CN112686481A - Runoff forecasting method and processor - Google Patents
Runoff forecasting method and processor Download PDFInfo
- Publication number
- CN112686481A CN112686481A CN202110303466.9A CN202110303466A CN112686481A CN 112686481 A CN112686481 A CN 112686481A CN 202110303466 A CN202110303466 A CN 202110303466A CN 112686481 A CN112686481 A CN 112686481A
- Authority
- CN
- China
- Prior art keywords
- runoff
- value
- hyper
- error
- error correction
- 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
- 238000013277 forecasting method Methods 0.000 title claims abstract description 18
- 238000012937 correction Methods 0.000 claims abstract description 73
- 238000000034 method Methods 0.000 claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 11
- 239000002245 particle Substances 0.000 claims description 73
- 230000003044 adaptive effect Effects 0.000 claims description 61
- 238000010606 normalization Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 9
- 238000003062 neural network model Methods 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 5
- 230000006403 short-term memory Effects 0.000 claims description 5
- 230000007774 longterm Effects 0.000 claims 1
- 230000006978 adaptation Effects 0.000 description 9
- 238000004088 simulation Methods 0.000 description 7
- 238000005457 optimization Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- 230000007787 long-term memory Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000015654 memory Effects 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 238000001556 precipitation Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 201000005665 thrombophilia Diseases 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a runoff forecasting method and a processor, wherein the method comprises the following steps: the method comprises the steps of collecting historical data of rainfall stations and hydrological stations in a drainage basin, wherein the historical data are daily rainfall and actually measured runoff observed by the stations, determining a runoff prediction sequence based on the historical data and a preset hydrological model, establishing an error correction model, training the error correction model based on the runoff prediction sequence and the actually measured runoff sequence, and forecasting the runoff based on the trained error correction model and the preset hydrological model, so that the error of runoff forecasting is reduced, and the accuracy of the runoff forecasting is improved.
Description
Technical Field
The present application relates to the technical field of runoff processing, and more particularly, to a runoff forecasting method and a processor.
Background
In a certain river basin, runoff forecasting is an effective flood control non-engineering measure, has important significance for guiding scientific dispatching and emergency disposal of reservoirs, ensuring flood control safety of power stations in the river basin, and improving social benefits and economic benefits, and accurate runoff forecasting can help to make comprehensive decisions on flood control, power generation, shipping and ecological regulation in advance.
So far, the error series of the radial flow rate predicted by the prior art and the measured radial flow rate are mostly unstable nonlinear sequences, and the reasons include:
1. observation errors influenced by natural conditions and observation technologies, 2 model structure errors incomplete and generalized to the actual hydrological process, and errors caused by model parameter errors, inaccurate precipitation in a forecast period and the like.
Therefore, how to reduce the error as much as possible and improve the accuracy of the runoff quantity forecast is a technical problem to be solved by those skilled in the art.
Disclosure of Invention
The invention discloses a runoff forecasting method, which is used for reducing the error of runoff forecasting and improving the accuracy of the runoff forecasting and comprises the following steps:
acquiring historical data of rainfall stations and hydrological stations in a drainage basin, wherein the historical data are the daily rainfall and the actually measured runoff observed by the stations;
determining a runoff prediction sequence based on historical data and a preset hydrological model;
establishing an error correction model, and training the error correction model based on the runoff prediction sequence and the measured runoff sequence;
and forecasting the runoff based on the trained error correction model and the preset hydrological model.
Preferably, before the error correction model is established, the method further comprises:
determining an error sequence according to the difference value of the actual measured runoff sequence and the runoff predicting sequence;
and carrying out normalization processing on the error sequence to obtain normalized data.
Preferably, the establishing an error correction model specifically includes:
optimizing the model parameters of the error correction model to be established;
determining all hyper-parameter sets of the error correction model to be established, wherein each group of hyper-parameter sets comprises a plurality of hyper-parameters, and each hyper-parameter has a corresponding value interval;
screening the hyper-parameter set to determine an optimal hyper-parameter set;
and establishing the error correction model based on the optimized model parameters and the optimal hyper-parameter set.
