Hydrological forecasting method and system, electronic equipment and medium

Abstract

The invention discloses a hydrological forecasting method and system, electronic equipment and a medium, and relates to the field of hydrological forecasting, the method comprises the steps of firstly obtaining observation data at the current moment, utilizing a current hydrological forecasting model to predict the runoff volume of the next time period, and updating the observation data with the set participation rate and the weight thereof in real time to avoid data degradation, so that the prediction accuracy of the runoff volume of the next time period is improved. The parameter correction result can more accurately reflect the hydrological element characteristics of the current drainage basin, and the real-time tracking of the parameters of the hydrological forecasting model is realized when the conditions of drainage basin rainfall, air temperature, underlying surface and the like are changed, so that the more accurate runoff volume is obtained, and the accuracy of hydrological forecasting is improved.

Description

Technical Field

[0001] The present invention relates to the field of hydrological forecasting, and in particular to a hydrological forecasting method, system, electronic equipment and medium. Background Art

[0002] The parameters of the hydrological model reflect the characteristic attributes of the hydrological elements of the basin and have a certain physical basis. There are many factors that cause uncertainty in the model parameters, such as the inherent randomness of the basin, the representativeness of the spatial random distribution characteristics of hydrological information and mathematical expectations, the homogenization problem of the temporal random distribution characteristics of hydrological information, and the lack of reliable information sources for hydrological elements such as soil moisture content. On the other hand, affected by climate change and human activities, the underlying surface characteristics of the basin, such as river width, vegetation cover and other conditions will change. When the model uses non-time-varying parameters, it will not be able to better reflect the dynamic changes of the basin, resulting in large deviations in its simulation or forecast results.

[0003] Studies have shown that even if the underlying surface conditions of a watershed are relatively stable, conditions such as precipitation and temperature within the watershed will change with climate change, and the optimization results of hydrological model parameters depend to a large extent on the sequence data involved in the calibration. Therefore, when the hydrological elements reflected by the input data and the data used for optimization differ greatly, the model simulation / forecast accuracy will also have large deviations.

[0004] The ultimate goal of hydrological model parameter optimization is to obtain the global optimal solution or probability distribution of the parameter group, so its parameter values ​​are "statistically" means. In other words, they should not be considered unchanging. If the time-varying nature of model parameters is ignored and they are used as fixed constants in practical work such as hydrological forecasting, large errors will inevitably occur. Hydrologists have realized that the existence of temporal uncertainty in model parameters must be faced squarely.

[0005] Some quantitative studies on the time-varying characteristics of parameters have found that when the water storage capacity parameters are formulated based on seasonal, annual changes and long-term trends, the model obtains better forecast results. This type of method is to achieve the transformation of parameters from static to dynamic by establishing the function form of the selected model parameters; another type of method is to divide the entire data series into multiple continuous subsets according to the data characteristics, and then calibrate the parameters of each continuous subset separately. These studies show that: (1) model parameters should be time-varying; (2) the acquisition of prior knowledge is the difficulty of time-varying parameters.

[0006] At present, the time scale of most time-varying parameter methods is quarterly or monthly, while real-time hydrological forecasting places higher requirements on the timeliness of time-varying parameter methods. How to construct a method that can meet the time scale of real-time hydrological forecasting, does not require prior knowledge of parameters obtained by human experience, and can realize automatic parameter correction is an urgent problem to be solved.

[0007] Recently, a particle learning algorithm (PL) based on the Bayesian principle has been proposed and has received continuous attention. The core idea of ​​the particle learning algorithm is to create a particle algorithm that samples directly from the joint posterior distribution according to the particle approximation and performs conditional sufficient statistics on the parameters in a resampling-propagation framework. Today, PL has been widely explored in many fields such as energy, medicine, mechanical automation, and social sciences. Summary of the invention

[0008] The purpose of the present invention is to provide a hydrological forecasting method, system, electronic equipment and medium to improve the accuracy of hydrological forecasting.

[0009] To achieve the above object, the present invention provides the following solutions:

[0010] A hydrological forecasting method, comprising:

[0011] Obtaining the current observation data of the hydrological station and the observation data after the first preset period; the observation data includes rainfall and evapotranspiration;

[0012] According to the observation data at the current moment and the first preset period, the runoff after the first preset period is determined by using the hydrological forecast model after the first preset period; wherein the hydrological forecast model after the first preset period is obtained by calibrating the parameters of the hydrological forecast model at the current moment using the historical observation data of the second preset period before the current moment and the historical runoff; the hydrological forecast model after the first preset period is Among them, Calq k+1 is the runoff after the first preset period; is the response function of the model; x k is the model parameter after the first preset period of calibration, y 1:k is the observation data set consisting of rainfall, evapotranspiration and runoff before the current moment, S0 is the initial value of the model intermediate variable, P k+1 is the rainfall after the first preset period; E k+1 is the evapotranspiration after the first preset period; the initial value of the intermediate variable of the model is updated according to the thickness of the saturated soil layer and the water inlet and outlet of the confluence tank.

[0013] Optionally, the parameters of the hydrological forecast model at the current moment are calibrated using the historical observation data of the second preset period before the current moment and the historical runoff, specifically including:

[0014] Obtain historical observation data and historical runoff for a second preset period before the current moment to form a retrospective observation particle set;

[0015] Using a nonlinear forgetting function to assign weights to the back-tracking observation particle set, to obtain a weighted back-tracking observation particle set;

[0016] According to the weighted retrospective observation particle set, the parameters of the hydrological forecast model at the current moment are calibrated to obtain the model parameters after the first preset period after calibration;

[0017] The hydrological forecast model after the first preset time period is determined according to the calibrated model parameters after the first preset time period.

[0018] Optionally, historical observation data and historical runoff of a second preset period before the current moment are obtained to form a retrospective observation particle set, specifically including:

[0019] From the current moment forward, historical observation data and historical runoff within a second preset period are selected to obtain a retrospective observation particle set.

[0020] Optionally, a nonlinear forgetting function is used to assign weights to the backtracking observation particle set to obtain a weighted backtracking observation particle set, specifically including:

[0021] Using the formula w i =e -δ(L-i) , determine the weight of each data in the backtracking observation particle set; where w i is the weight of the data at the i-th moment in the second preset period of time in the retrospective observation particle set, L is the end time of the second preset period, δ is the forgetting coefficient, and Li is the interval between the i-th moment and the end time of the second preset period;

[0022] According to the weight of each data and the back-tracking observation particle set, a weighted back-tracking observation particle set is determined.

[0023] Optionally, according to the weighted retrospective observation particle set, the hydrological forecast model at the current moment is calibrated to obtain the model parameters after the first preset period after calibration, specifically including:

[0024] Using the formula x k = r(y w k-L:k ,x0,S k-L-1 , w i ), determine the model parameters after the first preset period of time after calibration; where x k is the model parameter after the first preset period of time after calibration, r(·) is the model parameter calibration function, y w k-L:k is the weighted back-tracking observation particle set, x0 is the optimal model parameter, S k-L-1 is the intermediate parameter of the model at the kL-1 moment, w iis the weight of the data at the i-th moment within the second preset time period in the back-tracking observation particle set; the optimal model parameters are obtained by calibrating the parameters of the initial hydrological forecast model using the observation data set.

[0025] Optionally, the process of determining the optimal model parameters specifically includes:

[0026] Acquire an observation data set; the observation data set includes rainfall, evapotranspiration and runoff in the third preset period before the current moment;

[0027] The observation data set is used to calibrate the parameters of the initial hydrological forecast model to obtain the optimal model parameters.

[0028] A hydrological forecasting system, comprising:

[0029] A data acquisition module is used to acquire the current observation data of the hydrological station and the observation data after the first preset period; the observation data includes rainfall and evapotranspiration;

[0030] The forecast module is used to determine the runoff after the first preset period of time based on the observation data at the current moment and the observation data after the first preset period of time, using the hydrological forecast model after the first preset period of time; wherein the hydrological forecast model after the first preset period of time is obtained by calibrating the parameters of the hydrological forecast model at the current moment using the historical observation data of the second preset period before the current moment and the historical runoff; the hydrological forecast model after the first preset period of time is Among them, Calq k+1 is the runoff after the first preset period; k is the current time; is the response function of the model; x k is the model parameter after the first preset period of calibration, y 1:k is the observation data set consisting of rainfall, evapotranspiration and runoff before the current moment, S0 is the initial value of the model intermediate variable, P k+1 is the rainfall after the first preset period; E k+1 is the evapotranspiration after the first preset period; the initial value of the intermediate variable of the model is updated according to the thickness of the saturated soil layer and the water inlet and outlet of the confluence tank.

