CN113592143A - Rainfall forecast-based medium and long term runoff ensemble forecasting method and system - Google Patents
Rainfall forecast-based medium and long term runoff ensemble forecasting method and system Download PDFInfo
- Publication number
- CN113592143A CN113592143A CN202110719851.1A CN202110719851A CN113592143A CN 113592143 A CN113592143 A CN 113592143A CN 202110719851 A CN202110719851 A CN 202110719851A CN 113592143 A CN113592143 A CN 113592143A
- Authority
- CN
- China
- Prior art keywords
- runoff
- forecast
- sample
- data
- rainfall
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention provides a medium and long term runoff ensemble forecasting method and system based on rainfall forecast, wherein the method comprises the following steps: constructing runoff aggregate forecast data of a target area according to a forecast period, historical runoff data and forecast factors corresponding to the historical runoff data, wherein rainfall factors in the forecast factors are obtained by correcting the rainfall forecast data; and inputting the runoff ensemble forecasting data into a trained runoff ensemble forecasting model to obtain a medium-term and long-term runoff ensemble forecasting result of the target area, wherein the trained runoff ensemble forecasting model is obtained by training a plurality of data-driven models through sample runoff ensemble forecasting data. The method and the system improve the accuracy and reliability of medium-and-long-term runoff forecasting and solve the problem that the traditional medium-and-long-term runoff forecasting method can only generate a deterministic runoff forecasting result.
Description
Technical Field
The invention relates to the technical field of medium-and-long-term runoff forecasting, in particular to a medium-and-long-term runoff ensemble forecasting method and system based on rainfall forecasting.
Background
In recent years, global climate is significantly affected due to continuous increase of carbon dioxide greenhouse gas emission caused by human activities, global earth surface temperature shows a warming trend as a whole, and precipitation amounts in different areas show different changing trends. The changing climate conditions bring more uncertainty to the future water resource amount, and the demand of human activities for water resources is continuously increased, which brings great pressure to the comprehensive development and utilization of water resources. Under the condition, the medium-and-long-term runoff forecast capable of providing future runoff information can provide decision support for comprehensive utilization of water resources, and comprehensive utilization efficiency of the water resources is improved, so that more and more attention is paid.
In the existing medium-long term runoff forecasting, two difficulties mainly exist. On one hand, because reliable weather forecast results (such as reliable medium-and-long-term rainfall forecast) are lacked and the time scale of the medium-and-long-term runoff forecast is mostly a month scale, a physical model based on a rainfall-runoff relation is difficult to establish; on the other hand, the river runoff sequence has obvious random and nonlinear characteristics under the influence of a plurality of factors such as climate, vegetation, terrain, human activities and the like. For the two problems, in the current research, a data driving model is usually used to establish a statistical-based relationship between self-correlation factors such as runoff and early-stage runoff and remote correlation factors such as climate factors. The common data driving model includes an Artificial Neural Network (ANN), Support Vector Regression (SVR), random forest and the like, and the common climate factors include an atmospheric circulation factor, a sea temperature factor and the like. In this way, reliable medium and long term runoff forecasting results can be generated to a certain extent.
However, in this method, since all the adopted forecasting factors are determined values, a deterministic forecasting result is usually generated, uncertainty of future runoff cannot be reflected, and there is a great risk in practical application. On the other hand, since the critical information of rainfall is lacked in the modeling process, the accuracy of the forecast result is affected by the lack of information in the forecast. Therefore, there is a need for a method and a system for forecasting runoff collection in medium and long periods based on rainfall forecast to solve the above problems.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a medium-long term runoff ensemble forecasting method and system based on rainfall forecasting.
The invention provides a medium and long term runoff ensemble forecasting method based on rainfall forecast, which comprises the following steps:
constructing runoff aggregate forecast data of a target area according to a forecast period, historical runoff data and forecast factors corresponding to the historical runoff data, wherein rainfall factors in the forecast factors are obtained by correcting the rainfall forecast data;
and inputting the runoff ensemble forecasting data into a trained runoff ensemble forecasting model to obtain a medium-term and long-term runoff ensemble forecasting result of the target area, wherein the trained runoff ensemble forecasting model is obtained by training a plurality of data-driven models through sample runoff ensemble forecasting data.
According to the medium-and-long-term runoff ensemble forecasting method based on rainfall forecast, provided by the invention, the trained runoff ensemble forecasting model is obtained through the following steps:
based on a preset forecast period and a lag period of each sample forecasting factor, constructing a sample forecast amount and a sample forecasting factor set corresponding to the sample forecast amount through sample runoff data and the sample forecasting factors; the sample forecast factor set at least comprises sample rainfall factors, and the sample rainfall factors are measured rainfall data corresponding to the sample runoff data at the same time period;
inputting the sample forecast quantity and the sample forecast factor into different data-driven models respectively for training to obtain a plurality of runoff forecast submodels;
acquiring sample rainfall forecast data of the sample runoff data at the same period, and inputting the sample rainfall forecast data into a rainfall data correction model to obtain corrected sample rainfall forecast data; the corrected sample forecast rainfall data comprises a plurality of rainfall forecast values in each month;
based on the corrected sample forecast rainfall data, replacing rainfall measured data of a corresponding month in the sample forecast factor set to obtain a replaced sample forecast factor set, and respectively inputting the replaced sample forecast factor set into each runoff forecast sub-model to obtain a sample runoff aggregate forecast result output by each runoff forecast sub-model;
and determining the model weight of each runoff forecasting sub-model according to the sample runoff ensemble forecasting result and the corresponding sample runoff ensemble actual measurement result to obtain the trained runoff ensemble forecasting model.
According to the medium-and-long-term runoff ensemble forecasting method based on rainfall forecast, provided by the invention, the rainfall data correction model is obtained through the following steps:
acquiring sample rainfall forecast data and sample rainfall measured data corresponding to the sample rainfall forecast data at the same time;
and inputting the sample rainfall forecast data and the sample rainfall measured data into a Bayes combined probability distribution model for training to obtain a rainfall data correction model.
According to the invention, the method for forecasting the medium and long term runoff ensemble based on rainfall forecast, which is provided by the invention, determines the model weight of each runoff forecasting sub-model according to the sample runoff ensemble forecasting result and the corresponding sample runoff ensemble actual measurement result to obtain the trained runoff ensemble forecasting model, comprises the following steps:
acquiring a corresponding sample runoff set actual measurement result according to the sample runoff set forecasting result output by each runoff forecasting sub-model;
obtaining an evaluation index of each runoff forecasting sub-model according to probability density estimation corresponding to the actual measurement result of each sample runoff set by a Bayesian model averaging method;
and obtaining the model weight of each runoff forecasting sub-model according to the evaluation indexes, and fusing all the runoff forecasting sub-models according to the model weights to obtain the trained runoff ensemble forecasting model.
