A plant water requirement prediction method and system based on LSTM and remote sensing data

Abstract

The application provides a plant water demand prediction method and system based on LSTM and remote sensing data, and relates to the technical field of data processing.The method comprises the following steps: obtaining vegetation characteristic data of a region to be predicted; establishing a plant water content evolution equation based on the vegetation characteristic data; introducing time memory based on fractional differential to determine the current cell state and the current hidden state of an LSTM model; updating the update gate weight and the forget gate weight of the LSTM model based on the hidden state; and combining the plant water content evolution equation and the forget gate weight to obtain the plant water demand of the region to be predicted through inversion.The application realizes accurate prediction of plant water demand in a complex environment by constructing a hybrid model combining physical mechanism and deep learning.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for predicting plant water demand based on LSTM and remote sensing data. Background Technology

[0002] Plant water refers to the water content of the area where a plant is located. Water is fundamental to plant growth and development, and is crucial for physiological processes such as photosynthesis, nutrient transport, and cell expansion.

[0003] Accurately predicting plant water requirements helps agricultural managers allocate water resources rationally, avoiding over-irrigation or water shortages, optimizing water use efficiency, and promoting healthy crop growth. Efficient water management is particularly crucial in arid regions, enhancing the sustainability and efficiency of agricultural production, reducing production costs, and ensuring crop yield and quality.

[0004] However, existing technologies for predicting plant water requirements typically rely on traditional empirical statistics or subjective judgments to supply water to plants, ignoring the complexity of time and space and failing to accurately capture the dynamic changes in plant water requirements. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a plant water demand prediction method based on LSTM and remote sensing data, which can solve the technical problem that the existing technology usually relies on traditional empirical statistics or subjective judgment to predict plant water demand, ignoring the complexity of time and space, and failing to accurately capture the dynamic changes in plant water demand.

[0006] A first aspect of this invention proposes a method for predicting plant water demand based on LSTM and remote sensing data, comprising:

[0007] S1: Obtain vegetation feature data for the area to be predicted;

[0008] S2: Establish a plant water content evolution equation based on vegetation characteristic data;

[0009] S3: Introduce time memory based on fractional derivatives to determine the current cell state and current hidden state of the LSTM model;

[0010] S4: Update gate weights and forget gate weights of the LSTM model based on hidden state;

[0011] S5: Combining the plant water content evolution equation and the forget gate weight, the plant water demand of the area to be predicted is obtained by inversion.

[0012] A second aspect of this invention provides a plant water demand prediction system based on LSTM and remote sensing data, comprising: a processor and a memory;

[0013] The memory stores a program or instruction that can run on the processor, and when the program or instruction is executed by the processor, it implements the steps of the LSTM-based plant water demand prediction method based on remote sensing data as described in the first aspect.

[0014] A third aspect of the present invention provides a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the plant water demand prediction method based on LSTM and remote sensing data as described in the first aspect.

[0015] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0016] In this embodiment of the invention, a physical evolution equation describing the dynamic changes in plant water is established using remote sensing data, providing crucial mechanistic constraints for the LSTM model and effectively overcoming the shortcomings of purely data-driven models, such as overfitting and neglecting physical laws. Simultaneously, fractional derivatives are introduced to characterize the long-term historical dependence of water changes, enabling the model to accurately capture the cumulative effects of slow evolution. These two factors together optimize the gating mechanism within the LSTM network, particularly the weight allocation of the forget gate, thereby constructing a hybrid model that integrates physical mechanisms and deep learning, achieving accurate prediction of plant water requirements under complex environments. Attached Figure Description

[0017] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0018] Figure 1 This is a flowchart illustrating a method for predicting plant water demand based on LSTM and remote sensing data, provided in an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of a plant water demand prediction system based on LSTM and remote sensing data provided in an embodiment of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0021] The following description, in conjunction with the accompanying drawings, details the plant water demand prediction method based on LSTM and remote sensing data provided by the present invention through specific embodiments and application scenarios.

[0022] Reference manual attached Figure 1 The diagram illustrates a flowchart of a plant water demand prediction method based on LSTM and remote sensing data provided by an embodiment of the present invention.

[0023] This invention provides a method for predicting plant water demand based on LSTM and remote sensing data, which may include the following steps:

[0024] S1: Obtain vegetation feature data for the area to be predicted.