[0031] An electronic device comprises: a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the above-mentioned hydrological forecasting method.

[0032] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the hydrological forecasting method described above is implemented.

[0033] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0034] The hydrological forecasting method, system, electronic device and medium of the present invention first obtain the observation data at the current moment, and use the current hydrological forecasting model to predict the runoff in the next time period. The present invention updates the observation data involved in the calibration and its weight in real time to avoid "data degradation", so that the parameter correction result can more accurately reflect the characteristics of the current watershed hydrological elements, and realizes real-time tracking of the hydrological forecasting model parameters when the conditions such as precipitation, temperature, and underlying surface in the watershed change, thereby obtaining a more accurate runoff and improving the accuracy of the hydrological forecast. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0036] Figure 1 This is a flow chart of the hydrological forecasting method provided by the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. 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.

[0038] The purpose of the present invention is to provide a hydrological forecasting method, system, electronic equipment and medium to improve the accuracy of hydrological forecasting.

[0039] Recently, a particle learning algorithm (PL) based on the Bayesian principle has been proposed and has received continuous attention. The core idea of ​​the particle learning algorithm is to create a particle algorithm that samples directly from the joint posterior distribution according to the particle approximation and performs conditional sufficient statistics on the parameters in a resampling-propagation framework. Today, PL has been widely explored in many fields such as energy, medicine, mechanical automation, and social sciences.

[0040] Based on Bayesian theory, this paper explores how to use PL technology to solve problems such as "importance sampling" and "resampling" in the hydrological process, and to track the parameter changes of the hydrological model in real time, so as to reduce the uncertainty caused by the insufficient expression of the temporal heterogeneity of the hydrological elements in the basin by the model, and improve the accuracy of real-time hydrological forecasting.

[0041] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] Embodiment 1

[0043] like Figure 1 As shown, the hydrological forecasting method provided by the present invention comprises:

[0044] Step 101: Obtain the current observation data of the hydrological station and the observation data after the first preset period; the observation data include rainfall and evapotranspiration.

[0045] Step 102: According to the observation data at the current moment and the observation data after the first preset period, the runoff after the first preset period is determined by using the hydrological forecast model after the first preset period; wherein the hydrological forecast model after the first preset period is obtained by calibrating the parameters of the hydrological forecast model at the current moment using the historical observation data and historical runoff of the second preset period before the current moment; the hydrological forecast model at the current moment is Among them, Calq k+1 is the runoff after the first preset period; is the response function of the model; x k is the model parameter after the first preset period of calibration, y 1:k is the observation data set consisting of rainfall, evapotranspiration and runoff before the current moment, S0 is the initial value of the model intermediate variable, P k+1 is the rainfall after the first preset period; E k+1 is the evapotranspiration after the first preset period; the initial value of the intermediate variable of the model is updated according to the thickness of the saturated soil layer and the water inlet and outlet of the confluence tank.

[0046] The invention discloses a real-time tracking technology of model parameters based on particle learning. The technology establishes parameter correction logic with Bayesian inference as a framework, and proposes a correction scheme that takes the optimal parameter group (optimal model parameter) as the starting point of parameter calibration; proposes a recursive correction equation for parameter correction based on Bayesian recursive estimation to realize the recursive calculation of model parameters; uses the Monte Carlo method to solve the problem that the posterior distribution of parameters cannot be obtained in the recursive calculation, and combines the automatic calibration function of model parameters to propose a sampling scheme that takes observation data instead of parameters as sampling objects; constructs a nonlinear forgetting function based on the "importance probability density function" in particle learning, establishes a weighting equation for observation data, so that the parameter correction result can more accurately reflect the characteristics of the current watershed hydrological elements; establishes a backtracking mechanism based on "resampling" in particle learning, establishes an observation data sampling equation, and updates the observation data and its weights involved in the calibration in real time to avoid "data degradation"; establishes a recursive calculation equation for model intermediate variables based on "state sufficient statistics" in particle learning, and determines the initial value of the intermediate variable at the start time of correction.

[0047] This technology achieves real-time tracking of model parameters when basin precipitation, temperature, underlying surface and other conditions change through the comprehensive application of the model's optimal parameter group and hydrological observation data, thereby obtaining more accurate runoff simulation / forecast and realizing an accurate hydrological forecast process.

