A runoff prediction method and device based on a Gaussian function and a medium

Abstract

The application discloses a runoff prediction method and device based on a Gaussian function and a medium, relates to the technical field of runoff prediction, and comprises the following steps: determining fitting values of a plurality of Gaussian parameters of each historical year based on monitored values of runoff of each month in each historical year; inputting the fitting values of each Gaussian parameter of all historical years into a Gaussian parameter prediction model of the corresponding Gaussian parameter respectively to obtain prediction values of each Gaussian parameter of a future year; determining Gaussian function values of each month in the future year based on the prediction values of each Gaussian parameter of the future year; and determining prediction values of runoff of each month in the future year based on the Gaussian function values of each month in the future year and a runoff prediction model. The runoff prediction model is obtained by training a back propagation neural network. The application improves the prediction accuracy of runoff.

Description

Technical Field

[0001] The present application relates to the technical field of runoff prediction, and in particular to a runoff prediction method, device and medium based on a Gaussian function. Background Art

[0002] In the problem of water resource scheduling in a river basin, it is necessary to predict the runoff of the monitoring section of the river basin and formulate a future water resource scheduling plan based on the results of water resource allocation and scheduling calculations. There are two main methods for predicting the relevant runoff: (1) Empirical method: based on past experience and supplemented by other related algorithms, the predicted value of the runoff is determined to formulate a scheduling plan. (2) Machine learning method: Current researchers have proposed many machine learning algorithms for runoff prediction, such as grayscale prediction method, autoregressive integral moving average model algorithm, support vector regression algorithm, random forest algorithm, and deep learning algorithms such as recurrent neural network, long short-term memory network, gated recurrent unit, Transformer framework, Informer framework, etc.

[0003] However, the principle of full-sequence fitting or window sliding prediction in machine learning methods only works well on the training set, but performs poorly on the test set. Generally speaking, most models fit well on the training set, and when using future real data on the test set to predict the next future prediction data, the error is small, which only means that the fit is good on the test set. When performing sliding prediction in actual applications, window sliding is required. After adding the future prediction value to the window and continuing to the next prediction, the error gradually accumulates, causing the model to perform worse and worse over long-term predictions. After a certain period of time, the model will give the average value as the prediction value, resulting in low prediction accuracy for runoff volume prediction. Summary of the Invention

[0004] The purpose of this application is to provide a runoff prediction method, device and medium based on Gaussian function to solve the problem of low runoff prediction accuracy.

[0005] To achieve the above objectives, this application provides the following solutions.

[0006] In a first aspect, the present application provides a runoff prediction method based on a Gaussian function, comprising:

[0007] Obtain the monitoring values ​​of runoff in each month of multiple historical years in the historical period of the section to be predicted;

[0008] Determining the fitting values ​​of multiple Gaussian parameters of each historical year of the section to be predicted based on the monitored values ​​of runoff in each month of each historical year of the section to be predicted;

[0009] The fitted values ​​of each Gaussian parameter of all historical years of the section to be predicted are respectively input into the Gaussian parameter prediction model of the corresponding Gaussian parameter to obtain the predicted values ​​of each Gaussian parameter of the future year of the section to be predicted; each Gaussian parameter prediction model is obtained by training the long short-term memory network; the future year is the next year of the historical period;

[0010] Determine the Gaussian function value of each month in the future year of the section to be predicted based on the predicted value of each Gaussian parameter in the future year of the section to be predicted;

[0011] The Gaussian function value of each month in the future year of the section to be predicted is input into the runoff prediction model to obtain the predicted value of the runoff of each month in the future year of the section to be predicted; the runoff prediction model is obtained by training the back propagation neural network.

