CN115879646B - Reservoir water level prediction method, device, medium and equipment - Google Patents
Reservoir water level prediction method, device, medium and equipment Download PDFInfo
- Publication number
- CN115879646B CN115879646B CN202310112898.0A CN202310112898A CN115879646B CN 115879646 B CN115879646 B CN 115879646B CN 202310112898 A CN202310112898 A CN 202310112898A CN 115879646 B CN115879646 B CN 115879646B
- Authority
- CN
- China
- Prior art keywords
- output
- convolution
- water level
- information
- network
- 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
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 82
- 238000000034 method Methods 0.000 title claims abstract description 34
- 239000013598 vector Substances 0.000 claims description 48
- 238000012549 training Methods 0.000 claims description 17
- 238000006073 displacement reaction Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 9
- 230000000295 complement effect Effects 0.000 claims description 6
- 238000012856 packing Methods 0.000 claims description 4
- 125000004122 cyclic group Chemical group 0.000 claims 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 230000007613 environmental effect Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 9
- 238000012360 testing method Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 241000288105 Grus Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000013505 freshwater Substances 0.000 description 1
- 239000003621 irrigation water Substances 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
- Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of water level prediction and artificial intelligence, and provides a method, a device, a medium and equipment for predicting the water level of a reservoir, which are used for solving the problem that the accuracy of the water level prediction of the reservoir is reduced because the change of the water level of the reservoir is a complex result of various environmental factors and is not fused with the characteristics in space dimension. The reservoir water level prediction method comprises the steps of obtaining rainfall historical information, rainfall average value historical information and reservoir water level historical information of a plurality of sites; based on the acquired historical information and the trained reservoir water level prediction model, predicting reservoir water level information at a plurality of moments in the future; the reservoir water level prediction model comprises a convolution network, a continuous GRU network and a full-connection network; the reservoir water level prediction method can extract low-level features of input data from time dimension and space dimension, and fuse the extracted features to finally realize accurate prediction of long-time water level.
Description
Technical Field
The invention belongs to the technical field of water level prediction and artificial intelligence, and particularly relates to a reservoir water level prediction method, a device, a medium and equipment.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The fluctuation of the reservoir water level is important not only in planning, designing and running of fresh water reservoirs, namely living, industrial, hydroelectric, irrigation water supply, flood control, shipping, water quality and quantity improvement and the like, but also in all hydraulic construction. The water level measurement or the water level prediction through the pre-estimated model is two direct ways for obtaining reservoir management decisions. Compared with water level measurement, the prediction model can obtain future reservoir prediction results and is more economical, so that the prediction model is more preferable for predicting the future reservoir water level.
With the rise of artificial intelligence in recent years, deep learning networks have been regarded as the best tool for modeling complex environments, wherein, GRU is a variant of LSTM networks, compared with LSTM networks, GRU network structures are simpler, the quantity of parameters to be trained is smaller, and the problem of long dependence in RNN networks can be solved, thus playing a good prediction effect on time sequence prediction. In 2019, the Xuan-Hien Le predicts the water level of Vietnam Luo Kehe for 1-4 time steps by using a GRU network, and obtains remarkable results, and the accuracy rate of the prediction results is 94% -96% by using Nash coefficients. In 2021, the Xuan-HienLe again predicts the water level for 1 hour, 3 hours, 6 hours and 9 hours in the future by using the GRU network, and the prediction accuracy is found to be reduced with the delay of the prediction time.
The inventor finds that the GRU network used in the prior art can only predict the water level of 4 time steps at most, does not test and try water level prediction at more moments, uses water level historical data, does not utilize priori knowledge such as rainfall, and the like, and only depends on the data of the previous time points, so that relevant characteristics in time are extracted.
Disclosure of Invention
In order to solve the technical problems in the background art, the invention provides a reservoir water level prediction method, a device, a medium and equipment, which can extract low-level features of input data from time dimension and space dimension, fuse the extracted features and finally realize accurate prediction of long-time water level.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
a first aspect of the invention provides a reservoir level prediction method.
In one or more embodiments, a reservoir level prediction method specifically includes the steps of:
acquiring rainfall historical information, rainfall average historical information and reservoir water level historical information of a plurality of sites;
based on the acquired historical information and the trained reservoir water level prediction model, predicting reservoir water level information at a plurality of moments in the future;
the reservoir water level prediction model comprises a convolution network, a continuous GRU network and a full-connection network; the convolution network is used for extracting time information and space information of the history information; the continuous GRU network is used for predicting states of a plurality of time points in the future based on information extracted by the convolution network and outputting the combination of the historical output state and the future output state; the fully-connected network is used for fusing the historical output state and the future output state to obtain reservoir water level information at a plurality of moments in the future.
As one embodiment, a sample for training a reservoir water level prediction model is formed by sequentially arranging rainfall historical information, rainfall average value historical information and reservoir water level historical information of a plurality of stations at a plurality of moments; one sample has two dimensions, one dimension is time information, and the other dimension is rainfall information of a plurality of sites at the same moment.
As an embodiment, the convolutional network is used for: and replacing the first vector operation of the current convolution input by convolution operation, convolution displacement and the last output result to finally obtain a heap vector.
The convolution network firstly carries out packing operation on an input sample, and then a convolution kernel convolves vectors of the first two moments and the complementary 0 of the sample at a first moment to obtain a first output vector; then the convolution kernel moves rightwards by one step, the convolution result replaces a first input value of the position of the convolution kernel in the sample, and then the output result at the previous moment and the other two vectors in the sample are sent into the convolution kernel together to obtain a second output vector; and by analogy, replacing the first vector operation of the current convolution input by convolution operation, convolution displacement and the last output result, and finally obtaining a heap vector.
As an embodiment, the output of the continuous GRU network at the current time depends on the output of the previous time and the hidden state of the previous time.
As an embodiment, the continuous GRU network uses Dropout method of variational reasoning, and for the same sequence, the same dropping method is adopted on the loop connection at all the time points.
A second aspect of the present invention provides a reservoir level prediction apparatus.
In one or more embodiments, a reservoir level prediction apparatus specifically includes the following modules:
a historical information acquisition module for acquiring rainfall historical information, rainfall average historical information and reservoir water level historical information of a plurality of sites;
the reservoir water level prediction module is used for predicting reservoir water level information at a plurality of moments in the future based on the acquired historical information and a trained reservoir water level prediction model;
the reservoir water level prediction model comprises a convolution network, a continuous GRU network and a full-connection network; the convolution network is used for extracting time information and space information of the history information; the continuous GRU network is used for predicting states of a plurality of time points in the future based on information extracted by the convolution network and outputting the combination of the historical output state and the future output state; the fully-connected network is used for fusing the historical output state and the future output state to obtain reservoir water level information at a plurality of moments in the future.
As one embodiment, a sample for training a reservoir water level prediction model is formed by sequentially arranging rainfall historical information, rainfall average value historical information and reservoir water level historical information of a plurality of stations at a plurality of moments; one sample has two dimensions, one dimension is time information, and the other dimension is rainfall information of a plurality of sites at the same moment.
As an embodiment, the convolutional network is used for: and replacing the first vector operation of the current convolution input by convolution operation, convolution displacement and the last output result to finally obtain a heap vector.
The convolution network firstly carries out packing operation on an input sample, and then a convolution kernel convolves vectors of the first two moments and the complementary 0 of the sample at a first moment to obtain a first output vector; then the convolution kernel moves rightwards by one step, the convolution result replaces a first input value of the position of the convolution kernel in the sample, and then the output result at the previous moment and the other two vectors in the sample are sent into the convolution kernel together to obtain a second output vector; and by analogy, replacing the first vector operation of the current convolution input by convolution operation, convolution displacement and the last output result, and finally obtaining a heap vector.
