Inplanatable artificial intelligence flood forecasting method based on drainage basin runoff and confluence mechanism

Abstract

The invention discloses an interpretable artificial intelligence flood forecasting method and device based on a drainage basin runoff and runoff mechanism, a storage medium and electronic equipment. The method comprises the following steps: acquiring original meteorological and hydrological data of a target watershed, preprocessing the original meteorological and hydrological data, and calculating watershed surface average rainfall data and watershed surface average evaporation data; constructing an XAJRNN layer based on a Xinanjiang model hydrological principle and a recurrent neural network unit structure; constructing an interpretable deep learning flood forecasting model based on an XAJRNN layer, a normalization layer and an LSTM layer; and inputting the basin surface average rainfall data and the basin surface average evaporation data into the trained interpretable deep learning flood forecasting model, and performing real-time flood forecasting. According to the method and the device, the problem of lack of interpretability and physical consistency of flood forecasting results of a deep learning model in the prior art is solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence security technology, and in particular, to an interpretable artificial intelligence flood forecasting method, device, storage medium, and electronic device based on the mechanism of basin runoff generation and concentration. Background Art

[0002] Flood forecasting relies on hydrological models, most of which are based on statistical methods and empirical formulas. However, due to the non-linear characteristics of flood events, these traditional models show obvious limitations in terms of accuracy and adaptability. In recent years, the rapid development of artificial intelligence, especially the progress of deep learning technology, has provided new perspectives and methods for flood forecasting. Deep learning, especially recurrent neural networks (RNNs) such as long short-term memory networks (LSTMs), has shown great potential in time series prediction. The core of deep learning technology lies in enabling computers to automatically learn algorithms, discover patterns by analyzing a large amount of historical sample data, and predict unknown data. This process is highly consistent with the method of hydrological models to obtain patterns and generalize hydrological processes by analyzing historical hydrometeorological data.

[0003] Although deep learning technology has been widely used in the hydrological field, demonstrating the potential to understand the laws of physical processes from the perspective of data mining. However, flood forecasting not only requires the model to have excellent performance, but also requires the model to conform to the laws of basin runoff generation and concentration, and the results to have good interpretability. Traditional hydrological models, such as the Xin'anjiang model, are based on the mechanism of basin runoff generation and concentration, and the forecasting results have good interpretability and consistency, and are relatively less dependent on data, which can effectively complement deep learning models. Currently, the integration of deep learning models and physical mechanisms is mainly limited to shallow coupling methods such as loss function adjustment and calibration of parameters, and the deep integration of the basin runoff generation and concentration mechanism and the neural network structure has not been achieved. This limitation makes it difficult to balance the physical consistency and forecasting performance of the model, and there is an urgent need to develop a new deep learning hydrological model architecture that can strictly follow physical laws and have excellent forecasting ability. Summary of the Invention

[0004] Embodiments of this application provide an interpretable artificial intelligence flood forecasting method, device, storage medium, and electronic device based on the mechanism of basin runoff generation and concentration, which can solve the problem that the forecasting results of deep learning models in the prior art lack interpretability and physical consistency.

[0005] Embodiments of this application provide an interpretable artificial intelligence flood forecasting method based on the mechanism of basin runoff generation and concentration, including: Obtain the original meteorological and hydrological data of the target basin, preprocess the original meteorological and hydrological data, and calculate the basin-average rainfall data and basin-average evaporation data; Construct the XAJRNN layer based on the hydrological principle of the Xin'anjiang model and the structure of the recurrent neural network unit; Build an interpretable deep learning flood forecasting model based on the XAJRNN layer, normalization layer, and LSTM layer; Input the basin-average rainfall data and the basin-average evaporation data into the trained interpretable deep learning flood forecasting model for real-time flood forecasting.

[0006] Furthermore, for the above-mentioned interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism, where the preprocessing of the original meteorological and hydrological data and the calculation of the basin-average rainfall data and basin-average evaporation data include: Identify the missing values in the original meteorological and hydrological data, and fill the missing values by interpolation using the nearest neighbor method; Identify the outliers in the original meteorological and hydrological data, and correct the outliers by the nearest neighbor method; Based on the preprocessed original meteorological and hydrological data, calculate the basin-average rainfall data and basin-average evaporation data by the Thiessen polygon method.

[0007] Furthermore, for the above-mentioned interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism, where the construction of the XAJRNN layer based on the hydrological principle of the Xin'anjiang model and the structure of the recurrent neural network unit includes: Replace the mathematical equations in the standard RNN unit with the equations of the Xin'anjiang model, and introduce the corresponding physical parameters to construct the XAJRNN layer.