Preferably, the screening of the hyper-parameter set to determine an optimal hyper-parameter set specifically includes:
taking each hyper-parameter set as a particle, and taking a preset number of particles as a particle swarm;
initializing the particle swarm and iterating the particles;
determining the adaptive value of the particle at each iteration according to the fitness function;
determining an optimal adaptive value based on a particle swarm algorithm;
and determining the optimal hyper-parameter set according to the optimal adaptive value.
Preferably, determining the optimal hyper-parameter set according to the optimal adaptive value specifically includes:
judging whether the optimal adaptive value of each particle is higher than the global optimal adaptive value;
if the optimal adaptive value is higher than the global optimal adaptive value, taking the optimal adaptive value as the global optimal adaptive value;
and determining the optimal hyper-parameter set according to the global optimal adaptive value.
Preferably, training the error correction model specifically includes:
importing the normalized data into an error correction model;
and training the error correction model by using the normalized data and determining an error series relation.
Preferably, the runoff volume forecast is performed, specifically:
determining an error value according to the error series relation;
determining an initial runoff forecast value according to the preset hydrological model;
and forecasting the runoff quantity based on the error value and the initial runoff quantity forecast value.
Preferably, the radial flow rate is forecasted based on the error value and the initial radial flow rate forecast value, specifically, the radial flow rate to be forecasted is obtained by adding the error value and the initial radial flow rate forecast value.
Preferably, the error correction model is a long-short term memory neural network model.
Correspondingly, the invention also discloses a processor, and the processor executes the runoff forecasting method when working.
Compared with the prior art, the invention has the following beneficial effects:
the invention discloses a runoff forecasting method and a processor, wherein the method comprises the following steps: the method comprises the steps of collecting historical data of rainfall stations and hydrological stations in a drainage basin, wherein the historical data are daily rainfall and actually measured runoff observed by the stations, determining a runoff prediction sequence based on the historical data and a preset hydrological model, establishing an error correction model, training the error correction model based on the runoff prediction sequence and the actually measured runoff sequence, and forecasting the runoff based on the trained error correction model and the preset hydrological model, so that the error of runoff forecasting is reduced, and the accuracy of the runoff forecasting is improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flow chart illustrating a runoff forecasting method according to a preferred embodiment of the present disclosure;
fig. 2 is a schematic flow chart illustrating a runoff forecasting method according to another embodiment of the present disclosure.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. 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 application.
As described in the background, runoff forecasting is a very effective non-engineering measure for flood control, and accurate runoff forecasting may help to make comprehensive decisions on flood control, power generation, shipping, and ecological regulation in advance.
However, in the prior art, when the related hydrological model is used for forecasting the radial flow, the error is large, and the radial flow forecasting is not accurate enough.
In order to solve the above technical problem, in a preferred embodiment of the present application, a runoff forecasting method is provided, as shown in fig. 1, the method includes:
step S101, collecting historical data of rainfall stations and hydrological stations in a drainage basin, wherein the historical data are rainfall amount and measured runoff amount observed by stations each day.
Specifically, historical data of all rainfall stations and hydrological stations in the drainage basin are collected, the historical data are rainfall and measured runoff of all stations in the drainage basin in a preset time period, and it should be noted that in order to guarantee model accuracy, the preset time period should be selected as long as possible, namely the historical data is as long as possible, and the historical data can be all the historical data of the rainfall stations and the hydrological stations since the rainfall stations and the hydrological stations are built.
In addition, the rainfall and the actual diameter measuring flow are collected through a rainfall station and a hydrological station, and the rainfall and the actual diameter measuring flow can be collected through a plurality of devices such as rain gauges, flow meters, radar level gauges and radar flow meters which are uniformly distributed in a flow area.
And S102, determining a runoff prediction sequence based on the historical data and a preset hydrological model.
Specifically, daily precipitation and flow in historical data are led into a preset hydrological model for runoff prediction, that is, the historical data are led into the preset hydrological model for simulation, and a runoff prediction sequence is obtained, the historical data can be divided into a rate period and a verification period, wherein the preset hydrological model can be a Xinanjiang model or other models capable of forecasting runoff, no specific limitation is made here, and a person skilled in the art can flexibly select a model capable of forecasting the runoff.
