CN109840587A - Reservoir reservoir inflow prediction technique based on deep learning - Google Patents
Reservoir reservoir inflow prediction technique based on deep learning Download PDFInfo
- Publication number
- CN109840587A CN109840587A CN201910007771.6A CN201910007771A CN109840587A CN 109840587 A CN109840587 A CN 109840587A CN 201910007771 A CN201910007771 A CN 201910007771A CN 109840587 A CN109840587 A CN 109840587A
- Authority
- CN
- China
- Prior art keywords
- reservoir
- inflow
- predicted
- data
- rain
- 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.)
- Granted
Links
Landscapes
- Feedback Control In General (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to basin regimen electric powder predictions, disclose a kind of reservoir reservoir inflow prediction technique based on deep learning, include the following steps: to obtain historical data, learnt using DBN model, obtain the corresponding relationship between the historical traffic data of each control station and reservoir and the reservoir inflow of reservoir, and then it obtains without the reservoir reservoir inflow predicted in the case of rain, difference delta under the reservoir reservoir inflow and rainy situation predicted in rainy situation, learnt by LSTM training, obtain the reservoir inflow difference predicted in rainy situation, the reservoir reservoir inflow finally predicted is obtained in turn.The present invention is based on the reservoir reservoir inflow prediction techniques of deep learning, depth confidence network and shot and long term memory network algorithm are blended into the prediction applied to reservoir inflow, the forecast precision to reservoir inflow is improved, and improves the reliability and scalability of model.
Description
Technical field
The present invention relates to basin regimen electric powder predictions, and in particular to a kind of reservoir reservoir inflow based on deep learning
Prediction technique.
Background technique
Currently, being broadly divided into two classes for the method for reservoir reservoir inflow prediction: based on traditional hydrology science study method
And the method based on the study of the conventional machines such as artificial neural network.In based on traditional hydrology science study method, generally all
It is to be produced by calculating a variety of magnitudes of physical quantity such as soil moisture content, rainfall, upstream and downstream reservoir filling sluicing situation, and by simulation
Confluence mechanism forecasts regimen.Traditional reservoir reservoir inflow prediction model generally has specific physical relation, but comes relatively
Say, there is also some problems: physical model is more complicated, is difficult to be collected into all temporally spatial distributions needed for establishing model
Detail data, such as terrain data, channel characteristics data, basin soil characteristic distribution and rainfall runoff data；In addition, conduct
The production confluence of non-linear process is excessively complicated, it is difficult to pass through physical model accurate simulation；Furthermore the case where different basins thousand poor ten thousand
Not, and hydrological model used is generally difficult all to fully consider the situation of various complexity.
It is representative to there is decision tree, various clusters to calculate in the method learnt based on conventional machines such as artificial neural networks
Method, support vector machines, neural network etc., these methods are generally started with from history hydrographic data, and are gone out from the relationship between data
Hair finds the mode of inner link, by historical data analysis and study to the reservoir inflow to certain following period carry out it is pre-
It surveys.Although these methods are promoted on precision of prediction relative to conventional method, these methods usually require logarithm
According to being analyzed in advance, then using analysis result as the input of model.And most of learnt using shallow Model, is held
The problems such as easily falling into local optimum, calculating overlong time etc., and the mathematical computations of large-scale complex cannot be simulated, so entering
It is also difficult to obtain ideal effect in terms of the volume forecasting of library.In addition, these network models are all more single, structure is relatively simple, lack
The weary analysis and design to particular problem, scalability are poor.
Summary of the invention
The purpose of the present invention is to the deficiencies of above-mentioned technology, provide a kind of reservoir reservoir inflow based on deep learning
Depth confidence network and shot and long term memory network algorithm are blended the prediction applied to reservoir inflow, improved by prediction technique
To the precision of prediction of reservoir inflow, and improve the reliability and scalability of model.
