Self-adaptive hydrological forecasting dynamic optimization method and system

Abstract

The invention provides a self-adaptive hydrological forecast dynamic optimization method and system, and relates to the technical field of hydrological forecast, and the method comprises the steps: obtaining basic data of a forecast target position; determining a hydrological forecasting scheme corresponding to the forecasting demand based on the basic data and the forecasting demand; based on the optimization objective function, parameters of the hydrological forecasting schemes are optimized, a hydrological forecasting scheme set is obtained, and the hydrological forecasting scheme set comprises at least two groups of hydrological forecasting schemes; determining an optimal scheme number sequence in the hydrological forecasting scheme set according to the input data and the measured data; based on the optimization decision vector, taking the optimal scheme number sequence as a target value, and constructing a classification optimization model; based on the classification optimization model, the optimal hydrological forecasting scheme of each time period is determined, the technical problems that the traditional hydrological set probability practical application effect is poor and the operability of wide popularization is lacked are solved, and the technical effect that the optimized forecasting result can be conveniently used in practice is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrological forecasting, and in particular to an adaptive hydrological forecasting dynamic optimization method and system. Background Art

[0002] Hydrological forecasting refers to making qualitative or quantitative estimates of the hydrological status of a river basin in a certain period of time in the future based on the current known information. It is often based on the objective laws of hydrology and the basic principles of hydrology, serves production practice, and has been widely used and achieved huge benefits in flood control, drought relief, comprehensive utilization of water resources, national economic construction and national defense. However, due to the common problems of input uncertainty, model structure uncertainty, model parameter uncertainty, etc. in the process of hydrological forecasting, there is bound to be a difference between the forecast results and the true value.

[0003] Most of the existing operational hydrological forecasting methods are deterministic hydrological forecasting, that is, predicting the future hydrological situation with a deterministic estimate. The advantage of this method is that the results are simple and easy to understand, easy to publicize and publish, and convenient for decision makers to use. However, it ignores the uncertainty factors from many sources in the forecasting process, which makes it impossible for decision makers to quantitatively analyze the potential risks or uncertainties of the forecast results. As a result, the decision-making behavior is not safe and reliable enough, and there are major technical defects. In recent years, the uncertainty hydrological forecasting methods that have gradually developed, such as ensemble forecasting and probabilistic forecasting, have made up for the defects of deterministic forecasting to a certain extent, and can achieve a more accurate quantitative analysis of the uncertainty of forecast success, which can improve the accuracy of forecast results as a whole. However, due to the complexity of the principles of the relevant methods, they require a wealth of probability and statistics knowledge, which makes the application effect in practice poor and lacks operability for widespread promotion. Summary of the invention

[0004] The purpose of the present invention is to provide an adaptive hydrological forecast dynamic optimization method and system to solve the technical problem of poor practical application of hydrological ensemble probability in the prior art.

[0005] In a first aspect, the present application provides an adaptive hydrological forecast dynamic optimization method, comprising:

[0006] Obtain basic data for forecast target location;

[0007] Based on basic data and forecast requirements, determine the hydrological forecast scheme corresponding to the forecast requirements;

[0008] Based on the optimization objective function, the parameters of the hydrological forecast scheme are optimized to obtain a hydrological forecast scheme set, which includes at least two groups of hydrological forecast schemes;

[0009] Determine the optimal scheme number sequence in the hydrological forecast scheme set based on input data and measured data;

[0010] Based on the optimization decision vector, the classification optimization model is constructed with the optimal solution number sequence as the target value;

[0011] Based on the classification optimization model, the optimal hydrological forecast scheme for each time period is determined.

[0012] In a possible embodiment, based on basic data and forecast requirements, determining a hydrological model corresponding to the forecast requirements includes:

[0013] Taking forecast demand as the goal, multiple hydrological models corresponding to the forecast demand are obtained;

[0014] Construct a hydrological forecast scheme based on multiple hydrological models.

[0015] In a possible embodiment, the optimization objective function includes:

[0016]

[0017]

[0018]

[0019] In the formula, Q obs,i is the measured value of the flow rate at the target location for the i-th period forecast, Q sim,i is the simulated value of the flow value at the target location predicted in the i-th period, n is the length of the flow value sequence used for model parameter calibration, i = 1, 2, ..., n.

[0020] In a possible embodiment, determining the optimal scheme number sequence in the hydrological forecast scheme set according to the input data and the measured data includes:

[0021] Based on the input data, the forecast data for the target period is obtained through the hydrological forecast scheme;

[0022] Based on the forecast data and measured data of the target period, the highest scheme number of the hydrological forecast scheme corresponding to the target period is obtained;

[0023] Based on the highest solution number in each target period, the optimal model solution number vector is constructed.

[0024] In a possible embodiment, based on the optimization decision variables, the optimization decision vector of the i-th calculation period is obtained by the following formula, and the optimization decision vector is the optimization decision matrix INDEX:

[0025] INDEX(i)=[P(i-τ),ET(i-τ),Q ave (i)]

[0026] In the formula, P is precipitation, ET is evapotranspiration, Qave (i) is the simulated mean value, τ is T c The ratio of the flow data time interval rounded up, Tc is the confluence time, and i is the i-th calculation period.