[0025] The area to be predicted refers to the specific geographical region where plant water demand forecasting is needed, typically a piece of farmland, forest, or other vegetated area. Vegetation characteristic data describes the vegetation status and environmental characteristics of this area. This vegetation characteristic data is usually obtained through remote sensing technology, utilizing high-resolution data acquired via satellite imagery or drones, which can provide detailed information about plant growth, soil moisture status, and climate change in the area to be predicted. This data provides an important foundation for subsequent development of plant water demand forecasting models.

[0026] In one possible implementation, vegetation characteristic data include normalized vegetation index, leaf area index, and soil volumetric water content.

[0027] The Normalized Difference Vegetation Index (NDVI), Leaf Area Index (LAI), and Soil Volumetric Water Content (SM) are all important indicators of vegetation growth. NDVI and LAI are derived from optical remote sensing data, while SM is derived from microwave remote sensing data. The NDVI is a crucial indicator of vegetation growth obtained through remote sensing technology. A higher NDVI value indicates more vigorous vegetation growth, while a lower value indicates sparse or unhealthy vegetation. The LAI measures the total area of ​​plant leaves per unit area of ​​land, reflecting the plant's photosynthetic capacity and health. Soil Volumetric Water Content (SM) represents the proportion of water contained in a unit volume of soil and is an important parameter for assessing soil moisture status.

[0028] S2: Establish a plant water content evolution equation based on vegetation characteristic data.

[0029] The plant water content evolution equation is a mathematical model used to describe the changes in plant water content over time and under environmental conditions. Based on vegetation characteristic data, this equation simulates the dynamic changes in plant water by considering processes such as water absorption, transpiration, and recharge, as well as soil moisture conditions. Quantifying the evolution of plant water makes the prediction model more accurate and adaptable. This equation can capture the dynamic changes in plant water requirements under different temporal and spatial conditions, improving prediction accuracy and avoiding the simplification assumptions found in traditional empirical models.

[0030] In one possible implementation, S2 specifically includes:

[0031] S201: Determine the time fraction order describing the dependence of plant water requirements based on the normalized vegetation index.

[0032] The fractional derivative refers to the fractional differential exponent that describes the change of plant water requirement over time. By introducing the fractional derivative, the model can capture the historical dependence and memory effect of plant water requirement, thus simulating the more complex dynamic evolution of water requirement over time. For example, the fractional derivative can represent the asymptotic changes and delay effects of the system, and is more suitable than the traditional integer derivative for describing the temporal nonlinear characteristics of plant water requirement.

[0033] S202: Determine the spatial fractional order of plant root system complexity based on leaf area index.

[0034] The spatial fractional order is used to describe the spatial characteristics of the complexity of plant root structure. Root distribution influences a plant's water absorption capacity, and the spatial fractional derivative helps describe the spatial distribution characteristics during plant growth, reflecting water flow and distribution in different regions. The spatial fractional order model can capture the diffusion behavior of water at different spatial scales, making water demand predictions more consistent with the real environment.

[0035] S203: Determine the fractional diffusion coefficient based on soil volumetric water content to describe the rate of water diffusion in the soil to which the plant is located.

[0036] The fractional diffusion coefficient describes the rate of water diffusion in the soil where the plant is located. It depends on the soil moisture content and the plant roots' ability to absorb water. The fractional diffusion coefficient can more accurately simulate the water diffusion process in the soil, taking into account the nonlinear characteristics of water migration.

[0037] S204: By combining the fractional time order, the fractional spatial order, and the fractional diffusion coefficient, a water content evolution equation is established.

[0038] The equation for the evolution of water content is as follows:

[0039] ;

[0040] ;

[0041] ;

[0042] ;

[0043] in, , and These represent the fractional time order, fractional spatial order, and fractional diffusion coefficient, respectively. This represents the maximum water diffusion rate associated with soil type, i.e., the maximum fractional-order diffusion coefficient. , and The values ​​at time t represent the normalized vegetation index, leaf area index, and soil volumetric water content, respectively. Represents the natural constant. This represents the preset evapotranspiration loss rate for the area to be predicted. This indicates the moisture content of the area to be predicted. This indicates the partial derivative. This represents the Caputo time fractional derivative. Indicates and The relevant Riesz space fractional order Laplacian, This represents the amount of water replenishment to the region at time t.

[0044] The water replenishment at time t includes both artificial and natural replenishment. Artificial replenishment can be directly obtained, while natural replenishment can be obtained from meteorological data. The maximum fractional diffusion coefficient is a soil type-related parameter representing the maximum rate of water diffusion in the soil. The transpiration loss rate refers to the proportion of water lost by plants due to transpiration. The Caputo time fractional derivative is a form of fractional derivative used to describe dynamic processes with memory effects and history dependence. It is a non-integer derivative widely used in physics, chemistry, and biology.

[0045] It should be noted that those skilled in the art can set the preset evaporation loss rate according to actual needs, and this invention does not limit it.

[0046] Specifically, by introducing a fractional-order differential model, dynamic prediction of plant water demand is achieved, overcoming the limitations of traditional methods in handling changes in plant water content. First, fractional-order parameters describing plant water demand, root complexity, and soil moisture diffusion are determined using vegetation characteristic data such as normalized difference vegetation index, leaf area index, and soil volumetric water content. The introduction of temporal fractional-order, spatial fractional-order, and fractional-order diffusion coefficients enables the model to effectively simulate the nonlinear evolution of plant water content over time and space. By combining these parameters, a water content evolution equation is established, further enhancing the model's ability to accurately predict plant water demand. This method fully considers the spatiotemporal dependence of plant water demand and complex environmental factors, thereby improving the accuracy and applicability of predictions. Compared with traditional simple statistical models, the fractional-order differential model better captures the dynamic changes in water demand, especially in complex ecological environments, providing more accurate and real-time water management data.

[0047] S3: Introduce time memory based on fractional derivatives to determine the current cell state and current hidden state of the LSTM model.

[0048] Fractional differentiation is a generalized calculus method that introduces fractional (non-integer) derivatives to describe the memory effect and historical dependence of dynamic systems. In traditional integer differentiation, the derivative represents the rate of change of the system's current state, while fractional differentiation can more accurately describe systems with complex memory effects, particularly suitable for describing nonlinear, time-varying, and complex biological or physical processes. By introducing fractional differentiation, processes with historical dependence that change over time, such as plant water requirements, can be better simulated. LSTM (Long Short-Term Memory) models are a special type of recurrent neural network with the ability to store long-term memories, making them particularly suitable for processing and predicting time-series data. LSTM models overcome the gradient vanishing problem in traditional RNNs when dealing with long-term dependencies by introducing multiple gate structures (forget gate, input gate, and output gate) to control the storage and forgetting of information. The current cell state is the information stored internally by the LSTM model; it records the model's long-term memory at the current time step (time t), encompassing information from all previous times. The cell state is updated through mechanisms such as the input gate and forget gate, helping the model retain long-term dependency information useful for future predictions. The current hidden state is the output value of the LSTM model at the current time step. It not only reflects the information of the current cell state but also integrates the hidden state from the previous time step. The hidden state contains the model's immediate understanding of the current input information and is the main basis for LSTM to make the next prediction and update.

[0049] Introducing fractional derivatives enables LSTM to effectively handle dynamic changes with long-term memory effects, thereby improving the prediction accuracy of plant water requirement changes. Through the temporal memory based on fractional derivatives, LSTM can more accurately capture the nonlinear and time-varying characteristics of plant water requirements, thus optimizing the updating of cell states and hidden states.

[0050] In one possible implementation, the LSTM model includes an input layer, an input gate, a forget gate, an output gate, a memory unit, and an output layer. The input layer is connected to the memory unit via the input gate. The forget gate is connected to the memory unit. The memory unit is connected to the output layer via the output gate.

[0051] The LSTM (Long Short-Term Memory) model is a special type of recurrent neural network (RNN) used to process and predict sequential data. By introducing memory units and a gate mechanism, it effectively solves the vanishing and exploding gradient problems faced by traditional RNNs when processing long sequences of data. The LSTM model consists of the following parts: Input layer: Responsible for receiving external input data. Input gate: Controls the importance of input information, determining which input information needs to be retained and passed to the memory unit. Forget gate: Determines which information in the memory unit needs to be forgotten. It calculates a value based on the current input and the hidden state of the previous time step, controlling the degree of forgetting. Output gate: Determines which information will be passed to the output of the next time step based on the contents of the memory unit. Memory unit: Stores information from the input gate, forget gate, and output gate, playing a role in memory and information retention; it is the core part of the LSTM network. Output layer: Finally, it outputs the model's prediction results.

[0052] In practical applications, the training process of an LSTM model typically includes two stages: forward propagation and backpropagation. In the forward propagation stage, input data enters the LSTM network through the input layer, passes through the input gate, forget gate, and output gate, and the information is stored in the memory unit, ultimately producing the output. In the backpropagation stage, the model calculates the error between the loss function (such as mean squared error) and the actual output, and then uses backpropagation to progressively adjust the parameters of each gate (input gate, forget gate, and output gate) and the memory unit to minimize the error. Backpropagation uses gradient descent to optimize the weights. Through multiple iterations of training, the LSTM model can effectively learn the relationship between the input sequence and the output, achieving good predictive performance.

[0053] In one possible implementation, S3 specifically includes:

[0054] S301: Establish a fractional differential form of the cell state update equation.

[0055] The specific formula for the update equation is as follows:

[0056] ;

[0057] in, This represents the cell state at time t. and These represent the updated gate weights at time t and the forgotten gate weights at time t, respectively. Indicates multiplication by the dominant element. This represents the candidate state at time t.

[0058] S302: Discretize the cell state update equation and extract the current cell state.

[0059] The current cell state, i.e., the cell state at time t, is expressed by the following formula:

[0060] ;

[0061] ;

[0062] in, This represents the total number of historical time steps referenced, where k represents the index of the historical time step. This represents the forget gate weight at time tk. This represents the cell state at time tk. Indicates and Fractional binomial coefficients related to k This represents the gamma function used to quantify the decay of historical state weights.

[0063] S303: Calculate the current hidden state based on the current cell state.

[0064] The formula for calculating the current hidden state is as follows:

[0065] ;

[0066] in, This represents the output gate weight at time t. express Activation function This represents the hidden state at time t, which is the current hidden state.

[0067] Specifically, this process enhances the memory capacity and temporal dependence of the LSTM model by introducing fractional derivatives. During cell state updates, establishing an update equation based on fractional derivatives allows for a more accurate description of the temporal evolution and memory effects of information. The introduction of fractional derivatives enables LSTM to capture not only current input information but also the complex dependencies of historical information. By discretizing the cell state update equation, the model can incorporate the memory of historical time steps to adjust the current state, thereby improving its ability to model complex time-series data. Simultaneously, by combining the output gate with the tanh activation function, the current hidden state is calculated, achieving effective output and updating of information. This approach accurately captures nonlinear changes and long-term dependencies in the data, making it suitable for handling data such as plant water requirements that are influenced by historical factors and time-varying conditions, thus improving the model's prediction accuracy and generalization ability.

[0068] S4: Update gate weights and forget gate weights of the LSTM model based on hidden state.

[0069] The update gate weights control the degree to which the current input information is updated in the cell state. They determine the magnitude of the impact of the current input data on the cell state at each time step. Through the update gate, LSTM can decide which information needs to be retained and passed to the next step. The forget gate weights control which information in the cell state should be forgotten or retained. They are calculated based on the hidden state from the previous time step and the current input, determining the degree to which past information is forgotten, thus effectively filtering out historical information that is no longer useful. By updating the weights of these two gates based on the hidden state, the model's memory of historical data and its focus on the current input can be dynamically adjusted, allowing the model to adjust its memory and predictive capabilities in real time according to new inputs and environmental changes.

[0070] In one possible implementation, S4 specifically includes:

[0071] S401: Obtain the current input features of the LSTM model.

[0072] The current input features specifically include the current normalized vegetation index, leaf area index, and soil volumetric water content.

[0073] S402: Input the input features into the LSTM model.

[0074] S403: Determine the update gate weight based on the hidden state and the water supply amount of the region to be predicted at time t.

[0075] The specific formula for calculating the updated gate weights is as follows:

[0076] ;

[0077] in, This represents the Sigmoid activation function. and Let represent the learnable weight parameters and learnable bias parameters of the update gate, respectively. Let represent the input vector at time t, composed of the current input features, and represent the hidden state at time t-1. This indicates taking the maximum value. This indicates the avoidance of small positive numbers with a denominator of zero.

[0078] S404: Determine the forget gate weight by combining the hidden state and the saturated water vapor pressure difference.

[0079] The specific formula for updating the forget gate weights is as follows:

[0080] ;

[0081] in, and Let represent the learnable weight parameters and learnable bias parameters with respect to the forget gate, respectively. Represents the integral variable. Indicates time The saturated water vapor pressure difference in the region to be predicted.

[0082] The saturated water vapor pressure difference can be obtained directly from meteorological data.

[0083] Specifically, the LSTM model dynamically predicts water demand by inputting real-time vegetation feature data (normalized difference in vegetation index, leaf area index, and soil volumetric water content). First, the input features are processed by the LSTM model, and its internal state is updated by incorporating historical data. By calculating the updated gate weights, the LSTM can adjust its memory capacity based on the current input, historical state, and water supply, enabling the model to learn and adapt to environmental changes in real time. The update of the forget gate weights combines the current hidden state with meteorological data (such as saturated vapor pressure difference), controlling which information should be forgotten to ensure the model effectively removes unimportant historical information. Through this dynamic update mechanism, the LSTM model can accurately predict plant water demand, avoiding the shortcomings of traditional methods in long-term memory and time-varying data processing. It can effectively combine time-series data and meteorological information to dynamically adjust model parameters, thereby improving prediction accuracy and adaptability under changing climate and environmental conditions.

[0084] S5: Combining the plant water content evolution equation and the forget gate weight, the plant water demand of the area to be predicted is obtained by inversion.

[0085] In one possible implementation, S5 specifically includes:

[0086] S501: Perform a Fourier transform on both sides of the equation for the evolution of plant water content.

[0087] S502: Perform Fourier transform on the water supply amount.

[0088] S503: Calculate the frequency domain plant water requirement based on the transformation results.

[0089] S504: Perform an inverse Fourier transform on the frequency domain plant water requirement to obtain the plant water requirement.

[0090] The specific formula for calculating the water requirement of plants is as follows:

[0091] ;

[0092] ;

[0093] ;

[0094] in, and These represent the Fourier transform and the inverse Fourier transform, respectively. This represents the frequency domain transformation result of the water replenishment at time t. This represents the plant's water requirement at time t. Represents the imaginary unit. Represents angular frequency. This represents the fractional diffusion coefficient.

[0095] Specifically, by extracting the frequency domain features of the plant water content evolution equation and water replenishment, the plant water requirement can be calculated more accurately. First, by performing a Fourier transform on the plant water content evolution equation and water replenishment, the originally complex time-domain problem is transformed into a frequency-domain problem, facilitating more efficient analysis and calculation. The Fourier transform represents the dynamic changes of the system as different frequency components, allowing the prediction of water requirement to be adjusted at different frequencies, capturing the changing patterns of plant water requirement at different time scales. Then, by combining the transformed results, the plant water requirement in the frequency domain is calculated, and the inverse Fourier transform is used to restore it to the time domain to obtain the final water requirement. This frequency domain processing method effectively removes noise, improves calculation accuracy, and further enhances the model's adaptability and accuracy to complex dynamic processes by introducing a fractional-order diffusion coefficient and forget gate weights, thus providing more accurate prediction results.

[0096] In one possible implementation, it also includes:

[0097] Retrain the LSTM model at preset intervals.

[0098] It should be noted that those skilled in the art can set the preset duration according to actual needs, and this invention does not limit this.

[0099] It should be noted that the purpose of retraining the LSTM model is to ensure that the model can adjust according to the latest input data and environmental changes, thereby improving prediction accuracy. By retraining after a preset period, the model can adapt to changing data trends, enhance its generalization ability, and, especially in long-term predictions, effectively handle new time-series information, ensuring the reliability and accuracy of prediction results.

[0100] In practical applications, combining LSTM models with remote sensing data and fractional-order differential mathematical models can accurately predict plant water requirements. Specifically, firstly, remote sensing technology is used to acquire vegetation characteristic data, such as normalized difference vegetation index (NDVI), leaf area index (LAI), and soil volumetric water content. This data provides the model with crucial information about plant growth, soil conditions, and climate change. Next, by establishing a plant water content evolution equation and introducing fractional-order differentials in time and space, the model can dynamically capture the nonlinear evolution of plant water requirements. In the LSTM model, based on input data and historical states, and utilizing update and forget gates for weight updates, the model can learn from environmental changes in real time, optimizing its ability to predict plant water requirements. Furthermore, the introduction of Fourier transform transforms the complex time-domain problem into a frequency-domain problem, effectively improving computational efficiency and accuracy, making the model more adaptable to time-varying data in complex environments. Finally, by retraining the model after a preset time period, it is ensured that the model can adjust to changes in input data and the environment, improving prediction accuracy and generalization ability.

[0101] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0102] In this embodiment of the invention, a physical evolution equation describing the dynamic changes in plant water is established using remote sensing data, providing crucial mechanistic constraints for the LSTM model and effectively overcoming the shortcomings of purely data-driven models, such as overfitting and neglecting physical laws. Simultaneously, fractional derivatives are introduced to characterize the long-term historical dependence of water changes, enabling the model to accurately capture the cumulative effects of slow evolution. These two factors together optimize the gating mechanism within the LSTM network, particularly the weight allocation of the forget gate, thereby constructing a hybrid model that integrates physical mechanisms and deep learning, achieving accurate prediction of plant water requirements under complex environments.

[0103] Reference manual attached Figure 2 The diagram shows a schematic representation of a plant water demand prediction system based on LSTM and remote sensing data provided in an embodiment of the present invention.

[0104] This invention provides a plant water demand prediction system 20 based on LSTM and remote sensing data, including: a processor 201 and a memory 202;

[0105] The memory 202 stores programs or instructions that can run on the processor 201. When the program or instructions are executed by the processor 201, they implement the steps of the above-described plant water demand prediction method based on LSTM and remote sensing data, and achieve the same technical effect. To avoid repetition, the present invention will not elaborate further.

[0106] It should be understood that the processor 201 in this embodiment of the invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0107] It should also be understood that the memory 202 in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DR RAM).

[0108] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0109] It should be understood that, in various embodiments of the present invention, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0110] Those skilled in the art will recognize that the units and algorithm steps of the various 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 these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0111] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0112] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0113] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0114] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0115] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0116] This invention provides a readable storage medium comprising: storing a program or instructions on the readable storage medium, wherein when the program or instructions are executed by a processor, the program or instructions implement the steps of the above-described plant water demand prediction method based on LSTM and remote sensing data, and can achieve the same technical effect. To avoid repetition, this invention will not elaborate further.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting plant water demand based on LSTM and remote sensing data, characterized in that, include: S1: Obtain vegetation feature data for the area to be predicted; S2: Establish a plant water content evolution equation based on the vegetation characteristic data; S3: Introduce time memory based on fractional derivatives to determine the current cell state and current hidden state of the LSTM model; S4: Update the update gate weights and forget gate weights of the LSTM model based on the hidden state; S5: Combining the plant water content evolution equation and the forget gate weight, the plant water requirement of the region to be predicted is obtained by inversion.

2. The plant water requirement prediction method based on LSTM and remote sensing data according to claim 1, characterized in that, The vegetation characteristic data include normalized vegetation index, leaf area index, and soil volumetric water content.

3. The plant water requirement prediction method based on LSTM and remote sensing data according to claim 1, characterized in that, S2 specifically includes: S201: Determine the time fraction order describing the dependence of plant water requirements based on the normalized vegetation index; S202: Determine the spatial fractional order of the plant root system complexity based on the leaf area index; S203: Determine the fractional diffusion coefficient describing the water diffusion rate of the soil to which the plant belongs based on the soil volumetric water content; S204: By combining the fractional time order, the fractional spatial order, and the fractional diffusion coefficient, the water content evolution equation is established.

4. The plant water requirement prediction method based on LSTM and remote sensing data according to claim 1, characterized in that, The LSTM model includes an input layer, an input gate, a forget gate, an output gate, a memory unit, and an output layer; the input layer is connected to the memory unit through the input gate; the forget gate is connected to the memory unit; and the memory unit is connected to the output layer through the output gate.

5. The method for predicting plant water demand based on LSTM and remote sensing data according to claim 1, characterized in that, S3 specifically includes: S301: Establish a fractional differential form of the cell state update equation; S302: Discretize the cell state update equation and extract the current cell state; S303: Calculate the current hidden state based on the current cell state.

6. The method for predicting plant water demand based on LSTM and remote sensing data according to claim 1, characterized in that, S4 specifically includes: S401: Obtain the current input features of the LSTM model; S402: Input the input features into the LSTM model; S403: Determine the update gate weight based on the hidden state and the water replenishment amount of the region to be predicted at time t; S404: Determine the forget gate weight by combining the hidden state and the saturated water vapor pressure difference.

7. The plant water requirement prediction method based on LSTM and remote sensing data according to claim 1, characterized in that, S5 specifically includes: S501: Perform a Fourier transform on both sides of the equation for the evolution of plant water content; S502: Perform Fourier transform on the water supply amount; S503: Calculate the frequency domain plant water requirement based on the transformation results; S504: Perform an inverse Fourier transform on the frequency domain plant water requirement to obtain the plant water requirement.

8. The method for predicting plant water demand based on LSTM and remote sensing data according to claim 1, characterized in that, Also includes: The LSTM model is retrained at preset intervals.

9. A plant water requirement prediction system based on LSTM and remote sensing data, characterized in that, include: Processor and memory; The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the plant water demand prediction method based on LSTM and remote sensing data as described in any one of claims 1 to 8.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the plant water demand prediction method based on LSTM and remote sensing data as described in any one of claims 1 to 8.

Images