To achieve the above object, the designed reservoir reservoir inflow prediction technique based on deep learning of the present invention, including such as
Lower step:
A) basin perimeter around reservoir is divided into the first monitoring section for being provided with several control stations and be provided with several rain
The second monitoring section for measuring station obtains the historical traffic data of reservoir, the historical traffic data and rain of control station according to time point
The history rainfall data at station is measured, data on flows is that each time point measures primary instantaneous flow, and rainfall data is two neighboring
Accumulative rainfall between time point, it is rainy situation that the sum of rainfall data of all precipitation stations, which is greater than 10mm, in each time point,
It is no rain condition condition less than or equal to 10mm；
B) using depth confidence network (DBN) model to the historical traffic data of each control station in the case of no rain
Practise, choose the historical traffic data of each control station and reservoir in no rain, input DBN model, obtain each control station and
Corresponding relationship between the historical traffic data of reservoir and the reservoir inflow of reservoir, and then obtain without the reservoir predicted in the case of rain
Reservoir inflow F1；
C the historical traffic data of each control station and reservoir in rainy situation) is chosen, DBN model is inputted, by described
Step B) obtain the historical traffic data and the reservoir inflow of reservoir without control station each in the case of rain and reservoir between it is corresponding
Relationship obtains the reservoir reservoir inflow F2 predicted in rainy situation, then allows in rainy situation the reservoir reservoir inflow F2 predicted
It is poor to make with the true reservoir reservoir inflow in rainy situation, obtains the difference delta in rainy situation；
D) by shot and long term memory network (LSTM) using in basin perimeter precipitation station history rainfall data as LSTM's
Input, and the flow difference delta in the rainy situation obtained using in the step C) is obtained as label by training study
The reservoir inflow difference F3 predicted in rainy situation；
E) will obtain in the step B) without reservoir reservoir inflow F1, the step C predicted in the case of rain) in obtain
Rainy situation under the reservoir reservoir inflow F2 and the step D that predict) in the reservoir inflow predicted in the rainy situation that obtains
Difference F3 fusion obtains the reservoir reservoir inflow F=F1+F2+F3 finally predicted.
Preferably, the step B) in, the historical traffic data of each control station and reservoir in no rain is chosen, is turned
An one-dimensional Input matrix DBN model is turned to, DBN model is a kind of generative probabilistic model, and constituent element is to be limited Bohr hereby
Graceful machine (Restricted Boltzmann Machine, RBM), the DBN model is by 1 RBM and 1 artificial neural network
(ANN) it forms, wherein RBM is made of an aobvious layer and a hidden layer, and ANN is made of a hidden layer and an output layer,
The output of each full articulamentum is defined as formula in DBN model:
yl=f (wlxl-1+bl)
In formula, ylFor l layers of output, wlFor l layers of the weight connected entirely, xl-1For l-1 layers of output, blIt is
L layers of biasing, f are that the activation primitive of hidden layer in activation primitive, wherein RBM and ANN is sigmoid activation primitive, are defined as
Following formula:
Output layer does not have activation primitive in ANN, is linear convergent rate.
Preferably, DBN model learning process is divided into two steps: firstly, during pre-training, by unsupervisedly trained
RBM ensure feature vector when being mapped to different characteristic space, keeping characteristics information as much as possible；Then by RBM output
Feature vector inputs ANN, to initialize the weighting parameter of ANN, then by backpropagation finely tunes entire DBN, and DBN is using averagely putting down
Square error calculates the difference of ground true value and estimated result as loss function, and is optimized by stochastic gradient descent method,
Assuming that giving N number of training sample, y ' and y respectively represents the velocity vector and true value of prediction, then loss function is defined as public affairs
Formula:
Study and feature extraction by DBN constantly to each control station and reservoir historical traffic data, find each control
It stands and the corresponding relationship between reservoir historical traffic data and reservoir inflow, finally obtains each time point without predicting in the case of rain
Reservoir inflow value F1, wherein the output of model is 1 × 1 matrix.