[0027] In a possible embodiment, the simulation result mean flow is obtained by the following formula to obtain the simulation mean Q of the i-th calculation period: ave (i)

[0028]

[0029] In the formula, E = 3·m·s, m is the number of hydrological models, s is the number of input data sets, Q sim,ord is a sequence of length n.

[0030] In a possible embodiment, the method further comprises:

[0031] Based on the optimization decision variables and simulated mean values, the optimization decision matrix is ​​updated in real time.

[0032] In a second aspect, the present invention provides an adaptive hydrological forecast dynamic optimization system, comprising:

[0033] The acquisition module is used to obtain basic data of the forecast target location;

[0034] A selection module is used to determine the hydrological forecast scheme corresponding to the forecast demand based on basic data and forecast demand;

[0035] An optimization module is used to optimize the parameters of the hydrological forecast scheme based on the optimization objective function to obtain a hydrological forecast scheme set, wherein the hydrological forecast scheme set includes at least two groups of hydrological forecast schemes; and to determine the optimal scheme number sequence in the hydrological forecast scheme set based on the input data and the measured data;

[0036] A construction module is used to construct a classification optimization model based on the optimization decision variables and the optimal solution number sequence as the target value;

[0037] The forecast module is used to determine the optimal hydrological forecast scheme for each time period based on the classification optimization model.

[0038] In a third aspect, the present application further provides an electronic 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 steps of any one of the methods in the first aspect are implemented.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to execute any one of the methods in the first aspect above.

[0040] The present invention provides an adaptive hydrological forecast dynamic optimization method and system, which obtains basic data of a forecast target location; determines a hydrological forecast scheme corresponding to the forecast demand based on the basic data and forecast demand; optimizes parameters of the hydrological forecast scheme based on an optimization objective function to obtain a hydrological forecast scheme set, wherein the hydrological forecast scheme set includes at least two groups of hydrological forecast schemes; determines an optimal scheme numbering sequence in the hydrological forecast scheme set based on input data and measured data; constructs a classification optimization model based on an optimization decision vector and taking the optimal scheme numbering sequence as a target value; determines the optimal hydrological forecast scheme for each time period based on the classification optimization model, thereby realizing dynamic hydrological forecasting, taking into account various sources of uncertainty in the forecasting process, and making full use of the information provided by the forecast result itself, so that the obtained optimized forecasting results are convenient for practical use and have obvious advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 A schematic diagram of a flow chart of an adaptive hydrological forecast dynamic optimization method provided by an embodiment of the present invention;

[0043] Figure 2 A vector diagram of the optimal model scheme sequence number of reservoir A provided by the embodiment of the present invention;

[0044] Figure 3 A schematic diagram of the structure of an adaptive hydrological forecast dynamic optimization system provided by an embodiment of the present invention;

[0045] Figure 4 A schematic diagram of the structure of an electronic device provided in this application. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0047] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] The existing uncertainty hydrological forecasting methods have complex principles and require rich knowledge of probability statistics, which results in poor practical application and lacks operability for widespread promotion. Based on this, the present application provides an adaptive dynamic optimization method and system for hydrological forecasting, which establishes a set of hydrological forecasting schemes, selects appropriate optimization decision variables, constructs a classification optimization model, and adaptively optimizes and judges the optimal hydrological forecasting scheme for each time period, thereby achieving dynamic optimization of hydrological forecasting. The present application not only comprehensively considers the various sources of uncertainty in the hydrological forecasting process, but also makes full use of the information provided by the hydrological forecast results themselves. The optimized forecasting results obtained can be convenient for practical use and have obvious advantages.

[0049] Figure 1 A flowchart of a method for adaptive hydrological forecasting dynamic optimization provided in this application. The execution subject is a server. Figure 1 As shown, the method comprises the following steps:

[0050] S100: Obtain basic data for forecasting target positions.

[0051] By analyzing the specific needs of the hydrological forecasting program, the basic data of the forecast target location are determined, including the spatial geographical location of the forecast target, as well as the hydrological, meteorological, soil, geological, socio-economic, and geographic information of the forecast target location.

[0052] In order to effectively extend the forecast period of the hydrological forecast scheme, it is necessary to improve the forecast accuracy and obtain numerical weather forecast products within the forecast target location as input variables of the hydrological forecast model. Usually, numerical meteorological and hydrological data within the forecast target location are obtained through ECMWF (European Medium-Range Weather Forecast Dataset), CMA (China Meteorological Administration Forecast Dataset), JMA (Japan Meteorological Agency Forecast Dataset), and NCEP (National Center for Environmental Prediction Forecast Dataset).

[0053] Since the current numerical weather forecast model is seriously affected by the "butterfly effect", medium- and long-term numerical prediction products such as GCM (Global Climate Model) often have large forecast errors and are therefore not suitable for the prediction method of this application.

[0054] S200: Based on basic data and forecast requirements, determine a hydrological forecasting plan corresponding to the forecast requirements.

[0055] In a possible embodiment, step S200 further includes the following steps:

[0056] S210, taking the forecast demand as the goal, obtaining multiple hydrological models corresponding to the forecast demand;

[0057] S220. Construct a hydrological forecasting scheme based on multiple hydrological models.

[0058] Taking into account the different applicable scopes of different hydrological models, combined with basic data, with the purpose of meeting the specific needs of the hydrological forecast scheme, the hydrological model suitable for the current hydrological forecast target location is selected to construct the hydrological forecast scheme, and the finally selected hydrological models are recorded as M(1), M(2), ..., M(m), where m is the number of selected models.