[0012] In one embodiment, based on the monitored values ​​of runoff volume in each month of each historical year of the section to be predicted, the fitting values ​​of multiple Gaussian parameters of each historical year of the section to be predicted are determined, including:

[0013] Determine the average value of the runoff volume of each month of the section to be predicted based on the monitored values ​​of the runoff volume of each month in all historical years of the section to be predicted;

[0014] Based on the average value of the runoff volume of each month of the section to be predicted, the monitored values ​​of the runoff volume of each month in each historical year of the section to be predicted are adjusted to obtain the adjusted value of the runoff volume of each month in each historical year of the section to be predicted;

[0015] Gaussian fitting is performed based on the adjusted values ​​of runoff in all months of each historical year of the section to be predicted, so as to obtain fitting values ​​of multiple Gaussian parameters of each historical year of the section to be predicted.

[0016] In one embodiment, the process of determining the Gaussian parameter prediction model for any current Gaussian parameter includes:

[0017] Obtain the fitted values ​​of the current Gaussian parameters of the first sample section for N+1 consecutive historical years; N≥1;

[0018] Building long short-term memory networks;

[0019] The long short-term memory network is trained with the fitting values ​​of the current Gaussian parameters of the first N historical years of the first sample section as input and the fitting values ​​of the current Gaussian parameters of the N+1th historical year of the first sample section as output to obtain a Gaussian parameter prediction model for the current Gaussian parameters.

[0020] In one embodiment, obtaining the fitted values ​​of the current Gaussian parameters of the first sample section for N+1 consecutive historical years includes:

[0021] Obtain the monitoring values ​​of runoff volume for each month in N+1 consecutive historical years at the first sample section;

[0022] Based on the monitored values ​​of the runoff in each month of the N+1 historical years of the first sample section, the fitting values ​​of the current Gaussian parameters of the first sample section in the N+1 historical years are determined.

[0023] In one embodiment, determining the fitted values ​​of the current Gaussian parameters of the first sample section in the N+1 historical years based on the monitored values ​​of the runoff in each month in the N+1 historical years of the first sample section includes:

[0024] Determine the average value of the runoff volume of each month of the first sample section based on the monitored values ​​of the runoff volume of each month in all historical years of the first sample section;

[0025] Based on the average value of the runoff volume of each month of the first sample section, the monitored values ​​of the runoff volume of each month in N+1 historical years of the first sample section are adjusted to obtain the adjusted value of the runoff volume of each month in each historical year of the first sample section;

[0026] Gaussian fitting is performed based on the adjusted values ​​of runoff in all months in the N+1 historical years of the first sample section to obtain the fitting values ​​of the current Gaussian parameters of the N+1 historical years of the first sample section.

[0027] In one embodiment, the process of determining the runoff prediction model includes:

[0028] Obtain the Gaussian function value of each month in the historical years of the second sample section and the monitoring value of the runoff volume of each month in the historical years of the second sample section;

[0029] Build a back-propagation neural network;

[0030] The back propagation neural network is trained using the Gaussian function values ​​of each month in the historical years of the second sample section and the monitored values ​​of the runoff of each month in the historical years of the second sample section to obtain the runoff prediction model.

[0031] In one embodiment, obtaining the Gaussian function value of each month in the historical year of the second sample section and the monitoring value of the runoff volume of each month in the historical year of the second sample section includes:

[0032] Obtain the monitoring values ​​of runoff volume for each month in the historical years of the second sample section;

[0033] Determine fitting values ​​of multiple Gaussian parameters of the second sample section in the historical years based on the monitored values ​​of runoff in each month of the historical years of the second sample section;

[0034] Based on the fitting values ​​of multiple Gaussian parameters of the historical years of the second sample section, the Gaussian function value of each month in the historical years of the second sample section is determined.

