Hydrological data prediction method, device, equipment and program product

Abstract

The invention relates to the field of hydrological data prediction, in particular to a hydrological data prediction method, device, equipment and program product. The method comprises the following steps: acquiring a latest preset quantity of hydrological historical observation data; the hydrological historical observation data are input into a pre-trained dynamic hidden variable model based on the Gaussian process, hydrological prediction data corresponding to the hydrological historical observation data are obtained, and the dynamic hidden variable model based on the Gaussian process comprises an observation model and a dynamic hidden variable model; the observation model is used for determining hydrological historical real data corresponding to the hydrological historical observation data, and the dynamic hidden variable model is used for determining hydrological prediction data corresponding to the hydrological historical real data. According to the method, resource consumption can be reduced through small sample high efficiency of a Gaussian process and simplification of a dynamic hidden variable model, large-scale iteration is replaced by analysis optimization, and resource-limited hydrological monitoring equipment can be effectively adapted while the prediction precision is ensured.

Description

Technical Field

[0001] The present application relates to the field of hydrological data prediction, and in particular to hydrological data prediction methods, devices, equipment and program products. Background Art

[0002] Hydrological forecasting refers to the use of the principles and methods of hydrology, meteorology, and hydraulics to make qualitative or quantitative predictions of the hydrological conditions of a water body, a region, or a hydrological station in the future based on current hydrological and meteorological data. It includes the prediction of hydrological elements such as rainfall, runoff, water level, and flow.

[0003] At present, the mainstream method for predicting discrete data is usually based on the design of neural network architecture, such as LSTM (Long Short-Term Memory) and other solutions. This method trains the neural network based on a certain amount of sample data to form a model for predicting future data. However, since neural networks require a large computational overhead and a large training sample set, they are generally used on hardware platforms with strong computing power, such as personal PCs or mobile terminals. Since hydrological monitoring equipment is more sensitive to power consumption, its computing power is often very limited and is only suitable for limited-scale network reasoning. For the reasoning and training of large-scale neural networks, hydrological detection equipment cannot provide sufficient hardware resources and power consumption. There are major obstacles to the use of traditional neural networks in hydrological detection equipment. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a hydrological data prediction method, device, equipment and program product to solve the problem that hydrological detection equipment in the prior art cannot provide sufficient hardware resources and power consumption, and there are major obstacles in using traditional neural networks in hydrological detection equipment.

[0005] A first aspect of an embodiment of the present application provides a hydrological data prediction method, the method comprising:

[0006] Obtain the latest predetermined amount of hydrological historical observation data;

[0007] The hydrological historical observation data are input into a pre-trained dynamic latent variable model based on Gaussian process to obtain hydrological prediction data corresponding to the hydrological historical observation data. The dynamic latent variable model based on Gaussian process includes an observation model and a dynamic latent variable model. The observation model is used to determine the hydrological historical real data corresponding to the hydrological historical observation data. The dynamic latent variable model is used to determine the hydrological prediction data corresponding to the hydrological historical real data according to the nonlinear dynamic characteristics of the hydrological historical real data.

[0008] In combination with the first aspect, in a first possible implementation manner of the first aspect, the Gaussian process-based dynamic latent variable model further includes a joint probability model;

[0009] Before inputting the hydrological historical observation data into the pre-trained dynamic latent variable model based on Gaussian process, the method further includes:

[0010] Initializing first hydrological historical real data corresponding to the hydrological historical observation data;

[0011] According to the first real hydrological historical data, the hyperparameters in the joint probability model are iteratively optimized by the maximum a posteriori method until the hyperparameters meet the preset requirements, thereby completing the training of the dynamic latent variable model based on the Gaussian process.

[0012] In combination with the first possible implementation manner of the first aspect, in a second possible implementation manner of the first aspect, iteratively optimizing the hyperparameters in the joint probability model by a maximum a posteriori method according to the first hydrological historical real data includes:

[0013] Determining weight matrix parameters of the joint probability model analytically using maximum a posteriori estimation;

[0014] The first covariance matrix parameters and the second covariance matrix parameters in the joint probability model are optimized using a scaled conjugate gradient method.

[0015] In combination with the first possible implementation manner of the first aspect, in a third possible implementation manner of the first aspect, the joint probability model includes:

[0016] Given a first covariance matrix parameter, a weight matrix and real hydrological historical data, a first probability function model for the occurrence of the hydrological historical observation data;

[0017] A second probability function model for the occurrence of the hydrological historical real data when the second covariance matrix parameters are given;

[0018] A third probability function model in which the first covariance matrix parameters appear;

[0019] A fourth probability function model in which the second covariance matrix parameters appear;

[0020] A fifth probability function model appears in the weight matrix.