[0059] S300: Based on the optimization objective function, the parameters of the hydrological forecast scheme are optimized to obtain a hydrological forecast scheme set, wherein the hydrological forecast scheme set includes at least two groups of hydrological forecast schemes.

[0060] In order to make the parameters of the model more in line with actual needs, the parameters of the hydrological forecast scheme are optimized based on the optimization objective function to obtain a hydrological forecast scheme set, which includes at least two groups of hydrological forecast schemes.

[0061] Taking into account that different optimization objective functions have different emphases on the forecasting performance of hydrological models, in the process of constructing the hydrological forecasting scheme, this application optimizes the selected hydrological models through three optimization objective functions. The smaller the value of the three optimization objective functions, the better the corresponding model parameters. The three optimization objective functions are as follows:

[0062] The optimization objective function F(1) focuses on the optimization effect of the high-level flow part, that is, it focuses on the fitting accuracy of the flood part data. Its mathematical expression is as follows:

[0063]

[0064] The optimization objective function F(2) takes into account the optimization effect of all calibration data, and its mathematical expression is as follows:

[0065]

[0066] The optimization objective function F(3) focuses on the optimization effect of the low-level flow part, that is, it focuses on the fitting accuracy of the low-flow part data. Its mathematical expression is as follows:

[0067]

[0068] In the formula, Q obs,i is the measured value of the flow rate at the target location for the i-th period forecast, Qsim,i is the simulated value of the flow value at the target location predicted in the i-th period, n is the length of the flow value sequence used for model parameter calibration, i = 1, 2, ..., n.

[0069] Furthermore, the optimization algorithm can be used to calibrate the parameters of each hydrological model and optimize it to obtain a set of hydrological forecast schemes. That is, the optimization algorithm is used to find the model parameter group that makes the optimization objective function value optimal under certain specific optimization objective functions, given the model, input, and output.

[0070] When a deterministic optimization algorithm is used, the parameters that minimize the value of the optimization objective function are taken as the final optimal parameter group, which is denoted as θ(j, k), where j = 1, 2..., m; k = 1, 2, 3;

[0071] When the evolutionary algorithm is used, considering that the evolutionary algorithm is affected by uncertain factors such as initial value selection, mutation, crossover, and elite strategy, when F(k) is used as the optimization objective function for the hydrological model M(i), the evolutionary algorithm is used to optimize N opt times, take N opt The parameter with the smallest value of the optimization objective function is taken as the final optimal parameter group, denoted as θ(j, k), where j = 1, 2..., m; k = 1, 2, 3.

[0072] Furthermore, in this application, an evolutionary algorithm is used to optimize N opt The number of times is a positive integer greater than or equal to 1.

[0073] Furthermore, the optimization algorithms used in this application include but are not limited to genetic algorithms, neural network algorithms, etc.

[0074] S400: Determine the optimal scheme number sequence in the hydrological forecast scheme set based on the input data and the measured data.

[0075] In one embodiment, step S400 further includes the following steps:

[0076] S410, based on the input data, obtaining forecast data for the target period through a hydrological forecasting scheme;

[0077] S420, based on the forecast data and measured data of the target period, obtaining the highest scheme number of the hydrological forecast scheme corresponding to the target period;

[0078] S430: construct an optimal model solution number vector based on the highest solution number of each target time period.

[0079] Specifically, in the process of hydrological forecasting, the different sources or methods of input data of the hydrological model used will affect the accuracy of the forecast results.

[0080] This application takes the most important rainfall input of the hydrological model as an example. In addition to the measured rainfall station data, commonly used numerical weather forecast products include ECMWF, CMA, JMA, NCEP, etc. In addition, the potential evaporation data required by the hydrological model can also be obtained through the statistics of long series observation data from meteorological stations, or approximate calculations through the Penman formula.

[0081] Assume that s sets of different input data sets INDATA1, INDATA2, ..., INDATA s , then considering the combination of model, parameter and input data, a total of E = 3·m·s groups of hydrological forecast results can be obtained. The forecast result obtained when the j-th model uses the θ(j, k) parameter group and the l-th group of input data to carry out hydrological forecast is Qs im,ord , when j, k, 1 are determined, a unique solution number ord can be obtained, and ord = 3·m·l-3m+3j-3+k, l = 1, 2..., s, and the range of ord is 1, 2..., 3·m·s. sim,ord is a sequence of length n, whose i-th element Q sim,ord (i) is the forecast result of the flow value of the target location for the i-th period by the ord-th group of schemes.

[0082] In this application, the principle of minimizing the absolute value of the forecast error is adopted, and the sequence number of the plan with the highest forecast accuracy in the i-th period ORD(i) can be determined by the following formula:

[0083] ORD(i)=arg min(|Q sim,ORD(i) (i)-Q obs (i)|)

[0084] In the formula, Q sim,ORD(i) (i) is the forecast result of the flow value of the target location at the i-th period by the ord-th group of solutions, Q obs (i) is the measured value of the flow rate at the target location for the forecast in the i-th period.

[0085] Therefore, the sequence numbers of the plans with the highest forecast accuracy in each time period can be used to form the optimal model plan sequence number vector ORD, which is a column vector of n×1.

[0086] Among them, the absolute value of the forecast error is the absolute difference between the forecast value obtained by the hydrological forecast scheme for a certain period of time and the actual measured value of that period.

[0087] S500: Based on the optimization decision vector and taking the optimal solution number sequence as the target value, a classification optimization model is constructed.

[0088] In a possible embodiment, based on the optimization decision variables, the optimization decision vector of the i-th calculation period is obtained by the following formula, and the optimization decision vector is the optimization decision matrix INDEX:

[0089] INDEX(i)=[P(i-τ),ET(i-τ),Q ave (i)]

[0090] In the formula, P is precipitation, ET is evapotranspiration, Q ave (i) is the simulated mean value.

[0091] Among them, the simulation result mean flow, through the following formula to obtain the simulation mean Q of the i-th calculation period ave (i)

[0092]

[0093] In the formula, E = 3·m·s, m is the number of hydrological models, s is the number of input data sets, Q sim,ord is a sequence of length n.

[0094] Specifically, in combination with the s groups of input data sets used in the process of building the forecast scheme, appropriate optimization decision variables are selected to build the classification optimization model.

[0095] For hydrological forecasting, the necessary input data include precipitation P and evapotranspiration data ET. Since precipitation P and evapotranspiration data ET directly affect the flow process in the hydrological cycle, this application uses the two input quantities of precipitation P and evapotranspiration data ET as optimization decision variables.

[0096] The simulation accuracy of each hydrological model scheme for different magnitudes of flow values ​​is also different. Considering that the measured flow value of the target period is unknown before simulation or forecast calculation, it is not appropriate to directly use the measured flow value or the simulated flow value with the highest accuracy in this period as the optimization decision variable. Therefore, the mean of the simulation results of each scheme is used as the optimization decision variable in this application. Therefore, it is necessary to average the model simulation results of each calculation period to obtain the simulation mean series Q ave , the simulated mean value of the i-th calculation period is Q ave (i);

[0097] Considering that the precipitation and evapotranspiration data used in the construction of forecast schemes in practice are all measured values, the input data sets INDATA sol , sol=1, 2, ..., s have the same P and ET values.

[0098] Due to the inherent laws of the basin's hydrological cycle, the process of precipitation converging to form the flow of the forecast section requires a certain confluence time T. c, assuming that τ is T c The ratio of the time interval of the flow data rounded up is P(i-τ) and ET(i-τ) as the optimization decision variables for the i-th calculation period, where τ is T c The ratio of the traffic data time interval rounded up. The optimization decision vector of the i-th calculation period is INDEX(i) = [P(i-τ), ET(i-τ), Q ave (i)], then the optimization decision matrix INDEX obtained by calculating the optimization decision vector for all calculation periods is an n×3 matrix.

[0099] In one embodiment, step S500 further includes: constructing a classification optimization model based on the nearest neighbor model, and the specific construction process is as follows:

[0100] S510, perform maximum and minimum normalization calculations on each column vector of the optimization decision matrix INDEX to obtain a vector P that maps each column vector to the [0,1] space 0-1 , E.T. 0-1 , Q ave,0-1 ;

[0101] S520, according to the importance of the optimization decision variables corresponding to each column vector, assign them a given weight [λ1, λ2, λ3] to obtain a weighted normalized optimization decision matrix:

[0102] INDEX uni =[λ1·P 0-1 λ2·ET 0-1 ,λ3·Q ave-1 ],in The weight value can be obtained through manual experience, principal component analysis, correlation analysis and other methods.

[0103] S530. For each solution number ord, extract its order R(ord) in the optimal model solution number vector ORD, so that ORD(R(ord))=ord.

[0104] S540, extract the corresponding normalized optimization decision matrix INDEX according to the order R(ord) of each solution number ord in the optimal model solution number vector ORD uni,ord =INDEX uni (R(ord), 1:3).

[0105] S550, according to the nearest neighbor model construction idea, when the optimization decision vector index corresponding to any calculation period tf is given, in (tf-τ), ET in (tf-τ), Q ave,in(tf)], the weighted normalized optimization decision vector index can be calculated according to the above steps uni Then, according to formula (6), the normalized optimization decision matrix INDEX corresponding to index and the ord of each solution is calculated. uni,ord The Euclidean distance DIS euc (ord):

[0106] DIS euc (ord)=||index uni -INDEX uni，ord ||2

[0107] S560, for the tfth calculation period, the optimal model solution number ordopt(tf) is set to DIS ew (ord) takes the smallest ord value. It should be pointed out that the calculation period tf of the optimization decision vector index is different from the value range of the above variable i, and is not limited to the period 1 to n used to build the model. It can be extended as the data series is extended or a data subset can be selected as the calculation period as needed. Therefore, the optimization model established by this method can be used for real-time forecasting.

[0108] S570, select the forecast result Q of the tfth period corresponding to the ordopt(tfth)th model scheme sim,ordopt(tf) (tf) is taken as the optimized prediction result.

[0109] Furthermore, in addition to the clustering model, alternative solutions for the classification optimization model in this step also include linear regression models, nonlinear regression models, artificial intelligence (including deep learning) models, decision tree models, Bayesian models, and support vector machines.

[0110] S600: Determine the optimal hydrological forecasting scheme for each time period based on the classification optimization model.

[0111] Specifically, the established optimization decision matrix INDEX is calculated based on the data set in the model construction phase. In actual work, due to the real-time and variability of hydrological forecasts, the classification optimization model must also be dynamically adjusted to adapt to the latest forecast situation. Therefore, the optimization decision matrix INDEX should be updated in real time in combination with the data set in the model construction phase and the operational forecast data set carried out after the forecast model is run, and the optimal model scheme number for the time period is determined accordingly to achieve adaptive optimization. Furthermore, considering that the forecast work is also constantly developing and improving over time, too old data can be eliminated, and data within a certain period of time when the forecast work is carried out can be selected to construct the optimization decision matrix INDEX, instead of using all historical data, so as to maintain good consistency in the data used for optimizing decisions.

[0112] In this application, taking the inflow forecast of reservoir A as an example, the flow forecast is performed through the hydrological forecast optimization method in this application. The specific steps are as follows:

[0113] Step 1: First, collect the data and information needed for hydrological forecasting within the scope of Reservoir A, including hydrological, meteorological, soil, geological, socio-economic, and geographic information. According to the data analysis, Reservoir A is a river-type reservoir with a catchment area of ​​530km 2 , the basin terrain is mostly mountainous, and there are no large tributaries entering the reservoir area. It is necessary to carry out hourly flow forecast for Reservoir A for a 24-hour forecast period; there are no hydrological stations in the basin of Reservoir A, there is 1 meteorological station that can provide daily evaporation data, and 8 rainfall stations that can provide hourly measured rainfall data; considering that the basin of Reservoir A is a humid basin with no obvious karst terrain or underground river distribution, the Sanshuiyuan Xin'anjiang model and the GR4J model can be used to construct a forecast scheme; the hourly inflow data of Reservoir A from 2011 to 2020 were collected; the meteorological stations in the basin of Reservoir A were collected The daily evaporation data from 2011 to 2020 were divided by 24 as the hourly evaporation of the day; according to the spatial positions of the 8 rainfall stations and the boundary of the A reservoir basin, the Thiessen polygon method was used to calculate the hourly measured surface rainfall in the A reservoir basin from 2011 to 2020; the A reservoir basin was dominated by rainstorms and floods, and the flood confluence time was relatively short, only about 5 hours. In order to achieve the forecast requirements of the 24-hour forecast period, the 24-hour rainfall control forecast results (control forecast) of the three medium-term weather forecast centers, ECMWF, NCEP and CMA, were used to extend the forecast period of the hydrological model.

[0114] Step 2: Construct a forecasting scheme for the hourly flow forecast of Reservoir A in the 24h forecast period. The number of hydrological models is m = 2, including M(1) Sanshuiyuan Xin'anjiang model and M(2) GR4J model. Three optimization objective functions F(1), F(2), and F(3) are used as parameter optimization objective functions of the selected models. The hourly surface rainfall and evaporation of Reservoir A are used as inputs, and the hourly inflow data of Reservoir A is used as the output fitting target. The genetic algorithm in the evolutionary algorithm is used to optimize the three groups of parameters θ(1, k) of the M(1) model and the three groups of parameters θ(2, k) of the M(3) model, k = 1, 2, 3. Among them, the rate period is 2011-2017, and the verification period is 2018-2020.

[0115] Step 3: In terms of model input data, in this example, only the rainfall forecast data from different sources in the A reservoir basin are considered, which are the 24h rainfall control forecasts from three medium-term weather forecast centers, ECMWF, NCEP and CMA. Therefore, the forecast input data set for each model is s = 3 groups. It can be seen that when considering the combination of models, parameters and input data, a total of E = 3·m·s = 3·2·3 = 18 groups of hydrological forecast results can be obtained for A reservoir. According to ord = 3·m·l-3m+3j-3+k, each group of hydrological forecast results is numbered, as shown in Table T1.

[0116] Table T1 Forecast structure number table

[0117]

[0118]

[0119] The simulation results of each calculation period are statistically analyzed. Taking the simulation results of July 6, 2012 as an example, the program number with the highest forecast accuracy in this period is analyzed. Table T2 gives the forecast results of 18 groups of model programs and the measured flow values ​​in this period. From the results of Table T2, it can be seen that the third group of simulation programs has the smallest absolute error value, so the program number with the highest forecast accuracy in this period is 3. By analogy, the program numbers with the highest forecast accuracy in all calculation periods can be obtained to form the ORD vector, see Figure 2 .

[0120] Table T2 Forecast accuracy evaluation table of each scheme for the period of July 6, 2021

[0121]

[0122]

[0123] Step 4: Construct the optimized decision vector for each time period. Since the confluence time of reservoir A is about 5h, the optimized decision vector for the tf period is:

[0124] INDEX(tf)=[P(tf-5h),ET(tf-5h),Q ave (tf)]

[0125] Taking July 6, 2012 as an example, the basin surface rainfall 5 hours before this period was 57 mm, and the basin period evapotranspiration 5 hours before this period was 0.75 mm. The simulated mean value of this period can be obtained from the data in Table T2 as 1766m 3 / s; therefore, the optimization decision vector for this period is [57,0.75,1766]. Similarly, the optimization decision matrix INDEX is constructed for all calculation periods.

[0126] Step 5: Build a classification optimization model based on the nearest neighbor model.

[0127] 1) Perform normalization calculations on the maximum and minimum values ​​of each column vector of the optimization decision matrix INDEX. Taking the period of July 6, 2012 as an example, in order to achieve the normalization of the optimization decision vector for this period, the maximum values ​​of each column of the optimization decision matrix INDEX are first obtained as 189mm, 1.27mm, 11101m 3 / s, the minimum values ​​are: 0mm, 0.26mm, 128m 3 / s, according to the maximum and minimum value normalization method, the normalized optimization decision vector of this period can be calculated as [(57-0) / (189-0), (0.75-0.26) / (1.27-0.26), (1766-128) / (11101-128)], that is, [0.302, 0.485, 0.164]. By analogy, the optimization decision vectors of all periods can be normalized.

[0128] 2) Select weight coefficients for each decision variable. In this specific implementation, the correlation coefficient method is used for calculation. Calculate P(tf-5h), ET(tf-5h), Q ave (tf) and Q obs The linear correlation coefficients of (tf) are ρ1, ρ2, ρ3, where Q obs (tf) is the measured flow value, and we get ρ1=0.68, ρ2=0.12, ρ3=0.93. The corresponding weights are determined according to the relative size of each correlation coefficient: We obtain [λ1, λ2, λ3] = [0.39, 0.07, 0.54], so the weighted normalized optimization decision vector for the period of July 6, 2012 is [0.302·λ1, 0.485·λ2, 0.164·λ3] = [0.119, 0.034, 0.088].

[0129] 3) For each solution number ord, extract its order R(ord) in the optimal model solution number vector ORD, so that ORD(R(ord))=ord. Taking the period of July 6, 2012 as an example, its solution number is ord=3, and the order of this period in ORD is 553, then 553∈R(3), and so on, the order numbers corresponding to each period in ORD can be attributed to the corresponding R(ord). From the above classification process, we can see that ORD(x) = ord, and satisfies

[0130] 4) Extract the order R(ord) of each solution number ord in the optimal model solution number vector ORD and extract the corresponding weighted normalized optimization decision matrix INDEXuni,ord =INDEX uni (R(ord), 1:3). Taking the period of July 6, 2012 as an example, its scheme number is 3, and its period number is tf=553. In addition to this period, there are 232 periods with scheme number 3 in this example, so we get INDEX uni，3 The weighted normalized optimization decision matrix of is shown in Table T3, and the weighted normalized optimization decision matrices corresponding to the other scheme numbers can be deduced in the same way. From the above classification process, it can be seen that

[0131] Table T3 INDEX uni,3 The weighted normalized optimization decision matrix results table

[0132]

[0133]

[0134] 5) For the tfth calculation period, the optimal model solution number ordopt(tf) is set to DIS ew (ord) takes the smallest ord value. It should be pointed out that the calculation period tf of the optimization decision vector index is different from the value range of the above variable i, and is not limited to the period 1 to n used to build the model. It can be extended as the data series is extended or a data subset can be selected as the calculation period as needed. Therefore, the optimization model established by this method can be used for real-time forecasting.

[0135] Specifically, the data period used for model construction in this example is 2011-2020, while the actual forecasting work may occur outside the adopted data period. For example, the forecaster conducted a forecast of the inflow of Reservoir A on July 6, 2021, with tf = 3840. According to the forecast results, the forecast optimization decision vector for this period is [89, 0.84, 3558]. The weighted normalized optimization decision vector can be obtained from the above steps as index uni =[0.184, 0.058, 0.175]. Calculate the index of the tf period respectively uni With each INDEX uni，ord The Euclidean distance between euc (ord):

[0136] DIS euc (ord)=||index uni -INDEX uni，ord ||2

[0137] remember

[0138]

[0139] The above formula can also be expressed as:

[0140]

[0141] Where: r ord Represents INDEX uni,ord The number of rows; the rest are the same as before.

[0142] After calculation, the forecast DIS for the inflow of reservoir A on July 6, 2021 in this example is euc (ord) Calculated values ​​are shown in Table T4.

[0143] Table T4 DIS forecast of the inflow of reservoir A on July 6, 2021 euc (ord) Calculated value

[0144] Scheme number ord 1 2 3 4 5 6 7 8 9 <![CDATA[DIS euc (word)]]> 0.184 0.514 0.236 0.118 0.412 0.321 0.344 0.398 0.214 Scheme number ord 10 11 12 13 14 15 16 17 18 <![CDATA[DIS euc (word)]]> 0.097 0.154 0.205 0.321 0.075 0.147 0.364 0.278 0.335

[0145] 6) According to the results in Table T4, for the forecast of the inflow of Reservoir A on July 6, 2021, when the scheme number ord is 14, there is DIS euc (ord)DIS euc The calculated value of (ord) is the smallest, which is 0.075, indicating that the results of the 14th forecast scheme should be recommended for this forecast.

[0146] 7) By querying the forecast results of the 18 forecast plans, the result of the 14th forecast plan is 3775m 3 / s, which means that the forecast value of the inflow of Reservoir A on July 6, 2021 should be 3775m 3 / s forecast results.

[0147] In addition to the clustering model, alternative solutions for the classification optimization model in step S105 also include linear regression model, nonlinear regression model, artificial intelligence (including deep learning) model, decision tree model, Bayesian model, support vector machine. For example, by establishing a nonlinear support vector machine classification model with a decision variable matrix and the sequence number of the optimal solution for each time period, the same purpose can also be achieved. Therefore, the appropriate transformation of some steps of this method still belongs to the invention category of this method in essence.

[0148] Step 6: Based on the constructed classification optimization model, adaptive optimization is performed to determine the optimal hydrological forecast scheme for each time period, thereby realizing dynamic optimization of the hydrological forecast. Specifically, the optimization decision matrix INDEX established in step S104 of the present invention is calculated based on the data set in the model construction phase. In actual work, due to the real-time and variability of hydrological forecasts, the classification optimization model must also be dynamically adjusted to adapt to the latest forecast situation. Therefore, the optimization decision matrix INDEX should be updated in real time in combination with the data set in the model construction phase and the operational forecast data set carried out after the forecast model is run, and the optimal model scheme sequence number for the time period is determined accordingly to realize adaptive optimization.

[0149] In this example, the forecast value of the inflow of reservoir A on July 6, 2021 corresponds to the time period number tf=3840, and the maximum time period number of the 2011-2020 data used in the model construction phase is n=3653. Obviously, as the forecast work progresses, the data available for building the optimization model will continue to increase, and as time goes by, the forecast results of the forecast model will be able to obtain the corresponding measured values, so the optimal solution combination corresponding to this forecast can also be determined. From the perspective of adapting to the latest forecast situation, the forecast results generated by the forecast work should be included in the optimization decision matrix INDEX, and the adaptive update of the forecast classification optimization model can be achieved by reconstructing the model using the latest data.

[0150] On the other hand, considering that forecasting work is constantly developing and improving over time, outdated data should be considered for elimination. Therefore, when constructing the optimization decision matrix INDEX, data within a certain period of time when the forecasting work is carried out can be selected for construction instead of using all data. For example, when forecasting the inflow of Reservoir A on July 6, 2024, data from the past 20 years (i.e., July 6, 2004 to July 5, 2024) are used to construct the optimization decision matrix INDEX, and the optimal model scheme number for this period is determined according to the aforementioned steps of the present invention to achieve adaptive optimization of forecast results.

[0151] Figure 3 The present invention provides a flow chart of an adaptive hydrological forecast dynamic optimization system. Figure 3 As shown, the execution subject of the system may be a server, and the system includes:

[0152] The acquisition module is used to obtain basic data of the forecast target location;

[0153] A selection module is used to determine the hydrological forecast scheme corresponding to the forecast demand based on basic data and forecast demand;

[0154] An optimization module is used to optimize the parameters of the hydrological forecast scheme based on the optimization objective function to obtain a hydrological forecast scheme set, wherein the hydrological forecast scheme set includes at least two groups of hydrological forecast schemes; and to determine the optimal scheme number sequence in the hydrological forecast scheme set based on the input data and the measured data;

[0155] A construction module is used to construct a classification optimization model based on the optimization decision variables and the optimal solution number sequence as the target value;

[0156] The forecast module is used to determine the optimal hydrological forecast scheme for each time period based on the classification optimization model.

[0157] In an optional embodiment, based on basic data and forecast requirements, determining a hydrological model corresponding to the forecast requirements includes:

[0158] Taking forecast demand as the goal, multiple hydrological models corresponding to the forecast demand are obtained;

[0159] Construct a hydrological forecast scheme based on multiple hydrological models.

[0160] In an optional embodiment, the optimization objective function includes:

[0161]

[0162]

[0163]

[0164] In the formula, Q obs,i is the measured value of the flow rate at the target location for the i-th period forecast, Q sim,i is the simulated value of the flow value at the target location predicted in the i-th period, n is the length of the flow value sequence used for model parameter calibration, i = 1, 2, ..., n.

[0165] In an optional embodiment, determining the optimal scheme number sequence in the hydrological forecast scheme set according to the input data and the measured data includes:

[0166] Based on the input data, the forecast data for the target period is obtained through the hydrological forecast scheme;

[0167] Based on the forecast data and measured data of the target period, the highest scheme number of the hydrological forecast scheme corresponding to the target period is obtained;

[0168] Based on the highest solution number in each target period, the optimal model solution number vector is constructed.

[0169] In an optional embodiment, based on the optimization decision variables, the optimization decision vector of the i-th calculation period is obtained by the following formula, and the optimization decision vector is the optimization decision matrix INDEX:

[0170] INDEX(i)=[P(i-τ),ET(i-τ),Q ave (i)]

[0171] In the formula, P is precipitation, ET is evapotranspiration, Q ave (i) is the simulated mean value, τ is T c The ratio of the traffic data time interval rounded up, T c is the confluence time, and i is the i-th calculation period.

[0172] In an optional embodiment, the simulation result mean flow is obtained by the following formula: ave (i)

[0173]

[0174] In the formula, E = 3·m·s, m is the number of hydrological models, s is the number of input data sets, Q sim,ord is a sequence of length n.

[0175] In an optional embodiment, the method further comprises:

[0176] Based on the optimization decision variables and simulated mean values, the optimization decision matrix is ​​updated in real time.

[0177] The implementation principle and technical effects of the system provided in the embodiment of the present application are the same as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the system embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0178] like Figure 4 As shown, an electronic device 600 provided in an embodiment of the present application includes: a processor 601, a memory 602 and a bus, the memory 602 stores machine-readable instructions executable by the processor 601, when the electronic device is running, the processor 601 and the memory 602 communicate through the bus, and the processor 601 executes the machine-readable instructions to perform the steps of the above-mentioned adaptive hydrological forecast dynamic optimization method.

