirements of rice at different growth stages, the minimum water level required for crayfish survival, and water quality constraints such as dissolved oxygen and water temperature. Water consumption in the fields is complex, involving four types of water transformation relationships: rice canopy evaporation and transpiration, evaporation from the free surface of the crayfish ditches, vertical seepage from the cultivated layer, and lateral seepage exchange between the rice paddy and the crayfish ditches. Meanwhile, crayfish farming is highly sensitive to water quality indicators such as dissolved oxygen and water temperature; water quality is prone to deterioration during high-temperature, high-density farming stages, necessitating periodic water exchanges for water quality control. The above factors together result in the distinctive characteristics of irrigation water use in shrimp-rice farming areas, including complex water demand mechanisms, dual constraints of water layer and water quality, and linkage between irrigation replenishment and water quality exchange, which poses a great technical challenge to field irrigation forecasting.
[0003] Existing irrigation forecasting methods are mostly designed for ordinary paddy fields or dryland crops, employing a uniform crop coefficient method and field water balance to calculate irrigation water demand. They fail to consider the water consumption between paddy fields and shrimp ditches in shrimp-rice farming areas, the lateral seepage exchange between ditches and fields, and the water quality exchange requirements, making them unsuitable for direct application in these areas. Furthermore, the water use planning data accumulated in irrigation district management practices contains the irrigation decision-making experience of grassroots managers. Existing methods lack effective mining of this management experience data and, more importantly, lack irrigation forecasting tools that couple mechanistic models with data-driven models, resulting in forecast accuracy that fails to meet actual needs. With the advancement of digital twin irrigation district construction, there is an urgent need to develop precise irrigation forecasting methods adapted to the water transformation characteristics of shrimp-rice farming areas, providing support for the construction of digital twin irrigation district models and smart irrigation district development.
[0004] Therefore, developing a method and system for irrigation forecasting in shrimp-rice farming areas that is driven by both mechanistic and data-driven models, and enabling accurate forecasting of irrigation time, irrigation quota, and water exchange volume, is a technical challenge that urgently needs to be addressed in the current irrigation management of shrimp-rice farming areas. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides an irrigation forecasting method and system for shrimp-rice farming areas based on both mechanism and data-driven approaches. This system solves the technical problems of existing irrigation forecasting methods being unable to adapt to the water consumption and lateral seepage characteristics of rice paddy-shrimp ditch zones in shrimp-rice farming areas, lacking coupled forecasting of water quality and water exchange, having low utilization of irrigation area management experience data, and lacking coupling between mechanism models and data models. This system enables accurate forecasting and intelligent decision-making regarding irrigation time, irrigation quotas, and water quality and water exchange volume in shrimp-rice farming areas.
[0006] To achieve the above objectives, the present invention adopts the following technical solution: A data- and mechanism-driven method for irrigation forecasting in shrimp-rice farming areas includes the following steps: Step S1. Obtain multi-source data of the irrigation area; based on the spatial structure information of the irrigation area fields, divide the shrimp-rice farming fields into rice fields and shrimp ditch areas, and determine the shrimp ditch area ratio; Step S2. Based on the meteorological forecast data and the current rice growth stage information, determine the forecast value of total evapotranspiration in the whole field according to the rice growth stage and non-rice growth stage; Step S3. Based on data on different growth stages of rice and / or field water layer and groundwater depth, predict the field seepage amount and obtain the predicted value of field seepage amount; Step S4. Based on the different water layer conditions in the rice-shrimp co-culture stage and the rice-shrimp separate culture stage, predict the lateral seepage volume between the paddy field and the shrimp ditch, and obtain the lateral seepage volume forecast value. Step S5. Construct a mechanism forecasting model based on the principle of water balance in the water layer. The mechanism forecasting model determines the lower limit of the field water layer and the upper limit of the suitable water layer according to the joint constraints of rice and crayfish. Based on the forecast of rainfall and water consumption in the foreseeable period, the field water layer is progressively advanced day by day. When the water layer drops to the lower limit of the field water layer, irrigation is triggered and the irrigation quota is determined. The irrigation time is dynamically extended in combination with the forecast of rainfall in the foreseeable period. Step S6. Based on the shrimp ditch water quality monitoring data or water quality forecast data, determine the timing of water exchange according to the water quality control index threshold, verify the water exchange volume, and incorporate the water exchange volume into the unified scheduling of water layer water balance. Step S7. Compile the historical irrigation water use plan reporting data of the irrigation district into a manager experience sample database, train the data-driven irrigation forecasting model, and couple the data-driven irrigation forecasting model with the mechanism forecasting model based on water balance to synthesize the final irrigation forecasting results; Step S8. Based on the final irrigation forecast results, release the forecast results of irrigation time, irrigation quota and water exchange volume in different levels, and drive irrigation and drainage operations; use measured data to continuously correct the data to drive the irrigation forecast model parameters and the coupling weight with the mechanism forecast model to form a closed loop operation.
[0007] Furthermore, the multi-source data includes meteorological monitoring data, meteorological forecast data, field water layer and groundwater depth data, water quality monitoring data, planting and breeding production data, and irrigation water use plan reporting data of the irrigation district over the years.
[0008] Further, step S2 includes the following sub-steps: The weather type correction coefficient is determined based on the weather type in the forecast period meteorological forecast, and the daily reference crop evapotranspiration forecast value is calculated based on the weather type correction coefficient and the multi-year average reference crop evapotranspiration. During the rice growing season, a crop coefficient is determined based on the rice's growing season, and the crop coefficient is multiplied by the predicted value of evapotranspiration of the reference crop to obtain the predicted value of evapotranspiration in the paddy field area. When forecasting the evaporation of the Xiagou area, the evaporation forecast value issued by the meteorological department is used; when there is no evaporation forecast value, the Dalton-type empirical model is used to calculate the evaporation forecast value of the Xiagou area. When the rice is in its growing season, the predicted values of evapotranspiration in the rice paddy area and the predicted values of water surface evaporation in the shrimp ditch area are weighted and summed according to the proportion of rice paddy area to shrimp ditch area to synthesize the total evapotranspiration of the whole field; when the rice is not in its growing season, the predicted value of water surface evaporation of the shrimp ditch is directly used for the total evapotranspiration of the whole field.
[0009] Furthermore, the methods for predicting field seepage include the empirical value method and the empirical model method; wherein the empirical value method is: using field experimental observation data, determining the empirical value of daily seepage in the field according to different growth stages of rice, and forming a parameter table of seepage in the field for different growth stages; The empirical model method involves establishing an empirical model of seepage related to surface water depth, groundwater depth, and soil properties for fields with basic monitoring conditions. Based on this empirical model, the predicted field seepage is calculated. The empirical model for seepage is as follows:
[0010] In the formula: This represents the daily seepage rate in the field (mm / d). The water depth on the field surface (the water depth on the field surface is based on the bottom of the shrimp ditch). Due to the depth of groundwater, , The reference water depth and the reference groundwater depth are used as the reference depths. As the benchmark leakage strength, It is an experience index. , Calibration was performed using field-measured seepage data and the least squares criterion.
[0011] Further, step S4 specifically includes: During the rice-shrimp co-culture stage, the difference in water level drop between the rice paddy area and the shrimp ditch area is calculated based on the water level drop depth caused by evaporation and vertical seepage in the rice paddy area and the water level drop depth caused by surface evaporation and vertical seepage in the shrimp ditch area. Combined with the proportion of rice paddy area, the proportion of shrimp ditch area and the total area of the fields, the amount of water exchanged by lateral seepage is generalized. During the rice-shrimp separation stage, the amount of lateral seepage is calculated according to Darcy's law. The calculation of Darcy's law involves the soil saturated permeability coefficient of the ditch-field contact zone, the average depth of the lateral seepage cross section, the water level difference between the paddy field and the shrimp ditch after conversion to the same elevation datum, the seepage path length, the length of the ditch-field contact zone, and the calculation period.
[0012] Further, step S5 includes: Based on the combined constraints of the suitable water layer for each growth stage of rice and the minimum water layer required for the survival of crayfish, the lower limit of the field water layer and the upper limit of the suitable water layer are determined. The lower limit of the field water layer is the larger value between the lower limit of the suitable water layer for each growth stage of rice and the lower limit of the water layer required for the survival of crayfish. The upper limit of the suitable water layer is the smaller value between the upper limit of the suitable water layer for each growth stage of rice and the upper limit of the water layer required for the survival of crayfish. Using the current measured water layer as the initial condition, the forecast of the expected rainfall period is input and coupled with the forecast values of total evapotranspiration, field seepage, and lateral seepage, and the field water layer is recursively calculated day by day using a two-level water balance calculation model: first, the water balance calculation is performed on the entire field of the shrimp-rice farming unit, and then the lateral seepage exchange between the paddy field and the shrimp ditch is introduced to obtain the actual water layer of the paddy field and the actual water layer of the shrimp ditch respectively; Irrigation is triggered when the predicted water level drops to the lower limit of the field water level, and the field water level is restored to the upper limit of the appropriate water level to determine the irrigation quota; if the prediction shows that the water level will not fall below the lower limit of the field water level after rainfall replenishment in the future forecast period, the irrigation time will be postponed. When it is determined that water needs to be replaced, the amount of water replaced will be included in the water volume item for water layer recursion.
[0013] Further, step S6 includes: Based on the early-stage temperature forecast, the daily water temperature of the shrimp ditch is predicted using a first-order heat exchange model. Based on the relationship between saturated dissolved oxygen and water temperature, and the reoxygenation-oxygen consumption balance equation, combined with the reoxygenation coefficient, the comprehensive oxygen consumption rate at the reference water temperature, and the temperature correction coefficient, the dissolved oxygen concentration in the shrimp ditch is predicted daily. Dissolved oxygen and water temperature are selected as water exchange control indicators, and water quality exchange thresholds are preset. If either indicator exceeds the limit, water exchange is determined to be required. Normalized exceedance limits for dissolved oxygen and water temperature are defined. For dissolved oxygen, the exceedance limit is the ratio of the difference between the threshold and the predicted value to the threshold. For water temperature, the exceedance limit is the ratio of the difference between the predicted value and the threshold to the threshold. For control indicators that exceed the limit, the theoretical water exchange volume is calculated by inverting the field water body index value before water exchange, the water source index value, and the target value after water exchange according to the principle of mixed dilution material balance. The shrimp ditch replacement efficiency coefficient is then introduced to correct the calculated water exchange volume. When multiple indicators exceed the limit at the same time, the required water exchange volume for each indicator is calculated separately and the largest value is taken for verification. When water quality monitoring data is insufficient, the method of proportional coefficient is used to determine the water exchange volume; the actual water exchange volume is determined by multiplying the basic water exchange ratio by the dynamic adjustment factor of the degree of exceeding the limit, and then multiplying it by the total amount of water stored in the field. The total water storage in the field is determined by the sum of the water storage in the paddy field area and the shrimp ditch area. During the rice-shrimp co-culture stage, the water storage is calculated based on the uniform water depth of the entire field surface and the depth of the shrimp ditch. During the rice-shrimp separate culture stage or when there is no water layer on the field surface during the drying period, the water storage in the paddy field area is taken as zero.
[0014] Furthermore, the training method for the data-driven irrigation forecasting model includes: Collect irrigation water use plan data submitted by irrigation users in the irrigation district over the years according to basic field information and the format of the declared water demand date and declared water demand, and build a sample database of managers' experience together with the corresponding rainfall, temperature and field water layer monitoring data; Using the manager experience sample library as training samples, a data-driven irrigation forecasting model based on neural networks is constructed. At the same time, a decision tree algorithm is used to extract a condition-irrigation rule library from the same sample library to verify the forecasting results of the data-driven irrigation forecasting model. During training, the predicted irrigation volume output by the data-driven irrigation forecasting model is converted into water layer depth according to the total area of the field and then incorporated into the water balance equation to recursively calculate the field water layer on a daily basis. The sum of the squares of the overflow depths of the recursively calculated water layer beyond the upper and lower limits of the water layer is used to calculate the theoretical constraint penalty. The theoretical constraint penalty is then weighted and summed with the sample fitting error to form the training objective function.
[0015] Furthermore, the coupling forecasting method of the data-driven irrigation forecasting model and the mechanism forecasting model includes: The difference between the output value of the data-driven model and the replay prediction value of the mechanism model corresponding to the historical samples is used as the mechanism prediction bias correction amount, and the sum of the mechanism model prediction value and the mechanism prediction bias correction amount is used as the final prediction value. When the data-driven model fails to converge, the coupling weights are determined by the inverse of the historical forecast error variances of the mechanistic model and the data-driven model and then weighted and fused. The coupling weights are dynamically adjusted according to the rolling forecast errors of the two models.
[0016] On the other hand, the present invention provides an irrigation forecasting system for shrimp-rice farming areas based on both data and mechanism-driven approaches, comprising: The data acquisition module is used to acquire multi-source data of the irrigation area; based on the spatial structure information of the irrigation area fields, the shrimp-rice farming fields are divided into rice fields and shrimp ditch areas, and the shrimp ditch area ratio is determined. The mechanism prediction model has built-in regional evaporation and transpiration prediction units, leakage prediction units, lateral seepage prediction units, water layer water balance irrigation prediction units, and water quality prediction and water exchange prediction units. The zonal evapotranspiration forecasting unit is used to determine the forecast value of total evapotranspiration of the whole field according to the meteorological forecast data and the current rice growth stage information, based on the rice growth period and non-rice growth period. The seepage forecasting unit is used to forecast field seepage based on data of different rice growth stages and / or field water layer and groundwater depth, and to obtain field seepage forecast values. The side seepage prediction unit is used to predict the side seepage between the paddy field and the shrimp ditch according to the different water layer differences in the shrimp-rice co-culture stage and the shrimp-rice separate culture stage, and obtain the side seepage prediction value. The water layer water balance irrigation forecasting unit is used to construct a mechanism forecasting model based on the principle of water layer water balance. The mechanism forecasting model determines the lower limit of the field water layer and the upper limit of the suitable water layer according to the joint constraints of rice and crayfish. Based on the forecast of rainfall and water consumption, the field water layer is progressively advanced day by day. When the water layer drops to the lower limit of the field water layer, irrigation is triggered and the irrigation quota is determined. The irrigation time is dynamically extended in combination with the forecast of rainfall. The water quality prediction and water exchange forecasting unit determines the timing of water exchange based on the shrimp ditch water quality monitoring data or water quality forecast data, according to the water quality control index threshold, and verifies the water exchange volume, and incorporates the water exchange volume into the unified scheduling of water layer water balance. The dual-drive coupling module is used to compile the annual irrigation water use plan reporting data of the irrigation district into a manager experience sample library, train the data-driven irrigation forecasting model, and couple the data-driven irrigation forecasting model with the mechanism forecasting model based on water balance to synthesize the final irrigation forecasting results. The rolling correction module is used to release irrigation time, irrigation quota and water exchange volume forecasts in stages according to the final irrigation forecast results, and drive irrigation and drainage operations; it uses measured data to roll correct the data to drive the irrigation forecast model parameters and the coupling weights with the mechanism forecast model to form a closed loop operation. The forecast release and execution module is used to push irrigation time, irrigation quota and water exchange volume to management units and water users in a tiered manner, and drive pumping stations and gates to perform irrigation and drainage operations.
[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention addresses the field structure of rice paddies and shrimp ditches coexisting in rice-shrimp farming areas by proposing a method for determining the area ratio of shrimp ditches and a regional evaporation and transpiration forecasting model. It unifies rice evaporation and transpiration and shrimp ditch water surface evaporation into the whole field water consumption forecast, overcoming the shortcomings of ordinary irrigation forecasting methods that do not distinguish between the differences in water consumption between ditches and fields. The forecasting mechanism is in line with the actual situation in rice-shrimp farming areas.
[0018] 2. The present invention provides a method for predicting the amount of lateral seepage in paddy fields and shrimp ditches based on the generalized rice-shrimp farming model: during the co-cultivation stage, the water exchange between ditches and fields is generalized according to the principle of "differential evaporation leading to uneven water levels, which recover to a level within one day"; during the separate cultivation stage, the amount of lateral seepage during the non-level period is calculated according to Darcy's law, thus filling the gap in the quantitative prediction of lateral seepage in paddy fields and ditches in shrimp-rice farming areas.
[0019] 3. This invention couples water volume balance irrigation forecasting with water quality prediction and water exchange volume forecasting to achieve unified scheduling of irrigation and water exchange. The water exchange volume is determined by inversion based on water quality restoration targets and correction based on shrimp ditch replacement efficiency, overcoming the arbitrariness of determining water exchange volume using the empirical ratio method. The forecast results simultaneously meet the dual requirements of rice water demand and shrimp water quality control, closely aligning with the actual production of rice-shrimp farming and demonstrating strong practicality.
[0020] 4. This invention proposes a forecasting mechanism that couples the residuals of the mechanistic model and the data-driven model. It makes the manager's experience contained in the irrigation district water use plan declaration data explicit and uses it to correct the mechanistic forecasting bias. The water balance mechanism is embedded as a soft constraint into the training objective function of the data-driven model. From the training mechanism, it suppresses the forecast output that violates the mechanism and significantly improves the accuracy and robustness of irrigation forecasting under the condition of data imbalance.
[0021] 5. This invention operates fully automatically, with daily rolling forecasts and closed-loop corrections. It can be directly embedded into a digital twin irrigation district platform, providing model support for smart irrigation in shrimp-rice farming areas, and has broad prospects for promotion and application. Attached Figure Description
[0022] Figure 1 This is a general technical roadmap of the method of the present invention; Figure 2 Figure 1 shows a schematic diagram of the zoning structure of a shrimp-rice farming field, where (a) is a plan view and (b) is a cross-sectional view. Figure 3 This is a schematic diagram of the structure of a mechanism-based irrigation forecasting model; Figure 4 Figure 1 shows a simplified schematic diagram of lateral seepage between paddy fields and shrimp ditches during the co-culture and separate culture stages. Figure 2 shows a simplified schematic diagram of lateral seepage between paddy fields and shrimp ditches during the co-culture stage, and Figure 3 shows a simplified schematic diagram of lateral seepage between paddy fields and shrimp ditches during the separate culture stage. Figure 5 A flowchart for water quality prediction and water exchange decision-making; Figure 6 A diagram illustrating the data-driven model and the dual-drive coupling architecture; Figure 7 This is a line graph illustrating the predictive process of field water layer during the forecast period, as described in an embodiment of the present invention. Figure 8 This is a process curve for predicting water temperature and dissolved oxygen, and a water change determination diagram, according to an embodiment of the present invention. Figure 9 This is a comparison chart of irrigation forecast accuracy in embodiments of the present invention; Figure 10 This is a structural diagram of the system composition of the present invention. Detailed Implementation
[0023] To facilitate understanding and implementation of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.
[0024] like Figure 1 As shown, the present invention provides Example 1 like Figure 1 As shown, this embodiment provides an irrigation forecasting method for shrimp-rice farming areas based on both data and mechanism-driven approaches, including the following steps: Step S1. Obtain multi-source data of the irrigation area; based on the spatial structure information of the irrigation area fields, divide the shrimp-rice farming fields into rice fields and shrimp ditch areas, and determine the shrimp ditch area ratio; Step S2. Based on the meteorological forecast data and the current rice growth stage information, determine the forecast value of total evapotranspiration in the whole field according to the rice growth stage and non-rice growth stage; Step S3. Based on data on different growth stages of rice and / or field water layer and groundwater depth, predict the field seepage amount and obtain the predicted value of field seepage amount; Step S4. Based on the different water layer conditions in the rice-shrimp co-culture stage and the rice-shrimp separate culture stage, predict the lateral seepage volume between the paddy field and the shrimp ditch, and obtain the lateral seepage volume forecast value. Step S5. Construct a mechanism forecasting model based on the principle of water balance in the water layer. The mechanism forecasting model determines the lower limit of the field water layer and the upper limit of the suitable water layer according to the joint constraints of rice and crayfish. Based on the forecast of rainfall and water consumption in the foreseeable period, the field water layer is progressively advanced day by day. When the water layer drops to the lower limit of the field water layer, irrigation is triggered and the irrigation quota is determined. The irrigation time is dynamically extended in combination with the forecast of rainfall in the foreseeable period. Step S6. Based on the shrimp ditch water quality monitoring data or water quality forecast data, determine the timing of water exchange according to the water quality control index threshold, verify the water exchange volume, and incorporate the water exchange volume into the unified scheduling of water layer water balance. Step S7. Compile the historical irrigation water use plan reporting data of the irrigation district into a manager experience sample database, train the data-driven irrigation forecasting model, and couple the data-driven irrigation forecasting model with the mechanism forecasting model based on water balance to synthesize the final irrigation forecasting results; Step S8. Based on the final irrigation forecast results, release the forecast results of irrigation time, irrigation quota and water exchange volume in different levels, and drive irrigation and drainage operations; use measured data to continuously correct the data to drive the irrigation forecast model parameters and the coupling weight with the mechanism forecast model to form a closed loop operation.
[0025] In step S1, the collected multi-source data includes daily temperature, humidity, wind speed, sunshine, and rainfall monitoring data, as well as gridded meteorological forecast data for the foreseeable period; data on water depth in paddy fields and shrimp ditches, and groundwater depth; water quality monitoring data such as water temperature, dissolved oxygen, and transparency; and simultaneously data on field topography, soil texture, rice varieties and growth stages, crayfish farming density and feeding amount, as well as the irrigation water use plan reporting ledger of the irrigation district over the years, to construct a basic database for irrigation forecasting in the rice-shrimp farming area.
[0026] The rice-shrimp farming plots are divided into rice paddy areas and shrimp ditch areas. The ratio of shrimp ditch area is determined by: The rice-shrimp farming plots are divided into two water consumption units: rice paddy areas and shrimp ditch areas. The shrimp ditch area ratio η is defined as the ratio of the shrimp ditch area to the total area of the plot. For regular rectangular plots with shrimp ditches surrounding the entire plot, the following formula is used to calculate the water consumption based on the plot's geometric dimensions (plot zoning structure as follows). Figure 2 (as shown) (1) In the formula: , These represent the areas of shrimp ditches and rice paddies, respectively. The total area of the field. The length and width of the outer contour of the field. The shrimp ditch is wide.
[0027] For irregularly shaped paddy fields, orthophotos from UAVs or high-resolution remote sensing images are used to extract pixels of shrimp ditches and paddy fields by segmenting water surface and vegetation index thresholds. The area ratio η of shrimp ditches is determined according to the ratio of the number of pixels, so as to achieve rapid acquisition of zoning parameters.
[0028] Step S2 includes the following sub-steps: S2.1 Reference Crop Evapotranspiration Forecast: Based on the weather types in the forecast period, the daily reference crop evapotranspiration forecast is calculated using the following formula: (2) In the formula: For reference, crop evapotranspiration (mm / d) The weather type correction factor is calculated by classifying and statistically analyzing four weather types: sunny, partly cloudy (partly cloudy / cloudy), overcast, and rainy. It is obtained by the ratio of the average ET0 value under the weather type to the multi-year average ET0 value for the same day (general rule: sunny > partly cloudy > overcast > rainy). The multi-year average ET0 value can be calculated using historical data and the FAO Penman-Monteith formula.
[0029] S2.2 Forecast of Evapotranspiration in Paddy Fields During Rice Growth Stages: Water consumption in paddy fields is mainly due to rice evapotranspiration. Forecasts are made using a crop coefficient model determined by growth stage. (3) In the formula: Evapotranspiration in paddy fields (mm / d) The crop coefficient can be determined by providing empirical values based on the rice growth stage, or by using a dynamic adjustment method.
[0030] S2.3 Forecast of Free Surface Evaporation in Shrimp Ditches: Water consumption in the shrimp ditch area is mainly due to surface evaporation. The surface evaporation forecast values issued by the meteorological department are preferentially adopted. If no surface evaporation forecast values are available, a Dalton-type empirical model is used for calculation. (4) (5) In the formula: The saturated vapor pressure (kPa) is calculated based on water temperature (air temperature is used if no water temperature data is available), where T is the water temperature (°C). If no water temperature monitoring data is available, the average air temperature at a height of 2m is used instead. Evaporation rate of shrimp ditch surface (mm / d). A, B The wind speed function coefficient was determined by calibration using measured evaporation data from the shrimp ditch. The wind speed at a height of 2m (m / s) This represents the actual water vapor pressure (kPa).
[0031] S2.4 Forecast of total evapotranspiration for the entire field: During the rice growing season, the total evapotranspiration is calculated by weighting the area of the paddy field and shrimp ditch together; during the non-rice growing season, since there is no water consumption from rice in the field, the total evapotranspiration is calculated based on the water surface evaporation. (6) In the formula: The total evapotranspiration (mm / d) of the entire field is used. Water consumption in paddy fields and shrimp ditches is weighted by area ratio to unify the total water consumption at the whole-field scale, overcoming the deficiency of ordinary irrigation forecasting methods that do not distinguish between water consumption differences between ditches and fields. The structure of the mechanistic irrigation forecasting model is as follows: Figure 3 As shown.
[0032] Step S3 includes the following sub-steps: S3.1, Empirical Value Method: Using field trial observation data, determine the empirical values of daily seepage in the field according to the greening stage, tillering stage, jointing and booting stage, heading and flowering stage, milk stage, yellowing stage, and non-growth stage. (mm / d), forming a table of seepage parameters for different growth stages; for fields with scarce monitoring data, the values can be directly obtained from the table. Table 1. Empirical values of daily field seepage at different growth stages.
[0033] S3.2, Empirical Model Method: For fields with basic monitoring conditions, establish an empirical model for seepage volume related to surface water depth, groundwater depth, and soil properties. (7) In the formula: This represents the daily seepage rate in the field (mm / d). The water depth on the field surface (the water depth on the field surface is based on the bottom of the shrimp ditch). Due to the depth of groundwater, , The reference water depth and the reference groundwater depth are respectively taken as the median value of the suitable water layer in the field and the local multi-year average groundwater depth. As the benchmark leakage strength, It is an experience index. , The model was calibrated using field-measured seepage data and the least squares criterion. It describes how increased surface water depth exacerbates seepage, while increased groundwater depth weakens soil recharge.
[0034] Step S4 includes the following sub-steps: Rice-shrimp farming fields are mostly converted from low-lying, waterlogged fields. Lateral seepage at the outer boundaries of the fields is negligible. The lateral seepage mentioned in this application specifically refers to the water exchange between the paddy field and the shrimp ditch, and is quantitatively generalized according to the stage of the farming model (lateral seepage generalization for co-culture and separate culture stages is as follows). Figure 4 (as shown) S4.1 Rice-Shrimp Co-culture Stage: During the co-culture period, the rice paddy and shrimp ditch water surfaces are connected and basically level. Due to the difference in evaporation and transpiration between the rice paddy and the shrimp ditch, an instantaneous water level difference will occur between the two areas, which will return to normal within one day through lateral seepage between the paddy and ditch. The designed drawdown depth of the rice paddy area during the calculation period is... ,satisfy The water level in the Xiagou area has dropped to a depth of [missing information]. ,satisfy The amount of water exchanged through lateral seepage that restores the water level in the ditch to a balanced state is: (8) In the formula: The volume of water exchanged through lateral seepage (m³). The total area of the field (m²) is given, and the water level drawdown depth is expressed in meters (from...). ET , mm / d multiplied by ×10 - (³conversion) , The figures represent the vertical seepage rates (mm / d) in paddy fields and shrimp ditches, respectively; the lateral seepage direction is from the unit with slower water loss to the unit with faster water loss, thus achieving a quantitative generalization of the water exchange between ditches and fields under the co-cultivation model.
[0035] S4.2, Rice-Shrimp Separation Stage: During the rice drying, milk-ripe, and yellow-ripe stages, the water levels in the rice paddy and shrimp ditch are inconsistent. Lateral seepage will occur from the side with higher water level to the side with lower water level. Darcy's law is used for generalized calculation: (9) In the formula: The amount of water seepage during the time period (m³) is the amount of water that seeps into the side. The saturated permeability coefficient of the soil in the contact zone between the ditch and the field. The average depth of the cross-section where lateral seepage occurs. , These represent the water depths in the paddy fields and shrimp ditches, respectively. The infiltration path length is taken as the horizontal distance between the ditch and the field. L The length of the contact zone between the ditch and the field. To calculate the length of the time period; 、 The water level difference should be calculated after converting to the same elevation datum (e.g., if the bottom of the shrimp ditch is used as the datum, then the water level on the paddy field side is the depth of the water layer on the field surface). With trench depth sum).
[0036] Step S5 includes the following sub-steps: S5.1 Determination of Water Layer Standards: Based on the combined constraints of the suitable water layer for each growth stage of rice and the minimum water layer required for crayfish survival, the lower limit of the field water layer is determined. and the upper limit of the suitable water layer Among them, the lower limit of field water layer , The minimum water level in the shrimp ditches that is necessary to support crayfish survival corresponds to the depth of the field water layer. The lower limit of the suitable water layer for different growth stages of rice; the upper limit of the suitable water layer , The maximum water level in the crayfish ditch corresponds to the depth of the field water layer. The upper limit of the suitable water layer for different growth stages of rice; S5.2 Daily Recursive Forecast of Field Water Layer: Using the current measured water layer as the initial condition for calculation, the forecast of the expected rainfall period is input, and combined with the water consumption forecast results of steps S2-S4, a two-level water balance calculation mode is adopted: First, the water balance calculation of the whole field of the shrimp-rice farming unit is carried out to solve the overall water storage change of the unit; then, the lateral seepage exchange volume of the field ditch obtained in step S4 is introduced to complete the internal water redistribution of the paddy field sub-area and the shrimp ditch sub-area, and the actual water layer of the paddy field and the actual water layer of the shrimp ditch are obtained respectively.
[0037] The water balance calculation for the entire aquaculture unit is as follows: (10) For paddy field sub-areas, the water balance needs to incorporate the lateral seepage exchange term between ditches and fields: (11) For the shrimp ditch area, the water balance equation is: (12) In the formula: The water depth in the field on day t is (mm). , The values represent the water depth (mm) of the paddy field and shrimp ditch on day t. Forecast rainfall (mm) for the lead time period. Water volume (mm). , , The total water consumption, leakage, and lateral seepage predicted by equations (6), (7), (8), or (9) are respectively. The impact of lateral seepage on water intake in paddy fields is calculated according to the direction of water exchange with positive and negative signs; the lateral seepage exchange term... / , / Convert from m³ / m² to mm (×10³) and then include it.
[0038] S5.3, Irrigation Time and Irrigation Quota Forecast: When the predicted water level drops to the lower limit of the water level... When irrigation is triggered, the irrigation quota is determined based on restoring the field water level to the upper limit of the suitable water level: (13) (14) In the formula: The irrigation quota is in m³, with water layer units in mm and area units in m²; and is dynamically adjusted based on the forecast of rainfall in the foreseeable future: if the recursion shows that the rainfall in the foreseeable future... If the irrigation time is delayed accordingly, the dynamic irrigation system based on weather conditions will be forecasted.
[0039] S5.4 Water Exchange Coupling Constraint: When step S6 determines that water exchange is needed, priority is given to using water exchange for irrigation, and the water exchange volume is included in the calculation. This system participates in water layer advance calculations, avoiding duplicate metering between irrigation and water exchange, and enabling unified scheduling of irrigation and water exchange. The water exchange volume is calculated as water layer depth (×10³ / ) based on the total field area. Then it is included in I(t).
[0040] Step S6 includes the following sub-steps: S6.1 Water Temperature Prediction: Based on the foreseeable air temperature forecast, the daily water temperature of the shrimp ditch is predicted using a first-order heat exchange model. (15) In the formula: , These are water temperature and air temperature (°C), respectively. The heat exchange coefficient is determined using measured water and air temperature data. The water quality prediction and water exchange decision-making process is as follows: Figure 5 As shown.
[0041] S6.2 Dissolved oxygen prediction: Daily dissolved oxygen is predicted based on the relationship between saturated dissolved oxygen and water temperature and the reoxygenation-oxygen consumption balance equation. (16) (17) In the formula: Water temperature The corresponding saturated dissolved oxygen concentration (mg / L). This is the reoxygenation coefficient. The comprehensive oxygen consumption rate at a baseline water temperature of 25℃ (covering bottom sediment oxygen consumption, aquaculture organism respiration and organic matter decomposition, and categorized by stocking density and feeding amount). This is a temperature correction factor (taken as 1.02~1.10). The calculation period is long.
[0042] S6.3 Water Quality Replacement Threshold Determination: Dissolved oxygen and water temperature are selected as water replacement control indicators. Preset water quality replacement thresholds are used; if either indicator exceeds the limit, water replacement is deemed necessary, and the degree of exceedance is graded. The threshold is defined as follows: i Normalization limits for the indicator: (18) In the formula: , The first i The predicted (or measured) values and threshold values of the indicators; preferably, the reference value for the dissolved oxygen threshold is 5.0 mg / L and the reference value for the water temperature threshold is 32℃. A value greater than 0 indicates that the indicator has exceeded the limit.
[0043] S6.4 Water Exchange Quantity Forecast: The water exchange quantity is determined by combining water quality restoration target inversion with shrimp ditch replacement efficiency correction. For control indicators exceeding the limit, the required water exchange quantity to achieve the water quality restoration target is determined by inverting the pre-water exchange field water quality index value, water source index value, and post-water exchange target value based on the mixed dilution material balance. Considering the characteristics of short-flow and actual replacement rate being lower than nominal replacement ratio in the circular shrimp ditch water exchange process, a replacement efficiency coefficient is introduced to correct the inverted water exchange quantity. When multiple indicators exceed the limit simultaneously, the required water exchange quantity is determined separately for each indicator, and the largest value is used for verification. (19) In the formula: The theoretical water exchange volume (m³) required to achieve the water quality restoration target. This refers to the total amount of water stored in the field (m³). The values of field water control indicators before water exchange (dissolved oxygen and water temperature are the predicted values from steps S6.1 and S6.2) are used. The corresponding index values for the water source are used (dissolved oxygen is taken from the near-saturation value or the measured value of the water temperature, and the water temperature is taken from the measured or predicted value of the water source). The target values for the control indicators after water exchange are defined as follows (determined with a safety margin reserved according to the control indicator thresholds; the target value for dissolved oxygen is taken as 0.5~1.0 mg / L above the threshold, and the target value for water temperature is taken as 1~2℃ below the threshold); ξ is the shrimp ditch replacement efficiency coefficient, which characterizes the reduction in the actual replacement rate caused by short-flow during water exchange in the annular shrimp ditch, and is calibrated using the water temperature response process before and after water exchange or tracer experiments (generally taken as 0.70~0.95). The corrected nominal water exchange volume (m³) is used; during water exchange, drainage precedes irrigation, and drainage and irrigation are measured simultaneously. The water exchange volume is included in the water layer balance calculation in step S5. For cases where water quality monitoring data is insufficient or cannot be determined... For fields with ξ, the water exchange volume is determined using the proportional coefficient method: =λ· , λ= (1+Σ ), The basic water exchange ratio is 0.20~0.30. Let be the weight coefficient of the i-th indicator. The limit of equation (18) is the limit of non-limit (the non-limit index is 0).
[0044] Total water storage in the field Determined by the sum of water storage in the paddy field area and the shrimp ditch area: (20) In the formula: , The areas (m²) of the paddy field area and the shrimp ditch area are respectively determined by step S1.2. , , The total area of the field (m²); , These represent the water depth (mm) of the paddy field surface and the shrimp ditch area on the accounting day. Based on the field surface, Using the bottom of the shrimp ditch as a benchmark, take the actual measured value of the day or the water layer extrapolation value from step S5.2; when there is no water layer on the field surface during the shrimp-rice separation stage or the field drying period. Take 0; during the shrimp-rice co-culture stage, the water level in ditches and fields remained basically the same. The total amount of water stored in the field is calculated using the following simplified formula: (twenty one) In the formula: h The water depth (mm) is uniform across the entire field. The depth of the shrimp ditch is the height difference (mm) between the bottom of the shrimp ditch and the surface of the field, which is determined by actual measurement of the completed cross-section of the field.
[0045] Step S7 includes the following sub-steps: S7.1 Compilation of Irrigation Water Use Plan Reporting Data: Collect irrigation water use plan data reported by irrigation users in the irrigation area over the years in the format of "plot number, planting and breeding mode, plot area, growth period, date of water demand declaration, and water demand declaration", and build a sample database of managers' experience together with rainfall, temperature and field water layer monitoring data for the corresponding period. S7.2 Data-Driven Irrigation Forecasting Model: Using a database of managers' experience samples as training samples, a BP neural network data-driven forecasting model is constructed to make managers' irrigation decision-making experience explicit in the forecasting model. (twenty two) In the formula: The predicted irrigation amount output by the data-driven model. The input feature vector includes the cropping and breeding pattern code, growth period code, current field water level, forecasted rainfall, forecasted evapotranspiration, and the date and amount of the last irrigation. , , , For network weights and thresholds, For activation function, This serves as the model parameter set; simultaneously, a decision tree algorithm is used to extract a "condition-flooding" rule base from the same sample database to verify the rationality of the neural network prediction results.
[0046] Furthermore, during the training of the data-driven model, the water balance mechanism is embedded as a soft constraint into the training objective function, and the mechanism constraint penalty term is constructed as follows: For each training sample, the data-driven model outputs the predicted irrigation amount according to equation (22). ; forecast irrigation volume The total area of the field is converted into the water depth of the entire field and then substituted into formula (10). The water depth in the field is then calculated day by day to obtain the calculated water depth. ; to advance the water level beyond the lower limit of the field water level and the upper limit of the suitable water layer The sum of squared out-of-bounds depths of the interval is used as the penalty for the computer-mandated constraint, and the daily penalty is [max(0, h]. 下 )]² and [max(0, h 上 The sum of ]²; the average of the sums of each day in the forecast period and all training samples, and weighted with the sample fitting error to form the training objective function: (twenty three) In the formula: The sample fitting error is the squared mean of the deviations between the network output forecast irrigation amount and the sample reported irrigation amount. To constrain the penalty by mechanism, the mean of the sum of squares of the out-of-bounds depths for each day of the forecast period of all training samples is taken; To constrain the weighting coefficients and harmonize the dimensions and magnitudes between the fitting error and the mechanism constraint penalty, the coefficients were optimized and determined within the range of 0.1 to 1.0 through grid trial calculations. To predict the irrigation volume, the total area of the field is converted into the water layer depth and then entered into formula (10) to calculate the field water layer (mm).
[0047] By embedding training with the aforementioned mechanistic constraints, the data-driven model learns from the irrigation decision-making experience of managers while satisfying the water layer regulation mechanism constraints in the shrimp-rice farming area. This suppresses forecast outputs that violate the water balance mechanism from the training mechanism perspective, overcoming the defect that pure data fitting models may output forecast results that violate the mechanism.
[0048] S7.3 Mechanism-Data Dual-Driven Coupled Forecasting: The forecast primarily relies on mechanistic model prediction, supplemented by data-driven model correction. A residual coupling method is used to generate the final irrigation forecast. The deviation correction amount of the residual coupling is based on the mechanistic model replay forecast values of historical samples, characterizing the systematic deviation of irrigation district managers' irrigation decision-making experience relative to the mechanistic forecast. (twenty four) In the formula: To provide the final forecast of irrigation volume, To predict irrigation amount using mechanistic models, The predicted values are replayed using the mechanistic model corresponding to historical samples. The mechanism forecast bias correction is determined by the difference between the data-driven model output and the historical sample mechanism replay forecast values. During forecasting, the statistical values (such as the mean or moving average) of the historical sample mechanism forecast bias corrections are used as... For the daily mechanistic model forecast value Corrections are made; when there are insufficient management experience samples or the data-driven model fails to converge, it degenerates into a weighted fusion of the inverse of the error variance: (25) In the formula: The coupling weights are determined by the reciprocal of the historical forecast error variances of the two models. , These represent the standard deviations of historical forecast errors for the mechanistic model and the data-driven model, respectively. The weights are dynamically adjusted based on the rolling errors of the two models to ensure the robustness of the dual-drive coupled forecast under data abundance / scarcity conditions. The data-driven model and dual-drive coupled architecture are as follows: Figure 6 As shown.
[0049] Step S8 includes the following sub-steps: A three-tiered forecasting and dissemination module is established, connecting the irrigation district, management unit, and water user. The final forecasted irrigation time, irrigation quota, and water exchange volume are pushed to the management unit and water user at these levels, driving pumping stations and gates to execute irrigation and drainage operations. The forecast results are verified daily using measured water level, water quality, and irrigation / drainage metering data, and the relative error of the irrigation volume forecast is analyzed. The forecast accuracy is assessed, and the model parameters and dual-drive coupling weights are continuously adjusted to form a closed-loop operation mechanism of "forecast-execution-feedback-correction".
[0050] This embodiment uses a shrimp-rice farming demonstration area in an irrigation district of the Jianghan Plain as the application object to implement the irrigation forecasting method and system of the present invention. The specific steps are as follows: Step 1: Access the daily meteorological monitoring and 7-day gridded meteorological forecast data of the meteorological station in the demonstration area, the automatic monitoring data of field water layer and groundwater depth, and the online monitoring data of water temperature and dissolved oxygen. Collect more than 1,260 irrigation water plan reports from the demonstration area in the past 5 years and build a basic database. The outer contour of the field in the demonstration area is 100m long and 67m wide, and the width of the shrimp ditch is 3.0m. Calculate the shrimp ditch area ratio η≈0.144 according to formula (1) and verify it with the segmentation results of UAV orthophoto images. The deviation between the two is less than 2%.
[0051] Step 2: Calculate daily from the forecast data for the foreseeable period according to formula (2). ET 0; Predict the evaporation and transpiration of paddy fields according to formula (3); The evaporation of shrimp ditch water surface shall be preferentially based on the water surface evaporation forecast value issued by the meteorological department (in Table 2 of this embodiment). That is, using the published value); if there is no published value, calculate according to formula (4) and formula (5), and calibrate to obtain A =0.26、B =0.14; The total evapotranspiration of the whole field is weighted according to formula (6), and the water surface evaporation is directly used during the non-rice growing season.
[0052] Step 3: The leakage rate is calculated according to the empirical model of formula (7) and calibrated. K 0 = 4.2 mm / d α =0.35、 β =0.25, reference water depth h 0. The appropriate water layer median is 1555mm (the depth relative to the bottom of the ditch, i.e., 55mm relative to the field surface), which is the benchmark groundwater depth. d g0 Take 1.2m; for fields with scarce data, directly use the empirical values in Table 1.
[0053] Step 4: In the co-cultivation stage, the lateral seepage exchange between the paddy field and the ditch is generalized according to formula (8). On a typical sunny day, the water level drop in the paddy field area is 11.2 mm / d and in the shrimp ditch area it drops by 9.5 mm / d, with a lateral seepage exchange of approximately 1.4 m³ / d. In the separate cultivation stage, during the drying period, the water level difference between the paddy field and the shrimp ditch is 0.4 m. The lateral seepage volume is calculated according to formula (9). K s Take 2.0 × 10 -6 m / s.
[0054] Step 5: Determine the lower limit of the water layer (30mm) and the upper limit of the suitable water layer (80mm) according to the joint constraints of the water layer (the shrimp ditch should maintain the water level for shrimp survival accordingly); advance the water layer daily according to formulas (10)-(14) and predict the irrigation time and irrigation quota, and postpone the irrigation in conjunction with the predicted rainfall. The typical irrigation quota for a single field in the demonstration area is about 350m³ / time; the water exchange volume during the water exchange period is included in the unified scheduling of the irrigation item. The daily water balance advance process for the 7-day predicted period is shown in Table 2, and the water layer advance process line is as follows. Figure 7 As shown.
[0055] Table 2. Daily Water Balance Forecast Process for 7 Days During the Forecast Period
[0056] Step 6: Predict the water temperature according to formula (15). k =0.18), according to equations (16) and (17), the dissolved oxygen ( =0.12 / d, The dosage is determined according to the stocking density; in this example, it is 0.53 mg / (L·d). =1.05); when the predicted dissolved oxygen is below 5.0 mg / L and the water temperature is above 32℃, water replacement is determined. The water replacement volume is verified by combining the water quality recovery target inversion with the replacement efficiency correction according to formula (19). The shrimp ditch replacement efficiency coefficient is calibrated to ξ=0.88 through the water temperature response process of water replacement. The daily water quality prediction and water replacement determination results are shown in Table 3, and the process line is as follows. Figure 8As shown. On the 3rd day, the predicted dissolved oxygen level dropped to 4.67 mg / L, below the 5.0 mg / L threshold, indicating a need for water replacement. Since the water temperature did not exceed the limit, the target dissolved oxygen value after water replacement was set at 5.5 mg / L (a safety margin of 0.5 mg / L above the threshold). The dissolved oxygen level of the source water was set at 8.3 mg / L, based on the near-saturation value of the source water temperature. The surface water level on that day was 39.8 mm. The total field water storage was determined based on the sum of the water storage in each zone. = (5734 × 39.8 + 966 × 1539.8) × 10 - Theoretical water exchange capacity is approximately 1716 m³. =1716×(5.5) 4.67) / (8.3 4.67)≈392m³, the nominal water exchange volume is obtained by correcting for the replacement efficiency coefficient ξ=0.88. ≈445m³; From the 6th day onwards, both water temperature and dissolved oxygen exceed their limits. The required water exchange volume for each indicator is calculated separately, and the larger value is used for verification. The water exchange volume is increased accordingly. According to step S5.4, water exchange is coupled and occurs adjacently with irrigation on the 4th day. They are combined and uniformly scheduled. Water exchange is performed by first draining and then irrigating: first drain 445m³, then introduce 795m³ of water (of which 445m³ is used for replacing the shrimp ditch water, and 350m³ is used to restore the field surface water layer from 27.7mm to the suitable upper limit of 80mm). The planned irrigation of 350m³ on the 4th day is incorporated into the water exchange operation and completed without repeated water introduction. In Table 3, column λ represents the water exchange ratio using the proportional coefficient method (degradation scheme when data is insufficient), calculated as λ= (1+Σ ) calculation, in this embodiment Take 0.25, Take 0.6, Take 0.4.
[0057] Table 3. Water Quality Forecast and Water Exchange Judgment Table for 7-Day Forecast Period
[0058] Step 7: Train a BP neural network data-driven model (8 hidden layers, training and validation sets divided in an 8:2 ratio) using water use planning data reported over the past 5 years, and output the data-driven forecast value according to equation (22). Extract 23 irrigation rules from the decision tree for rationality verification. Output the final forecast result by residual coupling according to equation (24). For periods with insufficient samples, use equation (25) for weighted fusion. Construct a training objective function with mechanism constraint penalties according to equation (23), and constrain the weight coefficients. After grid trial calculation with a value of 0.3, the root mean square error of the irrigation volume forecast in the validation set decreased from 8.6 m³ without embedded constraints to 7.3 m³ (this error represents the fitting deviation between the network output and the sample reported irrigation volume). The rate of exceeding the limit when substituting the predicted irrigation volume into the water layer recursion decreased from 6.4% to 0.8%. Examples of the 23 rules extracted from the decision tree are as follows: ① If the tillering stage is underway, there is no effective rainfall for 3 days in the forecast period, and the current field water layer is ≤35 mm, then irrigate the next day; ② If the predicted water temperature is ≥32℃ or dissolved oxygen is ≤5.0 mg / L, then initiate water exchange. The consistency rate between the neural network output and the rule base judgment is approximately 91%. Historical sample playback statistics yielded the standard deviation of the forecast error between the mechanistic model and the data-driven model. ≈38m³ ≈42m³ (the forecast error compared to the actual irrigation volume), coupling weight κ≈0.55.
[0059] Step 8: Establish a three-level forecasting and dissemination platform to continuously release irrigation and water exchange forecasts for the next 7 days at 8:00 AM daily. Calculation results in 2025 show that the irrigation time forecast deviation of the dual-drive coupled forecast does not exceed 1 day, and the relative error of the irrigation volume forecast does not exceed 10%, fully meeting the irrigation forecasting and scheduling needs of the shrimp-rice farming area. Example of dissemination results: A 7-day forecast is continuously released at 8:00 AM daily, such as "Water exchange will begin on the 3rd day, with an exchange volume of approximately 445 m³; irrigation will be integrated into the water exchange implementation on the 4th day, and separate water diversion will no longer be required." 2025 flood season application statistics: Irrigation time forecast deviation ≤ 1 day accounted for 11 / 12 (91.7%), the average relative error of the dual-drive coupled irrigation volume forecast was 7.1%, with a maximum of 8.8%, and the average relative error of the water exchange volume forecast was 8.5%. A comparison of the accuracy of the mechanism model's individual forecast and the dual-drive coupled forecast is shown in Table 4. Figure 9 .
[0060] Table 4. Statistics on Irrigation Forecast Accuracy in 2025
[0061] Example 2 like Figure 10 As shown, this embodiment provides an irrigation forecasting system for shrimp-rice farming areas based on both data and mechanism-driven approaches, including: The data acquisition module is used to acquire multi-source data of the irrigation area; based on the spatial structure information of the irrigation area fields, the shrimp-rice farming fields are divided into rice fields and shrimp ditch areas, and the shrimp ditch area ratio is determined. The mechanism prediction model has built-in regional evaporation and transpiration prediction units, leakage prediction units, lateral seepage prediction units, water layer water balance irrigation prediction units, and water quality prediction and water exchange prediction units. The zonal evapotranspiration forecasting unit is used to determine the forecast value of total evapotranspiration of the whole field according to the meteorological forecast data and the current rice growth stage information, based on the rice growth period and non-rice growth period. The seepage forecasting unit is used to forecast field seepage based on data of different rice growth stages and / or field water layer and groundwater depth, and to obtain field seepage forecast values. The side seepage prediction unit is used to predict the side seepage between the paddy field and the shrimp ditch according to the different water layer differences in the shrimp-rice co-culture stage and the shrimp-rice separate culture stage, and obtain the side seepage prediction value. The water layer water balance irrigation forecasting unit is used to construct a mechanism forecasting model based on the principle of water layer water balance. The mechanism forecasting model determines the lower limit of the field water layer and the upper limit of the suitable water layer according to the joint constraints of rice and crayfish. Based on the forecast of rainfall and water consumption, the field water layer is progressively advanced day by day. When the water layer drops to the lower limit of the field water layer, irrigation is triggered and the irrigation quota is determined. The irrigation time is dynamically extended in combination with the forecast of rainfall. The water quality prediction and water exchange forecasting unit determines the timing of water exchange based on the shrimp ditch water quality monitoring data or water quality forecast data, according to the water quality control index threshold, and verifies the water exchange volume, and incorporates the water exchange volume into the unified scheduling of water layer water balance. The dual-drive coupling module is used to compile the annual irrigation water use plan reporting data of the irrigation district into a manager experience sample library, train the data-driven irrigation forecasting model, and couple the data-driven irrigation forecasting model with the mechanism forecasting model based on water balance to synthesize the final irrigation forecasting results. The rolling correction module is used to release irrigation time, irrigation quota and water exchange volume forecasts in stages according to the final irrigation forecast results, and drive irrigation and drainage operations; it uses measured data to roll correct the data to drive the irrigation forecast model parameters and the coupling weights with the mechanism forecast model to form a closed loop operation. The forecast release and execution module is used to push irrigation time, irrigation quota and water exchange volume to management units and water users in a tiered manner, and drive pumping stations and gates to perform irrigation and drainage operations.
[0062] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.
[0063] It should be understood that any parts not described in detail in this specification belong to the prior art.
[0064] It should be understood that the above description of the preferred embodiments is quite detailed, but it should not be considered as a limitation on the scope of protection of this invention. Those skilled in the art, under the guidance of this invention, can make substitutions or modifications without departing from the scope of protection of the claims of this invention, and all such substitutions or modifications fall within the scope of protection of this invention. The scope of protection of this invention should be determined by the appended claims.
Claims
1. A data- and mechanism-driven method for irrigation forecasting in shrimp-rice farming areas, characterized in that, Includes the following steps: Step S1. Obtain multi-source data of the irrigation area; based on the spatial structure information of the irrigation area fields, divide the shrimp-rice farming fields into rice fields and shrimp ditch areas, and determine the shrimp ditch area ratio; Step S2. Based on the meteorological forecast data and the current rice growth stage information, determine the forecast value of total evapotranspiration in the whole field according to the rice growth stage and non-rice growth stage; Step S3. Based on data on different growth stages of rice and / or field water layer and groundwater depth, predict the field seepage amount and obtain the predicted value of field seepage amount; Step S4. Based on the different water layer conditions in the rice-shrimp co-culture stage and the rice-shrimp separate culture stage, predict the lateral seepage volume between the paddy field and the shrimp ditch, and obtain the lateral seepage volume forecast value. Step S5. Construct a mechanism forecasting model based on the principle of water balance in the water layer. The mechanism forecasting model determines the lower limit of the field water layer and the upper limit of the suitable water layer according to the joint constraints of rice and crayfish. Based on the forecast of rainfall and water consumption in the foreseeable period, the field water layer is progressively advanced day by day. When the water layer drops to the lower limit of the field water layer, irrigation is triggered and the irrigation quota is determined. The irrigation time is dynamically extended in combination with the forecast of rainfall in the foreseeable period. Step S6. Based on the shrimp ditch water quality monitoring data or water quality forecast data, determine the timing of water exchange according to the water quality control index threshold, verify the water exchange volume, and incorporate the water exchange volume into the unified scheduling of water layer water balance. Step S7. Compile the historical irrigation water use plan reporting data of the irrigation district into a manager experience sample database, train the data-driven irrigation forecasting model, and couple the data-driven irrigation forecasting model with the mechanism forecasting model based on water balance to synthesize the final irrigation forecasting results; Step S8. Based on the final irrigation forecast results, release the forecast results of irrigation time, irrigation quota and water exchange volume in different levels, and drive irrigation and drainage operations; The measured data is used to continuously correct the parameters of the irrigation forecast model and the coupling weights with the mechanism forecast model, forming a closed-loop operation.
2. The irrigation forecasting method for shrimp-rice farming areas based on a dual-driven approach of data and mechanism, as described in claim 1, is characterized in that... The multi-source data includes meteorological monitoring data, meteorological forecast data, field water layer and groundwater depth data, water quality monitoring data, planting and breeding production data, and irrigation water use plan reporting data of the irrigation district over the years.
3. The irrigation forecasting method for shrimp-rice farming areas based on a dual-driven approach of data and mechanism, as described in claim 1, is characterized in that... Step S2 includes the following sub-steps: The weather type correction coefficient is determined based on the weather type in the forecast period meteorological forecast, and the daily reference crop evapotranspiration forecast value is calculated based on the weather type correction coefficient and the multi-year average reference crop evapotranspiration. During the rice growing season, a crop coefficient is determined based on the rice's growing season, and the crop coefficient is multiplied by the predicted value of evapotranspiration of the reference crop to obtain the predicted value of evapotranspiration in the paddy field area. When forecasting the evaporation of the Xiagou area, the evaporation forecast value issued by the meteorological department is used; when there is no evaporation forecast value, the Dalton-type empirical model is used to calculate the evaporation forecast value of the Xiagou area. When the rice is in its growing season, the predicted values of evapotranspiration in the rice paddy area and the predicted values of water surface evaporation in the shrimp ditch area are weighted and summed according to the proportion of rice paddy area to shrimp ditch area to synthesize the total evapotranspiration of the whole field; when the rice is not in its growing season, the predicted value of water surface evaporation of the shrimp ditch is directly used for the total evapotranspiration of the whole field.
4. The irrigation forecasting method for shrimp-rice farming areas based on a dual-driven approach of data and mechanism, as described in claim 1, is characterized in that... The methods for predicting field seepage include the empirical value method and the empirical model method. The empirical value method involves using field experiment observation data to determine the empirical value of daily seepage in the field according to different growth stages of rice, and forming a parameter table of seepage in the field for different growth stages. The empirical model method involves establishing an empirical model of seepage related to surface water depth, groundwater depth, and soil properties for fields with basic monitoring conditions. Based on this empirical model, the predicted field seepage is calculated. The empirical model for seepage is as follows: In the formula: This represents the daily seepage rate in the field (mm / d). The water depth on the field surface (the water depth on the field surface is based on the bottom of the shrimp ditch). Due to the depth of groundwater, , The reference water depth and the reference groundwater depth are used as the reference depths. As the benchmark leakage strength, It is an experience index. , Calibration was performed using field-measured seepage data and the least squares criterion.
5. The irrigation forecasting method for shrimp-rice farming areas based on a dual-driven approach of data and mechanism, as described in claim 1, is characterized in that... Step S4 specifically includes: During the rice-shrimp co-culture stage, the difference in water level drop between the rice paddy area and the shrimp ditch area is calculated based on the water level drop depth caused by evaporation and vertical seepage in the rice paddy area and the water level drop depth caused by surface evaporation and vertical seepage in the shrimp ditch area. Combined with the proportion of rice paddy area, the proportion of shrimp ditch area and the total area of the fields, the amount of water exchanged by lateral seepage is generalized. During the rice-shrimp separation stage, the amount of lateral seepage is calculated according to Darcy's law. The calculation of Darcy's law involves the soil saturated permeability coefficient of the ditch-field contact zone, the average depth of the lateral seepage cross section, the water level difference between the paddy field and the shrimp ditch after conversion to the same elevation datum, the seepage path length, the length of the ditch-field contact zone, and the calculation period.
6. The irrigation forecasting method for shrimp-rice farming areas based on a dual-driven approach of data and mechanism, as described in claim 1, is characterized in that... Step S5 includes: Based on the combined constraints of the suitable water layer for each growth stage of rice and the minimum water layer required for the survival of crayfish, the lower limit of the field water layer and the upper limit of the suitable water layer are determined. The lower limit of the field water layer is the larger value between the lower limit of the suitable water layer for each growth stage of rice and the lower limit of the water layer required for the survival of crayfish. The upper limit of the suitable water layer is the smaller value between the upper limit of the suitable water layer for each growth stage of rice and the upper limit of the water layer required for the survival of crayfish. Using the current measured water layer as the initial condition, the forecast of the expected rainfall period is input and coupled with the forecast values of total evapotranspiration, field seepage, and lateral seepage, and the field water layer is recursively calculated day by day using a two-level water balance calculation model: first, the water balance calculation is performed on the entire field of the shrimp-rice farming unit, and then the lateral seepage exchange between the paddy field and the shrimp ditch is introduced to obtain the actual water layer of the paddy field and the actual water layer of the shrimp ditch respectively; Irrigation is triggered when the predicted water level drops to the lower limit of the field water level, and the field water level is restored to the upper limit of the appropriate water level to determine the irrigation quota; if the prediction shows that the water level will not fall below the lower limit of the field water level after rainfall replenishment in the future forecast period, the irrigation time will be postponed. When it is determined that water needs to be replaced, the amount of water replaced will be included in the water volume item for water layer recursion.
7. The irrigation forecasting method for shrimp-rice farming areas based on a dual-driven approach of data and mechanism, as described in claim 1, is characterized in that... Step S6 includes: Based on the early-stage temperature forecast, the daily water temperature of the shrimp ditch is predicted using a first-order heat exchange model. Based on the relationship between saturated dissolved oxygen and water temperature, and the reoxygenation-oxygen consumption balance equation, combined with the reoxygenation coefficient, the comprehensive oxygen consumption rate at the reference water temperature, and the temperature correction coefficient, the dissolved oxygen concentration in the shrimp ditch is predicted daily. Dissolved oxygen and water temperature are selected as water exchange control indicators, and water quality exchange thresholds are preset. If either indicator exceeds the limit, water exchange is determined to be required. Normalized exceedance limits for dissolved oxygen and water temperature are defined. For dissolved oxygen, the exceedance limit is the ratio of the difference between the threshold and the predicted value to the threshold. For water temperature, the exceedance limit is the ratio of the difference between the predicted value and the threshold to the threshold. For control indicators that exceed the limit, the theoretical water exchange volume is calculated by inverting the field water body index value before water exchange, the water source index value, and the target value after water exchange according to the principle of mixed dilution material balance. The shrimp ditch replacement efficiency coefficient is then introduced to correct the calculated water exchange volume. When multiple indicators exceed the limit at the same time, the required water exchange volume for each indicator is calculated separately and the largest value is taken for verification. When water quality monitoring data is insufficient, the method of proportional coefficient is used to determine the water exchange volume; the actual water exchange volume is determined by multiplying the basic water exchange ratio by the dynamic adjustment factor of the degree of exceeding the limit, and then multiplying it by the total amount of water stored in the field. The total water storage in the field is determined by the sum of the water storage in the paddy field area and the shrimp ditch area. During the rice-shrimp co-culture stage, the water storage is calculated based on the uniform water depth of the entire field surface and the depth of the shrimp ditch. During the rice-shrimp separate culture stage or when there is no water layer on the field surface during the drying period, the water storage in the paddy field area is taken as zero.
8. The irrigation forecasting method for shrimp-rice farming areas based on a dual-driven approach of data and mechanism, as described in claim 1, is characterized in that... The training methods for the data-driven irrigation forecasting model include: Collect irrigation water use plan data submitted by irrigation users in the irrigation district over the years according to basic field information and the format of the declared water demand date and declared water demand, and build a sample database of managers' experience together with the corresponding rainfall, temperature and field water layer monitoring data; Using the manager experience sample library as training samples, a data-driven irrigation forecasting model based on neural networks is constructed. At the same time, a decision tree algorithm is used to extract a condition-irrigation rule library from the same sample library to verify the forecasting results of the data-driven irrigation forecasting model. During training, the predicted irrigation volume output by the data-driven irrigation forecasting model is converted into water layer depth according to the total area of the field and then incorporated into the water balance equation to recursively calculate the field water layer on a daily basis. The sum of the squares of the overflow depths of the recursively calculated water layer beyond the upper and lower limits of the water layer is used to calculate the theoretical constraint penalty. The theoretical constraint penalty is then weighted and summed with the sample fitting error to form the training objective function.
9. The irrigation forecasting method for shrimp-rice farming areas based on a dual-driven approach of data and mechanism, as described in claim 1, is characterized in that... The coupling forecasting methods of the data-driven irrigation forecasting model and the mechanism forecasting model include: The difference between the output value of the data-driven model and the replay prediction value of the mechanism model corresponding to the historical samples is used as the mechanism prediction bias correction amount, and the sum of the mechanism model prediction value and the mechanism prediction bias correction amount is used as the final prediction value. When the data-driven model fails to converge, the coupling weights are determined by the inverse of the historical forecast error variances of the mechanistic model and the data-driven model and then weighted and fused. The coupling weights are dynamically adjusted according to the rolling forecast errors of the two models.
10. An irrigation forecasting system for shrimp-rice farming areas based on both data and mechanism-driven approaches, characterized in that... include: The data acquisition module is used to acquire multi-source data from the irrigation area; Based on the spatial structure information of the irrigation area, the shrimp-rice farming plots are divided into rice paddy areas and shrimp ditch areas, and the area ratio of shrimp ditch is determined. The mechanism prediction model has built-in regional evaporation and transpiration prediction units, leakage prediction units, lateral seepage prediction units, water layer water balance irrigation prediction units, and water quality prediction and water exchange prediction units. The zonal evapotranspiration forecasting unit is used to determine the forecast value of total evapotranspiration of the whole field according to the meteorological forecast data and the current rice growth stage information, based on the rice growth period and non-rice growth period. The seepage forecasting unit is used to forecast field seepage based on data of different rice growth stages and / or field water layer and groundwater depth, and to obtain field seepage forecast values. The side seepage prediction unit is used to predict the side seepage between the paddy field and the shrimp ditch according to the different water layer differences in the shrimp-rice co-culture stage and the shrimp-rice separate culture stage, and obtain the side seepage prediction value. The water layer water balance irrigation forecasting unit is used to construct a mechanism forecasting model based on the principle of water layer water balance. The mechanism forecasting model determines the lower limit of the field water layer and the upper limit of the suitable water layer according to the joint constraints of rice and crayfish. Based on the forecast of rainfall and water consumption, the field water layer is progressively advanced day by day. When the water layer drops to the lower limit of the field water layer, irrigation is triggered and the irrigation quota is determined. The irrigation time is dynamically extended in combination with the forecast of rainfall. The water quality prediction and water exchange forecasting unit determines the timing of water exchange based on the shrimp ditch water quality monitoring data or water quality forecast data, according to the water quality control index threshold, and verifies the water exchange volume, and incorporates the water exchange volume into the unified scheduling of water layer water balance. The dual-drive coupling module is used to compile the annual irrigation water use plan reporting data of the irrigation district into a manager experience sample library, train the data-driven irrigation forecasting model, and couple the data-driven irrigation forecasting model with the mechanism forecasting model based on water balance to synthesize the final irrigation forecasting results. The rolling correction module is used to release the forecast results of irrigation time, irrigation quota and water exchange volume in stages according to the final irrigation forecast results, and drive irrigation and drainage operations. The measured data is used to continuously correct the parameters of the irrigation forecast model and the coupling weights with the mechanism forecast model, forming a closed-loop operation. The forecast release and execution module is used to push irrigation time, irrigation quota and water exchange volume to management units and water users in a hierarchical manner, and drive pumping stations and gates to perform irrigation and drainage operations; The data- and mechanism-driven irrigation forecasting method for shrimp-rice farming areas is used to perform the steps in the data- and mechanism-driven irrigation forecasting method for shrimp-rice farming areas as described in any one of claims 1-9.