[0035] In one embodiment, the back propagation neural network is trained using the Gaussian function values ​​of each month in the historical years of the second sample section and the monitored values ​​of the runoff of each month in the historical years of the second sample section to obtain the runoff prediction model, including:

[0036] Normalizing the Gaussian function values ​​of each month in the historical years of the second sample section and the monitored values ​​of the runoff of each month in the historical years of the second sample section respectively to obtain the normalized Gaussian function values ​​of each month in the historical years of the second sample section and the normalized values ​​of the runoff of each month in the historical years of the second sample section;

[0037] The back propagation neural network is trained with the normalized Gaussian function value of each month in the historical years of the second sample section as input and the normalized value of the runoff of each month in the historical years of the second sample section as output to obtain the runoff prediction model.

[0038] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned runoff prediction method based on the Gaussian function.

[0039] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned runoff prediction method based on the Gaussian function.

[0040] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0041] The present application discloses a method, device, and medium for predicting runoff based on a Gaussian function. First, the monitored runoff values ​​for each month in multiple historical years are used to determine the fitted values ​​of the Gaussian parameters for each historical year. Then, a Gaussian parameter prediction model obtained by training a long short-term memory network is used to predict the predicted values ​​of the Gaussian parameters for future years based on the fitted values ​​of the Gaussian parameters for each historical year. Then, based on the predicted values ​​of the Gaussian parameters for future years, the Gaussian function values ​​for each month in future years are determined. Finally, a runoff prediction model obtained by training a back-propagation neural network is used to predict the predicted values ​​of the runoff for each month in future years. In response to the problem of error accumulation in step-by-step prediction during long-term runoff prediction, by converting the problem into a prediction problem for multiple Gaussian parameters for each year, the error in long-term runoff prediction can be effectively reduced, thereby improving the accuracy of runoff prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 A schematic flow chart of a runoff prediction method based on a Gaussian function provided in one embodiment of the present application.

[0044] Figure 2 Schematic diagram of runoff monitoring values.

[0045] Figure 3 Schematic diagram of the adjusted value of runoff.

[0046] Figure 4 Schematic diagram of the long short-term memory network structure.

[0047] Figure 5 Schematic diagram of the structure of the back-propagation neural network.

[0048] Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0050] The purpose of this application is to provide a runoff prediction method, device and medium based on Gaussian function, aiming to improve the prediction accuracy of runoff.

[0051] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0052] In an exemplary embodiment, Figure 1 As shown in FIG, a runoff prediction method based on a Gaussian function is provided, including:

[0053] Step 1: Obtain the monitoring values ​​of runoff in each month of multiple historical years in the historical period of the section to be predicted.

[0054] Specifically, the monitoring values ​​of runoff are collected by sensors. There are abnormal values ​​in the data collected by sensors. These abnormal values ​​include null values ​​and non-numeric symbols caused by sensor failure or signal failure. At the same time, the huge mutation values ​​caused by extreme climate are also regarded as noise, because these noises will greatly affect the distribution state of the monitoring values ​​of runoff and destroy the distribution law of the overall data. For the monitoring values ​​of runoff in each month of multiple historical years in the historical period of the section to be predicted, if there are null values ​​or abnormal values, linear interpolation is directly used to fill in the data. The complete monitoring values ​​of runoff in each month of multiple historical years in the historical period of the section to be predicted are as follows: Figure 2 shown. Figure 2 In the figure, the vertical axis represents the monitored value of runoff, and the horizontal axis represents the number of each month in each historical year in the historical period. In the horizontal axis, 1 represents January of the first historical year in the historical period of the section to be predicted, 2 represents February of the first historical year in the historical period of the section to be predicted, 13 represents January of the second historical year in the historical period of the section to be predicted, 14 represents February of the second historical year in the historical period of the section to be predicted, and so on.

[0055] Step 2: Based on the monitored values ​​of runoff volume in each month of each historical year of the section to be predicted, the fitting values ​​of multiple Gaussian parameters of each historical year of the section to be predicted are determined.

[0056] As an optional implementation, step 2 includes steps 21 to 23.

[0057] Step 21: Based on the monitored values ​​of runoff in each month of all historical years of the section to be predicted, determine the average value of runoff in each month of the section to be predicted.