Preferably, the step D) in, according to the positional relationship of precipitation station, precipitation station is carried out from spatial position angle thick
Granular cluster, is divided into the apparent region of several boundaries, then at every point of time on to the history at each area of rainfall station
Rainfall data is summed, then using the value of this several summation as the input of LSTM, and with the step C) in acquirement have
Label of the flow difference delta as LSTM network training in the case of rain, LSTM is made of 1 cell, and carries out the time
Step expansion, timing persistence length are four times of history number of days, often do primary prediction, first input four times of history number of days number time points
History rainfall data forms one-dimensional matrix, and when training LSTM, history rainfall data and label when by no rain condition condition are all set to 0,
And the delta value by value less than 0 is also set to 0, and to remove unreasonable data, the loss function of LSTM optimization is following formula:
It is N number of training sample in formula, y ' and y respectively represent the velocity vector and true value of prediction, finally by training study
Obtain the reservoir inflow difference F3 predicted in rainy situation.
Preferably, time point chooses,, in one day, i.e. adjacent time point interval 6 hours at 2 points at 8 points at 14 points at 20 points, is anti-
The variation of flow and rainfall, historical traffic data and history rainfall data are reflected for 9 days before predicted time point to preceding 1 day flow
Data and rainfall data.
Compared with prior art, the present invention having the advantage that
1, by combining two kinds of deep learning models and to the classification of data, substantially effectively to related hydrographic data into
It has gone and has analyzed and excavate, established most suitable reservoir flux prediction model；
2, in data plane, classified by rain, without rain condition condition to data on flows, combined, fully considered rainfall
Influence to reservoir inflow；In model level, firstly, the performance outstanding using depth confidence network, overcomes general nerve net
Network is easily trapped into the disadvantage of local optimum and training time length because of random initializtion weighting parameter, can be in rainy, nothing
The rule and feature implied in data on flows is accurately excavated in the case of two kinds of rain, to obtain rain, without in the case of two kinds of rain
Secondly the predicted value of reservoir inflow is handled using shot and long term memory network performance outstanding on fitting time series data in rain
The fitting problems of situation rainfall and difference in flow delta obtain the prediction difference of storage quantity amount in rainy situation；
3, by the reservoir inflow value for organically fusion of result that above three part obtains getting up to predict to the end,
The advantage of each model is played to the greatest extent and takes full advantage of existing data resource, so that the effect of the prediction of model is more
Add accurately, also makes model that there is better robustness and applicability.
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow charts of the reservoir reservoir inflow prediction technique of deep learning；
Fig. 2 is depth confidence network architecture figure；
Fig. 3 is shot and long term memory network model structure；
Fig. 4 is the matched curve figure of example predicted value and true value.
Specific embodiment
The following further describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
A kind of reservoir reservoir inflow prediction technique based on deep learning, characterized by the following steps:
A basin perimeter around reservoir) is divided into setting there are three the first monitoring section of control station and is provided with 600
Second monitoring section of precipitation station, according to time point obtain the historical traffic data of reservoir, control station historical traffic data and
The history rainfall data of precipitation station, time point choose,, in one day at 2 points at 8 points at 14 points at 20 points, i.e. adjacent time point interval 6 is small
When, data on flows is the instantaneous flow primary every measurement in 6 hours, and rainfall data is 6 hours rainfall datas, historical traffic number
According to being 9 days before predicted time point to preceding 1 day data on flows and rainfall data with history rainfall data, define in each time point
It is rainy situation that the sum of rainfall data of all precipitation stations, which is greater than 10mm, and being less than or equal to 10mm is no rain condition condition；
B) using depth confidence network (DBN) model to the historical traffic data of each control station in the case of no rain
Practise, choose the historical traffic data of each control station and reservoir in no rain, input DBN model, obtain each control station and
Corresponding relationship between the historical traffic data of reservoir and the reservoir inflow of reservoir, and then obtain without the reservoir predicted in the case of rain
Reservoir inflow F1,
Wherein, it as shown in Fig. 2, choosing the historical traffic data of each control station and reservoir in no rain, is converted into