[0021] In combination with the third possible implementation of the first aspect, in a fourth possible implementation of the first aspect, the first probability function value in the first probability function model is determined based on a weight matrix, hydrological historical observation data, a dimension of hydrological historical observation data, the number of hydrological historical observation data, and a first covariance matrix.

[0022] In combination with the third possible implementation of the first aspect, in the fifth possible implementation of the first aspect, the second probability function value in the second probability function model is determined based on the number of historical hydrological observation data, the Markov term used to describe the dynamic characteristics of the hydrological prediction data, and the fourth probability function model.

[0023] In combination with any one of the first aspect to the fifth possible implementation manner of the first aspect, in the sixth possible implementation manner of the first aspect, the observation model determines the hydrological observation data based on the system observation noise and the system observation function, the system observation function is determined based on the weight matrix and the dynamic latent variable model, and the weight matrix is ​​determined based on the variance of the weight matrix probability distribution, the dimension of the hydrological historical observation data, and the elements on the diagonal of the weight matrix.

[0024] A second aspect of an embodiment of the present application provides a hydrological data prediction device, the device comprising:

[0025] A data acquisition unit, used to acquire the latest predetermined amount of hydrological historical observation data;

[0026] A hydrological prediction data determination unit is used to input the hydrological historical observation data into a pre-trained dynamic latent variable model based on a Gaussian process to obtain hydrological prediction data corresponding to the hydrological historical observation data. The dynamic latent variable model based on a Gaussian process includes an observation model and a dynamic latent variable model. The observation model is used to determine the hydrological historical real data corresponding to the hydrological historical observation data. The dynamic latent variable model is used to determine the hydrological prediction data corresponding to the hydrological historical real data based on the nonlinear dynamic characteristics of the hydrological historical real data.

[0027] A third aspect of an embodiment of the present application provides a hydrological data prediction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the hydrological data prediction device implements a method as described in any one of the first aspects.

[0028] A fourth aspect of the embodiments of the present application provides a computer program product, which, when executed on a computer, enables the computer to execute the method in the first aspect or its various implementations.

[0029] A fifth aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of the first aspects are implemented.

[0030] The sixth aspect of the embodiment of the present application provides a chip for implementing the methods in each implementation of the first aspect. Specifically, the chip includes: a processor for calling and running a computer program from a memory, so that a device equipped with the chip executes the method in the first aspect or its implementation.

[0031] Compared with the prior art, the embodiments of the present application have the following beneficial effects: the embodiments of the present application input the latest predetermined number of hydrological historical observation data obtained into a pre-trained dynamic latent variable model based on a Gaussian process, determine the hydrological historical real data corresponding to the hydrological historical observation data through the observation model, and determine the hydrological prediction data corresponding to the hydrological historical real data according to the nonlinear dynamic characteristics of the hydrological historical real data through the dynamic latent variable model. Compared with the neural network model, the embodiments of the present application are beneficial to reducing resource consumption through the small sample efficiency of the Gaussian process and the simplification of the dynamic latent variable model, and can effectively adapt to resource-constrained hydrological monitoring equipment while ensuring prediction accuracy by replacing large-scale iterations with analytical optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 paying any creative labor.

[0033] Figure 1 It is a schematic diagram of the implementation process of a hydrological data prediction method provided in an embodiment of the present application;

[0034] Figure 2 It is a schematic diagram of a framework of a hydrological data prediction method provided in an embodiment of the present application;

[0035] Figure 3 It is a schematic diagram comparing the prediction results of the hydrological data prediction method provided in the embodiment of the present application and the Kalman filtering method;

[0036] Figure 4 is a schematic diagram comparing prediction results of another hydrological data prediction method provided in an embodiment of the present application and a Kalman filter method;

[0037] Figure 5 This is a schematic diagram of the effect of correcting abnormal data provided by an embodiment of the present application;

[0038] Figure 6 is a schematic diagram of a hydrological data prediction device provided in an embodiment of the present application;

[0039] Figure 7It is a schematic diagram of a hydrological data prediction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0040] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0041] In order to illustrate the technical solution described in this application, a specific embodiment is provided below for illustration.

[0042] Hydrological forecasting is to use current hydrological and meteorological data, combined with the principles and methods of hydrology, meteorology and hydraulics, to make a qualitative or quantitative prediction of the future hydrological conditions of a certain water body, region or hydrological station. The forecast objects include hydrological elements such as rainfall, runoff, water level and flow.