[0058] Specifically, the calculation formula for the average runoff volume is:

[0059] .

[0060] in, For cross-section The average monthly runoff volume, ; is the number of historical years; For the Historical years Monitoring values ​​of monthly runoff, .

[0061] Step 22: Based on the average value of the runoff volume of each month of the section to be predicted, the monitored values ​​of the runoff volume of each month in each historical year of the section to be predicted are adjusted to obtain the adjusted values ​​of the runoff volume of each month in each historical year of the section to be predicted.

[0062] Specifically, in step 22, according to the adjustment formula, the monitored values ​​of the runoff of each month in each historical year of the section to be predicted are adjusted based on the average value of the runoff of each month in the section to be predicted, so as to obtain the adjusted value of the runoff of each month in each historical year of the section to be predicted. The adjustment formula is:

[0063] .

[0064] in, For the Historical years Adjusted value of monthly runoff volume; is the outlier scaling factor, which scales the larger outliers to the standard range. The smaller the value of The stronger the noise suppression, It is necessary to make corresponding adjustments based on the actual data. The adjusted values ​​of runoff in each month of each historical year in the predicted section are as follows: Figure 3 shown. Figure 3 In the figure, the vertical axis represents the monitored value of runoff, and the horizontal axis represents the number of each month in each historical year in the historical period.

[0065] Step 23: Perform Gaussian fitting based on the adjusted values ​​of runoff in all months of each historical year of the section to be predicted, and obtain fitting values ​​of multiple Gaussian parameters of each historical year of the section to be predicted.

[0066] Specifically, the Gaussian parameter is a characteristic parameter in the Gaussian function, and step 23 includes the following content.

[0067] The function between the set month and the adjustment value of runoff volume is expressed as:

[0068] .

[0069] in, For the Historical years month; For the The mapping relationship between historical years.

[0070] The inverse function is:

[0071] .

[0072] Furthermore, the formula for Gaussian fitting is as follows.

[0073] .

[0074] .

[0075] .

[0076] .

[0077] in, For the The fitted value of the first Gaussian parameter for each historical year; For the The maximum value of the adjusted runoff volume for all months in a historical year; For the The minimum value of the adjusted runoff volume for all months in a historical year; For the The fitted value of the second Gaussian parameter for each historical year; for The mean of For the The runoff distribution in each historical year, The determination process includes: and Divide the interval of the endpoints evenly to obtain the runoff after 10,000 divisions. Substitute the runoff after 10,000 divisions into the inverse function to obtain ; for The standard deviation of For the The fitted value of the third Gaussian parameter for each historical year; For the The fitted value of the fourth Gaussian parameter for each historical year.

[0078] Step 3: Input the fitting values ​​of each Gaussian parameter of all historical years of the section to be predicted into the Gaussian parameter prediction model of the corresponding Gaussian parameter to obtain the predicted values ​​of each Gaussian parameter of the section to be predicted in future years.

[0079] Among them, each Gaussian parameter prediction model is obtained by training the long short-term memory network; the future year is the next year of the historical period.

[0080] Specifically, for each Gaussian parameter, Long Short-Term Memory (LSTM) is used to construct Gaussian parameter prediction models with different Gaussian parameters.

[0081] As an optional implementation, in step 3, the process of determining the Gaussian parameter prediction model of any current Gaussian parameter includes steps 31 to 33.

[0082] Step 31: Obtain the fitted values ​​of the current Gaussian parameters of the first sample section for N+1 consecutive historical years; N≥1.

[0083] As an optional implementation, step 31 includes steps 311 and 312.

[0084] Step 311: Obtain monitoring values ​​of runoff volume of the first sample section in each month of N+1 consecutive historical years.

[0085] Step 312: Determine the fitting values ​​of the current Gaussian parameters of the first sample section in the N+1 historical years based on the monitored values ​​of the runoff in each month in the N+1 historical years of the first sample section.

[0086] As an optional implementation, step 312 includes steps 3121 to 3123.

[0087] Step 3121: Determine the average value of the runoff volume of each month of the first sample section based on the monitored values ​​of the runoff volume of each month in all historical years of the first sample section.

[0088] Step 3122: Based on the average value of the runoff of each month of the first sample section, the monitoring value of the runoff of each month in the N+1 historical years of the first sample section is adjusted to obtain the adjusted value of the runoff of each month in each historical year of the first sample section.

[0089] Step 3123: Perform Gaussian fitting based on the adjusted values ​​of runoff in all months in the N+1 historical years of the first sample section to obtain the fitting values ​​of the current Gaussian parameters of the N+1 historical years of the first sample section.

[0090] Step 32: Build a long short-term memory network.

[0091] Step 33: Using the fitting values ​​of the current Gaussian parameters of the first N historical years of the first sample section as input and the fitting values ​​of the current Gaussian parameters of the N+1th historical year of the first sample section as output, the long short-term memory network is trained to obtain a Gaussian parameter prediction model for the current Gaussian parameters.

[0092] Specifically, with the first Gaussian parameter For example, the long short-term memory network constructed is as follows Figure 4 As shown, the input is within the selected window length N The first layer is a fully connected layer that takes the input (i.e., the fitted value of the first Gaussian parameter of the first historical year) , the fitted value of the first Gaussian parameter in the second historical year , ..., the fitted value of the first Gaussian parameter in the Nth historical year ) is converted into a specified number of neurons, and the output of each neuron must pass through the activation function Relu ( Figure 4 The second layer is an LSTM layer, the number of which is the same as the first layer. Each LSTM module is Figure 4 The third layer is the same as the first layer, which is used to convert the output into a single value (i.e., the predicted value of the first Gaussian parameter in the N+1th historical year). ), calculated by the loss function and the fitting value of the first Gaussian parameter in the N+1th historical year The error between the two is updated using the Adam optimizer. The Gaussian parameter prediction model is fitted successfully by adjusting the parameters to ensure that the LSTM network performs well on both the training and test sets. These parameters include the window size, the number of neurons in each layer, the number of LSTM layers, the learning rate, and the number of training cycles.

[0093] Step 4: Based on the predicted values ​​of the Gaussian parameters of the future years of the section to be predicted, determine the Gaussian function values ​​of each month in the future years of the section to be predicted.

[0094] Specifically, the Gaussian function value is calculated by the Gaussian function, which is:

[0095] .

[0096] in, For the Years Gaussian function value of the month; For the The first Gaussian parameter of the year; is a natural constant; For the the second Gaussian parameter of the year; For the The third Gaussian parameter of the year; For the The fourth Gaussian parameter for the year.

[0097] Step 5: Input the Gaussian function value of each month in the future years of the section to be predicted into the runoff prediction model to obtain the predicted value of the runoff of each month in the future years of the section to be predicted.

[0098] Among them, the runoff prediction model is obtained by training the back propagation neural network.

[0099] As an optional implementation, in step 5, the process of determining the runoff prediction model includes steps 51 to 53.

[0100] Step 51: Obtain the Gaussian function value of each month in the historical years of the second sample section and the monitoring value of the runoff volume of each month in the historical years of the second sample section.

[0101] As an optional implementation, step 51 includes steps 511 to 513.

[0102] Step 511: Obtain the monitoring value of runoff volume of each month in the historical years of the second sample section.

[0103] Step 512: Based on the monitored values ​​of the runoff volume of each month in the historical years of the second sample section, the fitting values ​​of multiple Gaussian parameters of the historical years of the second sample section are determined.

[0104] Step 513: Based on the fitting values ​​of multiple Gaussian parameters of the historical years of the second sample section, determine the Gaussian function value of each month in the historical years of the second sample section.

[0105] Step 52: Construct a back-propagation neural network.