[0179] Specifically, the above-mentioned memory 602 and processor 601 can be general-purpose memories and processors, which are not specifically limited here. When the processor 601 runs the computer program stored in the memory 602, it can execute the above-mentioned adaptive hydrological forecast dynamic optimization method.

[0180] The processor 601 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 601. The above processor 601 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module may be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 602, and the processor 601 reads the information in the memory 602 and completes the steps of the above method in combination with its hardware.

[0181] Corresponding to the above-mentioned adaptive hydrological forecasting dynamic optimization method, an embodiment of the present application also provides a computer-readable storage medium, which stores machine-executable instructions. When the computer-executable instructions are called and executed by the processor, the computer-executable instructions prompt the processor to execute the steps of the above-mentioned adaptive hydrological forecasting dynamic optimization method.

[0182] An adaptive hydrological forecast dynamic optimization system provided in the embodiment of the present application can be specific hardware on the device or software or firmware installed on the device. The device provided in the embodiment of the present application, its implementation principle and the technical effect produced are the same as those in the aforementioned method embodiment. For the sake of brief description, the parts not mentioned in the device embodiment can refer to the corresponding contents in the aforementioned method embodiment. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices and units described above can all refer to the corresponding processes in the aforementioned method embodiment, and will not be repeated here.

[0183] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, 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 communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0184] For another example, the flowchart and block diagram in the accompanying drawings show the possible architecture, functions and operations of the device, method and computer program product according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and a part of the module, program segment or code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0185] 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.

[0186] In addition, each functional unit in the embodiments provided in 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.

[0187] If the function 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 technical solution of the present application can essentially or in other words, the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions for an electronic device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store program codes.

[0188] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0189] In the description of the present invention, it should be noted that the terms "first", "second", "third", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0190] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive hydrological forecast dynamic optimization method, characterized in that: include: Obtain basic data for forecast target location; Based on the basic data and forecast requirements, determine a hydrological forecast scheme corresponding to the forecast requirements; Based on the optimization objective function, the parameters of the hydrological forecast scheme are optimized to obtain a hydrological forecast scheme set, wherein the hydrological forecast scheme set includes at least two groups of hydrological forecast schemes; Determine the optimal scheme number sequence in the hydrological forecast scheme set according to the input data and the measured data; Based on the optimization decision vector, a classification optimization model is constructed with the optimal solution number sequence as the target value; Based on the classification optimization model, the optimal hydrological forecast scheme for each time period is determined.

2. The method according to claim 1, characterized in that Based on the basic data and the forecast requirement, determining a hydrological model corresponding to the forecast requirement includes: Taking forecasting demand as a goal, obtaining multiple hydrological models corresponding to the forecasting demand; Based on the multiple hydrological models, a hydrological forecasting scheme is constructed.

3. The method according to claim 1, characterized in that The optimization objective function includes: In the formula, Q obs,i is the measured value of the flow rate at the target location for the i-th period forecast, Q sim,i is the simulated value of the flow value at the target location predicted in the i-th period, n is the length of the flow value sequence used for model parameter calibration, i = 1, 2, ..., n.

4. The method according to claim 1, characterized in that: According to the input data and the measured data, the optimal scheme number sequence in the hydrological forecast scheme set is determined, including: Based on the input data, obtaining forecast data for a target period through the hydrological forecasting scheme; Based on the forecast data and measured data of the target period, obtaining the highest scheme number of the hydrological forecast scheme corresponding to the target period; Based on the highest solution number of each target time period, an optimal model solution number vector is constructed.

5. The method according to claim 1, characterized in that: Based on the optimization decision variables, the optimization decision vector of the i-th calculation period is obtained by the following formula, and the optimization decision vector is the optimization decision matrix INDEX: INDEX(i)=[P(i-τ),ET(i-τ),Q ave (i)] In the formula, P is precipitation, ET is evapotranspiration, Q ave (i) is the simulated mean value, τ is T c The ratio of the flow data time interval rounded up, Tc is the confluence time, and i is the i-th calculation period.

6. The method according to claim 5, characterized in that The simulated mean value is obtained by the following formula: ave (i) In the formula, E = 3·m·s, m is the number of hydrological models, s is the number of input data sets, Q sim,ord is a sequence of length n.

7. The method according to claim 6, characterized in that The method further comprises: Based on the optimization decision variables and the simulation mean value, the optimization decision matrix is ​​updated in real time.

8. An adaptive hydrological forecast dynamic optimization system, characterized in that: include: The acquisition module is used to obtain basic data of the forecast target location; A selection module, used for determining a hydrological forecasting scheme corresponding to the forecasting requirement based on the basic data and the forecasting requirement; An optimization module, configured to optimize the parameters of the hydrological forecast scheme based on an optimization objective function to obtain a hydrological forecast scheme set, wherein the hydrological forecast scheme set includes at least two groups of hydrological forecast schemes; and to determine the optimal scheme number sequence in the hydrological forecast scheme set based on input data and measured data; A construction module is used to construct a classification optimization model based on the optimization decision variables and taking the optimal solution number sequence as the target value; The forecast module is used to determine the optimal hydrological forecast scheme for each time period based on the classification optimization model.

9. An electronic 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 steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to execute the method according to any one of claims 1 to 7.

Images