According to the method for forecasting the medium-and-long-term runoff ensemble provided by the invention, before the runoff ensemble forecasting data of the target area is constructed according to the forecast period, the historical runoff data and the forecasting factor corresponding to the historical runoff data, the method further comprises the following steps:
and preprocessing the historical runoff data and the corrected rainfall forecast data based on log-sinh transformation to obtain target historical runoff data and target rainfall forecast data.
The invention also provides a medium-and-long-term runoff ensemble forecasting system based on rainfall forecast, which comprises:
the system comprises a runoff forecast data construction module, a rainfall forecast data correction module and a rainfall forecast data correction module, wherein the runoff forecast data construction module is used for constructing runoff aggregate forecast data of a target area according to a forecast period, historical runoff data and a forecast factor corresponding to the historical runoff data, and a rainfall factor in the forecast factor is obtained by correcting the rainfall forecast data;
and the medium-and-long-term runoff ensemble forecasting module is used for inputting the runoff ensemble forecasting data into a trained runoff ensemble forecasting model to obtain a medium-and-long-term runoff ensemble forecasting result of the target area, wherein the trained runoff ensemble forecasting model is obtained by training a plurality of data-driven models through sample runoff ensemble forecasting data.
According to the invention, the system for the medium and long term runoff ensemble forecasting based on rainfall forecast further comprises:
the sample construction module is used for constructing a sample forecast amount and a sample forecast factor set corresponding to the sample forecast amount through sample runoff data and the sample forecast factors based on a preset forecast period and a lag period of each sample forecast factor; the sample forecast factor set at least comprises sample rainfall factors, and the sample rainfall factors are measured rainfall data corresponding to the sample runoff data at the same time period;
the training module is used for inputting the sample forecast amount and the sample forecast factor into different data driving models respectively for training to obtain a plurality of runoff forecast submodels;
the rainfall forecast data correction module is used for acquiring sample rainfall forecast data of the sample runoff data in the same period, inputting the sample rainfall forecast data into the rainfall data correction model and obtaining corrected sample rainfall forecast data; the corrected sample forecast rainfall data comprises a plurality of rainfall forecast values in each month;
the rainfall forecast data replacement module is used for replacing rainfall measured data of a corresponding month in the sample forecast factor set based on the corrected sample forecast rainfall data to obtain a replaced sample forecast factor set, and inputting the replaced sample forecast factor set into each runoff forecast sub-model respectively to obtain a sample runoff set forecast result output by each runoff forecast sub-model;
and the model weight acquisition module is used for determining the model weight of each runoff forecasting submodel according to the sample runoff ensemble forecasting result and the corresponding sample runoff ensemble actual measurement result to obtain the trained runoff ensemble forecasting model.
According to the invention, the model weight acquisition module comprises:
the processing unit is used for acquiring a corresponding sample runoff aggregate actual measurement result according to the sample runoff aggregate forecasting result output by each runoff forecasting submodel;
the evaluation unit is used for obtaining the evaluation index of each runoff forecasting submodel according to the probability density estimation corresponding to the actual measurement result of each sample runoff set by using a Bayesian model averaging method;
and the multi-model fusion unit is used for acquiring the model weight of each runoff forecasting sub-model according to the evaluation indexes, and fusing all the runoff forecasting sub-models according to the model weights to obtain the trained runoff ensemble forecasting model.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the steps of the rainfall forecast-based medium-and-long-term runoff ensemble forecasting method.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method for forecasting runoff ensemble for medium and long periods based on rainfall forecasts as described in any of the above.
According to the rainfall forecast-based medium-and-long-term runoff ensemble forecasting method and system, rainfall forecast products are effectively applied to medium-and-long-term runoff forecasting, effective information of the rainfall forecast is reserved, influence of rainfall forecast errors is reduced, the accuracy and reliability of the medium-and-long-term runoff forecasting can be improved through information gain brought by the rainfall forecast, and the problem that a traditional medium-and-long-term runoff forecasting method can only generate a deterministic runoff forecasting result is solved.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of a medium-and-long-term runoff ensemble forecasting method based on rainfall forecast according to the present invention;
fig. 2 is a schematic overall flow chart of the medium-and long-term runoff forecasting provided by the invention;
fig. 3 is a schematic diagram illustrating the effect of the medium-and-long term runoff ensemble forecasting result based on a certain hydrological site according to the present invention;
fig. 4 is a schematic structural diagram of a rainfall forecast-based medium-and-long-term runoff ensemble forecasting system provided by the invention;
fig. 5 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the current medium-long term runoff forecasting, a forecasting model based on statistics is usually established through the remote correlation relationship of runoff, climate factors and the like. However, in this case, only deterministic runoff forecast can be generated, uncertainty of future runoff cannot be reflected, and the forecast accuracy of the model is affected by the absence of rainfall information.
In recent years, the forecasting effect of medium-and-long-term rainfall forecasting products is continuously improved, and more possibility is brought to medium-and-long-term runoff forecasting. Although the reliability and precision of the initial medium-and-long-term rainfall forecast product are still poor, and the forecast rainfall has larger deviation compared with the actually measured rainfall, the forecast rainfall can be effectively corrected through error correction, so that a relatively more reliable rainfall forecast result is generated. Thus, the corrected forecasted rainfall can be considered as an input for medium and long term runoff forecasting to introduce this critical information of rainfall.
In addition, there are studies that have shown that an aggregate forecast of future rainfall can be generated by error correction, which can effectively reflect the uncertainty of future rainfall. Therefore, each set value in the corrected rainfall ensemble forecasting data is used as the input of the medium-and-long-term runoff forecasting, uncertain factors can be introduced into the forecasting, a corresponding runoff ensemble forecasting result is generated, and the uncertainty of future runoff is reflected. Although theoretically, the introduction of rainfall forecast products into medium-and long-term runoff forecast can bring information gain and generate corresponding runoff aggregate forecast, rainfall forecast errors can also cause greater runoff forecast errors, and in such a case, how to introduce the rainfall forecast products into the medium-and long-term runoff forecast, retain effective information of the rainfall forecast products and reduce errors brought by the rainfall forecast products still remains a problem. Therefore, the invention establishes a medium-and-long-term runoff ensemble forecasting method frame introduced with rainfall forecasting products, effectively maintains information gain brought by rainfall forecasting, reduces the influence of rainfall forecasting errors as much as possible, improves the precision of runoff forecasting and generates reliable runoff ensemble forecasting results to reflect the uncertainty of future runoff. In the method framework, corrected rainfall ensemble forecasting results are used as input, corresponding runoff ensemble forecasting is generated through various rainfall-runoff models, and finally ensemble forecasting results of different models are fused by a model fusion method, so that more reliable and accurate runoff ensemble forecasting results are generated.