[0008] Furthermore, for the above-mentioned interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism, where the replacement of the mathematical equations in the standard RNN unit with the equations of the Xin'anjiang model and the introduction of the corresponding physical parameters to construct the XAJRNN layer includes: The mathematical equations in the standard RNN unit are:

[0009] Where, h t 、 x t and y t represent the state, input, and output at time t respectively, is the state at time t -1, Wxh , W hh and W hy are weight parameters, b h and b y are bias parameters, σ is a non - linear activation function.

[0010] The Xin'anjiang model equation is:

[0011] Wherein, h ( t ) represents the state variable in the Xin'anjiang model, x ( t ) represents the input variable of the Xin'anjiang model, y ( t ) represents the output of the Xin'anjiang model, φ represents the parameters of the Xin'anjiang model, F and G respectively represent the mathematical equations and functions in the Xin'anjiang model; Convert the Xin'anjiang model equation into an implicit discrete equation set, and replace the mathematical equation in the standard RNN unit with the implicit discrete equation set. The implicit discrete equation set is: .

[0012] Furthermore, in the above - mentioned interpretable artificial - intelligence flood forecasting method based on the basin runoff - yield and concentration mechanism, the interpretable deep - learning flood forecasting model includes an XAJRNN layer, a normalization layer, and an LSTM layer; Inputting the basin average rainfall data and the basin average evaporation data into the trained interpretable deep - learning flood forecasting model for real - time flood forecasting includes: Inputting the basin average rainfall data and the basin average evaporation data into the XAJRNN layer to obtain variable factors; Inputting the variable factors, the basin average rainfall data, and the basin average evaporation data into the normalization layer for normalization processing; Inputting the normalized data into the LSTM layer to obtain the flood forecasting result.

[0013] Furthermore, in the above - mentioned interpretable artificial - intelligence flood forecasting method based on the basin runoff - yield and concentration mechanism, the variable factors include actual evapotranspiration intensity, basin average free water storage, basin average tension water storage in soil, and basin outlet flow.

[0014] Furthermore, in the above-mentioned interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism, the training and testing process of the interpretable deep learning flood forecasting model includes: Divide the basin-average rainfall data and basin-average evaporation data into k sub-datasets of equal size, traverse these k sub-datasets in sequence, and each time use the current sub-dataset as the test dataset, and all the remaining samples as the corresponding training datasets; Input the k training sets into the interpretable deep learning flood forecasting model for training in sequence, and select the interpretable deep learning flood forecasting model when the composite loss function value of the training dataset is the smallest based on the composite loss function as the optimal model, so as to obtain k trained interpretable deep learning flood forecasting models; Use the k optimal interpretable deep learning flood forecasting models obtained by training to test the corresponding k groups of test datasets, and calculate the average value of the k groups of calculation results as the final test result; Use the Nash-Sutcliffe efficiency coefficient, relative error, and root mean square error to evaluate the performance of the test results.

[0015] Furthermore, in the above-mentioned interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism, the composite loss function is:

[0016] Among them, is the number of samples, is the flood forecast value, is the observed value, is the average value of the observed values.

[0017] The embodiment of the present application also provides an interpretable artificial intelligence flood forecasting device based on the basin runoff generation and concentration mechanism, including: A data acquisition and calculation module, which is used to obtain the original meteorological and hydrological data of the target basin, preprocess the original meteorological and hydrological data, and calculate the basin-average rainfall data and basin-average evaporation data; A forecast model construction module, which is used to construct an XAJRNN layer based on the hydrological principle of the Xin'anjiang model and the structure of the recurrent neural network unit; And construct an interpretable deep learning flood forecasting model based on the XAJRNN layer, the normalization layer, and the LSTM layer; A real-time flood forecasting module, which is used to input the basin-average rainfall data and the basin-average evaporation data into the trained interpretable deep learning flood forecasting model for real-time flood forecasting.

[0018] An embodiment of the present application also provides a computer-readable storage medium, in which multiple instructions are stored, and the instructions are adapted to be loaded by a processor to execute any one of the above-mentioned interpretable artificial intelligence flood forecasting methods based on the basin runoff generation and concentration mechanism.

[0019] An embodiment of the present application also provides an electronic device, including a processor and a memory. The processor is electrically connected to the memory. The memory is used to store instructions and data, and the processor is used for the steps in any one of the above-mentioned interpretable artificial intelligence flood forecasting methods based on the basin runoff generation and concentration mechanism.