[0106] Step 53: The back propagation neural network is trained using the Gaussian function values ​​of each month in the historical years of the second sample section and the monitored values ​​of the runoff of each month in the historical years of the second sample section to obtain a runoff prediction model.

[0107] As an optional implementation, step 53 includes steps 531 and 532.

[0108] Step 53: Normalize the Gaussian function values ​​of each month in the historical years of the second sample section and the monitored values ​​of the runoff of each month in the historical years of the second sample section respectively to obtain the normalized Gaussian function values ​​of each month in the historical years of the second sample section and the normalized values ​​of the runoff of each month in the historical years of the second sample section.

[0109] Specifically, the calculation formula for normalization processing includes:

[0110] .

[0111] in, For the Years Normalized Gaussian function value of the month; For the All the years The mean of the Gaussian function values ​​for the month; For the All the years The standard deviation of the Gaussian function values ​​for the month; For the Years Normalized value of monthly runoff; For the Years Monitoring value of monthly runoff; For the The minimum value of the monitored runoff volume in all months of a year; For the The maximum value of the monitored runoff values ​​for all months in a year.

[0112] Step 53: Using the normalized Gaussian function values ​​of each month in the historical years of the second sample section as input and the normalized runoff values ​​of each month in the historical years of the second sample section as output, the back propagation neural network is trained to obtain a runoff prediction model.

[0113] Specifically, the network structure of the backpropagation (BP) neural network is as follows: Figure 5 As shown, during training, the normalized Gaussian function values ​​of each month in the historical year of the second sample section (expressed as , including the normalized Gaussian function value in January of the historical year of the second sample section , the normalized Gaussian function value in February of the historical year of the second sample section ,…, the normalized Gaussian function value of December in the historical year of the second sample section ) is input to the input layer, and activation functions are set in n hidden layers .

[0114] Output after hidden layer 1 , .

[0115] Output after hidden layer 2 , .

[0116] After hidden layer n output , .

[0117] Output through the output layer , .

[0118] in, is the output matrix of hidden layer 1; is the weight of hidden layer 1; is the bias of hidden layer 1; is the output of the first unit of hidden layer 1; is the output of the second unit of hidden layer 1; is the output of the third unit of hidden layer 1; is the first hidden layer The output of each unit; is the output matrix of hidden layer 2; is the weight of hidden layer 2; is the bias of hidden layer 2; is the output of the first unit of hidden layer 2; is the output of the second unit of hidden layer 2; is the output of the third unit of hidden layer 2; is the hidden layer 2 The output of each unit; is the output matrix of hidden layer n; is the weight of hidden layer n; is the output matrix of hidden layer n-1; is the bias of hidden layer n; is the output of the first unit of hidden layer n; is the output of the second unit of the hidden layer n; is the output of the third unit of hidden layer n; is the hidden layer n The output of each unit; The output matrix of the output layer is the matrix composed of the predicted values ​​of runoff in all months of the historical years of the second sample section; is the weight of the output layer; is the bias of the output layer; is the predicted value of runoff in January in the historical years of the second sample section; is the predicted value of runoff in February in the historical year of the second sample section; is the predicted value of runoff in December of the historical year for the second sample section.

[0119] The loss function is mean square error loss , expressed as:

[0120] .

[0121] in, for Normalized value of monthly runoff; for The predicted value of runoff for the month.

[0122] Furthermore, during a certain verification, this application collected data from 84 monitoring sections in a certain river basin. This paper randomly selected data from one monitoring section for testing and predicted the runoff for the next three years to evaluate the prediction effect of the method of this application. At the same time, a gated recurrent unit (GRU) network was used for comparison with this application.

[0123] After debugging, the main parameters involved in the method of this application are shown in Table 1 and Table 2.

[0124] Table 1 LSTM parameters

[0125]

[0126] Table 2 BP parameter table

[0127]

[0128] In the GRU network for comparison, the window size is set to 100, the number of gate units is 1000, followed by 3 layers of fully connected layers with 2000 neurons, the learning rate is set to 0.01, and the number of training times is 150.

[0129] The verification results of the method of this application and the GRU network are shown in Table 3.

[0130] Table 3 Results

[0131]

[0132] The indicators in Table 3 are commonly used indicators for evaluating the fitting effect of neural networks. MSE (Mean Squared Error) is the mean square error, RMSE (Root Mean Squared Error) is the root mean square error, MAE (Mean Absolute Error) is the mean absolute error, ME (Mean Error) is the mean error, R 2The MAE of the test set in the table can be used as a direct indicator to judge the performance of the two algorithms. The smaller the MAE value, the smaller the prediction error of the algorithm on the test set. This shows that the prediction accuracy of this application is higher.

[0133] In an exemplary embodiment, a computer device is provided, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a runoff prediction method based on a Gaussian function.

[0134] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, a runoff prediction method based on a Gaussian function is implemented.

[0135] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a runoff prediction method based on a Gaussian function is implemented.

[0136] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0137] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0138] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0139] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0140] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0141] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A runoff prediction method based on Gaussian function, characterized in that: The runoff prediction method based on Gaussian function includes: Obtain the monitoring values ​​of runoff in each month of multiple historical years in the historical period of the section to be predicted; Determining the fitting values ​​of multiple Gaussian parameters of each historical year of the section to be predicted based on the monitored values ​​of runoff in each month of each historical year of the section to be predicted; The fitted values ​​of each Gaussian parameter of all historical years of the section to be predicted are respectively input into the Gaussian parameter prediction model of the corresponding Gaussian parameter to obtain the predicted values ​​of each Gaussian parameter of the future year of the section to be predicted; each Gaussian parameter prediction model is obtained by training the long short-term memory network; the future year is the next year of the historical period; Determine the Gaussian function value of each month in the future year of the section to be predicted based on the predicted value of each Gaussian parameter in the future year of the section to be predicted; Inputting the Gaussian function value of each month in the future year of the section to be predicted into the runoff prediction model to obtain the predicted value of the runoff of each month in the future year of the section to be predicted; the runoff prediction model is obtained by training the back propagation neural network; Gaussian parameters are characteristic parameters in the Gaussian function; The formula for Gaussian fitting includes: ； ； ； ； in, For the The fitted value of the first Gaussian parameter for each historical year; For the The maximum value of the adjusted runoff volume for all months in a historical year; For the The minimum value of the adjusted runoff volume for all months in a historical year; For the The fitted value of the second Gaussian parameter for each historical year; for The mean of For the Runoff distribution in each historical year; The Gaussian function value is calculated by the Gaussian function, which is: ； in, For the Years Gaussian function value of the month; For the The first Gaussian parameter of the year; is a natural constant; For the the second Gaussian parameter of the year; For the The third Gaussian parameter of the year; For the The fourth Gaussian parameter for the year.

2. The runoff prediction method based on Gaussian function according to claim 1, characterized in that: Based on the monitored values ​​of runoff volume in each month of each historical year of the section to be predicted, the fitting values ​​of multiple Gaussian parameters of each historical year of the section to be predicted are determined, including: Determine the average value of the runoff volume of each month of the section to be predicted based on the monitored values ​​of the runoff volume of each month in all historical years of the section to be predicted; Based on the average value of the runoff volume of each month of the section to be predicted, the monitored values ​​of the runoff volume of each month in each historical year of the section to be predicted are adjusted to obtain the adjusted value of the runoff volume of each month in each historical year of the section to be predicted; Gaussian fitting is performed based on the adjusted values ​​of runoff in all months of each historical year of the section to be predicted, so as to obtain fitting values ​​of multiple Gaussian parameters of each historical year of the section to be predicted.