Fig. 1 is a schematic flow chart of a rainfall forecast-based medium-and-long-term runoff ensemble forecasting method, as shown in fig. 1, the invention provides a rainfall forecast-based medium-and-long-term runoff ensemble forecasting method, which includes:
In the invention, firstly, the month needing runoff forecasting is determined based on the forecast period, and thus, the data input for forecasting the medium and long term runoff is constructed according to the selected forecasting factor and the data processing method. The invention explains the aggregate forecast of the runoff from 1 month in 2010 to 12 months in 2014 of a certain hydrological site, and the runoff from 1 month in 2010 to 12 months in 2014 is forecast sequentially according to the aggregate forecast data from 12 months in 2009 to 11 months in 2014 of the hydrological site (namely, the aggregate forecast data at 12 months in 2009 is used for forecasting the runoff data at 1 month in 2010, and the like). Since the forecasting steps of different months are similar, the invention takes forecasting performed in 12 months in 2009 as an example to explain the process of generating the collective forecasting result, specifically, to forecast the runoff in 1 month in 2010 in 12 months in 2009, the first step is to forecast the runoff in 1 month in 2010Rainfall data for month 1 2010 (as one of the forecast factors), runoff for month 1 2010, runoff for month 12 2009 (i.e., historical runoff data), and other corresponding climate factors (e.g., radiation factor, circulation factor, geographic factor, etc.) are obtained, and in addition to the rainfall data, other factors may be obtained for month 12 2009. Further, for historical runoff data, a log-sinh method can be adopted for preprocessing the historical runoff data; for the climate monitoring index in the forecasting factor, a monthly standardized processing method is adopted; for month data, it is normalized to between-1 and 1. After pre-processing the data, an input x may be constructednewWherein the rainfall data is temporarily empty.
Further, the corresponding rainfall forecasting value is obtained, and after the rainfall forecasting value is corrected, a corrected rainfall collective forecasting result is obtained. In the present invention, a rainfall forecast value P 'for 1 month in 2010 generated at 12 months in 2009 is acquired'newAnd obtaining a rainfall aggregate forecast result g (P 'of 1 month in 2010 after correction'new) The rainfall forecast data comprises 1000 rainfall forecast data.
Preferably, on the basis of the above embodiment, before the building of the runoff aggregate forecast data of the target area according to the forecast period, the historical runoff data and the forecast factor corresponding to the historical runoff data, the method further includes:
and preprocessing the historical runoff data and the corrected rainfall forecast data based on log-sinh transformation to obtain target historical runoff data and target rainfall forecast data.
In the invention, data in the rainfall ensemble forecasting result is preprocessed, and the rainfall ensemble forecasting result g (P ') is subjected to log-sinh transformation'new) The data in (1) is converted, and the converted data is recorded as g '(P'new). Similarly, historical runoff data was also processed using the log-sinh method.
Further, g '(P'new) Is placed at xnewThe position of the middle rainfall factor corresponding to the month is obtained to obtain an input vector xnew,jThereby obtaining a runoff set based on rainfall forecast dataForecast data xnew,jIn the present invention, the runoff ensemble prediction data xnew,jThe rainfall forecast data in 1 month 2010 includes 1000 forecast values.
In the invention, any runoff ensemble forecast data x is combinednew,jInputting the prediction into a trained runoff ensemble prediction model to generate a predicted runoff value fm(xnew,j) (ii) a Further, in the invention, the runoff forecast values corresponding to 1000 rainfall forecast values are aggregated to form a runoff aggregate forecast result containing 1000 members, and the result is recorded as fm(xnew,g′(P′new))。
In the invention, the trained runoff ensemble forecasting model is obtained by training and constructing a plurality of data-driven models. Different models are adopted, runoff ensemble forecasting results corresponding to the different models can be obtained, and the runoff ensemble forecasting results of the models are fused through the weight of the models. It should be noted that the runoff aggregate forecasting results are all results in a data space after log-sinh conversion, and the aggregated forecasting results after fusion need to be converted into actual runoff aggregate forecasting results.
According to the rainfall forecast-based medium and long-term runoff ensemble forecasting method, a rainfall forecast product is effectively applied to medium and long-term runoff forecasting, effective information of the rainfall forecasting is reserved, the influence of rainfall forecast errors is reduced, the medium and long-term runoff forecasting precision and reliability can be improved through information gain brought by the rainfall forecasting, and the problem that a traditional medium and long-term runoff forecasting method can only generate a deterministic runoff forecasting result is solved.
On the basis of the embodiment, the trained runoff ensemble forecasting model is obtained through the following steps:
step 201, constructing a sample forecast amount and a sample forecast factor set corresponding to the sample forecast amount through sample runoff data and the sample forecast factors based on a preset forecast period and a lag period of each sample forecast factor; the sample forecast factor set at least comprises sample rainfall factors, and the sample rainfall factors are measured rainfall data corresponding to the sample runoff data at the same time.
In the invention, the original sample data is firstly obtained and preprocessed. In one embodiment, the monthly runoff data (i.e. the monthly sample runoff data) of a certain hydrological site from 1 month to 12 months in 1959 and the sample rainfall actual measurement data in the same period are collected, and meanwhile, month information corresponding to the monthly runoff data is also naturally obtained; in addition, 130 weather monitoring indexes (as sample forecasting factors) published by the national climate center at the same time are collected, including 88 atmospheric circulation indexes, 26 sea temperature indexes and 16 other indexes (such as sun blackson index). Processing the sample runoff data and the sample rainfall actual measurement data by adopting a log-sinh method; for 130 climate monitoring indexes, a treatment method of monthly standardization is adopted; and normalizing the month data to be between-1 and 1, and recording the processed sample runoff data, sample rainfall actual measurement data, month information and climate factor data as Q, R, M and F respectively.
Further, by setting a forecast period and a lag period of influence of different factors on runoff, a sample forecast amount and a corresponding candidate forecast factor set are constructed. Specifically, the month information and the sample measured rainfall data both adopt data in the same period as the sample runoff data, the runoff factor in the early stage adopts the latest runoff data which is currently available, and the climate factor data adopts the climate factor data in the current stage and the early stage. In the present invention, the forecast period is set to 1 month, the maximum lag period of the climate factor data is set to 12, and when the runoff Q of the t month (1 month in 1959 is taken as 1 month)tAt first, prepareSelecting forecasting factors including rainfall measured data R of contemporaneous samplestDate of the same month MtCurrent sample runoff data Qt-1And early climate factor data [ Ft-1,Ft-2,…,Ft-12]. Based on the above, a candidate forecast factor set and a corresponding sample forecast amount can be constructed, wherein the corresponding sample forecast amount QtIs [ F ] as the candidate predictor sett-1,Ft-2,…,Ft-12,Qt-1,Mt,Rt]. Considering the climate factor in the first 12 months as the candidate forecasting factor, t should be between 13 and 612, then 600 sample points can be obtained in total, namely the runoff from 1 month to 12 months in 1960 is forecasted respectively, the 600 sample points are ranked as 1 to 600, wherein the forecast amount y of the ith sample point is from 1 month to 600iRunoff Q of i +12 monthsi+1Alternative predictor ziIs corresponding to [ Fi+11,Fi+10,…,Fi,Qi+11,Mi+12,Ri+12]Then 600 yiCan form forecast quantity Y, 600 ziA set of candidate predictors Z can be constructed.
In the present invention, the rainfall factor is used as the selected forecasting factor, and a suitable forecasting factor is selected from other forecasting factors in the candidate forecasting factor set (the forecasting factor that can generate the maximum gain for forecasting is selected from other factors in the candidate forecasting factor set by using the factor selection method in the present invention), so as to form the forecasting factor set X. Specifically, by adopting a factor selection method based on partial mutual information, selecting forecasting factors from the candidate forecasting factor set Z, wherein the selected factors comprise rainfall factors (determined in the early stage), early runoff factors, months and partial climate factors, and selecting rows corresponding to the forecasting factors from Z to form a sample forecasting factor set X.
Step 202, inputting the sample forecast quantity and the sample forecast factor into different data-driven models respectively for training to obtain a plurality of runoff forecast submodels;
in the invention, different data-driven models are selected, and bases are respectively established based on sample prediction quantity Y and prediction factor XAnd (4) a runoff forecasting submodel of rainfall. The data driving model adopted by the invention is 3 data driving models of Support Vector Regression (SVR), Multi-Layer perceptron (MLP) and Long-Short Term Memory network (LSTM), thereby establishing the relation between the prediction quantity and the prediction factor. For any input vector xiThe fitting value of the output of the 3 models is fm(xi) M is 1,2, …, M is the total number of models selected, M is between 1 and 3 in the invention, and represents SVR, MLP and LSTM models respectively. By training the 3 models respectively, when a preset training condition is reached (for example, a preset training number is reached), a plurality of runoff forecast submodels are obtained.
Step 203, acquiring sample rainfall forecast data of the sample runoff data at the same period, and inputting the sample rainfall forecast data into a rainfall data correction model to obtain corrected sample rainfall forecast data; the corrected sample forecast rainfall data comprises a plurality of rainfall forecast values in each month.
On the basis of the above embodiment, before rainfall data correction is performed, a rainfall data correction model needs to be constructed first, and the rainfall data correction model is obtained through the following steps:
acquiring sample rainfall forecast data and sample rainfall measured data corresponding to the sample rainfall forecast data at the same time;
and inputting the sample rainfall forecast data and the sample rainfall measured data into a Bayes combined probability distribution model for training to obtain a rainfall data correction model.
In the invention, the forecast rainfall data P' with the forecast period of the drainage basin and the runoff forecast consistent is collected. In this embodiment, sample rainfall forecast data of 1 month forecast duration of the SEAS5 version of the European mid-Range Weather forecast center (ECMWF for short) is downloaded and interpolated to the drainage basin scale. The sample forecast rainfall data comprises forecast data generated from 1 month to 12 months in 1993, namely forecast rainfall from 2 months to 1 month in 2010 in 1993.
And further, taking sample rainfall forecast data and corresponding sample actually-measured rainfall data as input of a data correction method, and establishing a rainfall data correction model. In the present invention, a Bayesian Joint Probability distribution (BJP for short) model is adopted, and measured rainfall data of a sample from 2 months 1993 to 1 month 2010 and sample rainfall forecast data of 2 months 1993 to 1 month 2010 generated based on 1 month 1993 to 12 months 2009 are input (for example, the forecast rainfall of 2 months 1993 generated in 1 month 1993 in 1993 and the like). The model can be established according to a forecast rainfall value P'jGenerating a rainfall forecast result recorded as g (P'j) In the present invention, the set forecast result includes 1000 set forecast values, i.e., g (P'j) Comprising 1000 values.
And then, correcting all the originally forecasted rainfall data by using the rainfall data correction model obtained by training. In the invention, the forecast rainfall of 2 months to 1 month in 1993 is corrected, and the set forecast rainfall of 2 months to 1 month in 1993 (216 months in total, 1000 forecast values in each month) is obtained after correction and is recorded as g (P').
And 204, replacing the rainfall measured data of the corresponding month in the sample forecasting factor set based on the corrected sample forecast rainfall data to obtain a replaced sample forecasting factor set, and respectively inputting the replaced sample forecasting factor set into each runoff forecasting sub-model to obtain a sample runoff ensemble forecasting result output by each runoff forecasting sub-model.
In the invention, from the monthly runoff data from 1 month to 12 months in 1959 and the sample forecast rainfall data in the same period, a month with the sample forecast amount and the sample forecast rainfall data is selected, and the corresponding sample forecast rainfall data, the corrected rainfall ensemble forecast data, the sample forecast amount and the corresponding sample forecast factor are intercepted. In the present invention, the month with both forecast volume and forecast rainfall data is 2 months 1993 to 12 months 2009 (215 months total). The sample forecasted rainfall data for this period of time is the top 215 values of P ', denoted as P2 ', and its corresponding corrected rainfall ensemble forecast is denoted as g (P2 '). The sample prediction amount and the last 215 sample data with sample prediction factors of Y and X corresponding to the period of time are respectively marked as Y2 and X2.
Further, all the data in g (P2') are converted, resulting in converted data. In the invention, rainfall ensemble forecast data (namely forecast data of each sample) are converted by adopting log-sinh transformation, and the converted data are recorded as g '(P2'). Then, each forecast value in g '(P2') is used for replacing a corresponding rainfall factor in the forecast factor X2, the replaced input vector is used as the input of each runoff forecast sub-model, a forecast runoff value is obtained, and then forecast runoff values corresponding to different forecast rainfall values in g '(P2') are collected, and a sample runoff collection forecast result is obtained.