[0020] For the interpretable artificial intelligence flood forecasting method, device, storage medium and electronic device provided by the present application, the present application first collects and processes the original meteorological and hydrological data of the target basin, supplements and corrects the missing values and abnormal values of the original data, and then uses the Thiessen polygon method to calculate rainfall and evaporation data to obtain average rainfall data and basin surface average evaporation data; replaces the original mathematical operations in the standard RNN unit with equations with physical meanings in the Xin'anjiang model and introduces corresponding parameters to obtain the XAJRNN unit, thereby constructing the XAJRNN layer; constructs an interpretable deep learning flood forecasting model using the proposed XAJRNN layer, normalization layer and LSTM layer. The XAJRNN layer proposed by the present invention uses key hydrological mechanisms such as runoff generation and concentration in the Xin'anjiang model to replace the mathematical operations in the standard RNN unit, enabling the model to strictly follow hydrological physical laws while retaining the sequence memory ability, thereby improving the physical consistency and forecasting accuracy of the model, and solving the problem that the forecasting results of deep learning models in the prior art lack interpretability and physical consistency. Description of the Drawings

[0021] The following combines the drawings and describes the specific embodiments of the present application in detail, and the technical solutions and other beneficial effects of the present application will be obvious.

[0022] Figure 1 It is a flowchart of the interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism provided by the embodiment of the present application.

[0023] Figure 2 It is a schematic structural diagram of the standard RNN unit provided by the embodiment of the present application.

[0024] Figure 3 It is a schematic structural diagram of the XAJRNN unit provided by the embodiment of the present application.

[0025] Figure 4 It is a structural diagram of the interpretable deep learning flood forecasting model provided by the embodiment of the present application.

[0026] Figure 5 This is a schematic structural diagram of an interpretable artificial intelligence flood forecasting device based on the basin runoff generation and concentration mechanism provided by an embodiment of the present application.

[0027] Figure 6 This is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0028] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0029] The embodiments of the present application provide an interpretable artificial intelligence flood forecasting method, device, storage medium, and electronic device based on the basin runoff generation and concentration mechanism. An interpretable artificial intelligence flood forecasting device based on the basin runoff generation and concentration mechanism provided by the embodiments of the present application can be integrated in an electronic device, and the electronic device can be a device such as a terminal or a server. Among them, the terminal can include a tablet computer, a notebook computer, a personal computer (PC), a microprocessing box, or other devices, etc.

[0030] Please refer to Figure 1 , Figure 1 This is a flowchart of an interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism provided by an embodiment of the present application, which is applied to an electronic device. The interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism includes the following steps: S1. Obtain the original meteorological and hydrological data of the target basin, preprocess the original meteorological and hydrological data, and calculate the basin-average rainfall data and the basin-average evaporation data.

[0031] Among them, the original meteorological and hydrological data includes rainfall data of rain gauges, evaporation data of weather stations, and flow data of hydrological stations.

[0032] In one embodiment, step S1 includes the following steps: S11. Identify the missing values in the original meteorological and hydrological data, and fill in the missing values by interpolation using the nearest neighbor method.

[0033] S12. Identify the outliers in the original meteorological and hydrological data, and correct the outliers by the nearest neighbor method.

[0034] Specifically, identify the missing values in the rainfall data of rain gauges within the target basin, interpolate and fill these missing values using the nearest neighbor method, detect the outliers in the rainfall data of rain gauges, and correct these outliers through nearest neighbor interpolation to obtain a complete rainfall data sequence of rain gauges.

[0035] Identify the missing values in the evaporation data of meteorological stations, interpolate and fill these missing values using the nearest neighbor method, detect the outliers in the evaporation data of meteorological stations, and correct these outliers through nearest neighbor interpolation to obtain a complete evaporation data sequence of meteorological stations.

[0036] Identify the missing values in the flow data of hydrological stations, interpolate and fill these missing values using the nearest neighbor method, detect the outliers in the flow data of hydrological stations, and correct these outliers through nearest neighbor interpolation to obtain a complete flow data sequence of hydrological stations.

[0037] Among them, the nearest neighbor method means first determining the k samples closest to the sample with missing data according to the Euclidean distance, and then weighted averaging these k values to estimate the missing data of this sample.

[0038] Unify the processed rainfall data sequence of rain gauges, evaporation data sequence of meteorological stations, and flow data sequence of hydrological stations to the same time scale according to the prediction requirements using the average value method, so as to obtain a rainfall data sequence of rain gauges, an evaporation data sequence of meteorological stations, and a flow data sequence of hydrological stations with the same time scale.