In order to reduce the error of the runoff volume prediction, in an embodiment of the present application, the determining a runoff volume prediction sequence based on the historical data and a preset hydrological model further includes:
determining an error sequence according to the actual runoff sequence and the runoff prediction sequence;
and carrying out normalization processing on the error sequence to obtain normalized data.
Specifically, the difference between the measured runoff sequence and the runoff prediction sequence determined by using the preset hydrological model is obtained as an error sequence, and the error sequence is normalized by using a maximum and minimum normalization processing method to obtain normalized data, wherein the normalization processing can be specifically represented by the following formula one:
wherein X is the original error; xmaxIs the maximum value of the error; xminIs the minimum value of the error; xnormIs normalized data. The data normalization processing also has the functions of improving the convergence rate, preventing gradient explosion and improving the calculation precision.
It should be noted that the above formula for normalization is only a specific implementation manner in the preferred embodiment of the present application, and it is also only for normalizing the error sequence to obtain normalized data, and other formulas or methods that can be used for normalization are all within the scope of the present application.
And S103, establishing an error correction model, and training the error correction model based on the runoff prediction sequence and the measured runoff sequence.
Specifically, the error correction model of the present application may be a long-short term memory neural network model, and then the error correction model is trained based on the runoff prediction sequence and the actual runoff sequence.
In order to make the error correction model more accurate, in a preferred embodiment of the present application, the establishing the error correction model specifically includes:
optimizing the model parameters of the error correction model to be established;
determining all hyper-parameter sets of the error correction model to be established, wherein each group of hyper-parameter sets comprises a plurality of hyper-parameters, and each hyper-parameter has a corresponding value interval;
screening the hyper-parameter set to determine an optimal hyper-parameter set;
and establishing the error correction model based on the optimized model parameters and the optimal hyper-parameter set.
Specifically, model parameters and hyper-parameters are needed for constructing the error correction model, and for the accuracy of the error correction model, the model parameters and the hyper-parameters need to be optimized and screened to determine the optimal model parameters and the hyper-parameters.
Firstly, selecting Mean Absolute Error (MAE) as a loss function, and performing iterative optimization on model parameters in an error correction model by using an Adam optimizer, wherein the iteration times are preset times, so as to determine optimal model parameters, and the iteration times can be set by a person skilled in the art according to actual conditions, and the loss function is specifically represented by the following formula II:
wherein n is the number of samples; y iso(i) The normalized real error value; y isf(i) Calculated for the model.
It should be noted that, the optimization of the model parameters by using the maximum and minimum normalization method in the foregoing embodiments is only a specific implementation manner in the present application, and a person skilled in the art may flexibly select a manner of determining the optimal model parameters according to actual situations.
In addition, the error correction model in the present application is specifically a Long Short-Term Memory neural network model (LSTM), which needs three hyper-parameters, the hyper-parameters and their value ranges are cell _ size [2,10], batch _ sizes [64,512], and time _ step [1,3], the three hyper-parameters for establishing the error correction model can be set as a group of hyper-parameter sets, all the hyper-parameter sets are determined by traversing all combinations, and then all the determined hyper-parameter sets are screened.
In order to determine an optimal hyper-parameter set, in a preferred embodiment of the present application, the selecting the hyper-parameter set to determine the optimal hyper-parameter set specifically includes:
taking each hyper-parameter set as a particle, and taking a preset number of particles as a particle swarm;
initializing the particle swarm and iterating the particles;
determining the adaptive value of the particle at each iteration according to the fitness function;
determining an optimal adaptive value based on a particle swarm algorithm;
and determining the optimal hyper-parameter set according to the optimal adaptive value.
Specifically, the particle swarm algorithm may be executed to process each hyper-parameter set as one particle and 10 particles as one particle swarm, where X is X, a particle swarm composed of 10 particles in a three-dimensional target search spacekDenotes the position of each particle in the population at the kth iteration (K ≦ 20), i.e.WhereinRepresents a particle PiIn a position ofThe velocity of each particle in the corresponding particle group isThat is to sayIn position whereRepresents a particle PiOf speed, i.e.。
In addition, the adaptive value of each set of hyper-parameter sets is determined, and the adaptive value can be determined by a fitness function, namely the following formula three:
wherein n is the number of samples;the normalized real error value;calculating a value for the model;is the true error mean.
And determining the optimal adaptive value of each particle, namely the optimal adaptive value of each hyper-parameter set according to the particle swarm algorithm, thereby determining the optimal hyper-parameter set.
In order to improve the accuracy of the optimal hyper-parameter set, in a preferred embodiment of the present application, determining the optimal hyper-parameter set according to the optimal adaptive value specifically includes:
judging whether the optimal adaptive value of each particle is higher than the global optimal adaptive value;
if the optimal adaptive value is higher than the global optimal adaptive value, taking the optimal adaptive value as the global optimal adaptive value;
and determining the optimal hyper-parameter set according to the global optimal adaptive value.
Specifically, according to the formula three, obtaining a current adaptive value of each particle, that is, each group of hyper-parameter sets, and performing size judgment on the current adaptive value and a historical individual optimal adaptive value, wherein in the particle swarm algorithm, each iteration obtains a value of r square distance 1, the smaller the value is, the better the smallest value is, the smallest value can be regarded as the current adaptive value, after a plurality of iterations, the smallest adaptive value can be set as the historical optimal adaptive value, the smallest value is taken as a new historical optimal adaptive value, the historical optimal adaptive value and the global optimal adaptive value of each group of hyper-parameter sets are subjected to size judgment, if the historical optimal adaptive value of a certain group of hyper-parameter sets is superior to the global optimal adaptive value, that is, the smallest value of all adaptive values is set as the global optimal adaptive value, the historical optimal adaptive value corresponding to the hyper-parameter set is taken as the new global optimal adaptive value, it should be noted that the hyper-parameter set is iterated in the above process.
In addition, in the iteration process, whether the global optimal adaptive value reaches the optimal value is judged, namely the r square distance 1 between the actual error series and the forecast error series is the minimum value, the iteration is continued when the actual error series and the forecast error series are not optimal according to the setting of actual conditions by a person skilled in the art, and when the optimal value is reached, a hyper-parameter set corresponding to the global optimal adaptive value is used as the hyper-parameter established by the error correction model.
It should be noted that the manner of determining the optimal hyper-parameter set in the above embodiments is only a specific implementation manner in the present application, and is not limited to the long-term and short-term memory neural network model in the present application, and the determination manner of the optimal hyper-parameter set may also be flexibly selected by a person skilled in the art according to practical situations, which does not affect the protection scope of the present application.
In order to further improve the accuracy of the error correction model, in the preferred embodiment of the present application, training the error correction model specifically includes:
importing the normalized data into an error correction model;
and training the error correction model by using the normalized data and determining an error series relation.
Specifically, after the error correction model is established, the normalized data, that is, the error series in the normalized data, needs to be imported into the error correction model for calibration, so as to determine the error series relationship.
It should be noted that, the model of the present application may be not only a long-term and short-term memory neural network model, but those skilled in the art may flexibly set different models according to actual situations, and the method for determining the hyper-parameter set is also only a specific implementation manner of the present application, and any other method for determining the optimal hyper-parameter set according to the adaptive value of the hyper-parameter set belongs to the protection scope of the present application.
And step S104, forecasting the runoff based on the trained error correction model and the preset hydrological model.
In order to accurately predict the radial flow, in the preferred embodiment of the present application, the radial flow prediction is specifically performed as follows:
determining an error value according to the error series relation;
determining an initial runoff forecast value according to the preset hydrological model;
and forecasting the runoff quantity based on the error value and the initial runoff quantity forecast value.
Specifically, an error value is determined according to the error series relationship, an initial radial flow prediction value is determined according to a preset hydrological model, then the radial flow is predicted based on the error value and the initial radial flow prediction value, specifically, the error value and the initial radial flow prediction value are added, and an obtained value is the final predicted radial flow prediction value.
The error value of the next m days can be obtained by inputting the error of the first n days after normalization into the trained error correction model, and m and n are determined by actual conditions.
By applying the technical scheme, historical data of the rainfall station and the hydrological station in the drainage basin are collected, daily rainfall and actual measurement runoff observed by specific stations of the historical data are determined based on the historical data and a preset hydrological model, an error correction model is established, the error correction model is trained based on the runoff prediction sequence and the actual measurement runoff sequence, and runoff forecast is performed based on the trained error correction model and the preset hydrological model, so that the runoff forecast error is reduced, and the accuracy of the runoff forecast is improved.
In order to further explain the technical idea of the present application, the technical solution of the present application is now described with reference to specific application scenarios.
The embodiment of the application provides a runoff forecasting method, which is characterized in that historical data of all rainfall stations and hydrological stations in a flow domain are collected, the collected historical data are processed, an error correction network model is established, parameters of the error correction network model are optimized, the processed historical data are led into the error correction network model to obtain an optimal combination, and runoff forecasting is carried out on the basis of the error correction network model and the preset hydrological model, so that the accuracy of the runoff forecasting is greatly improved.
As shown in fig. 2, the method comprises the following specific steps:
step S201, data collection and processing.
Historical data of all rainfall stations and hydrological stations in the river basin, such as historical data from 2009 to 2019, is collected, and the historical data is divided into a rate period and a verification period, wherein the rate period can be from 2009 to 2017, and the verification period can be from 2018 to 2019.
The method comprises the following steps of simulating by using a preset hydrological model to obtain a simulation result, namely a runoff prediction sequence, carrying out difference between the simulation result and a measured runoff sequence to obtain an error sequence, and carrying out maximum and minimum normalization processing on the error sequence to obtain normalized data, wherein the formula I is as follows:
wherein X is the original error; xmaxIs the maximum value of the error; xminIs the minimum value of the error; xnormIs normalized data. The data normalization processing also has the functions of improving the convergence rate, preventing gradient explosion and improving the calculation precision.
And S202, setting an error correction model.
Specifically, the error correction model in the present application may be a Long Short-Term Memory neural network model (LSTM), the neural network has a 1-2-1 structure, wherein the hidden layer is composed of one LSTM layer and one fully-connected layer, an average absolute error (MAE) is first selected as a loss function, and an Adam optimizer is used to perform iterative optimization on parameters in the error correction model, the iteration number is a preset number, and the number is set to 300 in the present application embodiment, or set by a person skilled in the art according to an actual situation, where the loss function is specifically as follows:
wherein n is the number of samples; y iso(i) The normalized real error value; y isf(i) Calculated for the model.
Meanwhile, each group of hyper-parameter sets for establishing the error correction model is regarded as a particle, and the ith particle PiCan be expressed asThe value ranges of the three over-parameters are as follows: cell _ size is [2,10]]The batch _ sizes is [64,512]]Time _ step is [1,3]]. Inertial weight at [0.4,1]And performing internal dynamic adjustment, wherein the learning factor c1 is 2, the c2 is 2, and the iteration is performed for 20 times.
Then initializing the particle group, determining the number of the particle group, the speed and the position of each particle, and forming a particle group by 10 particles in a three-dimensional target search space, XkDenotes the position of each particle in the population at the kth iteration (K ≦ 20), i.e.WhereinRepresents a particle PiIn a position ofThe velocity of each particle in the corresponding particle group isThat is to sayIn position whereRepresents a particle PiOf speed, i.e.。
Meanwhile, fitness judgment can be performed on each particle to select the optimal particle, namely the optimal hypercoagulable set is used for establishing an error correction model, and the fitness can be determined by the following formula three:
wherein n is the number of samples; y iso(i) The normalized real error value; y isf(i) Calculating a value for the model; is the true error mean.
Simultaneously comparing the current adaptive value of each particle with the historical individual optimal adaptive value, if the current adaptive value is superior to the historical optimal adaptive value, taking the current adaptive value as a new historical optimal adaptive value, and recording the optimal position of each particle, wherein the position of the individual optimal particle is expressed asWhereinIs a particle PiAt an optimum position, i.e.。
After the historical optimal adaptation value of each particle is determined, comparing the historical optimal adaptation value of each particle with the global optimal adaptation value, if the historical optimal adaptation value of a certain particle is superior to the global optimal adaptation value, taking the historical optimal adaptation value of the particle as a new global optimal adaptation value, and simultaneously recording the position of the global optimal particle,representing the global optimal particle position.
In the iterative process of the particles, the positions and the speeds of the particles are updated through the positions of the individual optimal particles and the positions of the global optimal particles, and the speeds and the positions of the particle swarm are updated according to a formula four to a formula six during k +1 iterations, wherein the formula four to the formula six can be specifically expressed as follows:
wherein, w is the inertia weight,is the minimum value of the inertial weight,is the maximum value of the inertial weight,is the iteration number;is the velocity of the particles and is the velocity of the particles,in order to learn the factors, the learning device is provided with a plurality of learning units,is [0, 1]]The random number of (a) is set,the position of the robot is the best position of the individual,is the global optimum position;is the position of the particle.
And judging whether the global optimal adaptation value reaches the optimal value, if not, continuing iteration until the optimal value is reached, determining a hyper-parameter set of the error correction model according to the optimal particles corresponding to the global optimal adaptation value reaching the optimal value, and establishing the error correction model.
And step S203, forecasting the radial flow.
The normalized data in step S201 is imported into the error correction model, an error series relationship may be established, then the radial flow rate is predicted according to the error value predicted by the error correction model and the initial radial flow rate predicted value of the preset hydrological model, specifically, the error value and the initial radial flow rate predicted value are added, that is, the radial flow rate predicted value, at this time, it may also be known that the hyper-parameter set corresponding to the optimal particle is combined as [ cell _ size =8, batch _ sizes =139, time _ step =1], and a final prediction result is obtained.
In addition, it can be further explained that the prediction accuracy of the present application can be verified by three indexes, namely, root mean square error RMSE, correlation coefficient R, and certainty coefficient NSE. Assuming a true error series of(ii) a The corresponding simulation series isN is the total number of the series time intervals, and each index is calculated as follows:
root mean square error: the root mean square error is the square root of the square of the average deviation between the simulated value and the measured value, and reflects the accuracy of the simulated value, and the smaller the value, the closer the simulated value and the measured value are, which can be determined by the following formula seven:
wherein, XobxIs the error value, X, in the normalized datasimIs a simulated value simulated by the error correction model
The correlation coefficient is used for describing the current degree of correlation between the simulation series and the measured series, the value of the correlation coefficient is closer to 1 and is uniformly distributed on the 45-degree line of the X-Y correlation diagram, which shows that the degree of correlation between the two is higher, and the correlation coefficient can be determined by the following formula eight:
The certainty factor represents the degree of coincidence between the simulated series and the measured series, and the closer the value is to 1, the higher the simulation accuracy is, and can be determined by the following formula:
through the three indexes and setting the preset hydrological model as the Xinanjiang model, the simulation precision table as shown in the following table 1 can be obtained:
TABLE 1
As can be seen from the above table, the accuracy of the prediction of the inner diameter flow of the drainage basin can be obviously improved.
By applying the technical scheme, historical data of all rainfall stations and hydrological stations in the flow field are collected, the collected historical data are processed, then an error correction network model is established, parameters of the error correction network model are optimized, the processed historical data are led into the error correction network model to obtain an optimal combination, and runoff is forecasted based on the error correction network model and the preset hydrological model, so that the accuracy of the runoff forecasting is greatly improved.
The present application also provides a processor for performing the runoff forecasting method according to any of the present application.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application 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; such modifications and substitutions do not necessarily depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (10)
1. A runoff forecasting method, the method comprising:
acquiring historical data of rainfall stations and hydrological stations in a drainage basin, wherein the historical data are daily rainfall and actually measured runoff observed by stations;
determining a runoff prediction sequence based on historical data and a preset hydrological model;
establishing an error correction model, and training the error correction model based on the runoff prediction sequence and the measured runoff sequence;
and forecasting the runoff based on the trained error correction model and the preset hydrological model.
2. The method of claim 1, wherein determining a sequence of runoff predictions based on historical data and a preset hydrological model, further comprises:
determining an error sequence according to the difference value of the actual measured runoff sequence and the runoff predicting sequence;
and carrying out normalization processing on the error sequence to obtain normalized data.
3. The method according to claim 1, wherein the creating of the error correction model specifically comprises:
optimizing the model parameters of the error correction model to be established;
determining all hyper-parameter sets of the error correction model to be established, wherein each group of hyper-parameter sets comprises a plurality of hyper-parameters, and each hyper-parameter has a corresponding value interval;
screening the hyper-parameter set to determine an optimal hyper-parameter set;
and establishing the error correction model based on the optimized model parameters and the optimal hyper-parameter set.
4. The runoff forecasting method according to claim 3, wherein the screening of the hyper-parameter set to determine an optimal hyper-parameter set comprises:
taking each hyper-parameter set as a particle, and taking a preset number of particles as a particle swarm;
initializing the particle swarm and iterating the particles;
determining the adaptive value of the particle at each iteration according to the fitness function;
determining an optimal adaptive value based on a particle swarm algorithm;
and determining the optimal hyper-parameter set according to the optimal adaptive value.
5. The runoff forecasting method according to claim 4, wherein the determining the optimal hyperparameter set according to the optimal adaptive value comprises:
judging whether the optimal adaptive value of each particle is higher than the global optimal adaptive value;
if the optimal adaptive value is higher than the global optimal adaptive value, taking the optimal adaptive value as the global optimal adaptive value;
and determining the optimal hyper-parameter set according to the global optimal adaptive value.
6. The method according to claim 2, wherein the training of the error correction model specifically comprises:
importing the normalized data into an error correction model;
and training the error correction model by using the normalized data and determining an error series relation.
7. The runoff volume forecasting method according to claim 6, wherein the runoff volume forecasting is performed by:
determining an error value according to the error series relation;
determining an initial runoff forecast value according to the preset hydrological model;
and forecasting the runoff quantity based on the error value and the initial runoff quantity forecast value.
8. The method as claimed in claim 7, wherein the runoff volume forecast is performed based on the error value and an initial runoff volume forecast value, and the sum of the error value and the initial runoff volume forecast value is a runoff volume to be forecasted.
9. A runoff forecasting method as claimed in any one of claims 1,3 and 6, wherein said error correction model is a long term short term memory neural network model.
10. A processor, wherein the processor is operative to perform a runoff forecasting method as claimed in any one of claims 1 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110303466.9A CN112686481A (en) | 2021-03-22 | 2021-03-22 | Runoff forecasting method and processor |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110303466.9A CN112686481A (en) | 2021-03-22 | 2021-03-22 | Runoff forecasting method and processor |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112686481A true CN112686481A (en) | 2021-04-20 |
Family
ID=75455751
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110303466.9A Pending CN112686481A (en) | 2021-03-22 | 2021-03-22 | Runoff forecasting method and processor |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112686481A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113642259A (en) * | 2021-10-14 | 2021-11-12 | 国能大渡河流域水电开发有限公司 | River flow field reconstruction method based on multilayer flow velocity measurement |
CN114169590A (en) * | 2021-11-22 | 2022-03-11 | 华中科技大学 | Reservoir warehousing runoff forecasting and correcting method and system based on multi-scenario division |
CN114491978A (en) * | 2022-01-04 | 2022-05-13 | 三峡大学 | Daily model real-time forecasting method based on time-varying parameter hydrological uncertainty processor |
CN114895336A (en) * | 2022-07-12 | 2022-08-12 | 开普勒卫星科技(武汉)有限公司 | Forecasting method for observed value of reference station in GNSS real-time dynamic positioning system |
CN115600749A (en) * | 2022-10-24 | 2023-01-13 | 北京市生态环境保护科学研究院(Cn) | Underground water level prediction method and device and electronic equipment |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112116149A (en) * | 2020-09-17 | 2020-12-22 | 河海大学 | Multi-station medium and long term runoff rolling probability prediction method considering forecast uncertainty associated evolution characteristics |
-
2021
- 2021-03-22 CN CN202110303466.9A patent/CN112686481A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112116149A (en) * | 2020-09-17 | 2020-12-22 | 河海大学 | Multi-station medium and long term runoff rolling probability prediction method considering forecast uncertainty associated evolution characteristics |
Non-Patent Citations (4)
Title |
---|
张璐: "基于组合模型的股市趋势预测方法研究", 《中国优秀硕士学位论文全文数据库 经济与管理科学辑》 * |
王宏力: "《惯性测量组合智能故障诊断及预测技术》", 31 May 2017 * |
薛文志等: "径流预测的生命旋回—拓扑模型", 《水利科技与经济》 * |
陶思铭等: "长短期记忆网络在中长期径流预报中的应用", 《武汉大学学报(工学版)》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113642259A (en) * | 2021-10-14 | 2021-11-12 | 国能大渡河流域水电开发有限公司 | River flow field reconstruction method based on multilayer flow velocity measurement |
CN113642259B (en) * | 2021-10-14 | 2021-12-14 | 国能大渡河流域水电开发有限公司 | River flow field reconstruction method based on multilayer flow velocity measurement |
CN114169590A (en) * | 2021-11-22 | 2022-03-11 | 华中科技大学 | Reservoir warehousing runoff forecasting and correcting method and system based on multi-scenario division |
CN114169590B (en) * | 2021-11-22 | 2024-05-28 | 华中科技大学 | Reservoir-in-storage runoff forecast correction method and system based on multi-scenario division |
CN114491978A (en) * | 2022-01-04 | 2022-05-13 | 三峡大学 | Daily model real-time forecasting method based on time-varying parameter hydrological uncertainty processor |
CN114491978B (en) * | 2022-01-04 | 2024-04-19 | 三峡大学 | Day model real-time forecasting method based on time-varying parameter hydrologic uncertainty processor |
CN114895336A (en) * | 2022-07-12 | 2022-08-12 | 开普勒卫星科技(武汉)有限公司 | Forecasting method for observed value of reference station in GNSS real-time dynamic positioning system |
CN115600749A (en) * | 2022-10-24 | 2023-01-13 | 北京市生态环境保护科学研究院(Cn) | Underground water level prediction method and device and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112686481A (en) | Runoff forecasting method and processor | |
CN110298501B (en) | Electrical load prediction method based on long-time and short-time memory neural network | |
CN111563706A (en) | Multivariable logistics freight volume prediction method based on LSTM network | |
CN110751318A (en) | IPSO-LSTM-based ultra-short-term power load prediction method | |
CN109919356B (en) | BP neural network-based interval water demand prediction method | |
CN113139329B (en) | Xinanjiang model parameter calibration method based on hydrological similarity and artificial neural network | |
CN114444378A (en) | Short-term power prediction method for regional wind power cluster | |
CN112396152A (en) | Flood forecasting method based on CS-LSTM | |
CN110969282A (en) | Runoff stability prediction method based on LSTM composite network | |
CN114492922A (en) | Medium-and-long-term power generation capacity prediction method | |
CN111967696A (en) | Neural network-based electric vehicle charging demand prediction method, system and device | |
CN109598052B (en) | Intelligent ammeter life cycle prediction method and device based on correlation coefficient analysis | |
CN115186923A (en) | Photovoltaic power generation power prediction method and device and electronic equipment | |
CN112711896A (en) | Complex reservoir group optimal scheduling method considering multi-source forecast error uncertainty | |
CN113991711B (en) | Capacity configuration method for energy storage system of photovoltaic power station | |
CN115948964A (en) | Road flatness prediction method based on GA-BP neural network | |
CN114372558A (en) | Residential electricity load prediction method, medium and equipment based on multi-model fusion | |
CN113762591A (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM counterstudy | |
CN117335425A (en) | Tidal current calculation method based on GA-BP neural network | |
CN113159395A (en) | Deep learning-based sewage treatment plant water inflow prediction method and system | |
CN116960962A (en) | Mid-long term area load prediction method for cross-area data fusion | |
CN116976227A (en) | Storm water increasing forecasting method and system based on LSTM machine learning | |
CN111914488A (en) | Data regional hydrological parameter calibration method based on antagonistic neural network | |
CN110852415A (en) | Vegetation index prediction method, system and equipment based on neural network algorithm | |
CN116307139A (en) | Wind power ultra-short-term prediction method for optimizing and improving extreme learning machine |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210420 |
|
RJ01 | Rejection of invention patent application after publication |