And step 205, determining the model weight of each runoff forecasting submodel according to the sample runoff ensemble forecasting result and the corresponding sample runoff ensemble actual measurement result to obtain the trained runoff ensemble forecasting model.
In the present invention, g '(P2') includes 215 × 1000 forecast values, and the converted rainfall ensemble forecast data is g '(P2') for the ith month of 215 months as an example2′i) Contains 1000 values, corresponding to the ith sample X2 in X2i. In g '(P2'i) The j-th value in (b) replaces x2iObtaining a new input vector by the corresponding value of the medium rainfall factor
The fitting value can be obtained by taking the data as the input of the data driving model
In the invention, if i is 1 to 215, j is 1 to 1000, 1000 forecast runoff values corresponding to 215 months can be obtained, and thus the runoff ensemble forecast corresponding to each runoff forecast sub-model can be generated on the basis of rainfall ensemble forecast.
Preferably, on the basis of the above embodiment, the determining the model weight of each runoff forecasting sub-model according to the sample runoff ensemble forecasting result and the corresponding sample runoff ensemble actual measurement result to obtain the trained runoff ensemble forecasting model includes:
acquiring a corresponding sample runoff set actual measurement result according to the sample runoff set forecasting result output by each runoff forecasting sub-model;
estimating the probability density corresponding to the actual measurement result of the sample runoff set by a Bayesian model averaging method to obtain the evaluation index of each runoff forecasting submodel;
and obtaining the model weight of each runoff forecasting sub-model according to the evaluation indexes, and fusing all the runoff forecasting sub-models according to the model weights to obtain the trained runoff ensemble forecasting model.
In the invention, a Bayesian Model Averaging (BMA for short) method is adopted to fuse the results of different models. In the BMA method, the corresponding sample runoff set actual measurement result is obtained according to the sample runoff set forecasting results of different models; and then, the probability density estimation based on the measured value is used as an evaluation index of the runoff ensemble forecasting result of each model. In the present invention, the larger the probability density estimation is, the more concentrated the prediction is near the measured value, and the more the weight of the model is. Specifically, when the method is fused by a BMA method, the forecasting results of different months adopt different weights, namely the weights corresponding to the SVR, MLP and LSTM models are different in 1-12 months, wherein the weights corresponding to 3 models in runoff forecasting of 1 month are 0.53, 0.34 and 0.13 respectively, and the sum of the model weights is 1. When fusion is carried out, corresponding number of runoff forecasting values are sampled from respective runoff ensemble forecasting results respectively according to the weight of each model, and a new aggregate runoff forecasting result is formed. For example, when forecasting for 1 month is performed, if the model weights are 0.53, 0.34 and 0.13, 530, 340 and 130 predicted values are sampled from the runoff ensemble forecasting of the SVR, MLP and LSTM models respectively, and a fused ensemble forecasting result is formed.
In the invention, a medium-and-long-term runoff ensemble forecasting method framework which takes the corrected rainfall ensemble forecasting result as input, generates runoff ensemble forecasting by various rainfall-runoff models and fuses different model ensemble forecasting results by a model fusion method is established, and rainfall forecasting products can be introduced into the medium-and-long-term runoff forecasting through the framework to generate accurate and reliable runoff ensemble forecasting results.
In an embodiment, fig. 2 is a schematic view of an overall flow of medium-and long-term runoff forecasting provided by the present invention, and as shown in fig. 2, in step S1, data processing and forecasting factor selection are first performed, so as to construct and obtain a sample forecast amount and a sample forecasting factor; then, in step S2, training different data-driven models based on the sample set obtained in the previous step to obtain multiple rainfall-runoff models, namely a runoff forecast submodel; in step S3, acquiring the corrected rainfall forecast data, and replacing a rainfall factor in the forecast factors; further, in step S4, determining the weight of each runoff forecasting submodel when performing model fusion based on the forecasting factor replacing rainfall forecasting data, so as to obtain a trained runoff ensemble forecasting model; finally, in step S5, the trained runoff ensemble forecasting model is applied to the actual medium-and-long-term runoff forecasting scenario.
The runoff ensemble forecasting model constructed by the method is used for carrying out ensemble forecasting on the runoff from 1 month in 2010 to 12 months in 2014 of the hydrological site in the embodiment to obtain the medium-long term runoff ensemble forecasting results from 1 month in 2010 to 12 months in 2014, fig. 3 is a schematic diagram of the effect of the medium-long term runoff ensemble forecasting results based on a certain hydrological site, as shown in fig. 3, a box line diagram is the distribution of the ensemble forecasting, and in each year, '+' is the outlier of the ensemble forecasting, it can be seen that about 90% of measured values fall in the ensemble forecasting results, and the forecasting results are high in reliability. On the other hand, the non-flood period and the flood period have different forecast characteristics. In the non-flood season, namely the months with smaller actual measuring runoff, the forecast interval range is also very small and basically consistent with the actual measuring value. In the flood season, namely the months with larger actual runoff, the uncertainty of the runoff is larger, so that the forecast interval range is larger to better reflect the forecast uncertainty.
In order to illustrate the improvement of the forecasting effect of the medium-long term runoff ensemble forecasting method adopted in the invention, the forecasting effect is evaluated from the quantitative perspective, and the forecasting effects under different conditions are compared. The evaluation index is a Continuous Probability ranking Score (CRPS for short) reflecting the overall ensemble prediction effect, and the calculation formula is as follows:
wherein, yiFor the measured value, N is the number of samples in the validation set, FiThe cumulative distribution function predicted for the set of ith sample points. In the invention, the smaller the CRPS value is, the better the forecasting effect is. In order to compare the merits of different models, the following CRPSS indexes are defined:
wherein, CRPSRefCRPS as an evaluation index for a reference modelNewFor the evaluation index of the new model, the CRPSS value reflects the improvement of the forecasting effect of the new model compared with the reference model.