[0039] S13, Based on the preprocessed original meteorological and hydrological data, calculate the basin-average rainfall data and basin-average evaporation data through the Thiessen polygon method.

[0040] Specifically, the embodiments of the present invention collect the daily rainfall data of 17 rain gauges in the target basin, the daily evaporation data of 1 meteorological station, and the daily flow data of 1 hydrological station at the basin outlet control. After interpolating the missing values and correcting the outliers of these data, 17 sets of daily rainfall data sequences of stations, 1 set of daily evaporation data sequences of stations, and 1 set of daily flow data sequences are obtained, and then the basin-average rainfall and evaporation data are calculated using the Thiessen polygon method.

[0041] S2, Based on the hydrological principles of the Xin'anjiang model and the structure of recurrent neural network units, construct the XAJRNN layer.

[0042] Replace the mathematical equations in the standard RNN unit with the Xin'anjiang model equations and introduce the corresponding physical parameters to construct the XAJRNN layer.

[0043] Specifically, the standard RNN is a neural network structure specialized for processing time series data. It can utilize the time-dependent characteristics in the sequence and assist the current calculation by memorizing the information of the previous state. Figure 2This is a schematic diagram of the standard RNN unit structure provided by the embodiments of the present application. As Figure 2 shown, at the t-th time step, the calculation of an RNN unit can be divided into the following two steps: The first is the update of the hidden state h t , and the second is the calculation of the output y t . The calculation formulas are as follows:

[0044] Among them, h t , x t and y t represent the state, input, and output at time t respectively, is the state at time t -1, W xh , W hh and W hy are weight parameters, b h and b y are bias parameters, σ is a non-linear activation function.

[0045] Establishing the Xin'anjiang model in the basin can be regarded as a complete system, which can characterize the changes in state variables in the basin (such as the change in the storage of the average tension water in the upper layer of soil). At the same time, the Xin'anjiang model describes the response of the basin model system under specific input conditions. These can be represented by the combination of differential equations and output equations:

[0046] Among them, h ( t ) represents the state variable in the Xin'anjiang model, x ( t ) represents the input variable in the Xin'anjiang model, y ( t ) represents the output of the Xin'anjiang model, φ represents the parameter of the Xin'anjiang model, F and G represent the mathematical equations and functions in the Xin'anjiang model respectively.

[0047] The above equations are explicit continuous equation sets. However, in the actual process, in order to obtain the numerical analytical solution of the equation sets, implicit discrete equation sets are generally used to calculate the results.

[0048] Convert the Xin'anjiang model equations into implicit discrete equation sets, and replace the mathematical equations in the standard RNN unit with the implicit discrete equation sets. The implicit discrete equation sets are as follows:

[0049] S3. Construct an interpretable deep learning flood forecasting model based on the XAJRNN layer, normalization layer, and LSTM layer.

[0050] The above equation sets show obvious similarities in form, all presenting a combined structure of differential equations and output equations. Based on this characteristic, under the guidance of the physical mechanism of the Xin'anjiang model, by replacing the mathematical equations in the standard RNN unit with the Xin'anjiang model equations with clear hydrological physical meanings and introducing corresponding physical parameters, a Xin'anjiang recurrent neural network (XAJRNN) unit is constructed. Figure 3 The structural schematic diagram of the XAJRNN unit provided by the embodiment of this application is as Figure 3 shown. Similar to the standard RNN layer, the XAJRNN layer is composed of multiple XAJRNN units, and these units retain the temporal memory ability of the sequence through physical constraints. In the architecture of the XAJRNN layer, the connection relationships between units (including input x t , state h t and output y t ) are characterized by implicit discrete equations.

[0051] Specifically, the evapotranspiration module of the Xin'anjiang model adopts a three-layer evapotranspiration formula, the runoff calculation uses the tension water storage capacity curve, the sub-source module is in the form of three sources, and the confluence module includes two sub-modules of overland and river network confluence, which adopt a linear reservoir and Nash unit hydrograph respectively. The Xin'anjiang model consists of the following parts: input variables, state variables, fluxes, output variables, and parameters, as well as the mathematical equations between them. The input variables include measured areal average precipitation and evaporation, and the output variable is the predicted flow. The state variables are physical quantities representing the state of the basin, and their dimensions are independent of time. The fluxes are used to describe the water exchange amount within the basin or between the basin and the outside, which can be expressed as a function of state variables or other fluxes, and their dimensions are related to time. The mathematical equations can be further divided into state variable control equations and flux constitutive equations. Among them, the control equations are used to describe the evolution of the state variables of the basin system over time; the constitutive equations are used to establish the relationship between unknown fluxes and state variables or other known fluxes.

[0052] S4. Input the basin areal average rainfall data and basin areal average evaporation data into the trained interpretable deep learning flood forecasting model for real-time flood forecasting.

[0053] In one embodiment, step S4 includes the following steps: S41, input the average rainfall data and average evaporation data of the basin area into the XAJRNN layer to obtain variable factors; S42, input the variable factors, average rainfall data and average evaporation data of the basin area into the normalization layer for normalization processing; S43, input the normalized data into the LSTM layer to obtain the flood forecast result.

[0054] Specifically, the interpretable deep learning flood forecasting model includes an input layer, three neural network layers and an output layer. Figure 4 For the structural diagram of the interpretable deep learning flood forecasting model provided by the embodiment of the present application, as Figure 4 shown, this neural network layer follows the principle of water balance and uses the physical sub-processes in the Xin'anjiang model to describe the runoff generation and concentration process. First, calculate the input data through the XAJRNN layer and output variable factors that have an obvious impact on the runoff generation and concentration process, and transfer these variable factors together with the input data to the normalization layer. The role of this layer is to normalize the data to help the model converge faster, improve training stability, and reduce the impact of differences between different features on model learning. Specifically, the normalization layer adjusts the mean of the data to 0 and the standard deviation to 1. Finally, input the standardized data into the LSTM layer to obtain the flood forecast result.

[0055] Among them, the variable factors include actual evapotranspiration intensity, average free water storage in the basin, average tension water storage in the basin soil, and basin outlet flow.

[0056] To avoid overfitting in the training process of the interpretable deep learning model, the k-fold cross-validation method is used to divide the training set and test set, that is, divide the average rainfall and evaporation on the surface into k sub-datasets of equal size, traverse these k sub-datasets in turn, and each time use the current sub-dataset as the test dataset and all the remaining samples as the corresponding training dataset to train the interpretable deep learning flood forecasting model. The training and testing process of the interpretable deep learning flood forecasting model includes: Divide the average rainfall data and average evaporation data of the basin area into k sub-datasets of equal size, traverse these k sub-datasets in turn, and each time use the current sub-dataset as the test dataset and all the remaining samples as the corresponding training dataset; Input the k training sets into the interpretable deep learning flood forecasting model for training in turn, and select the interpretable deep learning flood forecasting model with the smallest composite loss function value of the training dataset based on the composite loss function as the optimal model, that is, k trained interpretable deep learning flood forecasting models can be obtained; Use the k optimal interpretable deep learning flood forecasting models obtained through training to test the corresponding k groups of test data sets, and calculate the average value of the calculation results of the k groups as the final test result; Use the Nash-Sutcliffe efficiency coefficient, relative error, and root mean square error to evaluate the performance of the test results.

[0057] Specifically, the embodiments of the present invention set in the k-fold cross-validation method to be 4, that is, the areal average rainfall, evaporation, and station flow data sequences are respectively divided into 4 equal-sized subsets. Traverse these 4 subsets in turn. Each time, use the current subset as the test data set, and all the remaining samples as the corresponding training data set to train the interpretable deep learning flood forecasting model. According to the 4 groups of training data sets divided, train the constructed interpretable deep learning flood forecasting model, and select the interpretable deep learning flood forecasting model when the composite loss function value of the training data set is the smallest as the optimal model, that is, 4 trained interpretable deep learning flood forecasting models can be obtained.

[0058] The embodiments of the present invention input 4 groups of test data sets into the corresponding 4 trained optimal interpretable deep learning flood forecasting models to obtain 4 groups of forecasting results, and calculate the average value of these 4 groups of results as the final forecasting result. Use , and to evaluate the results of the interpretable deep learning model for forecasting flow.

[0059] Among them, during the training process, it is necessary to set the hyperparameters of the model, including the number of input time steps, the number of neurons, the learning rate, the batch size, and the number of training iterations.

[0060] Specifically, the embodiments of the present invention set the number of model input time steps to 15, the number of convolutional kernels to 64, the learning rate to 0.001, the batch size to 64, and the maximum number of training iterations to 300.

[0061] In one embodiment, the composite loss function (CLF) during the training process comprehensively considers the mean square error (MSE), Nash-Sutcliffe efficiency coefficient, and relative error. The calculation formula is as follows:

[0062] Among them, is the number of samples, is the flood forecast value, is the observed value, is the average value of the observed values.

[0063] In one embodiment, the Nash-Sutcliffe efficiency coefficient ( ), relative error ( ), and root mean square error ( ) are used to evaluate the overall performance of the ConvLSTM-SSA flood forecasting model, and the calculation formulas are as follows.

[0064]

[0065]

[0066]

[0067] Among them, is the number of samples, and respectively represent the observed flow data and its mean value, represents the forecast flow data, the closer the value is to 1, the closer the value and

[0068] value are to 0, the better the model effect.

[0069] Figure 5 Please refer to Figure 5 which specifically describes the interpretable artificial intelligence flood forecasting device based on the basin runoff generation and concentration mechanism provided in the embodiments of the present application. Applied to an electronic device, the interpretable artificial intelligence flood forecasting device based on the basin runoff generation and concentration mechanism may include: A data acquisition and calculation module, configured to obtain the original meteorological and hydrological data of the target basin, preprocess the original meteorological and hydrological data, and calculate the basin average rainfall data and basin average evaporation data; A forecast model construction module, configured to construct an XAJRNN layer based on the hydrological principle of the Xin'anjiang model and the structure of the recurrent neural network unit; and construct an interpretable deep learning flood forecasting model based on the XAJRNN layer, normalization layer, and LSTM layer; A real-time flood forecasting module for inputting the basin average rainfall data and the basin average evaporation data into a trained interpretable deep learning flood forecasting model for real-time flood forecasting.

[0070] In specific implementation, each of the above modules and / or units can be implemented as an independent entity, or can be combined arbitrarily to be implemented as the same or several entities. For the specific implementation of each of the above modules and / or units, reference can be made to the foregoing method embodiments. The beneficial effects that can be specifically achieved can also be referred to the beneficial effects in the foregoing method embodiments, which will not be elaborated herein.

[0071] In addition, an embodiment of the present application further provides an electronic device, which can be a device such as a computer or a tablet computer. The electronic device can implement the steps in any of the embodiments of the interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism provided by the embodiments of the present application. Therefore, the beneficial effects that can be achieved by any of the interpretable artificial intelligence flood forecasting methods based on the basin runoff generation and concentration mechanism provided by the embodiments of the present invention can be achieved. For details, refer to the foregoing embodiments, which will not be elaborated herein.

[0072] Figure 6 The specific structural block diagram of the electronic device provided by the embodiment of the present invention is shown. The electronic device can be used to implement the interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism provided in the above embodiments. The electronic device 500 can be a device such as a terminal or a server. Among them, the terminal can include a tablet computer, a notebook computer, a personal computer (PC), a microprocessing box, or other devices, etc.

[0073] The RF circuit 510 is used to receive and transmit electromagnetic waves, realizing the mutual conversion between electromagnetic waves and electrical signals, so as to communicate with a communication network or other devices. The RF circuit 510 may include various existing circuit elements for performing these functions. For example, antennas, radio frequency transceivers, digital signal processors, encryption / decryption chips, subscriber identity module (SIM) cards, memories, and so on. The RF circuit 510 can communicate with various networks such as the Internet, enterprise intranets, wireless networks, or communicate with other devices through wireless networks. The above-mentioned wireless networks may include cellular phone networks, wireless local area networks, or metropolitan area networks. The above-mentioned wireless networks can use various communication standards, protocols, and technologies, including but not limited to Global System for Mobile Communication (GSM), Enhanced Data GSM Environment (EDGE), Wideband Code Division Multiple Access (WCDMA), Code Division Access (CDMA), Time Division Multiple Access (TDMA), Wireless Fidelity (Wi-Fi) (such as Institute of Electrical and Electronics Engineers standards IEEE 802.11a, IEEE 802.11b, IEEE 802.11g, and / or IEEE 802.11n), Voice over Internet Protocol (VoIP), Worldwide Interoperability for Microwave Access (Wi-Max), other protocols for email, instant messaging, and short messages, and any other suitable communication protocols, and may even include those protocols that have not yet been developed currently.

[0074] The memory 520 can be used to store software programs and modules, such as the corresponding program instructions / modules in the above embodiments. The processor 580 executes various functional applications and data processing by running the software programs and modules stored in the memory 520, that is, realizes functions such as taking pictures with the front camera, processing the captured images, and switching the display colors of the display content on the display screen. The memory 520 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory 520 may further include a memory remotely provided with respect to the processor 580, and these remote memories can be connected to the electronic device 500 through a network. Examples of the above network include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and combinations thereof.

[0075] The input unit 530 can be used to receive input digital or character information, and generate a keyboard and a mouse related to user settings and function controls. The display unit 540 can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces, and these graphical user interfaces can be composed of graphics, texts, icons, videos, and any combination thereof. The display unit 540 may include a display panel 541. Optionally, the display panel 541 can be configured in forms such as LCD (Liquid Crystal Display) and OLED (Organic Light-Emitting Diode).

[0076] The audio circuit 560, the speaker 561, and the microphone 562 can provide an audio interface between the user and the electronic device 500. The audio circuit 560 can transmit the electrical signal converted from the received audio data to the speaker 561, and the speaker 561 converts it into a sound signal for output; on the other hand, the microphone 562 converts the collected sound signal into an electrical signal, which is received by the audio circuit 560 and then converted into audio data. After the audio data is output to the processor 580 for processing, it is sent to another terminal, such as through the RF circuit 510, or the audio data is output to the memory 520 for further processing. The audio circuit 560 may further include an earphone jack to provide communication between the peripheral earphone and the electronic device 500.

[0077] The electronic device 500 can help the user receive requests, send information, etc. through the transmission module 570 (such as a Wi-Fi module), and it provides the user with wireless broadband Internet access. Although the transmission module 570 is shown in the figure, it can be understood that it does not belong to the essential components of the electronic device 500 and can be omitted completely as needed without changing the essence of the invention.

[0078] The processor 580 is the control center of the electronic device 500, connecting various parts of the entire mobile phone through various interfaces and circuits. By running or executing software programs and / or modules stored in the memory 520, and calling the data stored in the memory 520, it executes various functions of the electronic device 500 and processes data, thereby monitoring the electronic device as a whole. Optionally, the processor 580 may include one or more processing cores; in some embodiments, the processor 580 may integrate an application processor and a modem processor, where the application processor mainly processes the operating system, user interface, and application programs, etc., and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor may not be integrated into the processor 580 either.

[0079] The electronic device 500 also includes a power source 590 (such as a battery) for supplying power to each component. In some embodiments, the power source can be logically connected to the processor 580 through a power management system, so as to realize functions such as management of charging, discharging, and power consumption management through the power management system. The power source 590 may also include any components such as one or more DC or AC power sources, a recharge system, a power failure detection circuit, a power converter or inverter, and a power status indicator.

[0080] Although not shown, the electronic device 500 also includes a camera (such as a front camera, a rear camera), a Bluetooth module, etc., which will not be elaborated here. Specifically, in this embodiment, the display unit of the electronic device is a touch screen display, and the mobile terminal also includes a memory, and one or more programs, where one or more programs are stored in the memory and are configured to be executed by one or more processors. The one or more programs include instructions for performing the following operations: In specific implementation, the above-mentioned each module can be implemented as an independent entity, or can be combined arbitrarily and implemented as the same or several entities. For the specific implementation of the above-mentioned each module, reference can be made to the method embodiments described above, which will not be elaborated here.

[0081] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions, or by controlling relevant hardware through instructions. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. For this purpose, an embodiment of the present invention provides a storage medium, in which multiple instructions are stored, and these instructions can be loaded by a processor to execute the steps of any one of the embodiments of the interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism provided by the embodiments of the present invention.

[0082] Among them, the computer-readable storage medium may include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, etc.

[0083] Since the instructions stored in the storage medium can execute the steps in any embodiment of the interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism provided by the embodiments of the present invention, the beneficial effects achievable by any of the interpretable artificial intelligence flood forecasting methods based on the basin runoff generation and concentration mechanism provided by the embodiments of the present invention can be realized. For details, see the previous embodiments and will not be elaborated here.

[0084] The above has introduced in detail an interpretable artificial intelligence flood forecasting method, device, storage medium and electronic device based on the basin runoff generation and concentration mechanism provided by the embodiments of the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those skilled in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. An interpretable artificial intelligence flood forecasting method based on the mechanism of basin runoff generation and concentration, characterized in that The method includes: Obtain the original meteorological and hydrological data of the target basin, preprocess the original meteorological and hydrological data, and calculate the basin-averaged rainfall data and the basin-averaged evaporation data; Construct an XAJRNN layer based on the hydrological principles of the Xin'anjiang model and the structure of recurrent neural network units; Build an interpretable deep learning flood forecasting model based on the XAJRNN layer, the normalization layer, and the LSTM layer; Input the basin-averaged rainfall data and the basin-averaged evaporation data into the trained interpretable deep learning flood forecasting model for real-time flood forecasting.

2. The interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism according to claim 1, wherein The preprocessing of the original meteorological and hydrological data and the calculation of the basin-averaged rainfall data and the basin-averaged evaporation data include: Identify the missing values in the original meteorological and hydrological data, and fill in the missing values by interpolation using the nearest neighbor method; Identify the outliers in the original meteorological and hydrological data, and correct the outliers by the nearest neighbor method; Based on the preprocessed original meteorological and hydrological data, calculate the basin-averaged rainfall data and the basin-averaged evaporation data by the Thiessen polygon method.

3. The interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism according to claim 1, characterized in that, The construction of the XAJRNN layer based on the hydrological principles of the Xin'anjiang model and the structure of recurrent neural network units includes: Replace the mathematical equations in the standard RNN unit with the Xin'anjiang model equations and introduce the corresponding physical parameters to construct the XAJRNN layer.

4. The interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism according to claim 3, characterized in that, The replacement of the mathematical equations in the standard RNN unit with the Xin'anjiang model equations and the introduction of the corresponding physical parameters to construct the XAJRNN layer include: The mathematical equations in the standard RNN unit are: Among them, h t 、 x t and y t represent the state, input, and output of time t respectively, is the state of time t -1, W xh 、 W hh and W hy are weight parameters, b h and b y are bias parameters, σ is a non-linear activation function. The Xin'anjiang model equations are: Among them, h ( t ) represents the state variables in the Xin'anjiang model, x ( t ) represents the input variables of the Xin'anjiang model, y ( t ) represents the output of the Xin'anjiang model, φ represents the parameters of the Xin'anjiang model, F and G respectively represent the mathematical equations and functions in the Xin'anjiang model; Convert the Xin'anjiang model equations into implicit discrete equation sets, and replace the mathematical equations in the standard RNN unit with the implicit discrete equation sets. The implicit discrete equation sets are: 。 5. The interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism according to claim 3, characterized in that, The interpretable deep learning flood forecasting model includes an XAJRNN layer, a normalization layer, and an LSTM layer; The input of the basin-averaged rainfall data and the basin-averaged evaporation data into the trained interpretable deep learning flood forecasting model for real-time flood forecasting includes: Input the basin-averaged rainfall data and the basin-averaged evaporation data into the XAJRNN layer to obtain variable factors; Input the variable factors, the basin-averaged rainfall data, and the basin-averaged evaporation data into the normalization layer for normalization processing; Input the normalized data into the LSTM layer to obtain the flood forecasting result.

6. The interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism according to claim 5, wherein The variable factors include the actual evapotranspiration intensity, the basin-averaged free water storage, the basin-averaged tension water storage in the soil, and the flow at the basin outlet.

7. The interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism according to claim 1, characterized in that, The training and testing process of the interpretable deep learning flood forecasting model includes: Divide the basin-averaged rainfall data and the basin-averaged evaporation data into k sub-datasets of equal size, traverse these k sub-datasets in turn, and each time use the current sub-dataset as the test dataset and all the remaining samples as the corresponding training dataset; Input k training sets into the interpretable deep learning flood forecasting model in sequence for training. Based on the composite loss function, select the interpretable deep learning flood forecasting model when the composite loss function value of the training data set is the smallest as the optimal model, and then k trained interpretable deep learning flood forecasting models can be obtained; Use the k optimal interpretable deep learning flood forecasting models obtained by training to test the corresponding k groups of test data sets, and calculate the average value of the k groups of calculation results as the final test result; Adopt Nash-Sutcliffe efficiency coefficient, relative error and root mean square error to evaluate the performance of the test results.

8. The interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism according to claim 7, characterized in that The composite loss function is: Among them, is the number of samples, is the flood forecast value, is the observed value, is the average value of the observed values.

9. An interpretable artificial intelligence flood forecasting device based on the mechanism of basin runoff generation and concentration, characterized in that, Including: A data acquisition and calculation module, which is used to obtain the original meteorological and hydrological data of the target basin, preprocess the original meteorological and hydrological data, and calculate the basin average rainfall data and basin average evaporation data; A forecast model construction module, which is used to construct an XAJRNN layer based on the hydrological principle of the Xin'anjiang model and the structure of the recurrent neural network unit; And construct an interpretable deep learning flood forecasting model based on the XAJRNN layer, the normalization layer and the LSTM layer; A real-time flood forecasting module, which is used to input the basin average rainfall data and the basin average evaporation data into the trained interpretable deep learning flood forecasting model for real-time flood forecasting.

10. A computer-readable storage medium, characterized in that, Multiple instructions are stored in the computer-readable storage medium, and the instructions are suitable for being loaded by a processor to execute the interpretable artificial intelligence flood forecasting method based on the basin runoff generation and concentration mechanism according to any one of claims 1 to 8.

Images