One one-dimensional Input matrix DBN model, DBN model are a kind of generative probabilistic models, and constituent element is limited Boltzmann machine
(Restricted Boltzmann Machine, RBM), the DBN model is by 1 RBM and 1 artificial neural network (ANN) group
At wherein RBM is made of an aobvious layer and a hidden layer, and ANN is made of a hidden layer and an output layer, DBN model
In the output of each full articulamentum be defined as formula:
yl=f (wlxl-1+bl)
In formula, ylFor l layers of output, wlFor l layers of the weight connected entirely, xl-1For l-1 layers of output, blIt is
L layers of biasing, f are that the activation primitive of hidden layer in activation primitive, wherein RBM and ANN is sigmoid activation primitive, are defined as
Following formula:
Output layer does not have activation primitive in ANN, is linear convergent rate, and DBN model learning process is divided into two steps: firstly, pre-
In training process, ensure that feature vector when being mapped to different characteristic space, is protected as much as possible by training RBM unsupervisedly
Stay characteristic information；Then the feature vector by RBM output inputs ANN, to initialize the weighting parameter of ANN, then by reversely passing
The entire DBN of fine tuning is broadcast, DBN calculates the difference of ground true value and estimated result using Mean Square Error as loss function, and
Optimized by stochastic gradient descent method, it is assumed that give N number of training sample, y ' and y respectively represent prediction velocity vector and
True value, then loss function is defined as formula:
Study and feature extraction by DBN constantly to each control station and reservoir historical traffic data, find each control
It stands and the corresponding relationship between reservoir historical traffic data and reservoir inflow, finally obtains each time point without predicting in the case of rain
Reservoir inflow value F1, wherein the output of model is 1 × 1 matrix；
C the historical traffic data of each control station and reservoir in rainy situation) is chosen, DBN model is inputted, by described
Step B) obtain the historical traffic data and the reservoir inflow of reservoir without control station each in the case of rain and reservoir between it is corresponding
Relationship obtains the reservoir reservoir inflow F2 predicted in rainy situation, then allows in rainy situation the reservoir reservoir inflow F2 predicted
It is poor to make with the true reservoir reservoir inflow in rainy situation, obtains the difference delta in rainy situation；
D) by shot and long term memory network (LSTM) using in basin perimeter precipitation station history rainfall data as LSTM's
Input, and the flow difference delta in the rainy situation obtained using in the step C) is obtained as label by training study
The reservoir inflow difference F3 predicted in rainy situation, wherein according to the positional relationship of precipitation station, from spatial position angle to rainfall
Stand carry out coarseness cluster, be divided into the apparent region of several boundaries, then at every point of time on to each area of rainfall
The history rainfall data stood is summed, then using the value of this several summation as the input of LSTM, and with the step C) in
Label of the flow difference delta as LSTM network training in the rainy situation of acquirement, LSTM are a kind of special circulation minds
Through network, crucial composed structure is cell cell, and a LSTM cell is by 3 threshold structures and 1 state vector transmission
Line composition, thresholding is to forget door, incoming door, out gate respectively, and wherein state vector transmission line is responsible for long-range memory, 3 doors
The selection for being responsible for short-term memory is limited, therefore thresholding setting can be deleted input vector or be added operation, to select
Property allow information to pass through, therefore LSTM can learn to rely on for a long time, therefore be highly suitable for that processing rainfall is this and time series
Highly relevant data, and forget that partial information can solve the problem of gradient disappears, such as Fig. 3 by forgeing door and out gate
Shown, LSTM is made of 1 cell, and carries out time step expansion, and timing persistence length is four times of history number of days, is often done primary
Prediction first inputs the history rainfall data at four times of history number of days number time points, forms one-dimensional matrix, when training LSTM, by nothing
History rainfall data and label when rain condition condition are all set to 0, and the delta value by value less than 0 is also set to 0, are not conformed to removal
Data are managed, the loss function of LSTM optimization is following formula:
It is N number of training sample in formula, y ' and y respectively represent the velocity vector and true value of prediction, finally by training study
Obtain the reservoir inflow difference F3 predicted in rainy situation；
E) will obtain in the step B) without reservoir reservoir inflow F1, the step C predicted in the case of rain) in obtain
Rainy situation under the reservoir reservoir inflow F2 and the step D that predict) in the reservoir inflow predicted in the rainy situation that obtains
Difference F3 fusion obtains the reservoir reservoir inflow F=F1+F2+F3 finally predicted.