[0043] At present, the mainstream methods for discrete data prediction are mostly based on neural network architectures, such as LSTM (Long Short-Term Memory Network). These methods train the neural network through a certain amount of sample data to build a model for predicting future data. However, neural networks require a large computational overhead and a large number of training samples, and are usually suitable for hardware platforms with strong computing power, such as personal PCs or mobile terminals. Hydrological monitoring equipment is sensitive to power consumption and has limited computing power. It can only perform network reasoning on a limited scale, which makes it difficult to meet the reasoning and training requirements of large-scale neural networks, and is not conducive to its effective application in hydrological detection equipment.

[0044] Secondly, traditional prediction and tracking methods include prediction and tracking methods based on Kalman filters. This method has been relatively maturely developed and verified in applications such as automotive radar and aviation radar. However, since the Kalman filter itself is designed based on the linear dynamic assumption, it has a strong tracking ability for linear motion in a short period of time. However, the changes in hydrological information (water level, flow) data generally have nonlinear characteristics. Therefore, the prediction and tracking method based on the Kalman filter cannot be effectively reflected in the application of hydrological information prediction.

[0045] To solve the above problems, the present invention proposes a hydrological data prediction method. Figure 1 The following is a schematic diagram of the implementation process of this method, which is described in detail as follows:

[0046] In S101, a predetermined amount of the latest historical hydrological observation data is obtained.

[0047] The hydrological historical observation data obtained in the embodiment of the present application can be used to input into the dynamic hidden variable model based on Gaussian process. The hydrological historical observation data used to input into the dynamic hidden variable model based on Gaussian process can be set to a predetermined number, for example, it can be set to no more than 10 hydrological historical observation data.

[0048] The latest predetermined number of hydrological historical observation data may be the latest predetermined number of hydrological historical observation data before the current time. The hydrological historical observation data may include flow data, water level data, and the like.

[0049] Before inputting the hydrological historical observation data into the pre-trained dynamic latent variable model based on Gaussian process, the embodiment of the present application can also pre-construct a dynamic latent variable model based on Gaussian process, and train it with a small amount of sample data, such as less than 1,000 sets of sample data, to obtain a trained dynamic latent variable model based on Gaussian process.

[0050] According to the design goal, assuming t n To observe hydrological data (including flow, water level and other hydrological data reported by hydrological detection equipment), x n For real hydrological data (including real flow, real water level and other hydrological data in the target area), the target problem is designed as an estimation problem of deducing real hydrological data based on observed hydrological data (i.e., inferring the real water level and flow based on the water level and flow measured by hydrological detection equipment, i.e., using t n Inference: x n , and predict the next set of t n and x n ), and through Figure 2 The factor graph shown builds the overall framework.

[0051] like Figure 2 As shown, black circles represent fixed value variables, white circles represent random variables, and boxes represent functions or probability distributions. In the figure, A and B are system hyperparameters. To provide sufficient model flexibility, the embodiment of the present application models hyperparameters A and B as random variables that obey beta distribution. Figure 2 As shown, the system observation model can be expressed as:

[0052]

[0053] where t n is the observed hydrological data (i.e., the hydrological data reported by the hydrological detection equipment, i.e., the observed hydrological data). Each set of hydrological observation data can be multidimensional data (multiple physical quantities) or scalar data (a single physical quantity). In formula (1), D represents the dimension of each set of hydrological observation data, i.e., the total number of physical quantities of each set of hydrological observation data. nis the current real hydrological data, which can also be multidimensional data or scalar. n is the system observation noise, and the system observation noise has a variance of β -1 independent Gaussian distribution. f(x n ; W) is the system observation function. For the convenience of subsequent writing, let The simplest observation function can be expressed as That is, the observed hydrological data is the same as the real hydrological data. In order to describe the target data, that is, the physical characteristics of the hydrological information, the observation function is given as follows:

[0054]

[0055] That is, the current hydrological observation data is affected by the historical hydrological observation data. n is Gaussian noise, W is the weight matrix, which can be given as:

[0056]

[0057] Among them, w d Represents the dth element on the diagonal of the weight matrix. κ represents the variance of the probability distribution of the weight matrix, which is a system hyperparameter. Since the algorithm design of the embodiment of the present application is based on the Gaussian process framework, the likelihood probability of the system, that is, the first probability function model can be expressed as:

[0058]

[0059] Where Tr(·) represents the sum of diagonal elements in matrix operation. Where T = [t1,…,t N ] T is the observed data set, B=[b0,b1,b2] T is the parameter of the first covariance matrix, K T is the system covariance, which is related to the target prediction quantity X. Since the true probability distribution of X is not of concern here, K can be directly T The design can describe the physical characteristics of X. Here, the RBF (Radial Basis Function) covariance matrix commonly used in Gaussian process can be used, that is, the third covariance matrix. The jth element in the i-th row of the third covariance matrix can be expressed as:

[0060]

[0061] where δ(·) is the Dirac function.