[0048] As an optional implementation, the parameters of the hydrological forecast model at the current moment are calibrated using the historical observation data of the second preset period before the current moment and the historical runoff, specifically including:

[0049] Step S1: Obtain historical observation data and historical runoff for a second preset period before the current moment to form a retrospective observation particle set.

[0050] Specifically, step S1 includes:

[0051] From the current moment forward, historical observation data and historical runoff within a second preset period are selected to obtain a retrospective observation particle set.

[0052] In practical applications, step S1 specifically includes:

[0053] Step 1: Collect relevant environmental data of the watershed, including but not limited to: watershed area, land use / cover type and its area proportion.

[0054] Step 2: Collect observation data of hydrological stations, including rainfall, evapotranspiration, and runoff, and interpolate missing data to obtain an observation data set. The observation data set includes rainfall, evapotranspiration, and runoff for the third preset period before the current moment.

[0055] Step 3: Submit the observation data set into the hydrological forecast model and calibrate to obtain the long-term optimal parameter set for the basin.

[0056] Step 4 (Observation data resampling): From the observation data set y 1:k =(P 1:k ,E 1:k ,Obq 1:k ) is traced back L time periods to obtain the backtracking observation particle set y k-L:k =(P k-L:k ,E k-L:k ,Obq k-L:k ). Among them, y 1:k is all the observed data from time 1 to time k, including the rainfall from time 1 to time k (P 1:k ), evaporation (E 1:k ) and runoff (Obq 1:k ). The lookback time L varies according to the basin and model, and can be obtained through trial and error. The resampling step is to avoid the phenomenon of "data degradation" in the process of parameter correction. For example, when the target time is January 11, 1991, the observation data on January 9, 1991 is selected by the model for parameter calibration. As the recursive calculation proceeds, when July 15, 2021 is used as the target time, the data on January 9, 1991 is still involved in the parameter correction, which not only wastes a lot of computing resources, but also the large state deviation may cause deviations in the calibration results. Therefore, the model needs to use the resampling method to re-range the "sample information" at each correction.

[0057] Step S2: using a nonlinear forgetting function to assign weights to the back-tracking observation particle set to obtain a weighted back-tracking observation particle set.

[0058] In practical applications, the nonlinear forgetting function is used to weight the backtracking observation particle set obtained in step S1: i =e -δ(L-i) (0≤i≤L). In the formula, w i The weight of the data at the i-th moment in the second preset period of the retrospective observation particle set, L is the end time of the second preset period, δ is the forgetting coefficient, and Li is the interval between the i-th moment and the end time of the second preset period. The use of the forgetting function (nonlinear monotonically decreasing) to weight the observation data involved in parameter calibration is actually a deformation of the residual series, so that when the model parameters are calibrated, its objective function will pay more attention to the errors of recent data.

[0059] According to the weight of each data and the back-tracking observation particle set, a weighted back-tracking observation particle set is determined.

[0060] Step S3: calibrating the parameters of the hydrological forecast model at the current moment according to the weighted retrospective observation particle set to obtain the calibrated model parameters after the first preset time period.

[0061] In practical applications, the weighted retrospective observation particle set is used to correct the parameters at the target time: k = r(y k-L:k ,x0,S k-L-1 ,w i ). In the formula, x k is the model parameter after the first preset period of time after calibration, r(·) is the model parameter calibration function, y k-L:k is the set of back-tracking observed particles, x0 is the optimal model parameter, S k-L-1 is the intermediate parameter of the model at the kL-1 moment, w i is the weight of the data at the i-th moment in the second preset period in the retrospective observation particle set. Step S3 takes the optimal parameter group as the starting point, ensuring that after the parameter correction is achieved, the improvement of the model prediction accuracy is carried out on the basis of the original prediction accuracy, and then uses the weighted retrospective observation particle set to obtain the correction result of the target moment parameter.

[0062] Update the intermediate variables of the model at the kL moment and calculate the initial values ​​of the intermediate variables when the next cycle parameter is corrected: S k-L =λ(S k-L-1 ,SS k-L-1 ,x0). In the formula, λ(·) is the calculation function of the intermediate variable of the model, ss k-L-1 It is the statistics of the model at the kL-1 moment, such as the thickness of the saturated soil layer, the water inlet and outlet of the confluence tank, etc.

[0063] Step S4: determining the hydrological forecast model after the first preset period of time according to the calibrated model parameters after the first preset period of time.

[0064] After that, after acquiring new observation data (rainfall and evapotranspiration), the runoff at the next moment is predicted using the hydrological forecast model after the first preset period determined in step S4 to obtain the predicted runoff (the predicted runoff is the process and scenario of the hydrological prediction to be achieved):

[0065] In steps S1 and S2, the present invention screens the sequence data involved in the calibration, and gives a greater weight to the data closer to the target time in time, which not only avoids the "data degradation" phenomenon in the process of parameter correction, but also makes the objective function more "focused" on the error of recent data, so that the calibration result can better reflect the characteristics of the current basin hydrological elements. Since the optimal parameter group is obtained by the model using the data information of the basin for decades or even hundreds of years, step S3 uses the optimal parameter group as the starting point of parameter correction. While integrating the long-term basic characteristics of the basin hydrological elements, it ensures that after the parameter correction is realized, the prediction accuracy of the model is improved based on the original accuracy. So far, steps S1, S2 and S3 realize the comprehensive use of "prior information" and "sample information" by the model when the parameters are corrected in real time. In addition, from the logical perspective of parameter correction, the present invention adopts the "resampling (steps S1, S2)-transfer (step S3)" logic, rather than the "transfer-resampling" logic, and incorporates the latest acquired observation data into the real-time parameter correction, which helps to improve the accuracy.

[0066] Without considering the input data error, the forecast of multi-year continuous hydrological process with a forecast period of one period (6 hours) was carried out. On average, the relative error of the flood volume predicted by this method was reduced by 24%; the Nash efficiency coefficient was increased by 21%. For secondary floods, the relative error of flood volume was reduced from 0.13 to 0.06 on average, the Nash efficiency coefficient was increased from 0.67 to 0.91 on average, the flood peak error was reduced from 0.21 to 0.15 on average, and the peak time error was reduced from 2 periods to 0 periods on average.

[0067] Embodiment 2

[0068] In order to execute the method corresponding to the above embodiment 1 to achieve corresponding functions and technical effects, a hydrological forecasting system is provided below, including:

[0069] The data acquisition module is used to obtain the current observation data of the hydrological station and the observation data after the first preset time period; the observation data includes rainfall and evapotranspiration.

[0070] The forecast module is used to determine the runoff after the first preset period of time based on the observation data at the current moment and the observation data after the first preset period of time, using the hydrological forecast model after the first preset period of time; wherein the hydrological forecast model after the first preset period of time is obtained by calibrating the parameters of the hydrological forecast model at the current moment using the historical observation data of the second preset period before the current moment and the historical runoff; the hydrological forecast model after the first preset period of time is Among them, Calq k+1 is the runoff after the first preset period; k is the current time; is the response function of the model; x kis the model parameter after the first preset period of calibration, y 1:k is the observation data set consisting of rainfall, evapotranspiration and runoff before the current moment, S0 is the initial value of the model intermediate variable, P k+1 is the rainfall after the first preset period; E k+1 is the evapotranspiration after the first preset period; the initial value of the intermediate variable of the model is updated according to the thickness of the saturated soil layer and the water inlet and outlet of the confluence tank.

[0071] Embodiment 3

[0072] The present invention provides an electronic device, comprising: a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the hydrological forecasting method of the first embodiment.

[0073] Embodiment 4

[0074] The present invention 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 hydrological forecasting method of the first embodiment is implemented.

[0075] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0076] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A hydrological forecasting method, characterized in that: include: Obtaining the current observation data of the hydrological station and the observation data after the first preset period; Observational data include rainfall and evapotranspiration; According to the observation data at the current moment and the observation data after the first preset period, the runoff after the first preset period is determined by using the hydrological forecast model after the first preset period; wherein the hydrological forecast model after the first preset period is obtained by calibrating the parameters of the hydrological forecast model at the current moment using the historical observation data of the second preset period before the current moment and the historical runoff; the hydrological forecast model after the first preset period is Among them, Calq k+1 is the runoff after the first preset period; k is the current time; is the response function of the model; x k y is the model parameter after the first preset period after calibration; 1:k is the observation data set consisting of rainfall, evapotranspiration and runoff before the current moment; S0 is the initial value of the model intermediate variable; P k+1 is the rainfall after the first preset period; E k+1 is the evapotranspiration after the first preset period; the initial value of the intermediate variable of the model is updated according to the thickness of the saturated soil layer and the water inlet and outlet of the confluence tank.

2. The hydrological forecasting method according to claim 1, characterized in that: The parameters of the hydrological forecast model at the current moment are calibrated using the historical observation data of the second preset period before the current moment and the historical runoff, including: Obtain historical observation data and historical runoff for a second preset period before the current moment to form a retrospective observation particle set; Using a nonlinear forgetting function to assign weights to the back-tracking observation particle set, to obtain a weighted back-tracking observation particle set; According to the weighted retrospective observation particle set, the parameters of the hydrological forecast model at the current moment are calibrated to obtain the model parameters after the first preset period after calibration; The hydrological forecast model after the first preset time period is determined according to the calibrated model parameters after the first preset time period.

3. The hydrological forecasting method according to claim 2, characterized in that: Obtain the historical observation data and historical runoff of the second preset period before the current moment to form a retrospective observation particle set, which specifically includes: From the current moment forward, historical observation data and historical runoff within a second preset period are selected to obtain a retrospective observation particle set.

4. The hydrological forecasting method according to claim 2, characterized in that: The back-tracking observation particle set is weighted by using a nonlinear forgetting function to obtain a weighted back-tracking observation particle set, which specifically includes: Using the formula w i =e -δ(L-i) , determine the weight of each data in the backtracking observation particle set; where w i is the weight of the data at the i-th moment in the second preset period of time in the retrospective observation particle set, L is the end time of the second preset period, δ is the forgetting coefficient, and Li is the interval between the i-th moment and the end time of the second preset period; According to the weight of each data and the back-tracking observation particle set, a weighted back-tracking observation particle set is determined.

5. The hydrological forecasting method according to claim 2, characterized in that: According to the weighted retrospective observation particle set, the parameters of the hydrological forecast model at the current moment are calibrated to obtain the model parameters after the first preset period after calibration, specifically including: Using the formula x k = r(y w k-L:k ,x0,S k-L-1 , w i ), determine the model parameters after the first preset period of time after calibration; where x k is the model parameter after the first preset period of time after calibration, r(·) is the model parameter calibration function, y w k-L:k is the weighted back-tracking observation particle set, x0 is the optimal model parameter, S k-L-1 is the intermediate parameter of the model at the kL-1 moment, w i is the weight of the data at the i-th moment within the second preset time period in the back-tracking observation particle set; the optimal model parameters are obtained by calibrating the parameters of the initial hydrological forecast model using the observation data set.

6. The hydrological forecasting method according to claim 5, characterized in that: The process of determining the optimal model parameters includes: Acquire an observation data set; the observation data set includes rainfall, evapotranspiration and runoff in the third preset period before the current moment; The observation data set is used to calibrate the parameters of the initial hydrological forecast model to obtain the optimal model parameters.

7. A hydrological forecasting system, characterized in that: include: A data acquisition module, used to acquire the current observation data of the hydrological station and the observation data after the first preset period; Observational data include rainfall and evapotranspiration; The forecast module is used to determine the runoff after the first preset period of time based on the observation data at the current moment and the observation data after the first preset period of time, using the hydrological forecast model after the first preset period of time; wherein the hydrological forecast model after the first preset period of time is obtained by calibrating the parameters of the hydrological forecast model at the current moment using the historical observation data of the second preset period before the current moment and the historical runoff; the hydrological forecast model after the first preset period of time is Among them, Calq k+1 is the runoff after the first preset period; k is the current time; is the response function of the model; x k is the model parameter after the first preset period of calibration, y 1:k is the observation data set consisting of rainfall, evapotranspiration and runoff before the current moment, S0 is the initial value of the model intermediate variable, P k+1 is the rainfall after the first preset period; E k+1 is the evapotranspiration after the first preset period; the initial value of the intermediate variable of the model is updated according to the thickness of the saturated soil layer and the water inlet and outlet of the confluence tank.

8. An electronic device, characterized in that: include: A memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the hydrological forecasting method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the hydrological forecasting method according to any one of claims 1 to 6 is implemented.

Images