3. The runoff prediction method based on Gaussian function according to claim 1, characterized in that: The process of determining the Gaussian parameter prediction model for any current Gaussian parameter includes: Obtain the fitted values ​​of the current Gaussian parameters of the first sample section for N+1 consecutive historical years; N≥1; Building long short-term memory networks; The long short-term memory network is trained with the fitting values ​​of the current Gaussian parameters of the first N historical years of the first sample section as input and the fitting values ​​of the current Gaussian parameters of the N+1th historical year of the first sample section as output to obtain a Gaussian parameter prediction model for the current Gaussian parameters.

4. The runoff prediction method based on Gaussian function according to claim 3 is characterized in that: Get the fitted values ​​of the current Gaussian parameters for N+1 consecutive historical years of the first sample section, including: Obtain the monitoring values ​​of runoff volume for each month in N+1 consecutive historical years at the first sample section; Based on the monitored values ​​of the runoff in each month of the N+1 historical years of the first sample section, the fitting values ​​of the current Gaussian parameters of the first sample section in the N+1 historical years are determined.

5. The runoff prediction method based on Gaussian function according to claim 4, characterized in that: Based on the monitored values ​​of runoff in each month of the N+1 historical years of the first sample section, the fitting values ​​of the current Gaussian parameters of the first sample section in the N+1 historical years are determined, including: Determine the average value of the runoff volume of each month of the first sample section based on the monitored values ​​of the runoff volume of each month in all historical years of the first sample section; Based on the average value of the runoff volume of each month of the first sample section, the monitored values ​​of the runoff volume of each month in N+1 historical years of the first sample section are adjusted to obtain the adjusted value of the runoff volume of each month in each historical year of the first sample section; Gaussian fitting is performed based on the adjusted values ​​of runoff in all months in the N+1 historical years of the first sample section to obtain the fitting values ​​of the current Gaussian parameters of the N+1 historical years of the first sample section.

6. The runoff prediction method based on Gaussian function according to claim 1, characterized in that: The process of determining the runoff prediction model includes: Obtain the Gaussian function value of each month in the historical years of the second sample section and the monitoring value of the runoff volume of each month in the historical years of the second sample section; Build a back-propagation neural network; The back propagation neural network is trained using the Gaussian function values ​​of each month in the historical years of the second sample section and the monitored values ​​of the runoff of each month in the historical years of the second sample section to obtain the runoff prediction model.

7. The runoff prediction method based on Gaussian function according to claim 6, characterized in that: Obtain the Gaussian function value of each month in the historical year of the second sample section and the monitoring value of the runoff volume of each month in the historical year of the second sample section, including: Obtain the monitoring values ​​of runoff volume for each month in the historical years of the second sample section; Determine fitting values ​​of multiple Gaussian parameters of the second sample section in the historical years based on the monitored values ​​of runoff in each month of the historical years of the second sample section; Based on the fitting values ​​of multiple Gaussian parameters of the historical years of the second sample section, the Gaussian function value of each month in the historical years of the second sample section is determined.

8. The runoff prediction method based on Gaussian function according to claim 6, characterized in that: The back propagation neural network is trained using the Gaussian function value of each month in the historical years of the second sample section and the monitored value of the runoff volume of each month in the historical years of the second sample section to obtain the runoff volume prediction model, including: Normalizing the Gaussian function values ​​of each month in the historical years of the second sample section and the monitored values ​​of the runoff of each month in the historical years of the second sample section respectively to obtain the normalized Gaussian function values ​​of each month in the historical years of the second sample section and the normalized values ​​of the runoff of each month in the historical years of the second sample section; The back propagation neural network is trained with the normalized Gaussian function value of each month in the historical years of the second sample section as input and the normalized value of the runoff of each month in the historical years of the second sample section as output to obtain the runoff prediction model.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the runoff prediction method based on the Gaussian function according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the runoff prediction method based on the Gaussian function described in any one of claims 1 to 8 is implemented.

Images