If Fig. 4 is the quasi- of the real traffic predicted with the model to Three Gorges Reservoir reservoir inflow and predicted flow rate
Curve graph is closed, wherein the longitudinal axis indicates that reservoir inflow, horizontal axis indicate time series, and what is selected here is from the 2 of on May 10th, 2017
Point is to 20 points of data on October 31, and dotted line indicates the prediction reservoir inflow value of model in figure, and solid line indicates corresponding true stream
Magnitude, it can be seen that fitting degree is very high between prediction reservoir inflow value curve and true flow rate value curve, thus
It can reflect out the high-accuracy of model prediction.
Claims (5)
1. a kind of reservoir reservoir inflow prediction technique based on deep learning, characterized by the following steps:
A) basin perimeter around reservoir is divided into the first monitoring section for being provided with several control stations and be provided with several precipitation stations
The second monitoring section, the historical traffic data and precipitation station of the historical traffic data of reservoir, control station are obtained according to time point
History rainfall data, data on flows is that each time point measures primary instantaneous flow, and rainfall data is the two neighboring time
Accumulative rainfall between point, it is rainy situation that the sum of rainfall data of all precipitation stations, which is greater than 10mm, in each time point, is less than
It is no rain condition condition equal to 10mm；
B) learnt using historical traffic data of depth confidence network (DBN) model to each control station in the case of no rain, selected
It takes the historical traffic data of each control station and reservoir in no rain, inputs DBN model, obtain each control station and reservoir
Corresponding relationship between historical traffic data and the reservoir inflow of reservoir, and then obtain without the reservoir storage stream predicted in the case of rain
Measure F1；
C the historical traffic data of each control station and reservoir in rainy situation) is chosen, DBN model is inputted, passes through the step
B the corresponding relationship between the reservoir inflow of the historical traffic data without control station each in the case of rain and reservoir and reservoir that) obtain,
The reservoir reservoir inflow F2 predicted in rainy situation is obtained, then allows in rainy situation the reservoir reservoir inflow F2 predicted and rainy
In the case of true reservoir reservoir inflow make it is poor, obtain rain situation under difference delta；
D) by shot and long term memory network (LSTM) using in basin perimeter precipitation station history rainfall data as LSTM input,
And the flow difference delta in the rainy situation obtained using in the step C) learns to obtain rainy as label by training
In the case of the reservoir inflow difference F3 that predicts；
E) will obtain in the step B) without reservoir reservoir inflow F1, the step C predicted in the case of rain) in obtain have
The reservoir reservoir inflow F2 and the step D predicted in the case of rain) in obtain rainy situation descend the reservoir inflow difference predicted
F3 fusion obtains the reservoir reservoir inflow F=F1+F2+F3 finally predicted.
2. the reservoir reservoir inflow prediction technique based on deep learning according to claim 1, it is characterised in that: the step
B in), the historical traffic data of each control station and reservoir in no rain is chosen, is converted into an one-dimensional Input matrix DBN
Model, DBN model are a kind of generative probabilistic models, and constituent element is limited Boltzmann machine (Restricted
Boltzmann Machine, RBM), which is made of 1 RBM and 1 artificial neural network (ANN), wherein RBM by
One aobvious layer and a hidden layer are constituted, and ANN is made of a hidden layer and an output layer, each full articulamentum in DBN model
Output be defined as formula:
yl=f (wlxl-1+bl)
In formula, ylFor l layers of output, wlFor l layers of the weight connected entirely, xl-1For l-1 layers of output, blIt is l layers
Biasing, f is that the activation primitive of hidden layer in activation primitive, wherein RBM and ANN is sigmoid activation primitive, is defined as
Formula:
Output layer does not have activation primitive in ANN, is linear convergent rate.
