CN109840587B - Reservoir warehousing flow prediction method based on deep learning - Google Patents
Reservoir warehousing flow prediction method based on deep learning Download PDFInfo
- Publication number
- CN109840587B CN109840587B CN201910007771.6A CN201910007771A CN109840587B CN 109840587 B CN109840587 B CN 109840587B CN 201910007771 A CN201910007771 A CN 201910007771A CN 109840587 B CN109840587 B CN 109840587B
- Authority
- CN
- China
- Prior art keywords
- reservoir
- flow
- historical
- under
- 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.)
- Active
Links
Images
Abstract
The invention relates to the technical field of watershed water regime prediction, and discloses a reservoir warehousing flow prediction method based on deep learning, which comprises the following steps of: obtaining historical data, learning by using a DBN (database-based network) model, obtaining the corresponding relation between the historical flow data of each control station and each reservoir and the warehousing flow of the reservoir, further obtaining the predicted warehousing flow of the reservoir under the rainless condition, the predicted warehousing flow of the reservoir under the rainy condition and the difference delta under the rainy condition, obtaining the predicted warehousing flow difference under the rainy condition through LSTM training learning, and further obtaining the finally predicted warehousing flow of the reservoir. According to the reservoir warehousing flow prediction method based on deep learning, a deep belief network and a long-short term memory network algorithm are fused and applied to the prediction of warehousing flow, the prediction precision of the warehousing flow is improved, and the reliability and the expandability of a model are improved.
Description
Technical Field
The invention relates to the technical field of watershed water regime prediction, in particular to a reservoir warehousing flow prediction method based on deep learning.
Background
At present, methods for predicting reservoir warehousing flow mainly fall into two categories: the method is based on a traditional hydrology research method and a traditional machine learning method such as an artificial neural network. In a traditional hydrology research method, the water regime is generally forecasted by calculating various physical quantities such as soil water content, rainfall, water storage and drainage conditions of upstream and downstream water reservoirs and simulating a production and confluence mechanism. The traditional reservoir warehousing flow prediction model generally has a definite physical relation, but relatively speaking, certain problems also exist: the physical model is complex, and it is difficult to collect all the detail data distributed according to time and space, such as terrain data, river channel characteristic data, drainage basin soil characteristic distribution and rainfall runoff data, required by the model; in addition, the production convergence as a nonlinear process is too complex to be accurately simulated by a physical model; moreover, the conditions of different watersheds are different, and the used hydrological model generally has difficulty in fully considering various complex conditions.
In traditional machine learning methods based on artificial neural networks and the like, a decision tree, various clustering algorithms, a support vector machine, a neural network and the like are typical, the methods generally start from historical hydrological data, search an internal connection mode from the relationship among data, and predict the warehousing flow of a certain period of time in the future by analyzing and learning the historical data. Although the prediction accuracy of the methods is improved compared with that of the traditional methods, the methods generally need to analyze data in advance, and then use the analysis result as the input of a model. Most of the methods use shallow models for learning, so that the problems of local optimization, overlong calculation time and the like are easily caused, and large-scale complex mathematical calculation cannot be simulated, so that an ideal effect is difficult to obtain in the aspect of warehouse entry flow prediction. In addition, the network models are single, simple in structure, lack of analysis and design on specific problems, and poor in expansibility.
Disclosure of Invention
The invention aims to provide a reservoir warehousing flow prediction method based on deep learning aiming at the defects of the technology, and the deep confidence network and the long-short term memory network algorithm are fused and applied to the prediction of warehousing flow, so that the prediction precision of the warehousing flow is improved, and the reliability and the expandability of a model are improved.
In order to achieve the aim, the reservoir warehousing flow prediction method based on deep learning comprises the following steps:
A) dividing a watershed range around a reservoir into a first monitoring area provided with a plurality of control stations and a second monitoring area provided with a plurality of rainfall stations, acquiring historical flow data of the reservoir, historical flow data of the control stations and historical rainfall data of the rainfall stations according to time points, wherein the flow data is instantaneous flow measured once at each time point, the rainfall data is accumulated rainfall between two adjacent time points, the rainfall data of all the rainfall stations at each time point is greater than 10mm in the case of rain, and the rainfall data of all the rainfall stations at each time point is less than or equal to 10mm in the case of no rain;
B) learning historical flow data of each control station under the condition of no rain by using a Deep Belief Network (DBN) model, selecting the historical flow data of each control station and the reservoir under the condition of no rain, inputting the historical flow data into the DBN model, obtaining the corresponding relation between the historical flow data of each control station and the reservoir and the warehousing flow of the reservoir, and further obtaining the reservoir warehousing flow F1 predicted under the condition of no rain;
C) selecting historical flow data of each control station and each reservoir under the condition of rain, inputting the historical flow data into a DBN model, obtaining reservoir warehousing flow F2 predicted under the condition of rain according to the corresponding relation between the historical flow data of each control station and each reservoir under the condition of no rain and the warehousing flow of the reservoir obtained in the step B), and then making the difference between the reservoir warehousing flow F2 predicted under the condition of rain and the real reservoir warehousing flow under the condition of rain to obtain a difference delta under the condition of rain;
D) taking historical rainfall data of rainfall stations in a watershed range as input of a long-short term memory network (LSTM) through the LSTM, taking the flow difference delta under the rainy condition obtained in the step C) as a label, and obtaining a predicted warehousing flow difference F3 under the rainy condition through training and learning;
E) fusing the reservoir warehousing flow rate F1 predicted under the no-rain condition obtained in the step B), the reservoir warehousing flow rate F2 predicted under the rain condition obtained in the step C) and the warehousing flow rate difference F3 predicted under the rain condition obtained in the step D) to obtain the final predicted reservoir warehousing flow rate F1+ F2+ F3.
Preferably, in step B), historical flow data of each control station and each reservoir in the rain-free condition is selected and converted into a one-dimensional matrix input DBN model, the DBN model is a probability generation model, the components of the model are Restricted Boltzmann Machines (RBMs), the DBN model is composed of 1 RBM and 1 Artificial Neural Network (ANN), wherein the RBM is composed of a display layer and a hidden layer, the ANN is composed of a hidden layer and an output layer, and the output of each fully-connected layer in the DBN model is defined as the following formula:
yl＝f(wlxl-1+bl)
in the formula, ylIs the output of the l-th layer, wlIs the weight of the full connection of the l-th layer, xl-1Is the output of layer l-1, blFor the bias of the l-th layer, f is an activation function, wherein the activation function of the hidden layer in the RBM and the ANN is a sigmoid activation function, and is defined as the following formula:
the output layer in the ANN has no activation function and is linear output.
Preferably, the DBN model learning process is divided into two steps: firstly, in the pre-training process, unsupervised RBM training is adopted to ensure that feature information is kept as much as possible when feature vectors are mapped to different feature spaces; then inputting the feature vector output by the RBM into the ANN to initialize the weight parameter of the ANN, then finely adjusting the whole DBN by back propagation, calculating the difference between the true value and the estimation result by the DBN by using the average square error as a loss function, and optimizing by a random gradient descent method, wherein if N training samples are given, y' and y respectively represent the predicted speed vector and the true value, the loss function is defined as the following formula:
and continuously learning and extracting characteristics of historical flow data of each control station and each reservoir through the DBN, finding out the corresponding relation between the historical flow data of each control station and each reservoir and the warehousing flow, and finally obtaining the predicted warehousing flow value F1 under the condition of no rain at each time point, wherein the output of the model is a matrix of 1 multiplied by 1.
Preferably, in the step D), based on the position relationship of the rainfall stations, the rainfall stations are subjected to coarse-grained clustering from the perspective of spatial position, and divided into several regions with distinct boundaries, then, the historical rainfall data of each regional rainfall station is summed at each time point, and the summed values are used as the input of the LSTM, and the flow difference delta under the rainy condition obtained in the step C) is taken as a label for LSTM network training, the LSTM is composed of 1 cell, and time step expansion is carried out, the time sequence duration is four times of the historical days, each time prediction is carried out, the historical rainfall data of the time point of four times of the historical days is firstly input to form a one-dimensional matrix, when the LSTM is trained, the historical rainfall data and the label under the condition of no rain are set to be 0, and setting the delta value less than 0 to remove unreasonable data, the loss function of LSTM optimization is as follows:
in the formula, N training samples are used, y' and y represent a predicted speed vector and a true value respectively, and finally, a predicted warehousing flow difference value F3 under a rainy condition is obtained through training and learning.
