CN110427663A - Face precipitation-water-level simulation method based on time series network - Google Patents
Face precipitation-water-level simulation method based on time series network Download PDFInfo
- Publication number
- CN110427663A CN110427663A CN201910645838.9A CN201910645838A CN110427663A CN 110427663 A CN110427663 A CN 110427663A CN 201910645838 A CN201910645838 A CN 201910645838A CN 110427663 A CN110427663 A CN 110427663A
- Authority
- CN
- China
- Prior art keywords
- data
- network
- water level
- time series
- website
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Strategic Management (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to water-level simulation technical fields, face precipitation-water-level simulation method based on time series network is disclosed, including obtains face rainfall input image, hydrology control module input image, waterlevel data, Yunnan snub-nosed monkey is carried out to Grid square product, chooses suitable short-term water level forecast time d1, choose suitable time series network delay parameter d2, to input data carry out PCA dimension-reduction treatment, time series network is trained and is calculated network analog precision.The present invention improves the precision of short-term water level forecast result, few and with a varied topography region is laid suitable for home position sensing, data integrity is stronger, evade ground home position sensing and is distributed sparse problem in region with a varied topography, short-term water level forecast precision is improved, data, method, feature and knowledge used in short-term water level forecast have scalability.
Description
Technical field
The present invention relates to water-level simulation technical fields, and in particular to face precipitation-water-level simulation based on time series network
Method.
Background technique
The big flood event due to caused by global extreme weather frequently occurs, and precipitation-water-level simulation is in hydrometeorological field
Study status it is higher and higher, by Interpretation Method of Area Rainfall can preresearch estimates flooding area range, flood control and disaster reduction is played an important role.
Existing Interpretation Method of Area Rainfall method is generally divided into model driven method and data-driven method, and model driven method is usual
There is a basis of the specific physical principle as model, but due to the complexity of real world, is often unable to measure complexity
Physical parameter causes precision of prediction relatively low, and model-driven at this stage receives assorted efficiency factor generally below 85%；With
The development of artificial intelligence, data-driven method be widely used, hydrologic forecasting method also from model-driven mode turn
Become data driven mode, data-driven model precision of prediction usually with higher, related coefficient generally can achieve 0.9 or
It is higher.
Currently, thering are many researchs to start using the grids such as TRMM, CHRIPS, GLDAS drop with the maturation of face Precipitation Products
Data source of the aquatic products as prediction model, using face Precipitation Products data can be with as the input data source of forecast model of water level
Ideal simulation precision is obtained, biggish work can be played by being distributed sparse area in ground station shortage of data or website
With, but data-driven model generally can not drive the interpretation prediction result of physical process from the hydrology, so existing research is more
It is improved from the angle of data fitting precision, the statistics approximating method of from " data " to " data " deviates from hydrology neck merely
The research contents in domain, and the soil due to not accounting for its dependent variable such as hydrologic process and vegetation element, it will usually occur
Additional error；In addition, SEA LEVEL VARIATION caused by precipitation is not an instant process, it may be necessary to for a period of time could be to water
Position variation shows to influence.In view of the time-lag effect during SEA LEVEL VARIATION, external model of nonlinear auto-companding (NARX) etc.
Recurrent neural network model theoretically can obtain better effect than artificial nerve network model.
The indefinite problem of physical mechanism existing for the above-mentioned data-driven model, this patent propose that a kind of combination is existing
Face precipitation-water-level simulation method based on time series network of physics hydrological model.
Summary of the invention
Based on the incomplete problem of modeling mechanism present in available data described above driving hydrological simulation, the present invention
A kind of face precipitation-water-level simulation method based on time series network is provided, from the angle of hydrologic cycle, considers face precipitation-water
The time delay effect and the polynary influence factor during precipitation-SEA LEVEL VARIATION of position process, propose that a set of physical significance is bright
True face precipitation-water-level simulation process, and solve the problems, such as that home position sensing spatial arrangement is discrete by using face Rainfall Products,
Simultaneously by with externally input nonlinear auto-companding (nonlinear autoregressive exogenous model,
NARX) time series network model simulates the cumulative effect of hydrologic process.
In order to solve the above technical problems, the present invention provides face precipitation-water-level simulation method based on time series network,
The following steps are included:
Step 1: it divides through basin water system, determines basin perimeter where water level control website C and water level control website C,
Water level control website C is referred to as website C below, the face rainfall of certain period T inputs shadow in basin perimeter where obtaining website C
As and required hydrology control module input image, and obtain with period T website C waterlevel data as target data, it is defeated
The temporal resolution for entering data and target data is 1 day, and remembers that water level to be predicted is the water level in the X day of website C；
Step 2: Yunnan snub-nosed monkey being carried out to Grid square product, using the vector shp file in basin where website C to net
Lattice data product is cut, and by all grid data product resamplings at unified resolution；
Step 3: choosing suitable short-term water level forecast time d1, and input data and target data are serialized, and makes
With being predicted d before day water level1It data are as training data；
Step 4: choosing suitable time series network delay parameter d2, d2Represent network think the forecast level in X day with
The preceding d in X day2Relationship between its water level, i.e., using the tested d a few days ago of water level2It recycles as time series network and inputs, and leads to
It crosses adjusting parameter and comes appropriate adjustment network structure, including time series network delay parameter and the network number of plies, the network number of plies is hidden
Number containing node layer；
Step 5: PCA dimension-reduction treatment being carried out to input data, to accelerate the convergence rate of network and remove part sample
Noise；
Step 6: using the sample data after progress PCA dimension-reduction treatment as network inputs, the waterlevel data of corresponding website C
It exports, and time series network is trained, network after being trained as network；
Step 7: taking the X-d in the basin perimeter in step 12~X days rainfall input image, hydrology module input image
With network inputs after training, network inputs are the X-d of website C after training2~X-1 days water level passes through the net after step 6 training
Network model calculates, and finally obtains the water level forecast value for being predicted the X day of website C, water level as to be predicted exports water to be predicted
Position simultaneously calculates network analog precision according to water level observation.
Further, the hydrology control module input image in step 1 includes: that the soil of grid form in basin perimeter is wet
Degree according to grid form in Vegetation canopy evapotranspiration data of grid form in, basin perimeter, basin perimeter mean wind speed number
According to the temperature on average data with grid form in basin perimeter.
Further, the step of carrying out PCA dimension-reduction treatment to input data in step 5 is as follows:
Step 5.1: setting input image sample set as D={ χ1,χ2,…,χm, dimension dim, to all input samples into
The following centralization operation of row:
Wherein χ represents image picture element；
Step 5.2: the covariance matrix XX of the sample after calculating centralizationT；
Step 5.3: to the covariance matrix XX of sampleTCarry out Eigenvalues Decomposition；
Step 5.4: feature vector corresponding to maximum dim characteristic value being taken to form projection matrix W*={ w1,w2,…,
wdim}。
Further, the time series network in step 6 is NARX time series network, NARX time series network model
Function representation it is as follows:
Y (t)=f (χ (t-1) ..., χ (t-d), y (t-1) ..., y (t-d))
Wherein, y (t) is t days output target values, and χ (t) is t days input variable values, and d is delay number of days.
Compared with prior art, the beneficial effects of the present invention are:
(1) precision for improving short-term water level forecast result has fully taken into account precipitation-SEA LEVEL VARIATION process physics meaning
Justice, the input variable of increased hydrology module make the physical significance of model definitely, while relative to traditional neural network
Analogy method, this method are simulated using Time Serial Neural Network, have fully considered the delay effect of SEA LEVEL VARIATION process,
Precision of prediction improves a lot with respect to BP neural network analogy method.
(2) it is suitable for home position sensing and lays few and with a varied topography region, the rainfall of this method and the input of hydrology module
Data use Grid square product, and data continuous uniform in basin is distributed and is easier to obtain, and data integrity is stronger, reduces
Due to ground home position sensing lay it is discrete caused by data distribution it is uneven or generate when using website interpolation in situ
Extra error, evaded ground home position sensing and be distributed sparse problem in region with a varied topography, it is pre- to improve short-term water level
Survey precision.
(3) data, method, feature and knowledge used in short-term water level forecast have scalability.
Detailed description of the invention
Fig. 1 is the schematic diagram of the selected survey region of the embodiment of the present invention；
Fig. 2 is the flow chart of the embodiment of the present invention；
Fig. 3 is the result of the water level observation of the BP neural network of the embodiment of the present invention；
Fig. 4 is the result of the water level observation of the NARX time series network of the embodiment of the present invention；
Fig. 5 is using face rainfall data of the embodiment of the present invention and the water level forecast result pair using ground station rainfall data
Than figure；
Fig. 6 be the embodiment of the present invention using increase the input of hydrology module with it is pre- using only rainfall data water level as input
Survey result and water level observation comparison diagram；
Fig. 7 be the embodiment of the present invention using increase the input of hydrology module with it is pre- using only rainfall data water level as input
Survey resultant error line.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below with reference to embodiment and attached drawing, to this
Invention is described in further detail, and exemplary embodiment of the invention and its explanation for explaining only the invention, are not made
For limitation of the invention.
Embodiment:
Face precipitation-water-level simulation method based on time series network, comprising the following steps:
Step 1: it divides through basin water system, determines basin perimeter where water level control website C and water level control website C,
Water level control website C is referred to as website C below, the face rainfall of certain period T inputs shadow in basin perimeter where obtaining website C
As data and required hydrology control module input image data, wherein hydrology control module input image data include: basin
The soil moisture data of grid form in range, in basin perimeter grid form Vegetation canopy evapotranspiration data, basin perimeter
The temperature on average data of grid form in the mean wind speed data and basin perimeter of interior grid form, and obtain the station with period T
For the waterlevel data of point C as target data, the temporal resolution of input data and target data is 1 day, and remembers water to be predicted
Position is the water level in the X day of website C；
Referring to Fig. 1, the present embodiment chooses the Drainage Area of Jinsha River of Upper Yangtze River as survey region, and the lower left corner in the region passes through
Latitude be (90.5354167877,24.4595834302), upper right corner longitude and latitude be (104.947083454,
35.7512500968), the time range of data selected by the present embodiment is 2016 to 2019；Input data packet
Pingshan hydrometric station water level day data, Drainage Area of Jinsha River rainfall data and hydrology control module variable data are included, can be obtained from Hydrological Bureau
The water level day data at Pingshan hydrometric station are obtained, which exists as water level true value；Remote sensing face rainfall data set and hydrology control
Module variable data collection uses GLDAS-2 data set, and spatial resolution is 0.25 ° × 0.25 °, while by the rainfall of remote sensing face
The result of data set and hydrology control module variable data collection and ground station rainfall data set and the face CHIRPS rainfall data set
Result carry out accuracy comparison；
In order to verify the implementation result of this method, the present embodiment has carried out multiple groups comparative experiments simultaneously, and the present embodiment will be refreshing
Input through network model is divided into the rainfall part for directly affecting SEA LEVEL VARIATION and influences the hydrology control mould of SEA LEVEL VARIATION indirectly
Block importation；The present embodiment use surface weather station observation data, CHIRPS Rainfall Products, GLDAS-2 rainfall product as
Input, surface weather station's observation data, CHIRPS Rainfall Products, GLDAS-2 rainfall product respectively represent input type I, input
Type II and input type III, the input of hydrology control module use the assimilation data in GLDAS-2, meanwhile, the present embodiment uses
The input of ground station is tested as a comparison, and the input of ground station represents input type IV, by input type I to inputting class
Type IV, the result of influence of the available different types of data source input to model prediction result；In addition, the present embodiment carries out
The experiment of water level forecast is carried out using only GLDAS-2 as input variable, which represent input type V, for inquiring into water
Literary control module inputs the influence to water level forecast result；The BP neural network mould that the present embodiment will use under identical input simultaneously
The result of the forecast model of water level of type is compared with the prediction result as NARX time series network, and comparing result is shown in figure
3 and Fig. 4；
Step 2: Yunnan snub-nosed monkey being carried out to Grid square product, using the vector shp file in basin where website C to net
Lattice data product is cut, and by all grid data product resamplings at unified resolution；
Step 3: choosing suitable short-term water level forecast time d1, and input data and target data are serialized, and makes
With being predicted d before day water level1It data choose d as training data, the present embodiment1=7, i.e., using being predicted before water level 7 days
The data waterlevel data of predicting the 8th day；
Step 4: choosing suitable time series network delay parameter d2, d2Represent network think the forecast level in X day with
The preceding d in X day2Relationship between its water level, i.e., using the tested d a few days ago of water level2It recycles as time series network and inputs, and leads to
It crosses adjusting parameter and comes appropriate adjustment network structure, including time series network delay parameter and the network number of plies, the network number of plies is hidden
Number containing node layer, the method for adjusting parameter include the methods of artificial experience tune ginseng and automatic tune ginseng, are used in this embodiment
The method of artificial parameter adjustment, the network delay parameter d of selection2=5, the network number of plies is 2, and node in hidden layer is respectively 6 Hes
10, network parameter adjustment can choose other algorithms, since this method does not discuss network reference services, repeat no more；
Step 5: PCA dimension-reduction treatment being carried out to input data, to accelerate the convergence rate of network and remove part sample
Noise, above-mentioned processing step are as follows: (1) setting input image sample set as D={ χ1,χ2,…,χm, dimension dim, to all defeated
Enter sample and carry out following centralization operation:Wherein χ represents image picture element；(2) centralization is calculated
The covariance matrix XX of sample afterwardsT；(3) to the covariance matrix XX of sampleTCarry out Eigenvalues Decomposition；(4) maximum dim is taken
Feature vector corresponding to a characteristic value forms projection matrix W*={ w1,w2,…,wdim}；
Step 6: using the sample data after progress PCA dimension-reduction treatment as network inputs, the waterlevel data of corresponding website C
It exports, and NARX time series network is trained, network after being trained as network, wherein NARX time series network
The function representation of model is as follows:
Y (t)=f (χ (t-1) ..., χ (t-d), y (t-1) ..., y (t-d))
Wherein, y (t) is t days output target values, and χ (t) is t days input variable values, and d is delay number of days；
Step 7: taking the X-d in the basin perimeter in step 12~X days rainfall input image, hydrology module input image
With network inputs after training, network inputs are the X-d of website C after training2~X-1 days water level passes through the net after step 6 training
Network model calculates, and finally obtains the water level forecast value for being predicted the X day of website C, water level as to be predicted exports water to be predicted
Position simultaneously calculates network analog precision according to water level observation；Repeat the operation of step 6~step 7 and assess water level forecast as a result,
Using heterogeneous networks input type and different neural network types, the results are shown in Table 1 for water level forecast, makes as can be seen from Table 1
RSME with the NARX network model of hydrology control module input (input type III) is minimum, is 0.5609m, related coefficient is most
A height of 98.93%, mean percent deviation is 0.0013% to be lower than other experimental groups, and assorted efficiency factor of receiving is 97.86% higher than institute
There is experimental group, illustrates model credibility height.
The water level forecast accuracy comparison that table 1 is inputted using heterogeneous networks and heterogeneous networks
Referring to Fig. 3 and Fig. 4, it can be seen that in four precision indexs, the average RMSE of BP neural network model is
0.999m, and the average RMSE of NARX model is 0.608m, the result of NARX time series network model is superior to BP nerve net
Network model；Referring to Fig. 5, carried out pair by the data of using face Rainfall Products and using the precision result of the data of ground station rainfall
Than the result that using face rainfall data carry out water level forecast as data source as the result is shown is more preferable；May be used also from table 1 simultaneously
Out, comparison is used only rainfall data input (input type V) and uses hydrology module input variable (input type III), input
The root-mean-square error of the NARX model of type-iii is 0.56m, and the root-mean-square error of input type V is 0.61m, uses BPNN mould
The root-mean-square error that type carries out water level forecast is respectively 0.98m and 1.17m, while the NASH efficiency factor of prediction model increases
About 3%.Water level forecast root-mean-square error can be effectively improved using the input of hydrology module and receives assorted efficiency factor.Referring to Fig. 6
And Fig. 7 can be preferably right using the time series network model for increasing the input of hydrology module when flood season level fluctuation is larger
Water level is simulated；The above results synthesis shows that this method can effectively improve short-term water level forecast precision, also demonstrates we
The feasibility of method.
It as above is the embodiment of the present invention.Design parameter in above-described embodiment and embodiment is merely to understand table
Invention verification process is stated, the scope of patent protection being not intended to limit the invention, scope of patent protection of the invention is still with it
It is all to change with equivalent structure made by specification and accompanying drawing content of the invention subject to claims, it should all similarly wrap
Containing within the scope of the present invention.
Claims (4)
1. face precipitation-water-level simulation method based on time series network, which comprises the following steps:
Step 1: dividing through basin water system, determine basin perimeter where water level control website C and water level control website C, below
Water level control website C is referred to as website C, where obtaining website C in basin perimeter the face rainfall input image of certain period T and
Required hydrology control module input image, and obtain with period T website C waterlevel data be used as target data, input number
It is 1 day according to the temporal resolution with target data, and remembers that water level to be predicted is the water level in the X day of website C；
Step 2: Yunnan snub-nosed monkey being carried out to Grid square product, using the vector shp file in basin where website C to grid number
It is cut according to product, and by all grid data product resamplings at unified resolution；
Step 3: choosing suitable short-term water level forecast time d1, and input data and target data are serialized, and use is pre-
D before survey day water level1It data are as training data；
Step 4: choosing suitable time series network delay parameter d2, d2Represent forecast level and X that network thinks X day
It preceding d2Relationship between its water level, i.e., using the tested d a few days ago of water level2It recycles as time series network and inputs, and passes through tune
Whole parameter comes appropriate adjustment network structure, including time series network delay parameter and the network number of plies, and the network number of plies is hidden layer
Number of nodes；
Step 5: PCA dimension-reduction treatment being carried out to input data, to accelerate the convergence rate of network and remove part sample noise；
Step 6: using the sample data after progress PCA dimension-reduction treatment as network inputs, the waterlevel data conduct of corresponding website C
Network output, and time series network is trained, network after being trained；
Step 7: taking the X-d in the basin perimeter in step 12~X days rainfall input images, hydrology module input image and instructions
Network inputs after white silk, network inputs are the X-d of website C after training2~X-1 days water level passes through the network mould after step 6 training
Type calculates, and finally obtains the water level forecast value for being predicted the X day of website C, water level as to be predicted exports water level to be predicted simultaneously
Network analog precision is calculated according to water level observation.
2. face precipitation-water-level simulation method as described in claim 1 based on time series network, which is characterized in that step 1
In hydrology control module input image include: the soil moisture data of grid form in basin perimeter, grid in basin perimeter
Grid form in the mean wind speed data and basin perimeter of grid form in the Vegetation canopy evapotranspiration data of form, basin perimeter
Temperature on average data.
3. face precipitation-water-level simulation method as described in claim 1 based on time series network, which is characterized in that step 5
In to input data carry out PCA dimension-reduction treatment the step of it is as follows:
Step 5.1: setting input image sample set as D={ χ1,χ2,…,χm, dimension dim carries out such as all input samples
Lower centralization operation:
Wherein χ represents image picture element；
Step 5.2: the covariance matrix XX of the sample after calculating centralizationT；
Step 5.3: to the covariance matrix XX of sampleTCarry out Eigenvalues Decomposition；
Step 5.4: feature vector corresponding to maximum dim characteristic value being taken to form projection matrix W*={ w1,w2,…,wdim}。
4. face precipitation-water-level simulation method as described in claim 1 based on time series network, which is characterized in that step 6
In time series network be NARX time series network, the function representation of NARX time series network model is as follows:
Y (t)=f (χ (t-1) ..., χ (t-d), y (t-1) ..., y (t-d))
Wherein, y (t) is t days output target values, and χ (t) is t days input variable values, and d is delay number of days.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910645838.9A CN110427663A (en) | 2019-07-17 | 2019-07-17 | Face precipitation-water-level simulation method based on time series network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910645838.9A CN110427663A (en) | 2019-07-17 | 2019-07-17 | Face precipitation-water-level simulation method based on time series network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110427663A true CN110427663A (en) | 2019-11-08 |
Family
ID=68410830
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910645838.9A Pending CN110427663A (en) | 2019-07-17 | 2019-07-17 | Face precipitation-water-level simulation method based on time series network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110427663A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112597647A (en) * | 2020-12-17 | 2021-04-02 | 河海大学 | Rapid-convergence ultrahigh-frequency microwave rainfall data discretization method |
| CN112989557A (en) * | 2021-01-14 | 2021-06-18 | 中国空间技术研究院 | Method for improving water reserve change prediction reliability based on neural network selectable model |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105046321A (en) * | 2015-06-25 | 2015-11-11 | 河海大学 | Combined model water level prediction method based on similarity search |
| CN107885951A (en) * | 2017-11-27 | 2018-04-06 | 河海大学 | A kind of Time series hydrological forecasting method based on built-up pattern |
| CN107992447A (en) * | 2017-12-13 | 2018-05-04 | 电子科技大学 | A kind of feature selecting decomposition method applied to river level prediction data |
| JP2018178405A (en) * | 2017-04-05 | 2018-11-15 | 株式会社日立製作所 | Control device and control program for sewage treatment plant |
| CN109991685A (en) * | 2019-04-03 | 2019-07-09 | 北京市天元网络技术股份有限公司 | A kind of precipitation prediction technique and device based on more LSTM Model Fusions |
| CN110010246A (en) * | 2019-04-16 | 2019-07-12 | 湖南大学 | A kind of disease Intelligent Diagnosis Technology based on neural network and confidence interval |
-
2019
- 2019-07-17 CN CN201910645838.9A patent/CN110427663A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105046321A (en) * | 2015-06-25 | 2015-11-11 | 河海大学 | Combined model water level prediction method based on similarity search |
| JP2018178405A (en) * | 2017-04-05 | 2018-11-15 | 株式会社日立製作所 | Control device and control program for sewage treatment plant |
| CN107885951A (en) * | 2017-11-27 | 2018-04-06 | 河海大学 | A kind of Time series hydrological forecasting method based on built-up pattern |
| CN107992447A (en) * | 2017-12-13 | 2018-05-04 | 电子科技大学 | A kind of feature selecting decomposition method applied to river level prediction data |
| CN109991685A (en) * | 2019-04-03 | 2019-07-09 | 北京市天元网络技术股份有限公司 | A kind of precipitation prediction technique and device based on more LSTM Model Fusions |
| CN110010246A (en) * | 2019-04-16 | 2019-07-12 | 湖南大学 | A kind of disease Intelligent Diagnosis Technology based on neural network and confidence interval |
Non-Patent Citations (2)
| Title |
|---|
| CHEN, ZEQIANG 等: "RFim: A Real-Time Inundation Extent Model for Large Floodplains Based on Remote Sensing Big Data and Water Level Observations", 《REMOTE SENSING》 * |
| 文明 等: "地铁车站施工过程中地表沉降的NARXNN时间序列预测模型", 《岩石力学与工程学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112597647A (en) * | 2020-12-17 | 2021-04-02 | 河海大学 | Rapid-convergence ultrahigh-frequency microwave rainfall data discretization method |
| CN112597647B (en) * | 2020-12-17 | 2021-06-01 | 河海大学 | Rapid-convergence ultrahigh-frequency microwave rainfall data discretization method |
| CN112989557A (en) * | 2021-01-14 | 2021-06-18 | 中国空间技术研究院 | Method for improving water reserve change prediction reliability based on neural network selectable model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112241814B (en) | Traffic prediction method based on reinforced space-time diagram neural network | |
| CN111222698B (en) | Internet of things-oriented ponding water level prediction method based on long-time and short-time memory network | |
| CN110084367B (en) | Soil moisture content prediction method based on LSTM deep learning model | |
| CN109978235B (en) | Waterlogging water level prediction method based on sample learning | |
| Ines et al. | Inverse modelling in estimating soil hydraulic functions: a Genetic Algorithm approach | |
| Liu et al. | Evaluating uncertainties in multi-layer soil moisture estimation with support vector machines and ensemble Kalman filtering | |
| CN109146162B (en) | A kind of probability wind speed forecasting method based on integrated Recognition with Recurrent Neural Network | |
| CN105825230A (en) | Soil moisture status prediction method and system based on DBN (Deep Belief Network) model | |
| CN106651012A (en) | Crop transpiration prediction method based on improved extreme learning machine | |
| CN112287294B (en) | Space-time bidirectional soil water content interpolation method based on deep learning | |
| CN103197538B (en) | Brooder temperature control method based on process neural network | |
| CN113837499A (en) | Ultra-short-term wind power prediction method and system | |
| CN110427663A (en) | Face precipitation-water-level simulation method based on time series network | |
| CN111767517A (en) | BiGRU multi-step prediction method and system applied to flood prediction and storage medium | |
| Han et al. | Probabilistic flood forecasting using hydrologic uncertainty processor with ensemble weather forecasts | |
| CN112785024A (en) | Runoff calculation and prediction method based on watershed hydrological model | |
| CN113378243A (en) | Personalized federal learning method based on multi-head attention mechanism | |
| CN110276477A (en) | A kind of Flood Forecasting Method based on layering Bayesian network and incremental learning | |
| Bergeron et al. | Combined assimilation of streamflow and snow water equivalent for mid-term ensemble streamflow forecasts in snow-dominated regions | |
| CN114818324A (en) | Method, device, medium and equipment for automatically regulating and controlling water quantity and water quality of basin scale | |
| CN109272144B (en) | BPNN-based prediction method for NDVI (normalized difference of variance) in northern grassland area of China | |
| CN108510097A (en) | Power transmission line corridor Snow Disaster monitoring method based on satellite remote sensing and system | |
| CN108920571B (en) | Indoor hydrological experiment model scale determination method based on distributed hydrological model | |
| CN109541729A (en) | A kind of prediction technique of the grassland in northern China area precipitation during growing season based on NARX | |
| Khatri et al. | Water and salt balances at farmer fields |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191108 |