Further, in order to illustrate the improvement of the medium-and-long-term runoff aggregate forecasting effect of the rainfall forecasting product, the corrected rainfall aggregate forecasting result in the embodiment is changed into the rainfall aggregate forecasting obtained by random sampling based on historical rainfall, and the rainfall aggregate forecasting result is used as the input of different data driving models, so that the medium-and-long-term runoff aggregate forecasting result in the case of no rainfall forecasting product can be obtained. The CRPS values under the two conditions and the forecast effect boost amplitudes (CRPSs) after the addition of the rainfall forecast product are shown in table 1:
TABLE 1
| Method | CRPSRef | CRPSP | CRPSS(％) |
| SVR | 0.446 | 0.378 | 15.4 |
| MLP | 0.451 | 0.391 | 13.3 |
| LSTM | 0.459 | 0.393 | 14.2 |
CRPS in Table 1RefIn the case of no rainfall forecast product (rainfall ensemble forecast obtained by random sampling based on historical data), the evaluation index of the forecast of 3 models in this example, CRPSPThe CRPSS is used for forecasting an evaluation index value when a rainfall forecast product exists (forecasting is carried out according to the specific implementation mode), and the forecasting effect of the introduced rainfall forecast product is improved. As can be seen from Table 1, the forecast results of different models are relatively small compared to the forecast product without rainfallThe forecasting effect is obviously improved when the rainfall forecast product exists, the obtained CRPSS value is between 13.3% and 15.4%, the runoff forecasting effect can be effectively improved by introducing the rainfall forecast product, and the method framework can play the beneficial role of the rainfall forecast product.
In order to illustrate the influence of model fusion on the runoff ensemble forecasting effect (namely the runoff ensemble forecasting effect promotion amplitude fused by the BMA method), the forecasting effects of different models are taken as reference, the forecasting effect promotion after fusion is calculated, and the result is shown in Table 2:
TABLE 2
| Reference method | CRPSP | CRPSBMA | CRPSS(％) |
| SVR | 0.378 | 2.1 | |
| MLP | 0.391 | 0.370 | 5.4 |
| LSTM | 0.393 | 5.9 |
CRPS in Table 2BMAThe fused CPRS value is obtained. As can be seen from table 2, compared with the runoff ensemble prediction generated by 3 single models, the prediction effect after fusion by the BMA method is further improved, and the CRPSS is between 2.1% and 5.9%, which indicates that the adopted model fusion method can effectively combine the advantages of different models to generate a better runoff prediction effect.
According to the comparative analysis, the rainfall forecast product can be effectively brought into medium-and-long-term runoff forecast through the medium-and-long-term runoff ensemble forecasting method based on rainfall forecast, and a reliable medium-and-long-term runoff ensemble forecasting result is generated. The introduction of the rainfall forecast product can bring more information gain, the beneficial information of the rainfall forecast product is effectively reserved, the accuracy of runoff forecast is improved, and the ensemble forecast is more reliable.
Fig. 4 is a schematic structural diagram of a system for forecasting the collection of medium-and long-term runoff based on rainfall forecast provided by the present invention, and as shown in fig. 4, the present invention provides a system for forecasting the collection of medium-and long-term runoff based on rainfall forecast, which includes a runoff forecast data construction module 401 and a medium-and long-term runoff collection forecast module 402, where the runoff forecast data construction module 401 is configured to construct runoff collection forecast data of a target area according to a forecast period, historical runoff data and forecast factors corresponding to the historical runoff data, and the rainfall factors in the forecast factors are obtained by correcting the rainfall forecast data; the medium-and-long-term runoff ensemble forecasting module 402 is configured to input the runoff ensemble forecasting data into a trained runoff ensemble forecasting model to obtain a medium-and-long-term runoff ensemble forecasting result of the target area, where the trained runoff ensemble forecasting model is obtained by training a plurality of data-driven models according to sample runoff ensemble forecasting data.
The rainfall forecast product is effectively applied to medium-and long-term runoff forecast, effective information of the rainfall forecast is reserved, influence of rainfall forecast errors is reduced, accuracy and reliability of the medium-and long-term runoff forecast can be improved through information gain brought by the rainfall forecast, and the problem that a traditional medium-and long-term runoff forecast method can only generate a deterministic runoff forecast result is solved.
On the basis of the embodiment, the system further comprises a sample construction module, a training module, a rainfall forecast data correction module, a rainfall forecast data replacement module and a model weight acquisition module, wherein the sample construction module is used for constructing a sample forecast amount and a sample forecast factor set corresponding to the sample forecast amount through sample runoff data and sample forecast factors based on a preset forecast period and a lag period of each sample forecast factor; the sample forecast factor set at least comprises sample rainfall factors, and the sample rainfall factors are measured rainfall data corresponding to the sample runoff data at the same time period; the training module is used for inputting the sample forecast amount and the sample forecast factor into different data driving models respectively for training to obtain a plurality of runoff forecast submodels; the rainfall forecast data correction module is used for acquiring sample rainfall forecast data of the sample runoff data in the same period, inputting the sample rainfall forecast data into the rainfall data correction model and obtaining corrected sample rainfall forecast data; the corrected sample forecast rainfall data comprises a plurality of rainfall forecast values in each month; the rainfall forecast data replacement module is used for replacing rainfall measured data of a corresponding month in the sample forecast factor set based on the corrected sample forecast rainfall data to obtain a replaced sample forecast factor set, and inputting the replaced sample forecast factor set into each runoff forecast sub-model respectively to obtain a sample runoff set forecast result output by each runoff forecast sub-model; and the model weight acquisition module is used for determining the model weight of each runoff forecasting submodel according to the sample runoff ensemble forecasting result and the corresponding sample runoff ensemble actual measurement result to obtain the trained runoff ensemble forecasting model.
On the basis of the embodiment, the model weight obtaining module comprises a processing unit, an evaluation unit and a multi-model fusion unit, wherein the processing unit is used for obtaining a corresponding sample runoff set actual measurement result according to the sample runoff set forecasting result output by each runoff forecasting sub-model; the evaluation unit is used for obtaining the evaluation index of each runoff forecasting submodel according to the probability density estimation corresponding to the actual measurement result of each sample runoff set by using a Bayesian model averaging method; the multi-model fusion unit is used for acquiring the model weight of each runoff forecasting sub-model according to the evaluation indexes, and fusing all the runoff forecasting sub-models according to the model weights to obtain the trained runoff ensemble forecasting model.
The system provided by the present invention is used for executing the above method embodiments, and for the specific processes and details, reference is made to the above embodiments, which are not described herein again.
Fig. 5 is a schematic structural diagram of an electronic device provided in the present invention, and as shown in fig. 5, the electronic device may include: a processor (processor)501, a communication interface (communication interface)502, a memory (memory)503 and a communication bus 504, wherein the processor 501, the communication interface 502 and the memory 503 are communicated with each other through the communication bus 504. The processor 501 may invoke logic instructions in the memory 803 to perform a method of medium and long term runoff ensemble forecasting based on rainfall forecasts, the method comprising: constructing runoff aggregate forecast data of a target area according to a forecast period, historical runoff data and forecast factors corresponding to the historical runoff data, wherein rainfall factors in the forecast factors are obtained by correcting the rainfall forecast data; and inputting the runoff ensemble forecasting data into a trained runoff ensemble forecasting model to obtain a medium-term and long-term runoff ensemble forecasting result of the target area, wherein the trained runoff ensemble forecasting model is obtained by training a plurality of data-driven models through sample runoff ensemble forecasting data.
In addition, the logic instructions in the memory 503 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
In another aspect, the present invention also provides a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, when the program instructions are executed by a computer, the computer being capable of executing the rainfall forecast-based medium-and long-term runoff ensemble forecasting method provided by the above methods, the method including: constructing runoff aggregate forecast data of a target area according to a forecast period, historical runoff data and forecast factors corresponding to the historical runoff data, wherein rainfall factors in the forecast factors are obtained by correcting the rainfall forecast data; and inputting the runoff ensemble forecasting data into a trained runoff ensemble forecasting model to obtain a medium-term and long-term runoff ensemble forecasting result of the target area, wherein the trained runoff ensemble forecasting model is obtained by training a plurality of data-driven models through sample runoff ensemble forecasting data.
In yet another aspect, the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is implemented by a processor to execute the method for forecasting the medium and long term runoff ensemble based on rainfall forecast provided in the foregoing embodiments, and the method includes: constructing runoff aggregate forecast data of a target area according to a forecast period, historical runoff data and forecast factors corresponding to the historical runoff data, wherein rainfall factors in the forecast factors are obtained by correcting the rainfall forecast data; and inputting the runoff ensemble forecasting data into a trained runoff ensemble forecasting model to obtain a medium-term and long-term runoff ensemble forecasting result of the target area, wherein the trained runoff ensemble forecasting model is obtained by training a plurality of data-driven models through sample runoff ensemble forecasting data.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A rainfall forecast-based medium and long term runoff ensemble forecasting method is characterized by comprising the following steps:
constructing runoff aggregate forecast data of a target area according to a forecast period, historical runoff data and forecast factors corresponding to the historical runoff data, wherein rainfall factors in the forecast factors are obtained by correcting the rainfall forecast data;
and inputting the runoff ensemble forecasting data into a trained runoff ensemble forecasting model to obtain a medium-term and long-term runoff ensemble forecasting result of the target area, wherein the trained runoff ensemble forecasting model is obtained by training a plurality of data-driven models through sample runoff ensemble forecasting data.
2. The rainfall forecast-based medium and long term runoff ensemble forecasting method according to claim 1, wherein the trained runoff ensemble forecasting model is obtained by the following steps:
based on a preset forecast period and a lag period of each sample forecasting factor, constructing a sample forecast amount and a sample forecasting factor set corresponding to the sample forecast amount through sample runoff data and the sample forecasting factors; the sample forecast factor set at least comprises sample rainfall factors, and the sample rainfall factors are measured rainfall data corresponding to the sample runoff data at the same time period;
inputting the sample forecast quantity and the sample forecast factor into different data-driven models respectively for training to obtain a plurality of runoff forecast submodels;
acquiring sample rainfall forecast data of the sample runoff data at the same period, and inputting the sample rainfall forecast data into a rainfall data correction model to obtain corrected sample rainfall forecast data; the corrected sample forecast rainfall data comprises a plurality of rainfall forecast values in each month;
based on the corrected sample forecast rainfall data, replacing rainfall measured data of a corresponding month in the sample forecast factor set to obtain a replaced sample forecast factor set, and respectively inputting the replaced sample forecast factor set into each runoff forecast sub-model to obtain a sample runoff aggregate forecast result output by each runoff forecast sub-model;
and determining the model weight of each runoff forecasting sub-model according to the sample runoff ensemble forecasting result and the corresponding sample runoff ensemble actual measurement result to obtain the trained runoff ensemble forecasting model.
3. The rainfall forecast-based medium and long term runoff ensemble forecasting method according to claim 2, wherein the rainfall data correction model is obtained by the following steps:
acquiring sample rainfall forecast data and sample rainfall measured data corresponding to the sample rainfall forecast data at the same time;
and inputting the sample rainfall forecast data and the sample rainfall measured data into a Bayes combined probability distribution model for training to obtain a rainfall data correction model.
4. The rainfall forecast-based medium and long term runoff ensemble forecasting method according to claim 2, wherein the determining the model weight of each runoff forecasting submodel according to the sample runoff ensemble forecasting result and the corresponding sample runoff ensemble actual measurement result to obtain the trained runoff ensemble forecasting model comprises:
acquiring a corresponding sample runoff set actual measurement result according to the sample runoff set forecasting result output by each runoff forecasting sub-model;
obtaining an evaluation index of each runoff forecasting sub-model according to probability density estimation corresponding to the actual measurement result of each sample runoff set by a Bayesian model averaging method;
and obtaining the model weight of each runoff forecasting sub-model according to the evaluation indexes, and fusing all the runoff forecasting sub-models according to the model weights to obtain the trained runoff ensemble forecasting model.
5. The rainfall forecast-based medium-and long-term runoff ensemble forecasting method according to claim 1, wherein before the construction of the runoff ensemble forecasting data of the target area according to the forecast period, the historical runoff data and the forecasting factors corresponding to the historical runoff data, the method further comprises:
and preprocessing the historical runoff data and the corrected rainfall forecast data based on log-sinh transformation to obtain target historical runoff data and target rainfall forecast data.
6. A rainfall forecast-based medium and long term runoff ensemble forecasting system is characterized by comprising:
the system comprises a runoff forecast data construction module, a rainfall forecast data correction module and a rainfall forecast data correction module, wherein the runoff forecast data construction module is used for constructing runoff aggregate forecast data of a target area according to a forecast period, historical runoff data and a forecast factor corresponding to the historical runoff data, and a rainfall factor in the forecast factor is obtained by correcting the rainfall forecast data;
and the medium-and-long-term runoff ensemble forecasting module is used for inputting the runoff ensemble forecasting data into a trained runoff ensemble forecasting model to obtain a medium-and-long-term runoff ensemble forecasting result of the target area, wherein the trained runoff ensemble forecasting model is obtained by training a plurality of data-driven models through sample runoff ensemble forecasting data.
7. The system of claim 6, further comprising:
the sample construction module is used for constructing a sample forecast amount and a sample forecast factor set corresponding to the sample forecast amount through sample runoff data and the sample forecast factors based on a preset forecast period and a lag period of each sample forecast factor; the sample forecast factor set at least comprises sample rainfall factors, and the sample rainfall factors are measured rainfall data corresponding to the sample runoff data at the same time period;
the training module is used for inputting the sample forecast amount and the sample forecast factor into different data driving models respectively for training to obtain a plurality of runoff forecast submodels;
the rainfall forecast data correction module is used for acquiring sample rainfall forecast data of the sample runoff data in the same period, inputting the sample rainfall forecast data into the rainfall data correction model and obtaining corrected sample rainfall forecast data; the corrected sample forecast rainfall data comprises a plurality of rainfall forecast values in each month;
the rainfall forecast data replacement module is used for replacing rainfall measured data of a corresponding month in the sample forecast factor set based on the corrected sample forecast rainfall data to obtain a replaced sample forecast factor set, and inputting the replaced sample forecast factor set into each runoff forecast sub-model respectively to obtain a sample runoff set forecast result output by each runoff forecast sub-model;
and the model weight acquisition module is used for determining the model weight of each runoff forecasting submodel according to the sample runoff ensemble forecasting result and the corresponding sample runoff ensemble actual measurement result to obtain the trained runoff ensemble forecasting model.
8. The system of claim 6, wherein the model weight obtaining module comprises:
the processing unit is used for acquiring a corresponding sample runoff aggregate actual measurement result according to the sample runoff aggregate forecasting result output by each runoff forecasting submodel;
the evaluation unit is used for obtaining the evaluation index of each runoff forecasting submodel according to the probability density estimation corresponding to the actual measurement result of each sample runoff set by using a Bayesian model averaging method;
and the multi-model fusion unit is used for acquiring the model weight of each runoff forecasting sub-model according to the evaluation indexes, and fusing all the runoff forecasting sub-models according to the model weights to obtain the trained runoff ensemble forecasting model.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the computer program implements the steps of the rainfall forecast based medium and long term runoff ensemble forecasting method according to any one of claims 1 to 5.
10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the steps of the rainfall forecast based medium and long term runoff ensemble forecasting method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110719851.1A CN113592143A (en) | 2021-06-28 | 2021-06-28 | Rainfall forecast-based medium and long term runoff ensemble forecasting method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110719851.1A CN113592143A (en) | 2021-06-28 | 2021-06-28 | Rainfall forecast-based medium and long term runoff ensemble forecasting method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113592143A true CN113592143A (en) | 2021-11-02 |
Family
ID=78244841
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110719851.1A Pending CN113592143A (en) | 2021-06-28 | 2021-06-28 | Rainfall forecast-based medium and long term runoff ensemble forecasting method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113592143A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114648181A (en) * | 2022-05-24 | 2022-06-21 | 国能大渡河大数据服务有限公司 | Rainfall forecast correction method and system based on machine learning |
| CN114676882A (en) * | 2022-03-01 | 2022-06-28 | 河海大学 | Hydrological multi-model time-varying weight combined forecasting method |
| CN115392128A (en) * | 2022-09-07 | 2022-11-25 | 黑河水资源与生态保护研究中心 | Method for simulating basin runoff by utilizing space-time convolution LSTM network |
-
2021
- 2021-06-28 CN CN202110719851.1A patent/CN113592143A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114676882A (en) * | 2022-03-01 | 2022-06-28 | 河海大学 | Hydrological multi-model time-varying weight combined forecasting method |
| CN114676882B (en) * | 2022-03-01 | 2022-12-13 | 河海大学 | Hydrological multi-model time-varying weight combined forecasting method |
| CN114648181A (en) * | 2022-05-24 | 2022-06-21 | 国能大渡河大数据服务有限公司 | Rainfall forecast correction method and system based on machine learning |
| CN115392128A (en) * | 2022-09-07 | 2022-11-25 | 黑河水资源与生态保护研究中心 | Method for simulating basin runoff by utilizing space-time convolution LSTM network |
| CN115392128B (en) * | 2022-09-07 | 2024-02-13 | 黑河水资源与生态保护研究中心 | Method for simulating river basin runoff by utilizing space-time convolution LSTM network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113592143A (en) | Rainfall forecast-based medium and long term runoff ensemble forecasting method and system | |
| CN108761574B (en) | Rainfall estimation method based on multi-source information fusion | |
| Dong et al. | Novel stochastic methods to predict short-term solar radiation and photovoltaic power | |
| CN102183621B (en) | Aquaculture dissolved oxygen concentration online forecasting method and system | |
| Wang et al. | An evaluation of ECMWF SEAS5 seasonal climate forecasts for Australia using a new forecast calibration algorithm | |
| Matsumura et al. | Maize yield forecasting by linear regression and artificial neural networks in Jilin, China | |
| Stathopoulos et al. | Wind power prediction based on numerical and statistical models | |
| Bouzgou et al. | Minimum redundancy–maximum relevance with extreme learning machines for global solar radiation forecasting: Toward an optimized dimensionality reduction for solar time series | |
| CN114254561A (en) | Waterlogging prediction method, waterlogging prediction system and storage medium | |
| CN116451879B (en) | Drought risk prediction method and system and electronic equipment | |
| Pourhabib et al. | Short-term wind speed forecast using measurements from multiple turbines in a wind farm | |
| CN112285807B (en) | Meteorological information prediction method and device | |
| CN111538935B (en) | Fine precipitation fusion method, system, electronic equipment and storage medium based on terrain features and multi-source mode products | |
| CN113204061B (en) | Method and device for constructing lattice point wind speed correction model | |
| CN113592144A (en) | Medium-and-long-term runoff probability forecasting method and system | |
| CN115906954A (en) | Multivariate time sequence prediction method and device based on graph neural network | |
| Bozorg et al. | Bayesian bootstrapping in real-time probabilistic photovoltaic power forecasting | |
| CN115329930A (en) | Flood process probability forecasting method based on mixed deep learning model | |
| CN116663719A (en) | Flood risk prediction method, system and electronic equipment | |
| CN115327662A (en) | Statistical learning-based regional ionized layer TEC prediction method | |
| CN115759445A (en) | Machine learning and cloud model-based classified flood random forecasting method | |
| JP2020171217A (en) | Method, apparatus and program for predicting yield | |
| WO2020228568A1 (en) | Method for training power generation amount prediction model of photovoltaic power station, power generation amount prediction method and device of photovoltaic power station, training system, prediction system and storage medium | |
| CN117131758A (en) | Training method and device of irrigation water estimation model and irrigation water estimation method | |
| CN108108860A (en) | A kind of four steps coupling MEDIUM OR LONG RANGE HYDROLOGIC FORECAST METHOD |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |