CN117421558A - Cascade reservoir operation rule extraction and model training method thereof - Google Patents
Cascade reservoir operation rule extraction and model training method thereof Download PDFInfo
- Publication number
- CN117421558A CN117421558A CN202311418406.7A CN202311418406A CN117421558A CN 117421558 A CN117421558 A CN 117421558A CN 202311418406 A CN202311418406 A CN 202311418406A CN 117421558 A CN117421558 A CN 117421558A
- Authority
- CN
- China
- Prior art keywords
- reservoir
- forecast
- period
- rule extraction
- cascade
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000605 extraction Methods 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012549 training Methods 0.000 title claims abstract description 23
- 238000001556 precipitation Methods 0.000 claims abstract description 52
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 16
- 238000003860 storage Methods 0.000 claims abstract description 13
- 230000008878 coupling Effects 0.000 claims abstract description 3
- 238000010168 coupling process Methods 0.000 claims abstract description 3
- 238000005859 coupling reaction Methods 0.000 claims abstract description 3
- 238000012360 testing method Methods 0.000 claims description 10
- 238000005259 measurement Methods 0.000 claims description 7
- 230000000694 effects Effects 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 230000009467 reduction Effects 0.000 claims description 4
- 238000012876 topography Methods 0.000 claims description 3
- 230000004931 aggregating effect Effects 0.000 claims description 2
- 230000009429 distress Effects 0.000 claims description 2
- 238000012544 monitoring process Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 abstract description 12
- 230000008859 change Effects 0.000 abstract description 2
- 238000002474 experimental method Methods 0.000 abstract 1
- 240000002834 Paulownia tomentosa Species 0.000 description 6
- 235000010678 Paulownia tomentosa Nutrition 0.000 description 6
- 238000004088 simulation Methods 0.000 description 6
- 238000011144 upstream manufacturing Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000004044 response Effects 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/047—Probabilistic or stochastic networks
-
- 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
-
- 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
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- Water Supply & Treatment (AREA)
- General Business, Economics & Management (AREA)
- Public Health (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Probability & Statistics with Applications (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a cascade reservoir operation rule extraction and model training method, and belongs to the technical field of water conservancy. Firstly, collecting historical operation data of a cascade reservoir, actually measured precipitation data of a river basin grid and forecast precipitation data of the river basin grid, establishing a hydrologic model, and determining the effective prediction period of the source reservoir storage flow forecast and the subinterval grid precipitation forecast; then, utilizing the source reservoir storage forecast flow and subinterval grid forecast precipitation of each period in the effective forecast period to construct input factor sets under different effective forecast periods; coupling ConvLSTM and LSTM, and constructing a cascade reservoir operation rule extraction model of associated hydrological space-time information; and finally, training by using the input factor set to obtain an optimal cascade reservoir operation rule extraction model. As shown by a comparison experiment, the cascade reservoir operation rule extraction model constructed by the method can simulate the outlet flow change process of reservoirs in different periods more accurately.
Description
Technical Field
The invention belongs to the technical field of water conservancy, and particularly relates to a cascade reservoir operation rule extraction and model training method thereof.
Background
To fully utilize water resources, many countries have been built and put into use to meet the needs of human production and life. About 10 thousands of reservoirs are built in China and are used for water supply, flood prevention, drought resistance, urban electricity utilization and the like of residents. However, the construction of large reservoirs inevitably disturbs the spatial-temporal distribution of water resources, making the original production and collection mechanisms unsuitable. Therefore, the scientific simulation of the operation rule of the reservoir has important significance for water circulation research, comprehensive benefit of the reservoir and sustainable development promotion.
Along with the continuous operation of the cascade reservoir, the historical operation data reflects decision information of a large number of operation scenes. Therefore, the data mining technology and the artificial intelligence model are utilized to mine the historical operation data, and the actual operation rules can be extracted. Common artificial intelligence models include support vector machines, neural networks, and tree models. In recent years, with the rise of deep learning, long-short-term memory neural networks and the like are introduced into the extraction of reservoir operation rules due to strong feature extraction and nonlinear fitting capability. Therefore, the artificial intelligent model has good performance and application prospect in reservoir operation rule extraction.
However, the artificial intelligence model has been rapidly researched and developed in reservoir operation, but some problems remain to be further discussed. The main problems are as follows: (1) The rule extraction method based on the artificial intelligent model mostly only focuses on the relationship between the reservoir self variable and the outlet flow, and neglects the influence of the upstream and downstream reservoir state and interval inflow; (2) With the continuous improvement of precipitation prediction technology, how to utilize effective prediction information to improve the accuracy of rule extraction still needs further research. Therefore, it is necessary to provide a step reservoir operation rule extraction method capable of fully considering the watershed hydrological weather space-time associated information, so that the simulation precision of the step reservoir scheduling operation process is improved, and service guidance is provided for reservoir operation decision makers.
Disclosure of Invention
Aiming at the defects or improvement demands of the prior art, the invention provides a cascade reservoir operation rule extraction and model training method thereof, which aims to solve the technical problem that the simulation precision is insufficient because the prior reservoir simulation technology does not consider the watershed hydrological weather space-time information.
To achieve the above object, in a first aspect, the present invention provides a training method for extracting a model of a running rule of a step reservoir, the method comprising:
collecting historical operation data of a cascade reservoir, actually measured drainage data of a drainage basin grid and forecast drainage data of the drainage basin grid, establishing a hydrologic model, and determining a source reservoir warehousing flow forecast and an effective forecast period of subinterval grid drainage forecast;
utilizing the source reservoir storage forecast flow and subinterval grid forecast precipitation of each period in the effective forecast period to construct input factor sets under different effective forecast periods; coupling ConvLSTM and LSTM, and constructing a cascade reservoir operation rule extraction model of associated hydrological space-time information;
and training by using the input factor set to obtain an optimal cascade reservoir operation rule extraction model.
Preferably, the effective distress is determined by the following method:
dividing a forecasting subinterval according to the distribution condition of the cascade reservoir and the river basin topography characteristic;
carrying out reduction treatment on the reservoir storage flow, dividing the monitoring flow data of each reservoir and the drainage basin grid actual measurement precipitation data into a rate period and a checking period, driving a hydrological model by using the drainage basin grid actual measurement precipitation data, and determining optimal model parameters of the hydrological model in each subinterval;
driving the hydrologic model trained in each subinterval by using drainage basin grid forecast precipitation data to obtain runoff forecast values of sections of reservoirs;
deriving a deterministic coefficient based on the predictive value, and selecting a predictive period with the deterministic coefficient being greater than or equal to a threshold value as a valid predictive period.
Preferably, the deterministic coefficient DC is:
wherein i=1, 2,3, …, n represents the number of foreseeable periods, O i The value of the observation is represented by a value,mean value of observed sequence, F i Representing the forecast values.
Preferably, the cascade reservoir operation rule extraction model comprises the following parts:
extracting grid precipitation data of all subintervals affecting reservoir delivery flow by using ConvLSTM;
and aggregating the extracted subinterval grid precipitation data and the operation data of the cascade reservoir to form an input factor driving LSTM, so as to obtain the outlet flow of the reservoir.
Preferably, the expression of the cascade reservoir operation rule extraction model is:
…
wherein F is ConvLSTM-LSTM () Representing a cascade reservoir operation rule extraction model coupled with ConvLSTM and LSTM; t represents an facing period; l represents the number of valid foreseeable periods; n represents the number of water reservoirs in the cascade reservoir;indicating the delivery flow of the nth reservoir during period t,>indicating the flow rate of the reservoir of the N-th reservoir, Z Initial,i Representing the initial water level of the reservoir i,grid precipitation representing the jth period of the ith subinterval; />Represents the jth period of the ith reservoirIs a delivery flow rate of the vehicle.
Preferably, the optimal cascade reservoir operation rule extraction model obtained by training the input factor set is specifically:
normalizing each input factor set;
dividing each input factor set into a training set and a testing set;
training the cascade reservoir operation rule extraction model under different input factor sets by using a training set, and optimizing super parameters by using a Bayesian optimization algorithm;
and (3) checking the performance of the cascade reservoir operation rule extraction model by using the test set, and evaluating the cascade reservoir operation rule extraction model by determining coefficients, root mean square errors and average relative errors to finally obtain the optimal cascade reservoir operation rule extraction model.
Preferably, the determining coefficients DC, the root mean square error RMSE and the average relative error MRE are respectively:
wherein i=1, 2,3, …, n represents the number of foreseeable periods, O i The value of the observation is represented by a value,mean value of observed sequence, F i Representing the forecast values.
In a second aspect, the invention provides a method for extracting operation rules of a cascade reservoir, which comprises the following steps:
the method comprises the steps of using the input forecast flow of a source reservoir in each period of the effective meeting period and the grid forecast precipitation of a subinterval as input factors to construct an input factor set;
driving a cascade reservoir operation rule extraction model through an input factor set to obtain a delivery forecast flow of each reservoir in the cascade reservoir;
evaluating the delivery forecast flow of each reservoir by determining the coefficient DC, the root mean square error RMSE and the average mean square error MRE, thereby determining the future time period number utilized when each reservoir reaches the optimal operation rule extraction effect, and then extracting the delivery forecast flow of the future time period number;
the cascade reservoir operation rule extraction model is trained according to any one of the methods in the first aspect.
In general, the above technical solutions conceived by the present invention have the following beneficial effects compared with the prior art:
(1) The reservoir operation rule extraction method model which is established by the invention and takes the hydrographic weather space-time associated information into consideration not only fully utilizes the future rainfall forecast information, but also fully considers the hydraulic connection between reservoirs in the cascade reservoirs and the production business requirements of each reservoir, and can simulate the operation process of the reservoirs more accurately;
(2) The cascade reservoir operation rule extraction model is coupled with ConvLSTM and LSTM, and ConvLSTM is utilized to extract grid precipitation data of all subintervals affecting reservoir delivery flow; then, the LSTM is utilized to gather the extracted subinterval grid precipitation data and the operation data of the cascade reservoir, and the extracted characteristics are simulated to obtain the reservoir outlet flow; therefore, the influence of space accumulated errors of the cascade reservoir forecast scheduling operation and mutual interference among multiple outputs can be effectively reduced, the cascade reservoir operation process under the condition of precipitation runoff uncertainty can be accurately simulated, the universality is high, the implementation is easy, and technical support can be provided for a cascade reservoir operation decision maker.
Drawings
FIG. 1 is a flow chart of a cascade reservoir operation rule extraction method taking into account hydrological and weather space-time associated information in an embodiment of the invention;
FIG. 2 is a schematic modeling diagram of a step reservoir operation rule extraction model in an embodiment of the invention;
FIG. 3 is a graph showing performance metrics of predicted runoff results at different forestation in an embodiment of the present invention;
FIG. 4 is a performance index of the result of extraction of the operating rules for each reservoir test set in an embodiment of the invention;
fig. 5 is a performance index of the step reservoir operation rule extraction result with forecast information as input in the embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The terms "first" and "second" and the like in the description and in the claims are used for distinguishing between different objects and not for describing a particular sequential order of objects. For example, the first response message and the second response message, etc. are used to distinguish between different response messages, and are not used to describe a particular order of response messages.
In embodiments of the invention, words such as "exemplary" or "such as" are used to mean serving as an example, instance, or illustration. Any embodiment or design described herein as "exemplary" or "e.g." in an embodiment should not be taken as preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "such as" is intended to present related concepts in a concrete fashion.
In the description of the embodiments of the present invention, unless otherwise specified, the meaning of "plurality" means two or more, for example, the meaning of a plurality of processing units means two or more, or the like; the plurality of elements means two or more elements and the like.
Next, the technical solution provided in the embodiment of the present invention will be described with reference to the accompanying drawings.
The embodiment of the invention discloses a cascade reservoir operation rule extraction method considering hydrological space-time associated information, which specifically comprises the following steps as shown in fig. 1:
(1) Collecting historical operation data, river basin grid actual measurement precipitation data and river basin grid forecast precipitation data of a cascade reservoir, and establishing a hydrologic model to determine the effective forecast period of source reservoir inflow forecast and interval precipitation forecast;
further, the step (1) includes the steps of:
the method comprises the steps of (1.1) collecting historical operation data of a cascade reservoir, actually measured precipitation data of a river basin grid and forecast precipitation data of the river basin grid;
according to actual business requirements, historical operation data of the cascade reservoirs are collected, wherein the historical operation data comprise the warehousing flow, the ex-warehouse flow, the initial water level process and the like of each reservoir; and collecting the actually measured precipitation data of the drainage basin grid with the spatial resolution of 0.1 degree multiplied by 0.1 degree and the forecast precipitation data of the drainage basin grid, wherein the time scale is the same as the reservoir operation data.
(1.2) establishing a hydrologic model to determine the effective prediction period of the source reservoir inflow prediction and subinterval precipitation prediction
The effective prediction period of the subinterval grid precipitation prediction can be reflected through the runoff prediction precision of the main control section of the river basin. According to the drainage basin topological structure, driving a hydrological model by subinterval grid precipitation data, determining the source reservoir storage flow forecast and the effective forecast period of interval precipitation forecast, and specifically comprising the following sub-steps:
step1: dividing a forecasting subinterval according to the distribution condition of the cascade reservoir and the river basin topography characteristic;
step2: carrying out reduction treatment on the reservoir storage flow, driving a hydrological model by using the river basin grid actual measurement precipitation data according to the monitored flow data of each reservoir and the river basin grid actual measurement precipitation data, and determining optimal model parameters of an upstream section and each section of a source reservoir;
step3: driving an upstream interval of a source reservoir and a hydrological model trained in each interval by using drainage basin grid forecast precipitation data to obtain runoff forecast values of sections of each reservoir;
step4: and selecting the foresight period with the certainty coefficient DC being more than or equal to 0.7 as the valid foresight period.
Wherein i=1, 2,3, …, n represents the number of sequences, O i The value of the observation is represented by a value,mean value of observed sequence, F i Representing the forecast values.
(2) Utilizing historical operation data of the cascade reservoir and actually measured precipitation data of the river basin grid, taking the reservoir outlet flow in the facing period of the reservoir as a model output variable, constructing input factor sets under different effective foreseeable periods, and adopting a multi-reservoir direct simulation strategy to establish a cascade reservoir operation rule extraction model considering hydrological weather space-time associated information;
further, the step (2) includes the steps of:
(2.1) constructing input/output factor sets under different valid foreseeing periods
And selecting the reservoir outlet flow in the facing period of the reservoir as a model output factor. Based on the hydraulic connection and scheduling requirements among the cascade reservoirs, the facing time period, the initial water level of each reservoir and the earlier-stage delivery flow of each reservoir are selected as fixed inputs. In addition, according to the source reservoir warehousing forecasting flow and the effective forecasting period of the subinterval grid forecasting precipitation, the source reservoir warehousing forecasting flow and the subinterval grid forecasting precipitation in each period in the effective forecasting period are sequentially added into the input factor set, and the model input and output factor sets under different effective forecasting periods are established. For example, the first factor set includes the source reservoir flow rate of a future period and grid precipitation information of each subinterval; the second factor set comprises the source reservoir storage flow of two future time periods and grid precipitation information of each subinterval; and so on, the number of factor sets is equal to the length of the valid foreseeing period.
(2.2) establishing a cascade reservoir operation rule extraction model considering hydrological space-time associated information
In the integral operation process of the cascade reservoir, the reservoir is dischargedThe reservoir flow is not only influenced by the current reservoir flow of the reservoir, but also influenced by the current states of the upstream reservoir and the downstream reservoir. The current state of the upstream and downstream reservoirs can be characterized by the current reservoir level and interval flow. Therefore, the reservoir operation rule extraction model F considering the hydrographic space-time related information ConvLSTM-LSTM Mainly comprises two processes:
extracting grid precipitation data of all subintervals affecting the delivery flow of a reservoir by using a convolutional long-short-term memory neural network, wherein the process is similar to a rainfall-runoff process;
2: and collecting the captured subinterval grid precipitation data and the operation data of the cascade reservoir to form an input factor-driven long-short-term memory neural network, so as to obtain the delivery flow of the reservoir.
On the basis, taking the step reservoirs as a whole, taking the facing time period, the primary water level of each reservoir, the earlier delivery flow of each reservoir, the source reservoir storage flow of each time period in the effective foreseeing period and grid precipitation of each subinterval as inputs, respectively predicting the delivery flow of each reservoir, wherein the expression is as follows:
...
wherein F is ConvLSTM-LSTM () Representing a cascade reservoir operation rule extraction model coupled with ConvLSTM and LSTM; t represents an facing period; l represents the number of valid foreseeable periods; n represents the number of water reservoirs in the cascade reservoir;representing the N reservoirDelivery flow in t period, +.>Indicating the flow rate of the reservoir of the N-th reservoir, Z Initial,i Representing the initial water level of the reservoir i,grid precipitation representing the jth period of the ith subinterval; />Indicating the delivery flow rate of the ith reservoir during the jth period.
The training process of the step reservoir operation rule extraction model comprises the following substeps:
step1: and carrying out normalization processing on each input factor set, wherein the normalization formula is as follows:
wherein x is * To process the input factor, x * ∈[0,1]The method comprises the steps of carrying out a first treatment on the surface of the x is an input factor, x max 、x min Maximum and minimum for each factor.
Step2: dividing each normalized input factor set into a training set and a testing set respectively;
step3: f under different factor sets by training set ConvLSTM-LSTM Training, namely optimizing super parameters such as model layer number, hidden layer number, characteristic dimension of an input vector and the like through a Bayesian optimization algorithm;
step4: and (3) utilizing the test set to test the performance of the model, evaluating the model simulation result through a Deterministic Coefficient (DC), a Root Mean Square Error (RMSE) and an average mean average error (MRE), and finally obtaining the optimal extraction model based on each reservoir operation rule under each factor set.
Wherein i=1, 2,3, …, n represents the number of sequences, O i The value of the observation is represented by a value,mean value of observed sequence, F i Representing the forecast values.
(3) And (3) taking effective forecast information as input, determining the number of forecast time periods utilized when each reservoir achieves the optimal operation rule extraction effect, and further realizing operation rule extraction of the step reservoirs under the facing time periods.
Further, the step (3) includes the steps of:
(3.1) taking the source reservoir storage forecast flow and subinterval grid forecast precipitation of each period in the effective meeting period as input factors to construct an input factor set;
(3.2) driving a cascade reservoir operation rule extraction model through an input factor set to obtain the delivery forecast flow of each reservoir in the cascade reservoir;
(3.3) evaluating the delivery forecast flow of each reservoir by determining the coefficient DC, the Root Mean Square Error (RMSE) and the average mean square error (MRE), so as to determine the number of future time periods utilized when each reservoir reaches the optimal operation rule extraction effect, and further realize the extraction of the cascade reservoir operation rule;
examples: five cascade reservoirs of elegance huller downstream brocade primary, brocade secondary, official, secondary beach and tung seed forest are taken as research objects, and a cascade reservoir operation rule extraction model for the hydrological weather time-space associated information is established. In this embodiment, the historical operation data and grid history of each reservoir and the forecast precipitation data cover 2015-01-2020-12-31 time periods, and data from 2019 to 2020 are used as test samples.
Firstly, carrying out reduction treatment on the storage flow of a cascade reservoir, driving the hydrologic model by 7 days of future forecast precipitation data through a hydrologic model of a first-level upstream of a first-level runoff brocade of a historical rainfall runoff sequence runoff and a first-level-tung forest zone of the brocade, and determining an effective forecast period;
secondly, selecting a facing period, the initial water level of each reservoir, the early delivery flow of each reservoir, the historical storage flow of the source reservoir in each period in the effective forecast period and the historical grid precipitation in each interval as inputs, and establishing model input and output factor sets under different effective forecast periods; on the basis, a direct strategy is adopted, a training sample set is taken as input, an operation rule extraction model considering the hydrological weather space-time associated information is established in each reservoir, and a Bayesian optimization algorithm is adopted to optimize the model super-parameters;
and finally, the forecast information in the effective forecast period is used as input to determine an optimal extraction model of each reservoir considering future forecast information, so that the operation rule extraction of the cascade reservoir is realized.
FIG. 3 shows the results of runoff forecasting accuracy for the first-order and Tung forest sections of the brocade screen at different forests. It can be seen that the accuracy of the runoff forecast decreases with the extension of the forecast period. When DC exceeds 0.7, the forestation periods corresponding to the first order of the brocade and the tung seed Lin Duanmian are 5 days and 3 days respectively, so that the forecast precipitation for 3 days in the future can be determined to be available for extracting the reservoir operation rules.
Fig. 4 and 5 are respectively the test set under the direct strategy and the extraction precision of the cascade reservoir operation rule with the forecast information as input. It can be seen from fig. 4 that the model performance of the mall secondary, official reservoir and the tung forest reservoir generally tended to decrease with increasing future information content, while the model performance of the mall primary and secondary reservoirs tended to increase. Therefore, the extraction accuracy of the primary and secondary reservoirs of the brocade, the official reservoir and the Tung forest reservoir are optimized when the forecast information of the first step in the future is added, and the model performance of the primary and secondary reservoirs of the brocade is optimized when the rainfall and inflow information of three steps in the future is increased. As can be seen from a comparison of fig. 4 and 5, when the forecast information is taken as input, the model performance of the first and second-order reservoirs of the brocade is optimal when using the future two-step forecast information, which is affected by uncertainty of the forecast information. Therefore, after the forecast information is checked, the effect of the extraction models of the first-order and second-beach reservoirs is optimal when the future two-step forecast information is considered, and the effect of the extraction models of the second-order, official reservoir and tung seed forest reservoirs is optimal when the future one-step forecast information is considered.
Table 1 compares the extraction accuracy of the step reservoir operating rules under different strategies. The invention can be seen that the direct strategy of the cascade reservoir operation rule extraction model is superior to the recursive strategy and the multi-input-multi-output strategy, can simulate the outlet flow change process of each reservoir in different periods more accurately, and can provide reliable decision reference values for a cascade reservoir operation decision maker.
TABLE 1
It will be appreciated that the various numerical numbers referred to in the embodiments of the present invention are merely for ease of description and are not intended to limit the scope of the embodiments of the present invention.
It will be readily appreciated by those skilled in the art that the foregoing is merely a preferred embodiment of the invention and is not intended to limit the invention, but any modifications, equivalents, improvements or alternatives falling within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (8)
1. A method for training a cascade reservoir operation rule extraction model, the method comprising:
collecting historical operation data of a cascade reservoir, actually measured drainage data of a drainage basin grid and forecast drainage data of the drainage basin grid, establishing a hydrologic model, and determining a source reservoir warehousing flow forecast and an effective forecast period of subinterval grid drainage forecast;
utilizing the source reservoir storage forecast flow and subinterval grid forecast precipitation of each period in the effective forecast period to construct input factor sets under different effective forecast periods; coupling ConvLSTM and LSTM, and constructing a cascade reservoir operation rule extraction model of associated hydrological space-time information;
and training by using the input factor set to obtain an optimal cascade reservoir operation rule extraction model.
2. The method of claim 1, wherein the effective distress period is determined by:
dividing a forecasting subinterval according to the distribution condition of the cascade reservoir and the river basin topography characteristic;
carrying out reduction treatment on the reservoir storage flow, dividing the monitoring flow data of each reservoir and the drainage basin grid actual measurement precipitation data into a rate period and a checking period, driving a hydrological model by using the drainage basin grid actual measurement precipitation data, and determining optimal model parameters of the hydrological model in each subinterval;
driving the hydrologic model trained in each subinterval by using drainage basin grid forecast precipitation data to obtain runoff forecast values of sections of reservoirs;
deriving a deterministic coefficient based on the predictive value, and selecting a predictive period with the deterministic coefficient being greater than or equal to a threshold value as a valid predictive period.
3. The method according to claim 2, characterized in that the deterministic coefficient DC is:
wherein i=1, 2,3, …, n represents the number of foreseeable periods, O i The value of the observation is represented by a value,mean value of observed sequence, F i Representing the forecast values.
4. The method of claim 1, wherein the step reservoir operating rule extraction model comprises the following:
extracting grid precipitation data of all subintervals affecting reservoir delivery flow by using ConvLSTM;
and aggregating the extracted subinterval grid precipitation data and the operation data of the cascade reservoir to form an input factor driving LSTM, so as to obtain the outlet flow of the reservoir.
5. The method of claim 1, wherein the expression of the cascade reservoir operating rule extraction model is:
…
wherein F is ConvLSTM-LSTM () Representing a cascade reservoir operation rule extraction model coupled with ConvLSTM and LSTM; t represents an facing period; l represents the number of valid foreseeable periods; n represents the number of water reservoirs in the cascade reservoir;indicating the delivery flow of the nth reservoir during period t,>indicating the flow rate of the reservoir of the N-th reservoir, Z Initial,i Representing the initial water level of the reservoir i,grid precipitation representing the jth period of the ith subinterval; />Indicating the delivery flow rate of the ith reservoir during the jth period.
6. The method of claim 1, wherein training the input factor set to obtain the optimal cascade reservoir operation rule extraction model is specifically:
normalizing each input factor set;
dividing each input factor set into a training set and a testing set;
training the cascade reservoir operation rule extraction model under different input factor sets by using a training set, and optimizing super parameters by using a Bayesian optimization algorithm;
and (3) checking the performance of the cascade reservoir operation rule extraction model by using the test set, and evaluating the cascade reservoir operation rule extraction model by determining coefficients, root mean square errors and average relative errors to finally obtain the optimal cascade reservoir operation rule extraction model.
7. The method of claim 6, wherein the determining coefficients DC, root mean square error RMSE and average relative error MRE are respectively:
wherein i=1, 2,3, …, n represents the number of foreseeable periods, O i The value of the observation is represented by a value,mean value of observed sequence, F i Representing the forecast values.
8. A method for extracting operational rules of a cascade reservoir, the method comprising:
the method comprises the steps of using the input forecast flow of a source reservoir in each period of the effective meeting period and the grid forecast precipitation of a subinterval as input factors to construct an input factor set;
driving a cascade reservoir operation rule extraction model through an input factor set to obtain a delivery forecast flow of each reservoir in the cascade reservoir;
evaluating the delivery forecast flow of each reservoir by determining the coefficient DC, the root mean square error RMSE and the average mean square error MRE, thereby determining the future time period number utilized when each reservoir reaches the optimal operation rule extraction effect, and then extracting the delivery forecast flow of the future time period number;
the cascade reservoir operation rule extraction model is trained according to the method of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311418406.7A CN117421558B (en) | 2023-10-26 | 2023-10-26 | Cascade reservoir operation rule extraction and model training method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311418406.7A CN117421558B (en) | 2023-10-26 | 2023-10-26 | Cascade reservoir operation rule extraction and model training method thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117421558A true CN117421558A (en) | 2024-01-19 |
| CN117421558B CN117421558B (en) | 2024-06-21 |
Family
ID=89522633
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311418406.7A Active CN117421558B (en) | 2023-10-26 | 2023-10-26 | Cascade reservoir operation rule extraction and model training method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117421558B (en) |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108647829A (en) * | 2018-05-16 | 2018-10-12 | 河海大学 | A kind of Hydropower Stations combined dispatching Rules extraction method based on random forest |
| CN111080152A (en) * | 2019-12-23 | 2020-04-28 | 华中科技大学 | Cascade reservoir power generation scheduling compensation electric quantity distribution method |
| CN111310968A (en) * | 2019-12-20 | 2020-06-19 | 西安电子科技大学 | LSTM neural network circulation hydrological forecasting method based on mutual information |
| KR20200087347A (en) * | 2018-12-31 | 2020-07-21 | 부산대학교 산학협력단 | Method and Apparatus for Rainfall Recognition from Road Surveillance Videos Using TSN |
| CN112182709A (en) * | 2020-09-28 | 2021-01-05 | 中国水利水电科学研究院 | Rapid prediction method for let-down water temperature of large-scale reservoir stop log door layered water taking facility |
| CN112232659A (en) * | 2020-10-15 | 2021-01-15 | 华中科技大学 | Cascade reservoir power generation scheduling method and system |
| CN115392128A (en) * | 2022-09-07 | 2022-11-25 | 黑河水资源与生态保护研究中心 | Method for simulating basin runoff by utilizing space-time convolution LSTM network |
| CN115456205A (en) * | 2022-07-25 | 2022-12-09 | 华南理工大学 | Large-scale multi-target unit maintenance optimization method |
| CN115640881A (en) * | 2022-10-09 | 2023-01-24 | 河海大学 | Reservoir water level correction method of reservoir warehousing runoff forecasting model based on LSTM |
| CN115906924A (en) * | 2022-12-14 | 2023-04-04 | 天津师范大学 | Error correction method for long-time sequence satellite remote sensing spectral characteristics |
| CN115933009A (en) * | 2022-12-13 | 2023-04-07 | 国网湖南省电力有限公司 | Reservoir precipitation forecasting method and system based on combination of dynamic statistics method |
| CN116542021A (en) * | 2023-04-04 | 2023-08-04 | 中国长江电力股份有限公司 | Hydrologic-hydrokinetic coupled river channel type reservoir flood regulating calculation method |
| CN116739286A (en) * | 2023-06-21 | 2023-09-12 | 河海大学 | Reservoir group optimal scheduling method, system and electronic equipment |
| US20230342392A1 (en) * | 2020-02-21 | 2023-10-26 | Brian MCCARSON | Generative ai systems and methods for economic analytics and forecasting |
-
2023
- 2023-10-26 CN CN202311418406.7A patent/CN117421558B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108647829A (en) * | 2018-05-16 | 2018-10-12 | 河海大学 | A kind of Hydropower Stations combined dispatching Rules extraction method based on random forest |
| KR20200087347A (en) * | 2018-12-31 | 2020-07-21 | 부산대학교 산학협력단 | Method and Apparatus for Rainfall Recognition from Road Surveillance Videos Using TSN |
| CN111310968A (en) * | 2019-12-20 | 2020-06-19 | 西安电子科技大学 | LSTM neural network circulation hydrological forecasting method based on mutual information |
| CN111080152A (en) * | 2019-12-23 | 2020-04-28 | 华中科技大学 | Cascade reservoir power generation scheduling compensation electric quantity distribution method |
| US20230342392A1 (en) * | 2020-02-21 | 2023-10-26 | Brian MCCARSON | Generative ai systems and methods for economic analytics and forecasting |
| CN112182709A (en) * | 2020-09-28 | 2021-01-05 | 中国水利水电科学研究院 | Rapid prediction method for let-down water temperature of large-scale reservoir stop log door layered water taking facility |
| CN112232659A (en) * | 2020-10-15 | 2021-01-15 | 华中科技大学 | Cascade reservoir power generation scheduling method and system |
| CN115456205A (en) * | 2022-07-25 | 2022-12-09 | 华南理工大学 | Large-scale multi-target unit maintenance optimization method |
| CN115392128A (en) * | 2022-09-07 | 2022-11-25 | 黑河水资源与生态保护研究中心 | Method for simulating basin runoff by utilizing space-time convolution LSTM network |
| CN115640881A (en) * | 2022-10-09 | 2023-01-24 | 河海大学 | Reservoir water level correction method of reservoir warehousing runoff forecasting model based on LSTM |
| CN115933009A (en) * | 2022-12-13 | 2023-04-07 | 国网湖南省电力有限公司 | Reservoir precipitation forecasting method and system based on combination of dynamic statistics method |
| CN115906924A (en) * | 2022-12-14 | 2023-04-04 | 天津师范大学 | Error correction method for long-time sequence satellite remote sensing spectral characteristics |
| CN116542021A (en) * | 2023-04-04 | 2023-08-04 | 中国长江电力股份有限公司 | Hydrologic-hydrokinetic coupled river channel type reservoir flood regulating calculation method |
| CN116739286A (en) * | 2023-06-21 | 2023-09-12 | 河海大学 | Reservoir group optimal scheduling method, system and electronic equipment |
Non-Patent Citations (3)
| Title |
|---|
| WEI FANG ET AL: "Extracting operation rule of cascade reservoirs using a novel framework considering hydrometeorological spatiotemporal information based on artificial intelligence models", 《JOURNAL OF CLEANER PRODUCTION》, vol. 437, no. 2024, 6 January 2024 (2024-01-06), pages 1 - 18 * |
| 潘信亮 等: "基于卷积长短期记忆网络的NDVI预测方法研究", 《地理信息世界》, vol. 27, no. 02, 30 April 2020 (2020-04-30), pages 60 - 67 * |
| 龚亮;: "深度学习法在水文中应用的现状", 《河南水利与南水北调》, vol. 49, no. 04, 30 April 2020 (2020-04-30), pages 33 - 34 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117421558B (en) | 2024-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113379109B (en) | Runoff forecasting method based on prediction model self-adaption | |
| Zhang et al. | Wind speed forecasting based on quantile regression minimal gated memory network and kernel density estimation | |
| CN109272146A (en) | A kind of Forecasting Flood method corrected based on deep learning model and BP neural network | |
| CN117236673B (en) | Urban water network multi-scale flood control and drainage combined optimization scheduling method and system | |
| CN104715292A (en) | City short-term water consumption prediction method based on least square support vector machine model | |
| CN113554466A (en) | Short-term power consumption prediction model construction method, prediction method and device | |
| CN117236199B (en) | Method and system for improving water quality and guaranteeing water safety of river and lake in urban water network area | |
| CN112801342A (en) | Adaptive runoff forecasting method based on rainfall runoff similarity | |
| CN109307159A (en) | A kind of pipe network model alarm method based on water consumption optimal prediction model | |
| CN101807045B (en) | Data-based urban sewage pumping station system modeling method | |
| CN116485584B (en) | Method and system for cooperative regulation and control of WEE of river with large bottom slope in alpine region | |
| CN111695666A (en) | Wind power ultra-short term conditional probability prediction method based on deep learning | |
| CN115374376A (en) | Small hydropower station ecological influence monitoring and evaluating method and system | |
| CN114819322B (en) | Forecasting method for flow of lake entering lake | |
| CN115169724A (en) | Runoff prediction method based on space-time graph convolutional neural network | |
| CN115186857A (en) | Neural network reservoir water level prediction method based on ensemble learning | |
| CN111598724A (en) | Time-interval integration method for day-ahead prediction of warehousing flow of small and medium-sized reservoirs | |
| CN111199298A (en) | Flood forecasting method and system based on neural network | |
| CN110135652B (en) | Long-term flood season runoff prediction method | |
| CN116205136A (en) | Large-scale river basin deep learning flood forecasting method based on runoff lag information | |
| CN108346009A (en) | A kind of power generation configuration method and device based on user model self study | |
| CN115329930A (en) | Flood process probability forecasting method based on mixed deep learning model | |
| Arsene et al. | Profiling consumers in a water distribution network using K-Means clustering and multiple pre-processing methods | |
| CN117421558B (en) | Cascade reservoir operation rule extraction and model training method thereof | |
| CN113836807B (en) | River and lake ecological flow forecasting and early warning method based on entropy method and long-term and short-term memory neural network |
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 |