3. the reservoir reservoir inflow prediction technique based on deep learning according to claim 2, it is characterised in that: DBN model
Learning process is divided into two steps: firstly, ensuring that feature vector is being mapped to by training RBM unsupervisedly during pre-training
When different characteristic space, keeping characteristics information as much as possible；Then the feature vector by RBM output inputs ANN, with initialization
The weighting parameter of ANN, then entire DBN is finely tuned by backpropagation, DBN is calculated using Mean Square Error as loss function
The difference of ground true value and estimated result, and optimized by stochastic gradient descent method, it is assumed that give N number of training sample, y ' and y
The velocity vector and true value of prediction are respectively represented, then loss function is defined as formula:
Study and feature extraction by DBN constantly to each control station and reservoir historical traffic data, find each control station with
And the corresponding relationship between reservoir historical traffic data and reservoir inflow, finally obtain each time point enters without what is predicted in the case of rain
Library flow value F1, wherein the output of model is 1 × 1 matrix.
4. the reservoir reservoir inflow prediction technique based on deep learning according to claim 1, it is characterised in that: the step
D in), according to the positional relationship of precipitation station, coarseness cluster is carried out to precipitation station from spatial position angle, is divided into several boundaries
Limit apparent region, then at every point of time on sum to the history rainfall data at each area of rainfall station, then by this
Inputs as LSTM of value of several summations, and with the step C) in acquirement rainy situation under flow difference delta
As the label of LSTM network training, LSTM is made of 1 cell, and carries out time step expansion, and timing persistence length is four
Times history number of days, often does primary prediction, first inputs the history rainfall data at four times of history number of days number time points, forms one-dimensional square
Battle array, when training LSTM, history rainfall data and label when by no rain condition condition are all set to 0, and the delta value by value less than 0
Also it is set to 0, to remove unreasonable data, the loss function of LSTM optimization is following formula:
It is N number of training sample in formula, y ' and y respectively represent the velocity vector and true value of prediction, finally obtain by training study
The reservoir inflow difference F3 predicted in rainy situation.
5. the reservoir reservoir inflow prediction technique based on deep learning according to claim 1, it is characterised in that: the time clicks
Take,, in one day, i.e. adjacent time point interval 6 hours, historical traffic data and history rainfall data at 2 points at 8 points at 14 points at 20 points
It is 9 days before predicted time point to preceding 1 day data on flows and rainfall data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910007771.6A CN109840587B (en) | 2019-01-04 | 2019-01-04 | Reservoir warehousing flow prediction method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910007771.6A CN109840587B (en) | 2019-01-04 | 2019-01-04 | Reservoir warehousing flow prediction method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109840587A true CN109840587A (en) | 2019-06-04 |
| CN109840587B CN109840587B (en) | 2022-07-05 |
Family
ID=66883693
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910007771.6A Active CN109840587B (en) | 2019-01-04 | 2019-01-04 | Reservoir warehousing flow prediction method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109840587B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111062464A (en) * | 2019-10-24 | 2020-04-24 | 中国电力科学研究院有限公司 | Power communication network reliability prediction and guarantee method and system based on deep learning |
| CN111324990A (en) * | 2020-03-19 | 2020-06-23 | 长江大学 | Porosity prediction method based on multilayer long-short term memory neural network model |
| CN111598724A (en) * | 2020-05-19 | 2020-08-28 | 四川革什扎水电开发有限责任公司 | Time-interval integration method for day-ahead prediction of warehousing flow of small and medium-sized reservoirs |
| CN112001556A (en) * | 2020-08-27 | 2020-11-27 | 华中科技大学 | Reservoir downstream water level prediction method based on deep learning model |
| CN112541839A (en) * | 2020-12-23 | 2021-03-23 | 四川大汇大数据服务有限公司 | Reservoir storage flow prediction method based on neural differential equation |
| CN112668773A (en) * | 2020-12-24 | 2021-04-16 | 北京百度网讯科技有限公司 | Method and device for predicting warehousing traffic and electronic equipment |
| CN112712209A (en) * | 2020-12-31 | 2021-04-27 | 润联智慧科技(西安)有限公司 | Reservoir warehousing flow prediction method and device, computer equipment and storage medium |
| CN112989705A (en) * | 2021-03-30 | 2021-06-18 | 海尔数字科技（上海）有限公司 | Method and device for predicting reservoir entry flow value, electronic device and medium |
| CN113592188A (en) * | 2021-08-06 | 2021-11-02 | 国网四川省电力公司 | Cascade power station hydraulic relationship coupling method based on deep learning theory |
| CN114611777A (en) * | 2022-03-04 | 2022-06-10 | 山东锋士信息技术有限公司 | Method for predicting reservoir warehousing flow of flood season based on stacked long and short memory networks |
| CN114707705A (en) * | 2022-03-14 | 2022-07-05 | 浙江大学 | Method, equipment and storage medium for predicting warehousing traffic |
| CN115358343A (en) * | 2022-09-06 | 2022-11-18 | 河海大学 | Small sample flood forecasting method based on confrontation and pseudo label learning |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014178116A1 (en) * | 2013-04-30 | 2014-11-06 | 中国電力株式会社 | Flow quantity forecasting device and flow quantity forecasting system |
| CN105224801A (en) * | 2015-10-08 | 2016-01-06 | 中国长江电力股份有限公司 | A kind of multiple-factor reservoir reservoir inflow short-time forecast evaluation method |
| CN107688871A (en) * | 2017-08-18 | 2018-02-13 | 中国农业大学 | A kind of water quality prediction method and device |
| CN108875161A (en) * | 2018-05-31 | 2018-11-23 | 长江勘测规划设计研究有限责任公司 | Flow grade prediction technique based on convolutional neural networks deep learning |
| CN108921279A (en) * | 2018-03-26 | 2018-11-30 | 西安电子科技大学 | Reservoir day enters water prediction technique |
| US20180373993A1 (en) * | 2017-06-23 | 2018-12-27 | University Of Alaska Fairbanks | Method Of Predicting Streamflow Data |
-
2019
- 2019-01-04 CN CN201910007771.6A patent/CN109840587B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014178116A1 (en) * | 2013-04-30 | 2014-11-06 | 中国電力株式会社 | Flow quantity forecasting device and flow quantity forecasting system |
| CN105224801A (en) * | 2015-10-08 | 2016-01-06 | 中国长江电力股份有限公司 | A kind of multiple-factor reservoir reservoir inflow short-time forecast evaluation method |
| US20180373993A1 (en) * | 2017-06-23 | 2018-12-27 | University Of Alaska Fairbanks | Method Of Predicting Streamflow Data |
| CN107688871A (en) * | 2017-08-18 | 2018-02-13 | 中国农业大学 | A kind of water quality prediction method and device |
| CN108921279A (en) * | 2018-03-26 | 2018-11-30 | 西安电子科技大学 | Reservoir day enters water prediction technique |
| CN108875161A (en) * | 2018-05-31 | 2018-11-23 | 长江勘测规划设计研究有限责任公司 | Flow grade prediction technique based on convolutional neural networks deep learning |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111062464B (en) * | 2019-10-24 | 2022-07-01 | 中国电力科学研究院有限公司 | Power communication network reliability prediction and guarantee method and system based on deep learning |
| CN111062464A (en) * | 2019-10-24 | 2020-04-24 | 中国电力科学研究院有限公司 | Power communication network reliability prediction and guarantee method and system based on deep learning |
| CN111324990A (en) * | 2020-03-19 | 2020-06-23 | 长江大学 | Porosity prediction method based on multilayer long-short term memory neural network model |
| CN111598724A (en) * | 2020-05-19 | 2020-08-28 | 四川革什扎水电开发有限责任公司 | Time-interval integration method for day-ahead prediction of warehousing flow of small and medium-sized reservoirs |
| CN111598724B (en) * | 2020-05-19 | 2022-07-22 | 四川革什扎水电开发有限责任公司 | Time-interval integration method for day-ahead prediction of warehousing flow of small and medium reservoirs |
| CN112001556A (en) * | 2020-08-27 | 2020-11-27 | 华中科技大学 | Reservoir downstream water level prediction method based on deep learning model |
| CN112001556B (en) * | 2020-08-27 | 2022-07-15 | 华中科技大学 | Reservoir downstream water level prediction method based on deep learning model |
| CN112541839A (en) * | 2020-12-23 | 2021-03-23 | 四川大汇大数据服务有限公司 | Reservoir storage flow prediction method based on neural differential equation |
