CN112700068A - Reservoir dispatching rule optimization method based on machine learning fusion of multi-source remote sensing data - Google Patents
Reservoir dispatching rule optimization method based on machine learning fusion of multi-source remote sensing data Download PDFInfo
- Publication number
- CN112700068A CN112700068A CN202110053303.XA CN202110053303A CN112700068A CN 112700068 A CN112700068 A CN 112700068A CN 202110053303 A CN202110053303 A CN 202110053303A CN 112700068 A CN112700068 A CN 112700068A
- Authority
- CN
- China
- Prior art keywords
- series
- runoff
- reservoir
- long
- data
- 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
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000005457 optimization Methods 0.000 title claims abstract description 29
- 230000004927 fusion Effects 0.000 title claims abstract description 15
- 238000010801 machine learning Methods 0.000 title claims abstract description 14
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 58
- 230000015654 memory Effects 0.000 claims abstract description 31
- 238000003062 neural network model Methods 0.000 claims abstract description 28
- 238000001556 precipitation Methods 0.000 claims abstract description 22
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 18
- 230000002068 genetic effect Effects 0.000 claims abstract description 8
- 238000003860 storage Methods 0.000 claims abstract description 8
- 238000013528 artificial neural network Methods 0.000 claims description 20
- 238000010248 power generation Methods 0.000 claims description 10
- 230000006403 short-term memory Effects 0.000 claims description 10
- 230000007787 long-term memory Effects 0.000 claims description 9
- 230000008901 benefit Effects 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 8
- 238000012950 reanalysis Methods 0.000 claims description 7
- 229920006395 saturated elastomer Polymers 0.000 claims description 6
- 230000008018 melting Effects 0.000 claims description 5
- 238000002844 melting Methods 0.000 claims description 5
- 238000012935 Averaging Methods 0.000 claims description 4
- 230000010354 integration Effects 0.000 claims description 4
- 238000009834 vaporization Methods 0.000 claims description 4
- 230000008016 vaporization Effects 0.000 claims description 4
- 238000000746 purification Methods 0.000 claims description 3
- 238000007619 statistical method Methods 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 2
- 238000013439 planning Methods 0.000 abstract description 3
- 238000004088 simulation Methods 0.000 description 15
- 238000010586 diagram Methods 0.000 description 7
- 239000000047 product Substances 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 238000005259 measurement Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000012546 transfer Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000003973 irrigation Methods 0.000 description 2
- 230000002262 irrigation Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012384 transportation and delivery Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 239000012467 final product Substances 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000006386 memory function Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
-
- 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)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Entrepreneurship & Innovation (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- General Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Tourism & Hospitality (AREA)
- Molecular Biology (AREA)
- Marketing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Feedback Control In General (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a reservoir dispatching rule optimization method based on machine learning and multi-source remote sensing data fusion, which comprises the following steps: collecting short series runoff observation data of a reservoir, and extracting precipitation series, meteorological data and land water storage quantity series of a basin where the reservoir is located; establishing a hydrological model of a basin where the reservoir is located according to the short series runoff observation data and the meteorological data and primarily simulating runoff; constructing a long-short term memory neural network model and correcting the simulated runoff by adopting the long-short term memory neural network model to obtain a corrected simulated runoff series; inputting the collected long series meteorological data into a hydrological model and a corrected simulated runoff system, and simulating a long series warehousing runoff process of the reservoir; and constructing a multi-objective optimized scheduling model according to the obtained long series of reservoir inlet runoff, and solving an optimized scheduling rule by adopting a genetic algorithm. The invention integrates multi-source remote sensing data for simulating the long series runoff process, and provides a reference basis for reservoir scheduling and water resource planning.
Description
Technical Field
The invention belongs to the technical field of reservoir scheduling, and particularly relates to a reservoir scheduling rule optimization method based on machine learning and multi-source remote sensing data fusion.
Background
The hydrological meteorological data is a basic basis for engineering planning, design, construction and operation management, is also an important data for evaluating flood control risks of watershed hydraulic engineering, is related to comprehensive management of watershed water resources and water safety guarantee, and has important significance for national economy and social development. However, some reservoirs in China only have a small amount of actually measured hydrographic meteorological monitoring data, so how to invert long series runoff processes and guide reservoir operation scheduling is a major challenge for hydrographic workers.
In recent years, satellite telemetry and data inversion algorithms are rapidly developed, precipitation quantitative observation products based on satellite remote sensing inversion have wider coverage range and higher space-time resolution, the defect of insufficient arrangement of meteorological stations is effectively overcome, and new data reference is provided for data-free areas. Meanwhile, as human observation means and data assimilation technologies become mature day by day, students perform quality control on observation data from various sources (ground, ships, radiosonde, anemometry balloons, airplanes, satellites and the like), and propose a data assimilation technology for numerical weather forecast to reconstruct a long-term historical climate process, namely a so-called reanalysis data set, which assimilates the numerical weather forecast and a large amount of ground observation data and satellite remote sensing information, and has the advantages of high spatial and temporal resolution precision, long time span and the like. The scholars at home and abroad apply satellite remote measurement and reanalysis data as meteorological input in scarce data areas, and invert long series runoff series through hydrological models, thereby obtaining certain application effect.
However, the hydrological model is suitable for simulating a runoff process in a natural state, and engineering measures such as dams, reservoirs, agricultural irrigation, water diversion, cross-basin water transfer and the like often damage the consistency of underlying surfaces, so that the hydrological model in the basin has large errors, and the hydrological simulation precision is restricted. The existing literature cannot fully utilize satellite telemetering meteorological information, cannot consider errors caused by human activity interference on runoff simulation, cannot solve the problem of long-series runoff simulation in scarce data areas, and is difficult to be used in application practice for optimizing reservoir dispatching rules.
Disclosure of Invention
The invention aims to provide a reservoir dispatching rule optimization method aiming at the defects of the prior art, which considers the error caused by human activity interference on runoff simulation, solves the problem of long series runoff simulation in scarce data areas and provides a reference basis for reservoir dispatching and water resource planning.
In order to solve the technical problems, the invention adopts the following technical scheme:
a reservoir dispatching rule optimization method based on machine learning and multi-source remote sensing data fusion comprises the following steps:
s1, collecting short series runoff observation data of a reservoir, collecting engineering characteristics and scheduling operation related data of the reservoir, extracting precipitation series of a basin where the reservoir is located, and analyzing data set meteorological data and land water storage quantity series;
s2, establishing a hydrological model of the basin where the reservoir is located according to the short series runoff observation data and the meteorological data obtained in the step S1, and simulating the primary runoff through the hydrological model;
s3, constructing a long-short term memory neural network model and correcting the simulated runoff in the step S2 by using the long-short term memory neural network model to obtain a corrected simulated runoff series model;
s4, inputting the long series meteorological data collected in the step S1 into the hydrological model established in the step S2 and the simulated runoff system model corrected in the step S3, and simulating the long series warehousing runoff process of the reservoir;
and S5, constructing a multi-objective optimization scheduling model according to the long series of warehousing runoff data of the reservoir obtained in the step S4, and solving by adopting a genetic algorithm to obtain an optimized scheduling rule.
Further, the short series of runoff observation data in the step S1 includes sunrise reservoir flow of the reservoir, and daily water level data of the reservoir area, and the reanalysis data set meteorological data includes near-earth air temperature, dew point temperature and horizontal wind speed in an hour scale.
Further, step S2 further includes the following sub-steps:
step 2.1, extracting a daily maximum air temperature and a daily minimum air temperature series according to the meteorological data acquired in the step S1, and calculating a daily average dew point temperature, a daily average air temperature and a daily average wind speed;
step 2.2, acquiring short series runoff observation data of the reservoir according to the step S1, and calculating a warehousing daily runoff series according to a water level-reservoir capacity curve and a water quantity balance principle;
step 2.3, constructing a hydrological model for calibrating and considering the snow melting module according to the warehousing daily runoff series of the reservoir and precipitation data, daily maximum air temperature and daily minimum air temperature data of the same period;
and 2.4, simulating to obtain a long-series warehousing runoff process by inputting the long-series precipitation and the temperature data in the step 2.1 by adopting the calibrated hydrological model.
Further, step S3 further includes the following sub-steps:
step 3.1, extracting the daily average air temperature and the dew point temperature according to the meteorological data obtained in the step S1, and then calculating a relative humidity series according to the daily average air temperature and the dew point temperature;
step 3.2, determining the time lag influencing the daily measured runoff by carrying out statistical analysis on the daily runoff process simulated in the step S2 and the actually measured daily runoff process;
and 3.3, constructing a long-short term memory neural network with a three-layer neural network architecture according to the data obtained in the steps, calibrating the long-short term memory neural network model by using meteorological data, the simulated runoff series in the step S2 and the actually-measured runoff series as input through a calibrated long-short term memory neural network model, and correcting the simulated runoff series in the step S2 through the calibrated long-short term memory neural network model to obtain a corrected simulated runoff series model.
Further, the method for solving the relative humidity in step 3.1 is as follows: solving through a Clausius-Clappe-long equation and a given air temperature T to obtain the atmospheric saturated vapor pressure es：
In the formula: t is0And es0Are integration constants 273.16K and 611Pa, L respectivelyvIs latent heat of vaporization (2.5X 10)6J kg-1)，RvThen is the vapor gas constant (461J kg)-1K-1)；
Relative humidity RH ═ es(Tdew)/es(Tmean) Wherein, TmeanIndicating the average daily temperature, TdewT represents dew point temperature to be obtainedmeanAnd TdewRespectively correspondingly substituting into atmospheric saturated vapor pressure esThe relative humidity RH can be solved.
Further, the model of the simulated runoff series after correction is as follows: qcor(t)＝FLSTM[QM(t),QM(t-1),QM(t-2),…,QM(t-N)]；
In the formula: qcor(t) represents runoff corrected at time t; QM (t) represents input variables of the long-term and short-term memory neural network model after calibration, including a runoff series, a precipitation series, a highest daily temperature series, a lowest daily temperature series, a relative humidity series of days obtained in the step 3.1 and a land water reserve series simulated by a hydrological model; QM (t-1) represents a simulated runoff and meteorological series at the t-1 moment; n represents the time lag determined by the long-term and short-term memory neural network model; fLSTMAnd representing the long-term and short-term memory neural network model after calibration.
Further, in step S5, a multi-objective optimization scheduling model is established with the maximum water supply benefit and the maximum power generation amount as scheduling objectives, where the multi-objective optimization scheduling model is:
in the formula: w*(T) and E*(T) water supply (m) in the planned period T3) And power generation (kW · h);
and PtAverage water supply flow rate (m) of t-th time period3/s) and average output (kW); k is the comprehensive output coefficient of the power station;
generating flow (m) for t-th period3/s)；HtAveraging the power generation water purification heads (m) for the t-th period; mtThe number of hours in the t period; t is the length of the scheduling period;
the upper water level (m) of the reservoir dam in the t-th time period;
the water level (m) under the reservoir dam in the t-th time period; qtThe lower discharge quantity of the reservoir in the t-th time period; Δ HtHydropower station head loss (m) for a t-th period; f. ofZQ(x) is a downstream water level flow relation function; f. ofΔH(. x) is a hydropower station head loss function.
Compared with the prior art, the invention has the beneficial effects that: the invention fully utilizes the advantages of satellite remote sensing observation data, reanalysis data and land assimilation system data to obtain long series meteorological data of scarce data areas, simulates long series runoff according to the long series meteorological data, a constructed hydrological model and a long and short term memory neural network model, and finally constructs a multi-objective optimization scheduling model based on long series runoff data.
Drawings
FIG. 1 is a flowchart illustrating a scheduling optimization method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a GR4J hydrological model structure according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of the change of correlation coefficients of actual measurement runoff and simulated runoff at different time lags in an embodiment of the invention;
FIG. 4 is a diagram of a long short term memory neural network (LSTM) model memory unit according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a flow chart of a second generation non-dominated sorting genetic algorithm according to an embodiment of the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the following embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
The present invention is further illustrated by the following examples, which are not to be construed as limiting the invention.
The invention provides a reservoir dispatching rule optimization method based on machine learning and multi-source remote sensing data fusion, which comprises the steps of firstly collecting short series runoff observation data of a reservoir, collecting relevant information of engineering characteristics and dispatching operation of the reservoir, extracting precipitation series of a basin where the reservoir is located, analyzing data set meteorological information including hourly air temperature, relative humidity and horizontal wind speed, and collecting land water reserve series inverted by a Global Land Data Assimilation System (GLDAS); establishing a hydrological model of a reservoir basin based on short series runoff observation data and multi-source remote sensing meteorological data of scarce data areas, and realizing the simulation of preliminary runoff through the hydrological model; constructing a long-short term memory neural network model and correcting the simulated runoff in the step S2 by adopting the long-short term memory neural network model to obtain a corrected simulated runoff series model so as to reduce simulation errors caused by engineering measures such as dams, reservoirs, agricultural irrigation, water diversion, cross-basin water transfer and the like; inputting the long-series meteorological data acquired in the step S1 into the hydrological model established in the step S2 and the simulated runoff system model corrected in the step S3, namely, realizing long-series runoff simulation by utilizing long-series satellite telemetering data, a well-calibrated watershed hydrological model and a long-short term memory neural network model; and finally, constructing a multi-objective optimization scheduling model based on long series runoff data, and solving by adopting a genetic algorithm to obtain an optimized scheduling rule, wherein the detailed flow is shown in figure 1.
The technical scheme of the invention is further explained in detail by the following embodiments and the accompanying drawings:
and S1, collecting short series runoff observation data of the reservoir, collecting relevant engineering characteristic and scheduling operation data of the reservoir, and extracting a GPM satellite precipitation series and an ERA5 reanalysis data gas collection image data of a basin where the reservoir is located and a land water storage quantity series of a GLDAS system.
In the step, the short series runoff observation data comprises daily ex-warehouse flow and warehouse area daily water level series, and the embodiment aims at the reservoir in the scarce data area, so the acquired daily ex-warehouse flow and warehouse area daily water level series are shorter, and a longer series warehousing flow process needs to be simulated by combining a satellite remote sensing means.