A third aspect of the present invention provides a computer-readable storage medium.
In one or more embodiments, a computer readable storage medium has stored thereon a computer program which when executed by a processor performs the steps in a reservoir level prediction method as described above.
A fourth aspect of the invention provides an electronic device.
In one or more embodiments, an electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor, which when executed implements the steps of the reservoir level prediction method as described above.
Compared with the prior art, the invention has the beneficial effects that:
the convolution network provided by the invention can extract low-level characteristics of input data from time dimension and space dimension, reduces network layer and convolution parameters, ensures that the output of the continuous GRU network can be divided into the output of a historical time period and a prediction time period, the output of the historical time period is the same as that of the traditional GRU network, the output of the future time period depends on the acquired historical data, the output result of the network at the last time and the output of the hidden layer of the network, and the fully connected network carries out further spatial fusion and characteristics on the characteristics extracted by the continuous GRU, thereby finally realizing the prediction of the water level after a longer time.
Additional aspects of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a flow chart of a reservoir level prediction method according to an embodiment of the present invention;
FIG. 2 is a training sample construction flow chart of an embodiment of the present invention;
FIG. 3 is a schematic diagram of a reservoir level prediction model according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a convolutional network of an embodiment of the present invention;
FIG. 5 is a schematic diagram of the result of a continuous GRU network model in accordance with an embodiment of the invention;
FIG. 6 is a schematic diagram of a fully connected network according to an embodiment of the invention;
fig. 7 is a graph comparing the predicted result with the true water level according to the embodiment of the present invention.
Detailed Description
The invention will be further described with reference to the drawings and examples.
It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
Term interpretation:
padding operation: for filling the area around the image so that the output feature map reaches a desired size.
Example 1
Referring to fig. 1, the embodiment provides a reservoir water level prediction method, which specifically includes the following steps:
step 1: and acquiring rainfall historical information, rainfall average historical information and reservoir water level historical information of a plurality of sites.
Step 2: and predicting reservoir water level information at a plurality of moments in the future based on the acquired historical information and the trained reservoir water level prediction model.
The reservoir water level prediction model comprises a convolution network, a continuous GRU network and a full-connection network; the convolution network is used for extracting time information and space information of the history information; the continuous GRU network is used for predicting states of a plurality of time points in the future based on information extracted by the convolution network and outputting the combination of the historical output state and the future output state; the fully-connected network is used for fusing the historical output state and the future output state to obtain reservoir water level information at a plurality of moments in the future.
In the specific implementation process, training samples of a reservoir water level prediction model by sequentially arranging rainfall historical information, rainfall average value historical information and reservoir water level historical information of a plurality of stations at a plurality of moments; one sample contains two aspects of information, one is time information, and the other is rainfall information of a plurality of sites at the same moment.
Specifically, rainfall data, average value and reservoir water level data of a plurality of sites of the reservoir for 10 years are collected. Further, data combination is performed, and missing data is complemented by an average value of related data at the upper and lower moments. Further, all data before 2020 is divided into training databases, and data after 2020 is divided into test databases. Further, the data are shaped respectively, and the process is as follows: the dimension of the data was set to 10, representing 8 sites of rainfall information, rainfall average information, and reservoir level information, respectively. Further, the time window is set to 56, and the data at 56 times are sequentially arranged together to form a sample, so that the sample has 2 dimensions in the form of [56,10 ]]Wherein 56 represents data of 56 continuous moments in a sample, 10 represents 8 sites rainfall information, rainfall average value and historical reservoir water level information, 1-8 of 10 represents 8 sites rainfall information, 9 represents rainfall average value, and 10 represents historical reservoir water level information. The window movement step size is set to 3, and a new step size is generated every 3 time sample points are moved in all data. The whole process is shown in fig. 2, wherein the time window represents the time window before movement, sample 1 is generated, the dotted line portion represents the time window after movement, and sample 2 is represented. Further, the training database and the test database are operated respectively according to the settings to obtain a training set and a test set for training and testing the subsequently designed model. Further, the method comprises the steps of,normalizing each dimension of the data set, wherein the normalization formula is thatRight side of formula +.>Representing the original data>Representing minimum data, +.>For maximum data, formula left +.>Is normalized data.
Then, the network model for predicting the water level is structurally designed, and a structural outline diagram is shown in fig. 3. The convolution model starts from two dimensions of time and space, low-level features of input historical data are extracted, the continuous GRU network is used for taking output of the convolution model as input, multi-step feature extraction tasks of a historical time period and a future time period are completed, the full-connection network takes output of the continuous GRU as input, feature information of the historical time period and the future time period is fused, feature extraction is further carried out, and finally multi-step prediction tasks of reservoir water level are completed.
Specifically, the convolution network firstly carries out padding operation on an input sample, and then a convolution kernel convolves vectors of the first two moments and the complementary 0 of the sample at a first moment to obtain a first output vector; then the convolution kernel moves rightwards by one step, the convolution result replaces a first input value of the position of the convolution kernel in the sample, and then the output result at the previous moment and the other two vectors in the sample are sent into the convolution kernel together to obtain a second output vector; and by analogy, replacing the first vector operation of the current convolution input by convolution operation, convolution displacement and the last output result to finally obtain a two-dimensional vector, wherein the length of the first dimension of the vector is the length of time and represents the time length to be predicted; the second dimension has a length of 8 and represents the key features extracted by the model.
The convolution model structure is shown in fig. 4, and the model is different from the conventional convolution model in that the size of the convolution kernel is set to 3, the step size is set to 1, the channels are set to 10, and the number is also set to 10. Firstly, carrying out padding operation on an input sample, and then, carrying out convolution on vectors of the first two moments and the complementary 0 of the sample at the moment 1 by a convolution kernel to obtain an output vectorThis operation may fuse the feature information in the spatial dimension. Then the convolution kernel is shifted one step to the right and the convolution result is +.>First input value of the position of the convolution kernel in the substitution sample +.>Then output the result at the last moment with +.>、/>The vectors are fed into the convolution kernel together to obtain an output vector +.>This operation may fuse the feature information in the time dimension. And so on, the first vector of the current convolution input is replaced by convolution operation, convolution displacement and the last output result to finally obtain a heap vector +.>The length of the heap vector is 56, and the number of channels is 10. The convolution model has the advantages that the time information and the space information of the input samples are fused at the same time, and the time convolution can be realized without constructing a plurality of network layers and convolution kernels.
The result of the continuous GRU network model is shown in FIG. 5, and the output of the convolution model is taken as input to the delayIn a continuous GRU. The output at time t depends on the input at time t and the hidden state at time t-1. Unlike conventional GRUs, a continuous GRU is a multi-step model that predicts states at 56 times into the future using states at the first 56 times, at the historical data t 1 -t 56 The time, the output of the current time depends on the state of the previous time; and at t 57 -t 112 The output of the model at the current time depends on the output of the previous time and the hidden state of the previous time. Finally, the output of the model is a combination of hidden states at the historical time and future time. Compared with the traditional GUP, the continuous GPU can predict states of a plurality of time points in the future, and the historical output states are combined with the future output states, so that longer-time state data can be seen, and accurate prediction of water levels at a plurality of time points in the future can be facilitated. To prevent model overfitting, the network uses a Dropout method of variational reasoning, with the same dropping method being used on the cyclical connection at all its moments for the same sequence. Compared with the common dropout, the memory capacity of the recurrent neural network can be ensured not to be reduced, and important information can be ensured not to be lost.
As shown in fig. 6, the output sequence of the continuous neural network is fed as an input into a fully-connected network containing 56 neurons, which outputs water level prediction results for a plurality of time points of the reservoir at 56 time points (7 days) in the future. The function of the fully-connected neural network is mainly to perform feature fusion on features extracted in a historical time period and features extracted in a future time period by the continuous GRU network, so that information extraction in input feature time and space is completed, and finally reservoir water level prediction tasks at 56 moments in the future are completed.
Next, the model is trained using a training set, one sample of which contains 8 site rainfall information, average rainfall information, and water level information for reservoirs at the past 56 times. The super parameters are set before training, the training round is set to 100, the initial training parameters are set to 0.001, and the gradient truncation is set to 15, and the loss function uses a mean square error loss function (MSE) with the formula ofIn the formula +.>Output result of representative model,/->Representing the true value of the sample,/->Representing the batch size of the sample. In the training process, an Adam optimizer is used, when the model is continuous for 5 rounds and the loss is not reduced, the learning rate is set to be 0.1 times of the original learning rate, and when the loss function is continuous for 20 rounds and the loss is not reduced, the network stops training.
And then, testing the trained model by using the test set, evaluating the test result, and sending the test set and the training set into the model together after the evaluation result accords with the expectation, and fine-tuning the model parameters again to finally obtain the model required in actual use. If the evaluation result does not meet the expectation, the model is retrained, tested and evaluated until the evaluation result meets the expectation.
And finally, deploying the model into a server, and returning the result to the design front-end interface after the model calculates the result by calling the data in the database and transmitting the data into the deployed model by the server. The model is tested by using data after 2020, and the predicted result is compared with the real water level, and the comparison result is shown in fig. 7.
Example two
The embodiment provides a reservoir water level prediction device, which specifically comprises the following modules:
a historical information acquisition module for acquiring rainfall historical information, rainfall average historical information and reservoir water level historical information of a plurality of sites;
the reservoir water level prediction module is used for predicting reservoir water level information at a plurality of moments in the future based on the acquired historical information and a trained reservoir water level prediction model;
the reservoir water level prediction model comprises a convolution network, a continuous GRU network and a full-connection network; the convolution network is used for extracting time information and space information of the history information; the continuous GRU network is used for predicting states of a plurality of time points in the future based on information extracted by the convolution network and outputting the combination of the historical output state and the future output state; the fully-connected network is used for fusing the historical output state and the future output state to obtain reservoir water level information at a plurality of moments in the future.
It should be noted that, each module in the embodiment corresponds to each step in the first embodiment one to one, and the implementation process is the same, which is not described here.
Example III
The present embodiment provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the reservoir level prediction method as described above.
Example IV
The embodiment provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the steps in the reservoir water level prediction method.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (6)
1. A reservoir level prediction method, comprising:
acquiring rainfall historical information, rainfall average historical information and reservoir water level historical information of a plurality of sites;
based on the acquired historical information and the trained reservoir water level prediction model, predicting reservoir water level information at a plurality of moments in the future;
the reservoir water level prediction model comprises a convolution network, a continuous GRU network and a full-connection network; the convolution network is used for extracting time information and space information of the history information; the continuous GRU network is used for predicting states of a plurality of time points in the future based on information extracted by the convolution network and outputting the combination of the historical output state and the future output state; the fully-connected network is used for fusing the historical output state and the future output state to obtain reservoir water level information at a plurality of moments in the future;
the convolutional network is used for: replacing the first vector operation of the current convolution input by convolution operation, convolution displacement and the last output result to finally obtain a heap vector; the convolution network firstly carries out packing operation on an input sample, and then a convolution kernel convolves vectors of the first two moments and the complementary 0 of the sample at a first moment to obtain a first output vector; then the convolution kernel moves rightwards by one step, the convolution result replaces a first input value of the position of the convolution kernel in the sample, and then the output result at the previous moment and the other two vectors in the sample are sent into the convolution kernel together to obtain a second output vector; and by analogy, replacing the first vector operation of the current convolution input by convolution operation, convolution displacement and the last output result to finally obtain a two-dimensional vector, wherein the length of the first dimension of the vector is the length of time and represents the time length to be predicted; the second dimension represents key features extracted by the model;
the output of the continuous GRU network is divided into the output of a historical time period and a predicted time period, wherein the output of the historical time period depends on collected historical data, and the output of the future time period depends on the output result of the network at the last moment and the hidden layer output thereof;
the continuous GRU network adopts the states of the first n times to predict the states of the next n times, and the states of the last n times are recorded in the historical data t 1 -t n The time, the output of the current time depends on the state of the previous time; and at t n+1 -t m The output of the current moment of the model depends on the output of the previous moment and the hidden state of the previous moment, and finally, the output of the model is the combination of the hidden states of the historical moment and the future moment;
the continuous GRU network uses a Dropout method of variational reasoning, and adopts the same discarding method on the cyclic connection of the same sequence at all the moments.
2. The reservoir level prediction method as set forth in claim 1, wherein the samples for training the reservoir level prediction model are sequentially arranged together by rainfall historical information, rainfall average historical information and reservoir level historical information of a plurality of sites at a plurality of times; one sample has two dimensions, one dimension is time information, and the other dimension is rainfall information of a plurality of sites at the same moment.
3. A reservoir level prediction apparatus, comprising:
a historical information acquisition module for acquiring rainfall historical information, rainfall average historical information and reservoir water level historical information of a plurality of sites;
the reservoir water level prediction module is used for predicting reservoir water level information at a plurality of moments in the future based on the acquired historical information and a trained reservoir water level prediction model;
the reservoir water level prediction model comprises a convolution network, a continuous GRU network and a full-connection network; the convolution network is used for extracting time information and space information of the history information; the continuous GRU network is used for predicting states of a plurality of time points in the future based on information extracted by the convolution network and outputting the combination of the historical output state and the future output state; the fully-connected network is used for fusing the historical output state and the future output state to obtain reservoir water level information at a plurality of moments in the future;
the convolutional network is used for: replacing the first vector operation of the current convolution input by convolution operation, convolution displacement and the last output result to finally obtain a heap vector; the convolution network firstly carries out packing operation on an input sample, and then a convolution kernel convolves vectors of the first two moments and the complementary 0 of the sample at a first moment to obtain a first output vector; then the convolution kernel moves rightwards by one step, the convolution result replaces a first input value of the position of the convolution kernel in the sample, and then the output result at the previous moment and the other two vectors in the sample are sent into the convolution kernel together to obtain a second output vector; and by analogy, replacing the first vector operation of the current convolution input by convolution operation, convolution displacement and the last output result to finally obtain a two-dimensional vector, wherein the length of the first dimension of the vector is the length of time and represents the time length to be predicted; the second dimension represents key features extracted by the model;
the output of the continuous GRU network is divided into the output of a historical time period and a predicted time period, wherein the output of the historical time period depends on collected historical data, and the output of the future time period depends on the output result of the network at the last moment and the hidden layer output thereof; the continuous GRU network adopts the states of the first n times to predict the states of the next n times, and the states of the last n times are recorded in the historical data t 1 -t n The time, the output of the current time depends on the state of the previous time; and at t n+1 -t m The output of the current moment of the model depends on the output of the previous moment and the hidden state of the previous moment, and finally, the output of the model is the combination of the hidden states of the historical moment and the future moment;
the continuous GRU network uses a Dropout method of variational reasoning, and adopts the same discarding method on the cyclic connection of the same sequence at all the moments.
4. A reservoir level prediction apparatus as set forth in claim 3, wherein the samples for training the reservoir level prediction model are sequentially arranged together by rainfall histories, average rainfall histories, and reservoir level histories at a plurality of sites at a plurality of times; one sample has two dimensions, one dimension is time information, and the other dimension is rainfall information of a plurality of sites at the same moment.
5. A computer readable storage medium having stored thereon a computer program, which when executed by a processor performs the steps of the reservoir level prediction method as claimed in any one of claims 1 to 2.
6. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor performs the steps in the reservoir level prediction method of any one of claims 1-2 when the program is executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310112898.0A CN115879646B (en) | 2023-02-15 | 2023-02-15 | Reservoir water level prediction method, device, medium and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310112898.0A CN115879646B (en) | 2023-02-15 | 2023-02-15 | Reservoir water level prediction method, device, medium and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115879646A CN115879646A (en) | 2023-03-31 |
| CN115879646B true CN115879646B (en) | 2023-11-07 |
Family
ID=85761113
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310112898.0A Active CN115879646B (en) | 2023-02-15 | 2023-02-15 | Reservoir water level prediction method, device, medium and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115879646B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2019094640A (en) * | 2017-11-20 | 2019-06-20 | 日本無線株式会社 | Water level prediction method, water level prediction program and water level prediction device |
| CN110991776A (en) * | 2020-03-04 | 2020-04-10 | 浙江鹏信信息科技股份有限公司 | Method and system for realizing water level prediction based on GRU network |
| CN111242344A (en) * | 2019-12-11 | 2020-06-05 | 大连海事大学 | Intelligent water level prediction method based on cyclic neural network and convolutional neural network |
| CN112561191A (en) * | 2020-12-22 | 2021-03-26 | 北京百度网讯科技有限公司 | Prediction model training method, prediction method, device, apparatus, program, and medium |
| CN113947259A (en) * | 2021-11-09 | 2022-01-18 | 吉林大学 | Method for predicting vehicle speeds of drivers in different styles based on GRU neural network |
| CN115034497A (en) * | 2022-06-27 | 2022-09-09 | 武汉理工大学 | Multi-site daily water level prediction method and device, electronic equipment and computer medium |
| CN115186857A (en) * | 2022-03-24 | 2022-10-14 | 国家能源集团西藏尼洋河流域水电开发有限公司 | Neural network reservoir water level prediction method based on ensemble learning |
| CN115238855A (en) * | 2022-05-23 | 2022-10-25 | 北京邮电大学 | Completion method of time sequence knowledge graph based on graph neural network and related equipment |
| CN115310536A (en) * | 2022-08-06 | 2022-11-08 | 福建中锐网络股份有限公司 | Reservoir water level prediction early warning method based on neural network and GCN deep learning model |
| CN115310532A (en) * | 2022-08-04 | 2022-11-08 | 福建中锐网络股份有限公司 | Basin multipoint prediction early warning method based on space-time association mixed deep learning model |
-
2023
- 2023-02-15 CN CN202310112898.0A patent/CN115879646B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2019094640A (en) * | 2017-11-20 | 2019-06-20 | 日本無線株式会社 | Water level prediction method, water level prediction program and water level prediction device |
| CN111242344A (en) * | 2019-12-11 | 2020-06-05 | 大连海事大学 | Intelligent water level prediction method based on cyclic neural network and convolutional neural network |
| CN110991776A (en) * | 2020-03-04 | 2020-04-10 | 浙江鹏信信息科技股份有限公司 | Method and system for realizing water level prediction based on GRU network |
| CN112561191A (en) * | 2020-12-22 | 2021-03-26 | 北京百度网讯科技有限公司 | Prediction model training method, prediction method, device, apparatus, program, and medium |
| CN113947259A (en) * | 2021-11-09 | 2022-01-18 | 吉林大学 | Method for predicting vehicle speeds of drivers in different styles based on GRU neural network |
| CN115186857A (en) * | 2022-03-24 | 2022-10-14 | 国家能源集团西藏尼洋河流域水电开发有限公司 | Neural network reservoir water level prediction method based on ensemble learning |
| CN115238855A (en) * | 2022-05-23 | 2022-10-25 | 北京邮电大学 | Completion method of time sequence knowledge graph based on graph neural network and related equipment |
| CN115034497A (en) * | 2022-06-27 | 2022-09-09 | 武汉理工大学 | Multi-site daily water level prediction method and device, electronic equipment and computer medium |
| CN115310532A (en) * | 2022-08-04 | 2022-11-08 | 福建中锐网络股份有限公司 | Basin multipoint prediction early warning method based on space-time association mixed deep learning model |
| CN115310536A (en) * | 2022-08-06 | 2022-11-08 | 福建中锐网络股份有限公司 | Reservoir water level prediction early warning method based on neural network and GCN deep learning model |
Non-Patent Citations (7)
| Title |
|---|
| Condition monitoring of wind turbines based on spatio-temporal fusion of SCADA data by convolutional neural networks and gated recurrent units;Ziqian Kong，Baoping Tang，，，Yan Han;《Renewable Energy》;第146卷;Pages 760-768 * |
| 人机混驾环境下基于LSTM的无人驾驶车辆换道行为模型;黄玲;郭亨聪;张荣辉;吴建平;;《中国公路学报》(第07期);全文 * |
| 基于GRU-BP组合模型的湖泊水位预测方法探索;刘惟飞，陈兵，余周;《中国农村水利水电》(第11期);全文 * |
| 基于GRU和LightGBM特征选择的水位时间序列预测模型;许国艳;周星熠;司存友;胡文斌;刘凡;;计算机应用与软件(第02期);全文 * |
| 基于长短时记忆神经网络的鄱阳湖水位预测;郭燕;赖锡军;;湖泊科学(第03期);全文 * |
| 基于门控循环单元—支持向量回归组合模型的湖泊水位预测方法探索;刘惟飞，陈兵，余周;《科学技术与工程》;第22卷(第33期);全文 * |
| 陈云霁等.《智能计算系统》.机械工业出版社,2020,第89-90页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115879646A (en) | 2023-03-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110647900B (en) | Intelligent safety situation prediction method, device and system based on deep neural network | |
| CN113313947B (en) | Road condition evaluation method of short-term traffic prediction graph convolution network | |
| Taghi Sattari et al. | M5 model tree application in daily river flow forecasting in Sohu Stream, Turkey | |
| US20210232376A1 (en) | Vectorized representation method of software source code | |
| CN111027686B (en) | Landslide displacement prediction method, device and equipment | |
| CN110046787A (en) | A kind of urban area charging demand for electric vehicles spatio-temporal prediction method | |
| CN111726243B (en) | Method and device for predicting node state | |
| CN106997474A (en) | A kind of node of graph multi-tag sorting technique based on deep learning | |
| CN103544496A (en) | Method for recognizing robot scenes on basis of space and time information fusion | |
| CN115935834B (en) | Historical fitting method based on deep autoregressive network and continuous learning strategy | |
| Qian et al. | Physics informed data driven model for flood prediction: Application of deep learning in prediction of urban flood development | |
| CN109787821B (en) | Intelligent prediction method for large-scale mobile client traffic consumption | |
| CN111523713A (en) | Method and device for predicting residual oil saturation distribution in oil field | |
| CN111738520A (en) | System load prediction method fusing isolated forest and long-short term memory network | |
| CN114492211B (en) | Residual oil distribution prediction method based on autoregressive network model | |
| CN113449919B (en) | Power consumption prediction method and system based on feature and trend perception | |
| CN114330101A (en) | Slope earthquake slip prediction method and system based on artificial neural network | |
| Xie et al. | Research of PM2. 5 prediction system based on CNNs-GRU in Wuxi urban area | |
| CN115660135A (en) | Traffic flow prediction method and system based on Bayes method and graph convolution | |
| CN111260082A (en) | Spatial object motion trajectory curve prediction method based on neural network | |
| KR20210093875A (en) | Video analysis methods and associated model training methods, devices, and devices | |
| CN113779882A (en) | Method, device, equipment and storage medium for predicting residual service life of equipment | |
| CN113807951A (en) | Transaction data trend prediction method and system based on deep learning | |
| CN115345222A (en) | Fault classification method based on TimeGAN model | |
| CN115062779A (en) | Event prediction method and device based on dynamic knowledge graph |
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 |