[0062] In the traditional Gaussian process, the target prediction amount X is treated as a deterministic amount. Due to the nonlinear dynamic characteristics of the hydrological data in the embodiment of the present application, the target prediction amount X is treated as a random variable and modeled. In order to describe the nonlinear dynamic characteristics, we model the target prediction amount X as follows:

[0063]

[0064] The probability model is constructed by constructing the Markov term (p(x n |x n-1 ,A)) describes the dynamic characteristics of the target prediction quantity X. Where A=[a0,…a M ] T is the parameter of the second covariance matrix, K X is the covariance of the probability model of the target prediction quantity X. The nonlinear characteristic description of the target prediction quantity X can be reflected by the design of the second covariance matrix. Assuming that the target prediction quantity X is a first-order nonlinear characteristic, the corresponding second covariance matrix form can be directly constructed by a polynomial. The jth element of the i-th row in the second covariance matrix can be expressed as:

[0065]

[0066] Among them, the first term in formula (7) is used to describe nonlinear characteristics, the second term is used to describe linear characteristics, and the third term is used to describe fixed deviations. According to actual usage requirements, the order of this second covariance matrix can be adjusted arbitrarily, and the jth element of the i-th row in the second-order covariance matrix can be expressed as:

[0067]

[0068] According to the above formula, the dynamic latent variable model based on Gaussian process can be determined. The core probability model of the model includes the first probability function model, the second probability function model, the third probability function model, the fourth probability function model and the fifth probability function model, which are respectively expressed as:

[0069]

[0070]

[0071] Among them, A, B, and W are all system hyperparameters. Based on the above model, the system joint probability model can be expressed as:

[0072] p(T,X,A,B,W)=p(T|X,B,W)p(X|A)p(A)p(B)p(W) (14)

[0073] This joint probability model indicates that there is a specific relationship between the statistical information of observed physical quantities (such as observed hydrological data) and the statistical information of real physical quantities (such as real hydrological data), and there is a correlation between the real physical quantities that are continuous multiple times before and after. After obtaining the joint probability of the system, the parameters {A, B, W} in the model can be automatically learned through maximum a posteriori training. Specifically, W can be directly obtained through analytical solutions, such as determining the weight matrix parameters through maximum a posteriori estimation. For example, the weight matrix W can be derived according to the likelihood function of the observation model (Formula 4), combined with the prior of the weight matrix (Formula 13), and the logarithm of the posterior distribution, to obtain a closed-form solution for W. Due to the combination of the linear observation model and the Gaussian prior, the posterior distribution is still a Gaussian distribution, and its mean can be directly calculated analytically, avoiding iterative optimization and significantly reducing the demand for computing resources.

[0074] {A, B} can be determined using the SCG (Scaled Conjugate Gradient) optimization method. The gradient of the log joint probability with respect to A and B is calculated, and the parameters are updated along the conjugate gradient direction until convergence. The convergence speed is faster than traditional gradient descent, making this method memory-efficient and suitable for edge devices.

[0075] In S102, the hydrological historical observation data is input into a pre-trained dynamic latent variable model based on Gaussian process to obtain hydrological prediction data corresponding to the hydrological historical observation data. The dynamic latent variable model based on Gaussian process includes an observation model and a dynamic latent variable model. The observation model is used to determine the hydrological historical real data corresponding to the hydrological historical observation data. The dynamic latent variable model is used to determine the hydrological prediction data corresponding to the hydrological historical real data according to the nonlinear dynamic characteristics of the hydrological historical real data.

[0076] The hydrological prediction data may be actual hydrological prediction data or observed hydrological prediction data.

[0077] After the model training is completed, the prediction method using the trained dynamic latent variable model based on Gaussian process can be expressed as follows:

[0078] Given a new hydrological data point to be predicted {t * ,x *}, its conditional probability is

[0079] p(t * ,x * |T,X,A,B,W)∝p(t * |T,X,x * ,B,W)p(X,x * |A)(15)

[0080] Assuming that the hydrological data of the predicted point needs to rely on the past K valid historical data, then:

[0081]

[0082] Where K T* is the covariance matrix of the data point to be predicted, μ t* That is, the observed quantity of hydrological data to be predicted t * The predicted value of . Similarly, we can get:

[0083]

[0084] Where K X* is the covariance matrix of the data point to be predicted, μ X* That is the true value x of the hydrological information to be predicted * The predicted value of can be calculated using the following formula:

[0085] μ X* = argmaxp(x * |T,X,A,B,W)∝argmin∫p(x * |T,X,A,B,W)dx * (18)

[0086] First, we need to define the posterior probability and find the most likely x by maximizing the posterior probability. ＊ value, use numerical optimization methods (such as gradient descent, Newton's method, etc.) to find the x that maximizes the posterior probability ＊ value.

[0087] The embodiment of the present application can construct a white box network to perform water level / flow prediction through covariance matrix design and probability model description. Compared with the black box system of neural network, it has better portability and maintainability.

[0088] And in the training stage, due to the combination of the linear observation model and the Gaussian prior, the posterior distribution is still a Gaussian distribution, and its mean can be directly calculated analytically, avoiding iterative optimization and significantly reducing the demand for computing resources. {A, B} can be determined using the SCG (scaled conjugate gradient) optimization method. The gradient of the logarithmic joint probability with respect to A and B is calculated, and the parameters are updated along the conjugate gradient direction until convergence. The convergence speed is faster than traditional gradient descent, making this method highly memory-efficient and suitable for edge devices.

[0089] like Figure 3 As shown, the hydrological data prediction method in the embodiment of the present application doubles the prediction accuracy of the water level value compared with the Kalman filtering method.

[0090] like Figure 4As shown, the hydrological data prediction method in the embodiment of the present application has a prediction accuracy more than twice as high as that of the Kalman filtering method, and the normalized mean square error is one thousandth, which meets the actual application requirements.

[0091] And if Figure 5 As shown, in the hydrological data prediction method in the embodiment of the present application, the training sample library used includes 2000 sample data with serial numbers 1-2000. During training, sample data can be obtained by equally spaced extraction. When prediction is performed after training is completed, when the sensor device is abnormal, the method can effectively correct the abnormal data, and the prediction value used to correct the abnormal data is a valid result.

[0092] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0093] Figure 6 A schematic diagram of a hydrological data prediction device provided in an embodiment of the present application, the device comprising:

[0094] The data acquisition unit 601 is used to acquire the latest predetermined amount of hydrological historical observation data;

[0095] The hydrological prediction data determination unit 602 is used to input the hydrological historical observation data into a pre-trained dynamic latent variable model based on Gaussian process to obtain hydrological prediction data corresponding to the hydrological historical observation data. The dynamic latent variable model based on Gaussian process includes an observation model and a dynamic latent variable model. The observation model is used to determine the hydrological historical real data corresponding to the hydrological historical observation data. The dynamic latent variable model is used to determine the hydrological prediction data corresponding to the hydrological historical real data based on the nonlinear dynamic characteristics of the hydrological historical real data.

[0096] Figure 6 The hydrological data prediction device shown in FIG. Figure 1 The hydrological data prediction method shown corresponds to the above.

[0097] Figure 7 Schematic diagram of a hydrological data prediction device provided in an embodiment of the present application. Figure 7 As shown, the hydrological data prediction device 7 of this embodiment includes: a processor 70, a memory 71, and a computer program 72 stored in the memory 71 and executable on the processor 70, such as a hydrological data prediction program. When the processor 70 executes the computer program 72, the steps in the above-mentioned various hydrological data prediction method embodiments are implemented. Alternatively, when the processor 70 executes the computer program 72, the functions of each module / unit in the above-mentioned various device embodiments are implemented.

[0098] Exemplarily, the computer program 72 may be divided into one or more modules / units, which are stored in the memory 71 and executed by the processor 70 to complete the present application. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 72 in the hydrological data prediction device 7.

[0099] The hydrological data prediction device 7 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The hydrological data prediction device may include, but is not limited to, a processor 70 and a memory 71. Those skilled in the art will appreciate that Figure 7 It is only an example of the hydrological data prediction device 7 and does not constitute a limitation of the hydrological data prediction device 7. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the hydrological data prediction device may also include input and output devices, network access devices, buses, etc.

[0100] The processor 70 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0101] The memory 71 may be an internal storage unit of the hydrological data prediction device 7, such as a hard disk or memory of the hydrological data prediction device 7. The memory 71 may also be an external storage device of the hydrological data prediction device 7, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the hydrological data prediction device 7. Further, the memory 71 may also include both an internal storage unit and an external storage device of the hydrological data prediction device 7. The memory 71 is used to store the computer program and other programs and data required by the hydrological data prediction device. The memory 71 may also be used to temporarily store data that has been output or is to be output.

[0102] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0103] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0104] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0105] In the embodiments provided in the present application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0106] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0107] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0108] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the process in the above-mentioned embodiment method, and can also be completed by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0109] In addition, an embodiment of the present application also provides a computer program product, which, when executed on a computer, enables the computer to execute the methods in the above-mentioned implementation modes.

[0110] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A hydrological data prediction method, characterized in that: The method comprises: Obtain the latest predetermined amount of hydrological historical observation data; The hydrological historical observation data are input into a pre-trained dynamic latent variable model based on Gaussian process to obtain hydrological prediction data corresponding to the hydrological historical observation data. The dynamic latent variable model based on Gaussian process includes an observation model and a dynamic latent variable model. The observation model is used to determine the hydrological historical real data corresponding to the hydrological historical observation data. The dynamic latent variable model is used to determine the hydrological prediction data corresponding to the hydrological historical real data according to the nonlinear dynamic characteristics of the hydrological historical real data.

2. The method according to claim 1, characterized in that The Gaussian process-based dynamic latent variable model also includes a joint probability model; Before inputting the hydrological historical observation data into the pre-trained dynamic latent variable model based on Gaussian process, the method further includes: Initializing first hydrological historical real data corresponding to the hydrological historical observation data; According to the first real hydrological historical data, the hyperparameters in the joint probability model are iteratively optimized by the maximum a posteriori method until the hyperparameters meet the preset requirements, thereby completing the training of the dynamic latent variable model based on the Gaussian process.

3. The method according to claim 2, characterized in that According to the first hydrological historical real data, the hyperparameters in the joint probability model are iteratively optimized by the maximum a posteriori method, including: Determining weight matrix parameters of the joint probability model analytically using maximum a posteriori estimation; The first covariance matrix parameters and the second covariance matrix parameters in the joint probability model are optimized using a scaled conjugate gradient method.

4. The method according to claim 2, characterized in that: The joint probability model includes: Given a first covariance matrix parameter, a weight matrix and real hydrological historical data, a first probability function model for the occurrence of the hydrological historical observation data; A second probability function model for the occurrence of the real hydrological historical data when the second covariance matrix parameters are given; A third probability function model in which the first covariance matrix parameters appear; A fourth probability function model in which the second covariance matrix parameters appear; A fifth probability function model appears in the weight matrix.

5. The method according to claim 4, characterized in that The first probability function value in the first probability function model is determined according to a weight matrix, hydrological historical observation data, a dimension of the hydrological historical observation data, the number of hydrological historical observation data, and a first covariance matrix.

6. The method according to claim 4, characterized in that The second probability function value in the second probability function model is determined according to the quantity of the historical hydrological observation data, the Markov term used to describe the dynamic characteristics of the hydrological prediction data, and the fourth probability function model.

7. The method according to any one of claims 1 to 6, characterized in that: The observation model determines the hydrological observation data according to the system observation noise and the system observation function. The system observation function is determined according to the weight matrix and the dynamic latent variable model. The weight matrix is ​​determined according to the variance of the probability distribution of the weight matrix, the dimension of the hydrological historical observation data, and the elements on the diagonal of the weight matrix.

8. A hydrological data prediction device, characterized in that: The device comprises: A data acquisition unit, used to acquire the latest predetermined amount of hydrological historical observation data; A hydrological prediction data determination unit is used to input the hydrological historical observation data into a pre-trained dynamic latent variable model based on a Gaussian process to obtain hydrological prediction data corresponding to the hydrological historical observation data. The dynamic latent variable model based on a Gaussian process includes an observation model and a dynamic latent variable model. The observation model is used to determine the hydrological historical real data corresponding to the hydrological historical observation data. The dynamic latent variable model is used to determine the hydrological prediction data corresponding to the hydrological historical real data based on the nonlinear dynamic characteristics of the hydrological historical real data.

9. A hydrological data prediction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the hydrological data prediction device implements the method according to any one of claims 1 to 7.

10. A computer program product comprising computer program instructions, characterized in that When the computer program is executed, the method according to any one of claims 1 to 7 is performed.

Images