Preferably, the time points are selected from 2, 8, 14 and 20 points in a day, namely, adjacent time points are separated by 6 hours, in order to reflect the change of the flow and the rainfall, and the historical flow data and the historical rainfall data are flow data and rainfall data from 9 days to 1 day before the predicted time point.
Compared with the prior art, the invention has the following advantages:
1. by combining two deep learning models and classifying data, related hydrological data are fully and effectively analyzed and mined, and a most suitable reservoir flow prediction model is established;
2. on the data layer, classifying and combining the flow data according to the rain condition and the rain condition, and fully considering the influence of rainfall on the warehousing flow; on the model level, firstly, the excellent performance of a deep confidence network is utilized, the defects that a general neural network is easy to fall into local optimization and long in training time due to random initialization of weight parameters are overcome, rules and characteristics implicit in flow data can be accurately mined under the two conditions of rain and no rain, so that predicted values of warehousing flow under the two conditions of rain and no rain are obtained, secondly, the excellent performance of a long and short-term memory network on fitting time sequence data is utilized to process the problem of delta fitting between the rainfall and the flow difference under the rain condition, and the prediction difference value of the warehousing flow under the rain condition is obtained;
3. the finally predicted warehousing flow value is obtained by organically fusing the results obtained by the three parts, so that the advantages of each model are furthest exerted, the existing data resources are fully utilized, the prediction effect of the model is more accurate, and the model has better robustness and applicability.
Drawings
FIG. 1 is a flow chart of a reservoir warehousing flow prediction method based on deep learning of the invention;
FIG. 2 is a diagram of a deep belief network model architecture;
FIG. 3 is a diagram of a long term short term memory network model architecture;
fig. 4 is a plot of a fit of example predicted values to true values.
Detailed Description
The invention is described in further detail below with reference to the figures and the specific embodiments.
A reservoir warehousing flow prediction method based on deep learning is characterized by comprising the following steps: the method comprises the following steps:
A) dividing a watershed range around a reservoir into a first monitoring area provided with three control stations and a second monitoring area provided with six hundred rainfall stations, acquiring historical flow data of the reservoir, historical flow data of the control stations and historical rainfall data of the rainfall stations according to time points, selecting 2 points, 8 points, 14 points and 20 points in one day from the time points, namely, the interval between adjacent time points is 6 hours, wherein the flow data is instantaneous flow measured every 6 hours, the rainfall data is 6-hour rainfall data, the historical flow data and the historical rainfall data are flow data and rainfall data of 9 days to 1 day before a predicted time point, and defining that the sum of the rainfall data of all the rainfall stations at each time point is greater than 10mm in the case of rain and is less than or equal to 10mm in the case of no rain;
B) learning historical flow data of each control station under the no-rain condition by using a Deep Belief Network (DBN) model, selecting the historical flow data of each control station and the reservoir under the no-rain condition, inputting the historical flow data into the DBN model, obtaining the corresponding relation between the historical flow data of each control station and the reservoir and the warehousing flow of the reservoir, further obtaining the reservoir warehousing flow F1 predicted under the no-rain condition,
as shown in fig. 2, historical flow data of each control station and the reservoir under a no-rain condition is selected and converted into a one-dimensional matrix to be input into a DBN model, the DBN model is a probability generation model, the components of the DBN model are limited Boltzmann machines (RBMs), the DBN model comprises 1 RBM and 1 Artificial Neural Network (ANN), the RBM comprises a display layer and a hidden layer, the ANN comprises a hidden layer and an output layer, and the output of each full-connection layer in the DBN model is defined as the following formula:
yl＝f(wlxl-1+bl)
in the formula, ylIs the output of the l-th layer, wlIs the weight of the full connection of the l-th layer, xl-1Is the output of layer l-1, blFor the bias of the l-th layer, f is an activation function, wherein the activation function of the hidden layer in the RBM and the ANN is a sigmoid activation function, and is defined as the following formula:
the output layer in the ANN has no activation function and is linear output, and the DBN model learning process is divided into two steps: firstly, in the pre-training process, the feature information is kept as much as possible when the feature vector is mapped to different feature spaces through unsupervised training of RBMs; then inputting the feature vector output by the RBM into an ANN to initialize weight parameters of the ANN, finely adjusting the whole DBN through back propagation, calculating the difference value of a true value and an estimation result by the DBN by using an average square error as a loss function, and optimizing through a random gradient descent method, wherein the loss function is defined as the following formula under the assumption that N training samples are given, and y' and y respectively represent a predicted speed vector and a true value:
continuously learning and extracting characteristics of historical flow data of each control station and the reservoir through a DBN (database server), finding a corresponding relation between the historical flow data and warehousing flow of each control station and the reservoir, and finally obtaining a predicted warehousing flow value F1 under the condition that no rain exists at each time point, wherein the output of the model is a matrix of 1 multiplied by 1;
C) selecting historical flow data of each control station and each reservoir under the condition of rain, inputting the historical flow data into a DBN model, obtaining reservoir warehousing flow F2 predicted under the condition of rain according to the corresponding relation between the historical flow data of each control station and each reservoir under the condition of no rain and the warehousing flow of the reservoir obtained in the step B), and then making the difference between the reservoir warehousing flow F2 predicted under the condition of rain and the real reservoir warehousing flow under the condition of rain to obtain a difference delta under the condition of rain;
D) taking the historical rainfall data of the rainfall stations in the watershed range as input of an LSTM through a long-short term memory network (LSTM), taking the flow difference delta under the raining condition obtained in the step C) as a label, and obtaining a storage flow difference F3 predicted under the raining condition through training and learning, wherein according to the position relation of the rainfall stations, the rainfall stations are subjected to coarse-grained clustering from the spatial position angle and are divided into a plurality of regions with obvious boundaries, then the historical rainfall data of each regional rainfall station are summed at each time point, the plurality of summed values are taken as input of the LSTM, the flow difference delta under the raining condition obtained in the step C) is taken as a label for LSTM network training, the LSTM is a special recurrent neural network, the key composition structure of the LSTM network is a cell, and one LSTM cell is composed of 3 threshold structures and 1 state vector transmission line, the thresholds are a forgetting gate, an input gate and an output gate respectively, wherein a state vector transmission line is responsible for long-range memory, 3 thresholds are responsible for selection of short-term memory, therefore, threshold setting can delete or add operation to an input vector so as to selectively pass information, so that the LSTM can learn long-term dependence, and is very suitable for processing rainfall, which is data highly related to a time sequence, and the problem of gradient disappearance can be solved by forgetting part of information through the forgetting gate and the output gate, as shown in FIG. 3, the LSTM comprises 1 cell and is expanded in time step, the time sequence duration is fourfold historical days, each time of prediction is made, the historical rainfall data of a time point of fourfold historical days is firstly input to form a one-dimensional matrix, when the LSTM is trained, the historical rainfall data and a label under the no rain condition are both set to be 0, and the value of delta smaller than 0 is also set to be 0, to remove unreasonable data, the loss function for LSTM optimization is the following equation:
in the formula, N training samples are used, y' and y respectively represent a predicted speed vector and a true value, and finally, a predicted warehousing flow difference value F3 under the condition of rain is obtained through training and learning;
E) fusing the reservoir warehousing flow rate F1 predicted under the no-rain condition obtained in the step B), the reservoir warehousing flow rate F2 predicted under the rain condition obtained in the step C) and the warehousing flow rate difference F3 predicted under the rain condition obtained in the step D) to obtain the final predicted reservoir warehousing flow rate F1+ F2+ F3.
For example, fig. 4 is a fitting curve graph of the actual flow and the predicted flow obtained by predicting the warehousing flow of the three gorges reservoir by using the model, wherein the vertical axis represents the warehousing flow, and the horizontal axis represents a time sequence, data from 2 points of 5, 10 and 10 months and 20 points of 31 and 10 months are selected, a dotted line in the graph represents the predicted warehousing flow value of the model, and a solid line represents the corresponding actual flow value, so that the fitting degree between the predicted warehousing flow value curve and the actual flow value curve is very high, and the high accuracy of model prediction can be reflected.
Claims (2)
1. A reservoir warehousing flow prediction method based on deep learning is characterized by comprising the following steps: the method comprises the following steps:
A) dividing a watershed range around a reservoir into a first monitoring area provided with a plurality of control stations and a second monitoring area provided with a plurality of rainfall stations, acquiring historical flow data of the reservoir, historical flow data of the control stations and historical rainfall data of the rainfall stations according to time points, wherein the flow data is instantaneous flow measured once at each time point, the rainfall data is accumulated rainfall between two adjacent time points, the rainfall data of all the rainfall stations at each time point is greater than 10mm in the case of rain, and the rainfall data of all the rainfall stations at each time point is less than or equal to 10mm in the case of no rain;
B) the method comprises the following steps of learning historical flow data of each control station under the condition of no rain by using a DBN model, selecting the historical flow data of each control station and a reservoir under the condition of no rain, inputting the historical flow data of each control station and the reservoir into the DBN model, obtaining the corresponding relation between the historical flow data of each control station and the reservoir and the warehousing flow of the reservoir, and further obtaining the predicted warehousing flow F1 of the reservoir under the condition of no rain, wherein the method comprises the following specific steps:
selecting historical flow data of each control station and each reservoir under the condition of no rain, converting the historical flow data into a one-dimensional matrix, inputting the one-dimensional matrix into a DBN model, wherein the DBN model is a probability generation model, and the component element of the DBN model is RBM (radial basis function), the DBN model consists of 1 RBM and 1 ANN, the RBM consists of a display layer and a hidden layer, the ANN consists of a hidden layer and an output layer, and the output of each full-connection layer in the DBN model is defined as the following formula:
yl＝f(wlxl-1+bl)
in the formula, ylIs the output of the l-th layer, wlIs the weight of the full connection of the l-th layer, xl-1Is the output of layer l-1, blFor the bias of the l-th layer, f is an activation function, wherein the activation function of the hidden layer in the RBM and the ANN is a sigmoid activation function, and is defined as the following formula:
the output layer in the ANN has no activation function and is linear output;
the DBN model learning process is divided into two steps: firstly, in the pre-training process, unsupervised RBM training is adopted to ensure that feature information is kept as much as possible when feature vectors are mapped to different feature spaces; then inputting the feature vector output by the RBM into an ANN to initialize weight parameters of the ANN, finely adjusting the whole DBN through back propagation, calculating the difference value of a true value and an estimation result by the DBN by using an average square error as a loss function, and optimizing through a random gradient descent method, wherein the loss function is defined as the following formula under the assumption that N training samples are given, and y' and y respectively represent a predicted speed vector and a true value:
continuously learning and extracting characteristics of historical flow data of each control station and each reservoir through a DBN (database server), finding out the corresponding relation between the historical flow data of each control station and each reservoir and warehousing flow, and finally obtaining a predicted warehousing flow value F1 under the condition of no rain at each time point, wherein the output of the model is a matrix of 1 multiplied by 1;
C) selecting historical flow data of each control station and each reservoir under the condition of rain, inputting the historical flow data into a DBN model, obtaining reservoir warehousing flow F2 predicted under the condition of rain according to the corresponding relation between the historical flow data of each control station and each reservoir under the condition of no rain and the warehousing flow of the reservoir obtained in the step B), and then making the difference between the reservoir warehousing flow F2 predicted under the condition of rain and the real reservoir warehousing flow under the condition of rain to obtain a difference delta under the condition of rain;
D) taking the historical rainfall data of the rainfall station in the watershed range as the input of the LSTM through the LSTM, taking the flow difference delta under the rainy condition obtained in the step C) as a label, and obtaining the predicted warehousing flow difference F3 under the rainy condition through training and learning, wherein the method specifically comprises the following steps:
according to the position relation of the rainfall stations, the rainfall stations are subjected to coarse-grained clustering from the perspective of spatial position and are divided into a plurality of regions with obvious boundaries, then, the historical rainfall data of each regional rainfall station is summed at each time point, and the summed values are used as the input of the LSTM, and the flow difference delta under the rainy condition obtained in the step C) is taken as a label for LSTM network training, the LSTM is composed of 1 cell, and time step expansion is carried out, the time sequence duration is four times of the historical days, each time prediction is carried out, the historical rainfall data of the time point of four times of the historical days is firstly input to form a one-dimensional matrix, when the LSTM is trained, the historical rainfall data and the label under the condition of no rain are set to be 0, and setting the delta value less than 0 to remove unreasonable data, the loss function of LSTM optimization is as follows:
in the formula, the training samples are N, y' and y respectively represent a predicted speed vector and a true value, and finally, a predicted warehousing flow difference value F3 under the condition of rain is obtained through training and learning;
E) fusing the reservoir warehousing flow rate F1 predicted under the no-rain condition obtained in the step B), the reservoir warehousing flow rate F2 predicted under the rain condition obtained in the step C) and the warehousing flow rate difference F3 predicted under the rain condition obtained in the step D) to obtain the final predicted reservoir warehousing flow rate F1+ F2+ F3.
2. The reservoir warehousing traffic prediction method based on deep learning of claim 1, characterized in that: the time points are selected from 2 points, 8 points, 14 points and 20 points in one day, namely adjacent time points are separated by 6 hours, and the historical flow data and the historical rainfall data are flow data and rainfall data from 9 days to 1 day before the predicted time point.
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 CN109840587A (en) | 2019-06-04 |
| CN109840587B true 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) |
Families Citing this family (10)
| 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 |
| 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 |
| 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 |
| 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 |
| CN112541839B (en) * | 2020-12-23 | 2022-02-11 | 国能大渡河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 |
| CN112712209B (en) * | 2020-12-31 | 2021-11-05 | 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 |
Citations (5)
| 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 |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11238356B2 (en) * | 2017-06-23 | 2022-02-01 | University Of Alaska Fairbanks | Method of predicting streamflow data |
-
2019
- 2019-01-04 CN CN201910007771.6A patent/CN109840587B/en active Active
Patent Citations (5)
| 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 |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109840587A (en) | 2019-06-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109840587B (en) | Reservoir warehousing flow prediction method based on deep learning | |
| CN111222698B (en) | Internet of things-oriented ponding water level prediction method based on long-time and short-time memory network | |
| CN111310968B (en) | LSTM neural network circulating hydrologic forecasting method based on mutual information | |
| Najafi et al. | Statistical downscaling of precipitation using machine learning with optimal predictor selection | |
| Ramirez et al. | Artificial neural network technique for rainfall forecasting applied to the Sao Paulo region | |
| CN111563706A (en) | Multivariable logistics freight volume prediction method based on LSTM network | |
| CN107292098A (en) | Medium-and Long-Term Runoff Forecasting method based on early stage meteorological factor and data mining technology | |
| Marofi et al. | Predicting spatial distribution of snow water equivalent using multivariate non-linear regression and computational intelligence methods | |
| CN111665575B (en) | Medium-and-long-term rainfall grading coupling forecasting method and system based on statistical power | |
| CN110705115B (en) | Weather forecast method and system based on deep belief network | |
| CN109143408B (en) | Dynamic region combined short-time rainfall forecasting method based on MLP | |
| Chen et al. | Groundwater level prediction using SOM-RBFN multisite model | |
| CN111126704A (en) | Multi-region precipitation prediction model construction method based on multi-graph convolution and memory network | |
| CN113139329B (en) | Xinanjiang model parameter calibration method based on hydrological similarity and artificial neural network | |
| CN113705877A (en) | Real-time monthly runoff forecasting method based on deep learning model | |
| CN112766603A (en) | Traffic flow prediction method, system, computer device and storage medium | |
| Ahmadi et al. | Input data selection for solar radiation estimation | |
| Lee et al. | Deep learning for hydrometeorology and environmental science | |
| CN116432828A (en) | Intelligent prediction method for runoff of data-missing river basin | |
| Mohd-Safar et al. | Integration of fuzzy c-means and artificial neural network for short-term localized rainfall forecasting in tropical climate | |
| CN114372631A (en) | Data-lacking area runoff prediction method based on small sample learning and LSTM | |
| Steyn | Short-term stream flow forecasting and downstream gap infilling using machine learning techniques | |
| Wu et al. | Hierarchical Bayesian network based incremental model for flood prediction | |
| Yaseen et al. | Artificial intelligence based flood forecasting for River Hunza at Danyor Station in Pakistan | |
| Al-Fawa’Reh et al. | Intelligent methods for flood forecasting in Wadi al Wala, Jordan |
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 |