| CN112668773A (en) * | 2020-12-24 | 2021-04-16 | 北京百度网讯科技有限公司 | Method and device for predicting warehousing traffic and electronic equipment |
| CN112668773B (en) * | 2020-12-24 | 2024-06-28 | 北京百度网讯科技有限公司 | Warehouse entry flow prediction method and device and electronic equipment |
| CN112712209A (en) * | 2020-12-31 | 2021-04-27 | 润联智慧科技(西安)有限公司 | Reservoir warehousing flow prediction method and device, computer equipment and storage medium |
| CN112989705A (en) * | 2021-03-30 | 2021-06-18 | 海尔数字科技（上海）有限公司 | Method and device for predicting reservoir entry flow value, electronic device and medium |
| CN113592188A (en) * | 2021-08-06 | 2021-11-02 | 国网四川省电力公司 | Cascade power station hydraulic relationship coupling method based on deep learning theory |
| CN114611777A (en) * | 2022-03-04 | 2022-06-10 | 山东锋士信息技术有限公司 | Method for predicting reservoir warehousing flow of flood season based on stacked long and short memory networks |
| CN114707705A (en) * | 2022-03-14 | 2022-07-05 | 浙江大学 | Method, equipment and storage medium for predicting warehousing traffic |
| CN115358343A (en) * | 2022-09-06 | 2022-11-18 | 河海大学 | Small sample flood forecasting method based on confrontation and pseudo label learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109840587B (en) | 2022-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109840587A (en) | Reservoir reservoir inflow prediction technique based on deep learning | |
| CN111310968B (en) | LSTM neural network circulating hydrologic forecasting method based on mutual information | |
| Fu et al. | Deep learning data-intelligence model based on adjusted forecasting window scale: application in daily streamflow simulation | |
| CN108304668B (en) | Flood prediction method combining hydrologic process data and historical prior data | |
| Ramirez et al. | Artificial neural network technique for rainfall forecasting applied to the Sao Paulo region | |
| Razavi et al. | An evaluation of regionalization and watershed classification schemes for continuous daily streamflow prediction in ungauged watersheds | |
| CN107292098A (en) | Medium-and Long-Term Runoff Forecasting method based on early stage meteorological factor and data mining technology | |
| Toth et al. | Multistep ahead streamflow forecasting: Role of calibration data in conceptual and neural network modeling | |
| Marofi et al. | Predicting spatial distribution of snow water equivalent using multivariate non-linear regression and computational intelligence methods | |
| Lee et al. | Interpolation of missing precipitation data using kernel estimations for hydrologic modeling | |
| Samantaray et al. | Prediction of suspended sediment concentration using hybrid SVM-WOA approaches | |
| Chen et al. | Groundwater level prediction using SOM-RBFN multisite model | |
| CN111723523B (en) | Estuary surplus water level prediction method based on cascade neural network | |
| Üneş et al. | Estimation of groundwater level using artificial neural networks: a case study of Hatay-Turkey | |
| CN107918166A (en) | More satellite fusion precipitation methods and system | |
| Terzi | Monthly Rainfall Estimation Using Data‐Mining Process | |
| Maddu et al. | Prediction of land surface temperature of major coastal cities of India using bidirectional LSTM neural networks | |
| Budamala et al. | Enhance the prediction of complex hydrological models by pseudo-simulators | |
| Agarwal et al. | Flood forecasting and flood flow modeling in a river system using ANN | |
| Aoulmi et al. | Runoff predictions in a semiarid watershed by convolutional neural networks improved with metaheuristic algorithms and forced with reanalysis and climate data | |
| Patra et al. | Regional groundwater sequential forecasting using global and local LSTM models | |
| Giang et al. | Monthly precipitation prediction using neural network algorithms in the Thua Thien Hue Province | |
| CN116960962A (en) | Mid-long term area load prediction method for cross-area data fusion | |
| Nizar et al. | Forecasting of temperature by using LSTM and bidirectional LSTM approach: case study in Semarang, Indonesia | |
| Huang et al. | Using Artificial Intelligence to Retrieve the Optimal Parameters and Structures of Adaptive Network‐Based Fuzzy Inference System for Typhoon Precipitation Forecast Modeling |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |