Comprehensive management system for zero-carbon water resources in agricultural irrigation area

Abstract

The invention relates to the field of water resource management, in particular to an agricultural irrigation district zero-carbon water resource comprehensive management system, which comprises an irrigation district basic data calculation module for respectively calculating soil latent heat flux of an irrigation district and basic transpiration of crops according to detected air temperature, air pressure, solar radiation and wind speed; the irrigation district data prediction module generates predicted water storage and utilization amount of the irrigation district through an irrigation district prediction model based on ConvLSTM architecture; the water demand fusion prediction module generates comprehensive predicted water demand of the irrigation area through a fusion mapping model based on a gating mechanism and multi-scale convolution; and the zero-carbon water resource management module calculates the predicted net irrigation amount of the irrigation area based on the comprehensive predicted water demand and performs irrigation management on a photovoltaic water pump and a photovoltaic valve of the irrigation area. According to the method, the water resource management accuracy of the irrigation area is improved, the scheduling energy consumption of the photovoltaic water pump and the photovoltaic valve with zero carbon emission is optimized, and zero carbon transformation of the agricultural irrigation area is optimized and promoted from an irrigation energy consumption source.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of water resource management, and particularly relates to a zero-carbon water resource comprehensive management system for agricultural irrigation areas. BACKGROUND

[0002] In the water resource management of agricultural irrigation areas, with the promotion of "double carbon" smart agriculture, the industry finds that the scheduling efficiency of irrigation equipment directly affects the water resource utilization benefit and carbon emission level. The scheduling of water pumps and valves in traditional irrigation areas relies on manual experience or simple timing control, and lacks consideration of the coordination of crop water demand dynamics and meteorological changes, resulting in prominent problems of high energy consumption and high emission.

[0003] Currently, clean energy equipment such as photovoltaic water pumps has been introduced in irrigation areas, but there are still obvious shortcomings in scheduling management.

[0004] The operation of photovoltaic water pumps relies on solar power generation, and solar power has strong intermittency and volatility. If the precise water demand of crops cannot be accurately predicted and scheduled, the irrigation will be interrupted when the zero-carbon energy supply is insufficient, which will affect crop growth. That is, if the crops are accurately irrigated according to the real-time predicted demand, water can be saved, energy consumption can be reduced, and accurate matching of photovoltaic pumping and photovoltaic power generation can be achieved.

[0005] Therefore, how to improve the water demand prediction of agricultural irrigation areas to realize precise energy-saving zero-carbon photovoltaic water resource management is a technical problem to be solved at present. SUMMARY

[0006] Therefore, the present application provides a zero-carbon water resource comprehensive management system for agricultural irrigation areas, which predicts the water demand of crops by fusing soil latent heat flux and basic evapotranspiration through a ConvLSTM model, improves the precision of water resource management in irrigation areas, reduces waste and imbalance, optimizes the scheduling energy consumption of zero-carbon emission photovoltaic water pumps and photovoltaic valves, and realizes the optimization of energy head from irrigation to promote the zero-carbon transformation of agricultural irrigation areas.

[0007] To achieve the above purpose, the present application provides a zero-carbon water resource comprehensive management system for agricultural irrigation areas, comprising:

[0008] An irrigation area basic data calculation module is used to calculate the soil latent heat flux of the irrigation area and the basic evapotranspiration of the crops according to the detected air temperature, air pressure, solar radiation and wind speed, respectively;

[0009] An irrigation area data prediction module is used to generate the predicted water storage and utilization amount of the irrigation area based on the multi-dimensional meteorological data and crop data through an irrigation area prediction model based on the ConvLSTM architecture;

[0010] The water demand fusion prediction module is configured to fuse the soil latent heat flux, the base transpiration and the predicted water storage utilization through a fusion mapping model based on a feature fusion gating mechanism and a multi-scale convolution to generate a comprehensive predicted water demand of the irrigation area.

[0011] The zero-carbon resource management module is configured to calculate a predicted net irrigation amount of the irrigation area based on the comprehensive predicted water demand, and perform irrigation management on a photovoltaic water pump and a photovoltaic valve of the irrigation area based on the predicted net irrigation amount of the irrigation area.

[0012] Further, the irrigation area prediction model comprises a ConvLSTM layer and a feature mapping layer, and the irrigation area data prediction module comprises:

[0013] The convolution gating mechanism generation unit is configured to take the multi-dimensional meteorological data and the crop data as input data, and generate an input gate, a forget gate, an output gate and a candidate state of an updated gating mechanism through a convolution operation of the ConvLSTM layer according to the input data and a previous time step hidden state.

[0014] The gating fusion unit is configured to generate a state update through an updated gating mechanism of the ConvLSTM layer based on the input gate weighting the candidate state and the forget gate weighting a previous time step state.

[0015] The ConvLSTM output unit is configured to generate a water demand spatio-temporal feature based on the state update and the output gate.

[0016] The crop growth stage enhancement unit is configured to generate the predicted water storage utilization through the feature mapping layer by taking the water demand spatio-temporal feature and a crop growth stage code.

[0017] Further, the crop growth stage enhancement unit comprises:

[0018] The encoded water demand feature mapping subunit is configured to map a crop growth stage of the irrigation area into the crop growth stage code through convolutional encoding.

[0019] The factor mapping subunit is configured to respectively generate a mapping factor and a growth stage translation correction factor through linear convolution based on the crop growth stage code.

[0020] The prediction subunit is configured to adjust the water demand spatio-temporal feature based on the mapping factor and the growth stage translation correction factor to generate the predicted water storage utilization.

[0021] Further, the predicted water storage utilization comprises a predicted transpiration, and the irrigation area data prediction module further comprises:

[0022] The transpiration loss function training unit is configured to construct a transpiration loss function based on the predicted transpiration amount, a true transpiration amount of a sample, a standard crop coefficient, and a prediction model parameter of an irrigation area, and train and optimize the prediction model of the irrigation area based on the transpiration loss function.

[0023] In particular, the ConvLSTM layer can effectively capture the dynamic correlation features of the irrigation water demand in time and space. The gating fusion unit updates the state by weighting, filters redundant information, and retains key features, so that the generated water demand spatiotemporal features are more in line with the actual change law, providing higher-quality basic data for subsequent prediction of water storage and utilization, integrating crop growth stage features, realizing dynamic adaptation of water demand prediction, and performing special optimization through the transpiration loss function to ensure stable prediction accuracy of the model under different crop types and complex environments of different irrigation areas.

[0024] Further, the water demand fusion prediction module comprises:

[0025] The gating weight calculation unit is configured to map and convolve the concatenation vector of the soil latent heat flux, the basic transpiration amount, and the predicted water storage and utilization amount through a feature fusion gating mechanism to generate a gating weight.

[0026] The gating fusion unit is configured to perform weighted fusion of the concatenation vector and the predicted water storage and utilization amount based on the gating weight through the feature fusion gating mechanism to generate a fusion water demand feature.

[0027] The multi-scale convolution mapping unit is configured to aggregate the fusion water demand feature through multi-scale convolution to generate the comprehensive predicted water demand.

[0028] Further, the multi-scale convolution mapping unit comprises:

[0029] The cavity convolution mapping subunit is configured to perform multi-cavity rate convolution on the fusion water demand feature to generate a multi-time scale water demand.

[0030] The cross-time step aggregation subunit is configured to perform cross-time step aggregation on the multi-time scale water demand through attention weight to generate the comprehensive predicted water demand.

[0031] Further, the water demand fusion prediction module further comprises:

[0032] The comprehensive loss function training unit is configured to construct a data fitting term based on regularization calculation of the comprehensive predicted water demand and a true water demand of a sample, construct a water vapor conservation term based on a water vapor change rate of an irrigation area, construct a comprehensive loss function based on weighted combination of the data fitting term and the water vapor conservation term, and train the fusion mapping model through the comprehensive loss function.

[0033] Further, the irrigation area basic data calculation module comprises:

[0034] A soil latent heat flux calculation unit is configured to determine a saturated water vapor pressure difference according to the air temperature conversion, determine an air constant-pressure specific heat capacity according to the air pressure conversion, and calculate the soil latent heat flux by improving the Penman-Monteith formula according to the water vapor pressure difference, the solar radiation, the air constant-pressure specific heat capacity, and the wind speed.

[0035] Further, the irrigation area basic data calculation module further comprises:

[0036] A basic transpiration calculation unit is configured to multiply the saturated water vapor pressure difference by the solar radiation, divide the product by a sum of the saturated water vapor pressure difference and a wet-dry surface constant, and generate an energy term, and calculate the basic transpiration by improving the extended Penman-Monteith formula.

[0037] Further, the zero-carbon water resource management module comprises:

[0038] A predicted net irrigation amount calculation unit is configured to calculate an irrigation area predicted net irrigation amount based on the detected irrigation area water holding capacity, irrigation efficiency, effective precipitation, and the comprehensive predicted water demand.

[0039] A predicted photovoltaic power calculation unit is configured to determine a hydraulic power demand based on the predicted net irrigation amount, and set a sum of operating powers of the photovoltaic water pump and the photovoltaic valve to be less than the hydraulic power demand.

[0040] In particular, by means of the gating mechanism precise weighting, the importance of different data in the water demand prediction is identified, and the interference of low-quality information on the fusion result is avoided, so that the generated fusion water demand feature can better reflect the real water demand condition of the irrigation area, and through the empty convolution mapping, the water demand change law of different time scales can be captured, so that the comprehensive predicted water demand can accurately respond to short-term water demand changes and grasp long-term water demand trends, and through the comprehensive loss function optimization, the prediction accuracy and physical rationality are ensured.

[0041] Compared with the prior art, the application has the beneficial effects that the application predicts crop water demand by fusing soil latent heat flux and basic transpiration through a ConvLSTM model, improves the precision of water resource management in an irrigation area, reduces waste and imbalance, optimizes the scheduling energy consumption of a zero-carbon photovoltaic water pump and a photovoltaic valve, and realizes the optimization of the zero-carbon transformation of an agricultural irrigation area from the head of irrigation energy consumption.

[0042] Especially, the application can effectively capture the dynamic correlation characteristics of the irrigation district water demand in time and space through the ConvLSTM layer, the gate fusion unit filters the redundant information and retains the key features by weighted updating the state, so that the generated water demand spatiotemporal characteristics are more in line with the actual change law, which provides higher quality basic data for subsequent prediction of water storage and utilization, integrates the characteristics of crop growth stages, realizes dynamic adaptation of water demand prediction, and performs special optimization through the transpiration loss function, so that the model can maintain stable prediction accuracy under different crop types and different complex environments of irrigation districts.

[0043] Especially, the application can effectively capture the dynamic correlation characteristics of the irrigation district water demand in time and space through the ConvLSTM layer, the gate fusion unit filters the redundant information and retains the key features by weighted updating the state, so that the generated water demand spatiotemporal characteristics are more in line with the actual change law, which provides higher quality basic data for subsequent prediction of water storage and utilization, integrates the characteristics of crop growth stages, realizes dynamic adaptation of water demand prediction, and performs special optimization through the transpiration loss function, so that the model can maintain stable prediction accuracy under different crop types and different complex environments of irrigation districts. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 It is a structure schematic diagram of the agricultural irrigation district zero-carbon water resource comprehensive management system of the embodiment of the application.

[0045] Figure 2 It is a flowchart of the agricultural irrigation district zero-carbon water resource comprehensive management system of the embodiment of the application.

[0046] Figure 3 It is a flowchart of the irrigation district prediction model of the agricultural irrigation district zero-carbon water resource comprehensive management system of the embodiment of the application.

[0047] Figure 4 It is a flowchart of the fusion mapping model of the agricultural irrigation district zero-carbon water resource comprehensive management system of the embodiment of the application. DETAILED DESCRIPTION

[0048] In order to make the purpose and advantages of the application clearer and more apparent, the application will be further described below with reference to the embodiments; it should be understood that the specific embodiments described herein are only used to explain the application, and do not limit the application.

[0049] The preferred embodiments of the application will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the application, and are not intended to limit the protection scope of the application.

[0050] It should be noted that in the description of the present application, the terms "upper", "lower", "left", "right", "inner", "outer" and the like indicate the direction or positional relationship of the terms based on the direction or positional relationship shown in the drawings, which is only for the convenience of description, and does not indicate or imply that the device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0051] In addition, it should be noted that in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0052] As shown in Figures 1 to 4 , the present application provides a zero-carbon water resource comprehensive management system for agricultural irrigation area, which predicts crop water demand by fusing soil latent heat flux and basic evapotranspiration through ConvLSTM model, improves the precision of water resource management in irrigation area, reduces waste and imbalance, optimizes the scheduling energy consumption of zero-carbon emission photovoltaic water pump and photovoltaic valve, and realizes the optimization of energy head from irrigation to promote the zero-carbon transformation of agricultural irrigation area.

[0053] As shown in Figure 1 and 2 , the present embodiment provides a zero-carbon water resource comprehensive management system for agricultural irrigation area, which comprises:

[0054] The irrigation area basic data calculation module is used to calculate the soil latent heat flux of the irrigation area and the basic evapotranspiration of the crops according to the detected air temperature, air pressure, solar radiation and wind speed, respectively;

[0055] The irrigation area data prediction module is used to generate the predicted water storage and utilization amount of the irrigation area based on the multi-dimensional meteorological data and crop data through the irrigation area prediction model based on the ConvLSTM architecture;

[0056] The water demand fusion prediction module is used to generate the comprehensive predicted water demand of the irrigation area by fusing the soil latent heat flux, the basic evapotranspiration and the predicted water storage and utilization amount through the fusion mapping model based on the feature fusion gating mechanism and multi-scale convolution;

[0057] The zero-carbon water resource management module is used to calculate the predicted net irrigation amount of the irrigation area based on the comprehensive predicted water demand, and to perform irrigation management on the photovoltaic water pump and photovoltaic valve of the irrigation area based on the predicted net irrigation amount of the irrigation area.

[0058] Specifically, the agricultural irrigation zero-carbon resource comprehensive management system can be applied to an Internet of Things agricultural irrigation management system, that is, multi-dimensional meteorological data of an irrigation area, stored crop data, and detected multi-modal data of air temperature, air pressure, solar radiation, and wind speed are fused for irrigation management through the Internet of Things.

[0059] It should be noted that the transpiration amount is crop evapotranspiration (ET), which represents the total amount of soil evaporation and plant transpiration, so as to make the predicted water demand meet the real-time crop water demand state of the irrigation area under various climate conditions. The soil latent heat flux (Latent heat flux, LE) represents the heat absorbed or released in the water vapor phase change process of the irrigation area. A positive value indicates that the heat is transported from the surface to the atmosphere (evaporation / transpiration process), and a negative value indicates that the heat is transported from the atmosphere to the surface (condensation process), reflecting whether the farmland soil of the irrigation area is in water absorption or water evaporation state, and the degree of water absorption and water evaporation, so as to make the predicted water demand meet the climate water consumption / water delivery situation of the irrigation area.

[0060] It should be noted that the predicted water storage utilization amount includes predicted soil moisture and predicted transpiration amount. The predicted soil moisture, that is, the predicted soil water content, is related to the crop growth condition. The predicted transpiration amount is used to predict the water demand of crops under the influence of various environmental factors, such as the crop LAI in the vigorous growth period, the active stomata, and the increase of transpiration amount due to the influence of environmental factors such as photosynthetic active radiation, thereby increasing the irrigation water demand for good growth of crops.

[0061] As shown in Figure 3 Further, the irrigation area prediction model includes a ConvLSTM layer and a feature mapping layer, and the irrigation area data prediction module includes:

[0062] The convolutional gating mechanism generation unit takes the multi-dimensional meteorological data and the crop data as input data, and generates an input gate, a forget gate, an output gate, and a candidate state of an updated gating mechanism through a convolution operation of the ConvLSTM layer according to the input data and a previous time step hidden state.

[0063] The gating fusion unit generates a state update through the updated gating mechanism of the ConvLSTM layer based on the weighting of the candidate state by the input gate and the weighting of the previous time step state by the forget gate.

[0064] The ConvLSTM output unit generates a water demand spatiotemporal feature based on the state update and the output gate.

[0065] a crop growth stage enhancement unit to encode the spatiotemporal water demand feature and the crop growth stage through a feature mapping layer to generate the predicted water storage utilization.

[0066] In particular, the process of generating spatiotemporal water demand feature by the ConvLSTM layer can be represented as:

[0067] i t = σ(W xi *X t +W hi *H t-1 +b i )

[0068] f t = σ(W xf *X t +W hf *H t-1 +b f )

[0069] o t = σ(W xo *X t +W ho *H t-1 +b o )

[0070]

[0071] H t = o t ⊙ tanh(C t )

[0072] where i t , f t , o t , represent input gate, forget gate, output gate, candidate state respectively, σ represents Sigmoid activation function, used to control the preservation or forgetting of information, X t represents input data, W xi , W hi , W xf , W hf , W xo , W ho , W xc , W hc represent the convolution kernel weights of different gate parameters (input gate, forget gate, output gate, candidate state) respectively, * represents 3D convolution operation (kernel = 3x3x3 / 5x5x5), b i , b f , b o , b cbias terms for different gating parameters, tanh denotes a hyperbolic tangent activation function for non-linear transformation, C t denotes state update, denotes Hadamard product, C t-1 denotes previous time step state, H t denotes water demand spatiotemporal characteristics.

[0073] Specifically, the multi-dimensional meteorological data includes temperature, humidity, wind speed, sunshine, and precipitation (history, forecast), and the multi-dimensional meteorological data is standardized as the input data, and the crop data includes crop type and crop wilting point, and the crop data is one-hot encoded as the input data. Therefore, the irrigation area prediction model can be an intelligent model that can perceive "how is the crop growing", "how much water can the soil store", and "what is the current growth stage".

[0074] In particular, the ConvLSTM layer (Convolutional LSTM) performs convolution operation on the multi-dimensional meteorological data and the crop data, generates key parameters such as input gate and forget gate in combination with the previous time step hidden state, and can effectively capture the dynamic correlation characteristics of the water demand in the irrigation area in time and space. The gating fusion unit further filters redundant information and retains key features by updating the state through weighting, so that the generated water demand spatiotemporal characteristics are more in line with the actual change law. The limitations of the traditional time series model in capturing spatial information of the irrigation area are avoided, higher quality basic data are provided for subsequent prediction of soil moisture content and prediction of transpiration, and the reliability of the prediction of water demand is improved.

[0075] As shown in Figure 3 Further, the crop growth stage enhancement unit includes:

[0076] The encoding water demand feature mapping subunit is configured to map the crop growth stage in the irrigation area to the crop growth stage encoding through convolutional encoding.

[0077] The factor mapping subunit is configured to generate a mapping factor and a growth stage translation correction factor through linear convolution of the crop growth stage encoding, respectively.

[0078] The prediction subunit is configured to adjust the water demand spatiotemporal characteristics based on the mapping factor and the growth stage translation correction factor to generate the predicted water storage and utilization amount.

[0079] Specifically, the process of the prediction subunit to generate the predicted water storage and utilization amount can be represented as:

[0080] e pheno = E one hot (s)

[0081] γ = W γ e pheno + b γ

[0082] δ = W δ e pheno + b δ

[0083]

[0084] wherein e pheno denotes the crop growth stage encoding, E denotes the phenologicalembedding matrix, one hot(s) denotes the one-hot encoding for the crop growth stage s, γ, δ denote the mapping factor and the growth stage translation correction factor respectively, W γ , W δ denote the convolution kernel weights for the mapping factor and the growth stage translation correction factor respectively, b γ , b δ denote the bias terms for the mapping factor and the growth stage translation correction factor respectively, denotes the model output at time step t including the predicted water storage utilization, the predicted water storage utilization including the predicted soil moisture and the predicted transpiration, ⊙ denotes the Hadamard product, H t denotes the spatio-temporal characteristics of water requirement.

[0085] In particular, the crop growth stage comprises at least the seedling stage, the heading stage, the grain filling stage, and the mature stage, so that the predicted transpiration of the predicted water storage utilization is more consistent with the growth stage with larger change in water requirement of the crop.

[0086] Especially, through the mapping factor, the soil feature of the channel related to the deep soil humidity and the root distribution area water in the enlarged convolution feature of the spatio-temporal characteristics of water requirement in the key period of water requirement (i.e. the heading stage and the grain filling stage) is amplified, so that the model pays more attention to the actual available water of the crop. Through the growth stage translation correction factor, the deep water feature is weakened in the seedling stage or the mature stage, the weight of the channel related to the surface soil evaporation and the precipitation infiltration is enhanced, the dominant influencing factor of the soil moisture at this time is matched, and especially before the harvest in the mature stage, the feature distribution is translated through the growth stage translation correction factor, the prediction baseline of the overall transpiration is reduced, and the low transpiration feature of the residual dry leaves is avoided from being misjudged as an abnormality.

[0087] Specifically, the parameters of the irrigation area prediction model based on the ConvLSTM architecture include: an input time step of 7 days, a spatial scale of 32x32, a channel number of 6, a number of ConvLSTM layers of 2, a number of convolution kernels of 64 and 32 respectively, a size of the convolution kernel of 3x3x3 and 5x5x5 respectively, a stride of 1, a learning rate of 1e-4, a batch size of 8, and a dropout rate of 0.3.

[0088] Further, the predicted water storage utilization amount includes a predicted transpiration amount, and the irrigation area data prediction module further includes:

[0089] a transpiration loss function training unit configured to construct a transpiration loss function based on the predicted transpiration amount, a real transpiration amount of a sample, a standard crop coefficient, and a parameter of the irrigation area prediction model, and train and optimize the irrigation area prediction model based on the transpiration loss function.

[0090] Specifically, the transpiration loss function can be represented as:

[0091]

[0092] In the formula, L represents the transpiration loss function, T represents a predicted time step, which is preferably 24 hours in the future, 1 step per hour, H and W represent spatial dimensions, which are preferably corresponding to the number of rows and columns of the irrigation farmland, represents the predicted transpiration amount at time step t and spatial position (i, j), ET 0,t,i,j represents the real transpiration amount of the sample at time step t and spatial position (i, j), represents the reference crop coefficient at time step t and spatial position (i, j), which is a standard crop coefficient lookup value, λ represents a regularization coefficient, which is preferably 1e-5, W ConvLSTM represents a set of gating parameters of the ConvLSTM layer, W γ , and W δ respectively represent the convolution kernel weights of the feature mapping layer and the growth stage translation correction factor, represents L2 regularization.

[0093] Especially, through the ConvLSTM layer, the dynamic correlation features of the irrigation water demand in time and space can be effectively captured, the gating fusion unit updates the state through weighting, filters redundant information, and retains key features, so that the generated water demand spatiotemporal features are more in line with the actual change law, providing higher quality basic data for subsequent prediction of water storage utilization amount, integrating crop growth stage features, realizing dynamic adaptation of water demand prediction, and performing special optimization through the transpiration loss function, ensuring that the model can maintain stable prediction accuracy under different crop types and different irrigation complex environments.

[0094] For example, Figure 4As shown, further, the water demand fusion prediction module comprises:

[0095] a gating weight calculation unit configured to generate gating weights by mapping convolution of the soil latent heat flux, the base evapotranspiration and the concatenation vector of the predicted water storage utilization through a feature fusion gating mechanism;

[0096] a gating fusion unit configured to perform weighted fusion of the concatenation vector, the predicted water storage utilization based on the gating weights through the feature fusion gating mechanism to generate a fusion water demand feature;

[0097] a multi-scale convolution mapping unit configured to aggregate the fusion water demand feature through multi-scale convolution to generate the comprehensive predicted water demand.

[0098] Specifically, the process of the feature fusion gating mechanism generating the fusion water demand feature can be represented as:

[0099]

[0100] G t = σ(W g *P t +b g )

[0101]

[0102] wherein P t represents the concatenation vector, represents the model output including the predicted water storage utilization at time step t, LE t represents the soil latent heat flux at time step t, ET t represents the base evapotranspiration at time step t, G t represents the gating weights, σ represents a Sigmoid activation function, W g represents a gating convolution matrix, b g represents a gating bias term, * represents a convolution operation, U t represents the fusion water demand feature, and represents a Hadamard product.

[0103] Especially, the gating weight calculation unit generates gating weights by mapping convolution of the spliced vector of the soil latent heat flux, the basic transpiration, and the predicted water storage utilization through the gating mechanism, and the gating fusion unit performs weighted fusion on the related data based on the weights, which can identify the importance of different input data in the prediction of water requirement, for example, when the soil moisture is low, the predicted soil moisture of the predicted water storage utilization is given higher weight, when the crop transpiration is vigorous, the weights of the predicted transpiration of the basic transpiration and the predicted water storage utilization are increased, thereby effectively avoiding the interference of irrelevant or low-quality information on the fusion result, and the generated fusion water requirement feature can better reflect the real water requirement condition of the irrigation area.

[0104] As shown in Figure 4 , further, the multi-scale convolution mapping unit comprises:

[0105] The cavity convolution mapping sub-unit is configured to generate a multi-time scale water requirement by multi-cavity rate convolution of the fusion water requirement feature.

[0106] The cross-time step aggregation sub-unit is configured to generate the comprehensive predicted water requirement by cross-time step aggregation of the multi-time scale water requirement through attention weights.

[0107] Specifically, the process of generating the comprehensive predicted water requirement by multi-scale convolution can be represented as:

[0108]

[0109] In the formula, Z t , Z k represents the multi-time scale water requirement at time t and time k, represents multi-cavity rate convolution with cavity rates of 1, 3, and 5, U t represents the fusion water requirement feature, a t,k represents the attention weight of the time step t to the historical time step k, which is used to measure the attention degree to the historical time step k, represents the transpose of the query vector of the time step t, which is generated by the model output, K j represents the key vector of the historical time step k, which is generated by the multi-time scale water requirement, represents the comprehensive predicted water requirement at future time t+τ, τ represents the prediction lead time step, preferably, τ=1 represents 1 hour in the future, * represents convolution operation, W y , b y represents the learnable parameters of the convolution kernel.

[0110] Specifically, the parameters of the fusion mapping model based on the feature fusion gating mechanism and the multi-scale convolution include: the number of convolution kernels of the feature fusion gating mechanism is 3 respectively, the size of the convolution kernel is 3*3, the number of convolution kernels of each path of the multi-scale convolution is 32, the size of the convolution kernel is 3*3, and the relu post-activation function is adopted.

[0111] Especially, the multi-hole rate convolution processing is adopted in the hollow convolution mapping subunit in the multi-scale convolution mapping unit to process the fused water demand features, so that the water demand change law in different time scales can be captured, such as the water demand fluctuation caused by meteorological mutation in a short time and the water demand trend caused by crop growth and seasonal replacement in a long time, thereby improving the comprehensiveness and adaptability of the prediction.

[0112] Further, the water demand fusion prediction module further includes:

[0113] The comprehensive loss function training unit is configured to calculate the regularization of the comprehensive predicted water demand and the sample true water demand to construct a data fitting term, construct a water vapor conservation term based on the water vapor change rate of the irrigation area, construct a comprehensive loss function based on the weighted data fitting term and the water vapor conservation term, and train the fusion mapping model through the comprehensive loss function.

[0114] Specifically, the comprehensive loss function can be represented as:

[0115]

[0116] In the formula, L data , L moisture respectively represent the data fitting term and the water vapor conservation term, τ represents the prediction lead time step, H and W represent the spatial dimensions, represents the comprehensive predicted water demand, Y true represents the sample true water demand, represents the square of the Frobenius norm, represents the local change rate of the water vapor mixing ratio, qv is the moisture flux (Moisture flux), v is the detected wind speed, q is the water vapor mixing ratio, i.e. the water vapor mass contained in unit mass of wet air, q is the local meteorological station monitoring data of the irrigation area, E represents the evaporation rate, P represents the precipitation rate, E and P are both set values of the irrigation area, || ||1 represents the L1 norm, L total represents the comprehensive loss function, λ1 and λ2 are two weight coefficients, and are preferably 0.6 and 0.4.

[0117] Especially, the water vapor conservation term strictly satisfies the water vapor balance in the model training, and is suitable for scenarios that need to simulate water cycle, i.e. the comprehensive predicted water demand of the irrigation area containing the prediction of the soil moisture content of the farmland, which can make up for the problem that pure data fitting may ignore the actual situation of the irrigation area scene, and improve the rationality of the prediction.

[0118] Especially, by the gating mechanism precise weighting, the importance of different data in water demand prediction is identified, avoiding the interference of low-quality information on the fusion result, making the generated fusion water demand feature better reflect the real water demand situation of the irrigation district, through the mapping of the hollow convolution, the water demand change law of different time scales can be captured, so that the comprehensive predicted water demand can accurately respond to short-term water demand changes and grasp long-term water demand trends, and through the optimization of the comprehensive loss function, the prediction accuracy and physical rationality are ensured.

[0119] Further, the irrigation district basic data calculation module comprises:

[0120] The soil latent heat flux calculation unit is configured to determine a saturated water vapor pressure difference according to the air temperature conversion, determine an air constant-pressure specific heat capacity according to the air pressure conversion, and calculate the soil latent heat flux according to the water vapor pressure difference, the solar radiation, the air constant-pressure specific heat capacity, and the wind speed by using an improved Penman-Monteith formula.

[0121] Specifically, the improved Penman-Monteith formula for calculating the soil latent heat flux is as follows:

[0122]

[0123] In the formula, LE t represents the soil latent heat flux at time step t, Δ represents the saturated water vapor pressure difference, which is determined by conversion, wherein e s represents the saturated water vapor pressure, T represents the air temperature, R n,t represents the solar radiation at time t, p represents the air density, which is preferably 1.225 kg / m 3 , c p represents the air constant-pressure specific heat capacity, which is determined by the current air pressure conversion, e s,t , e a,t respectively represent the saturated water vapor pressure and the actual water vapor pressure at time t, which are respectively determined by the detected air temperature conversion and the local meteorological station monitoring of the irrigation district, r s , r a respectively represent the wind speed resistance and the crop resistance, wherein wherein the observation height z is 2 meters, the zero plane displacement d is 0.67h, the momentum roughness length z 0m is 0.1h, the water vapor roughness length z 0h is 0.01h, h is the current average height of the crop, the von Karman constant k is 0.41, u z is the detected wind speed, r s is preferably 70s / m, and γ represents the psychrometer constant, which is about 0.0665 kPa / ℃ under standard atmospheric pressure.

[0124] Especially, the soil latent heat flux is used to back-calculate the water requirement (irrigation amount).

[0125] Further, the irrigation area basic data calculation module further comprises:

[0126] The basic transpiration amount calculation unit is configured to multiply the saturated water vapor pressure difference by the solar radiation, divide the product by the sum of the saturated water vapor pressure difference and the wet and dry table constant, and generate an energy term, and calculate the basic transpiration amount by using the improved and extended Penman-Monteith formula.

[0127] Specifically, the improved and extended Penman-Monteith formula for calculating the basic transpiration amount is as follows:

[0128]

[0129] In the formula, ET t represents the basic transpiration amount at time step t, Δ represents the saturated water vapor pressure difference, R n represents the solar radiation, γ represents the wet and dry table constant, ρ represents the air density, c p represents the air constant-pressure specific heat capacity, r a represents the crop resistance, and γ, ρ, and c p , r a are constants, κ represents the extinction coefficient, and LAI represents the leaf area index (Leaf Area Index). The leaf area index LAI is determined according to the crop data stored by the Internet of Things.

[0130] Especially, the actual evapotranspiration is used to back-calculate the water requirement and the irrigation amount by using the soil water balance equation.

[0131] It can be understood that the standard form of the Penman-Monteith formula is as follows:

[0132]

[0133] The embodiment improves and extends the standard Penman-Monteith formula, and fuses the prediction data, to accurately predict the soil water requirement and the crop water requirement of the irrigation area, and further more accurately evaluate the irrigation water amount and the water delivery power, so as to ensure that the power generated by the photovoltaic water pump and the photovoltaic valve is as small as possible under the condition of meeting the soil and crop water requirements, and further more realize the zero-carbon water resource management of the irrigation area.

[0134] Further, the zero-carbon water resource management module comprises:

[0135] a predicted net irrigation amount calculation unit configured to calculate a predicted net irrigation amount of the irrigation area based on the detected water holding capacity of the irrigation area, the irrigation efficiency, the effective precipitation, and the comprehensive predicted water demand;

[0136] a predicted photovoltaic power calculation unit configured to determine a hydraulic power demand based on the predicted net irrigation amount, and set the sum of the operating power of the photovoltaic water pump and the photovoltaic valve to be less than the hydraulic power demand.

[0137] Specifically, the process of calculating the predicted net irrigation amount of the irrigation area is as follows:

[0138]

[0139] wherein, I irr represents the predicted net irrigation amount of the irrigation area, θ fc represents the water holding capacity of the irrigation area, Z r represents the root depth, η represents the irrigation efficiency, P e represents the effective precipitation, represents the sum of the predicted net irrigation amounts of the irrigation area in a period of time.

[0140] Specifically, the process of calculating the hydraulic power demand is as follows:

[0141]

[0142] wherein, P hyd represents the hydraulic power demand, ρ water represents the density of water, g represents the acceleration of gravity, H tot represents the total head, which is determined according to historical measurement data of different irrigation amounts, I irr represents the predicted net irrigation amount of the irrigation area, A represents the irrigation area, t irr represents the planned irrigation time, ρ water , g, A, t irr are all given values or set values.

[0143] In particular, the maximum operating power of the photovoltaic water pump and the valve is dynamically adjusted by the predicted net irrigation amount, so that clear scheduling management control is achieved.

[0144] In this embodiment, the ConvLSTM model is used to fuse soil latent heat flux and basic transpiration to predict crop water demand, improve the accuracy of water resource management in irrigation areas, reduce waste and imbalance, optimize the scheduling energy consumption of zero-carbon photovoltaic water pumps and photovoltaic valves, and realize the optimization of energy head from irrigation to promote the zero-carbon transformation of agricultural irrigation areas. Through the ConvLSTM layer, the dynamic correlation characteristics of the irrigation water demand in time and space can be effectively captured. The gating fusion unit updates the state by weighting, filters redundant information, and retains key features, so that the generated water demand spatiotemporal features are more in line with the actual change law, providing higher quality basic data for subsequent prediction of water storage and utilization, integrating crop growth stage characteristics, realizing dynamic adaptation of water demand prediction, and through the transpiration loss function for special optimization, ensuring that the model can maintain stable prediction accuracy under different crop types and different complex environments of irrigation areas. Through the gating mechanism precise weighting, the importance of different data in water demand prediction is identified, avoiding the interference of low-quality information on the fusion result, making the generated fusion water demand features better reflect the real water demand situation of the irrigation area, and through the hollow convolution mapping, the water demand change law of different time scales can be captured, so that the comprehensive predicted water demand can accurately respond to short-term water demand changes and grasp long-term water demand trends. Through comprehensive loss function optimization, the prediction accuracy and physical rationality are ensured.

[0145] Those skilled in the art can appreciate that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0146] So far, the technical solutions of the present application have been described in conjunction with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to related technical features without departing from the principles of the present application, and the technical solutions after such changes or replacements will fall within the protection scope of the present application.

Claims

1. A zero-carbon water resource integrated management system for agricultural irrigation areas, characterized in that, The method comprises the following steps: a basic data calculation module of an irrigation area is used to calculate the soil latent heat flux and the basic transpiration of crops in the irrigation area according to the detected air temperature, air pressure, solar radiation and wind speed; an irrigation area data prediction module is used to generate the predicted water storage and utilization amount of the irrigation area by an irrigation area prediction model based on the multi-dimensional meteorological data and crop data and based on the ConvLSTM architecture; a water demand fusion prediction module is used to generate the comprehensive predicted water demand of the irrigation area by a fusion mapping model based on the feature fusion gating mechanism and the multi-scale convolution, by fusing the soil latent heat flux, the basic transpiration and the predicted water storage and utilization amount; a zero-carbon water resource management module is used to calculate the predicted net irrigation amount of the irrigation area based on the comprehensive predicted water demand, and to perform irrigation management on the photovoltaic water pump and the photovoltaic valve of the irrigation area based on the predicted net irrigation amount of the irrigation area.

2. The agricultural irrigation zero-carbon resource integrated management system according to claim 1, characterized in that, The irrigation area prediction model comprises a ConvLSTM layer and a feature mapping layer, and the irrigation area data prediction module comprises: a convolution gating mechanism generation unit is used to take the multi-dimensional meteorological data and the crop data as input data, to generate the input gate, the forget gate, the output gate and the candidate state of the updated gating mechanism by the convolution operation of the ConvLSTM layer according to the input data and the previous time step hidden state, and to generate the water demand spatio-temporal feature based on the state update and the output gate; a gating fusion unit is used to generate the state update by the updated gating mechanism of the ConvLSTM layer based on the weighting of the candidate state by the input gate and the weighting of the previous time step state by the forget gate; a ConvLSTM output unit is used to generate the water demand spatio-temporal feature based on the state update and the output gate; a crop growth stage enhancement unit is used to generate the predicted water storage and utilization amount by the feature mapping layer through the water demand spatio-temporal feature and the crop growth stage code.

3. The agricultural irrigation zero-carbon resource integrated management system according to claim 2, characterized in that, The crop growth stage enhancement unit comprises: an encoded water feature mapping subunit is used to map the crop growth stage of the irrigation area to the crop growth stage code by convolution encoding; a factor mapping subunit is used to generate the mapping factor and the growth stage translation correction factor by linear convolution respectively through the crop growth stage code; a prediction subunit is used to adjust the water demand spatio-temporal feature based on the mapping factor and the growth stage translation correction factor to generate the predicted water storage and utilization amount.

4. The integrated zero-carbon water resource management system for agricultural irrigation districts of claim 2, wherein, The predicted water storage and utilization amount comprises a predicted transpiration amount, and the irrigation area data prediction module further comprises: a transpiration loss function training unit is used to construct a transpiration loss function based on the predicted transpiration amount, the true transpiration amount of the sample, the standard crop coefficient and the irrigation area prediction model parameters, and to train and optimize the irrigation area prediction model based on the transpiration loss function.

5. The agricultural irrigation zero-carbon resource integrated management system according to claim 1, wherein, The water demand fusion prediction module comprises: a gating weight calculation unit is used to generate the gating weight by the mapping convolution of the feature fusion gating mechanism through the splicing vector of the soil latent heat flux, the basic transpiration and the predicted water storage and utilization amount. a gated fusion unit configured to perform weighted fusion on the splicing vector and the predicted water storage utilization amount based on the gating weight to generate a fused water demand feature; a multi-scale convolution mapping unit configured to aggregate the fused water demand feature through multi-scale convolution to generate the comprehensive predicted water demand.

6. The agricultural irrigation zero-carbon resource integrated management system according to claim 5, characterized in that, The multi-scale convolution mapping unit comprises: a cavity convolution mapping sub-unit configured to perform multi-cavity rate convolution on the fused water demand feature to generate a multi-time scale water demand; a cross-time step aggregation sub-unit configured to perform cross-time step aggregation on the multi-time scale water demand through attention weight to generate the comprehensive predicted water demand.

7. The agricultural irrigation zero-carbon resource integrated management system according to claim 5, characterized in that, The water demand fusion prediction module further comprises: a comprehensive loss function training unit configured to construct a data fitting term based on regularization calculation of the comprehensive predicted water demand and a sample true water demand, construct a water vapor conservation term based on a water vapor change rate of an irrigation area, construct a comprehensive loss function based on weighted combination of the data fitting term and the water vapor conservation term, and train the fusion mapping model through the comprehensive loss function.

8. The agricultural irrigation zero-carbon resource integrated management system according to claim 1, characterized in that, The irrigation area basic data calculation module comprises: a soil latent heat flux calculation unit configured to determine a saturated water vapor pressure difference according to the air temperature conversion, determine an air constant pressure specific heat capacity according to the air pressure conversion, and calculate the soil latent heat flux through an improved Penman-Monteith formula based on the water vapor pressure difference, the solar radiation, the air constant pressure specific heat capacity, and the wind speed.

9. The agricultural irrigation zero-carbon resource integrated management system according to claim 8, characterized in that, The irrigation area basic data calculation module further comprises: a basic transpiration calculation unit configured to multiply the saturated water vapor pressure difference by the solar radiation, divide the product by a product of the saturated water vapor pressure difference and a wet and dry table constant to generate an energy term, and calculate the basic transpiration through an improved extended Penman-Monteith formula based on the energy term.

10. The integrated zero-carbon water resource management system for an agricultural irrigation district of any of claims 1 to 9, wherein, The zero-carbon water resource management module comprises: a predicted net irrigation amount calculation unit configured to calculate an irrigation area predicted net irrigation amount based on detected irrigation area water holding capacity, irrigation efficiency, effective precipitation, and the comprehensive predicted water demand; a predicted photovoltaic power calculation unit configured to determine a hydraulic power demand based on the predicted net irrigation amount, and set a sum of operating powers of the photovoltaic water pump and the photovoltaic valve to be less than the hydraulic power demand.

Images