IMERG is the latest generation multi-satellite fusion inversion precipitation data of the global precipitation plan GPM, is a grade 3 product of the GPM, fully utilizes data (including active and passive microwave sensors, various infrared data sensors and the like) provided by all satellite sensors on a GPM platform, and fully references various satellite precipitation inversion algorithms which are basically mature in the TRMM era before for organic fusion; IMERG currently provides three types of satellite precipitation data, namely Early, Late, Final versions, where the Final product incorporates global rainfall sites for calibration. The GPM satellite precipitation product adopted by the embodiment is an IMERG-Final data set.
In addition, the re-analyzed weather data set adopted in the present embodiment is the fifth generation re-analyzed weather product ERA5 of the european mid-term weather forecast center. The horizontal resolution of the hourly analysis field of the data set is 31km, 137 layers are vertically layered, and the top layer reaches the height of 0.01 hPa; the ERA5 adopts a Cycle31r2 model version of a comprehensive forecasting system, based on spectral harmonic resolution T255, and interpolates simplified Gaussian grid (N128) data to grids with different resolutions of 0.25-2.5 degrees and the like by a bilinear interpolation technology, so that the data is one of global reanalysis data with the highest space-time resolution at present. The variables used in the ERA5 re-analysis dataset in this example include near earth air temperature, dew point temperature, and horizontal wind speed on an hourly scale. Land water reserve data of the second generation products of the GLDAS employed in this embodiment.
And S2, establishing a hydrological model of the basin where the reservoir is located according to the short series runoff observation data and the meteorological data obtained in the step S1, and realizing the simulation of the primary runoff through the hydrological model. The step further comprises the sub-steps of:
step 2.1, extracting daily maximum air temperature (T) based on the hourly scale meteorological data of ERA5max) And daily minimum temperature (T)min) Series, and calculate the average daily dew point temperature (T)dew) Average daily temperature (T)mean) And the average daily wind speed;
step 2.2, based on the daily delivery flow of the reservoir and the daily water level data of the reservoir area collected in the step 1, calculating daily runoff series of warehousing according to a water level-reservoir capacity curve and a water balance principle;
step 2.3, according to the warehousing daily runoff series of the reservoir and GPM satellite precipitation data, the daily maximum air temperature and the daily minimum air temperature data of ERA5, calibrating a GR4J-9 hydrological model considering the snow melting module;
and 2.4, simulating to obtain a long-series warehousing runoff process by inputting long-series precipitation and the air temperature data in the step 2.1 by adopting the calibrated GR4J-9 hydrological model.
The GR4J hydrological model is a lumped conceptual hydrological model with only 4 parameters, has the characteristics of simple structure, fewer parameters, high precision and the like, is widely used, mainly comprises two nonlinear reservoirs which are respectively a production flow reservoir and a confluence reservoir, and has a structure shown in FIG. 2; in the present embodiment, a snow melting module is further considered on the basis of the GR4J model, wherein the snow melting module is a CemaNeige module, so as to improve the accuracy of the hydrological simulation. The GR4J-9 hydrological model is a common model in the field, the input of the model is short series of precipitation and air temperature, and the output is warehousing runoff, so that after the model is calibrated, a long series of warehousing runoff process can be simulated and obtained by inputting long series of precipitation and air temperature data.
In the embodiment, a composite mixed evolution (SCE-UA) algorithm is adopted to optimize hydrological model parameters, the algorithm is a global optimization algorithm, integrates the advantages of methods such as a random search algorithm, a simplex method, cluster analysis and biological competitive evolution, can effectively solve the problems of insensitivity, no protrusion and the like of the reflecting surface of an objective function, and is not interfered by local optimal points.
S3, constructing a long-short term memory neural network model and correcting the simulated runoff of the S2 by using the long-short term memory neural network model to obtain a corrected simulated runoff series model; this step further comprises the sub-steps of:
step 3.1, extracting the daily average air temperature (T) based on ERA5mean) And dew point temperature (T)dew) Calculating a relative humidity series; specifically, the atmospheric saturated vapor pressure e can be obtained by the Clausius-Clapperland equation and the given air temperature Ts：
In the formula: t is0And es0Are integration constants 273.16K and 611Pa, L respectivelyvIs latent heat of vaporization (2.5X 10)6J kg-1)，RvThen is the vapor gas constant (461J kg)-1K-1)；
In the formula: t is0And es0Are integration constants 273.16K and 611Pa, L respectivelyvIs latent heat of vaporization (2.5X 10)6J kg-1)，RvThen is the vapor gas constant (461J kg)-1K-1)；
Relative humidity RH ═ es(Tdew)/es(Tmean) Wherein, TmeanIndicating the average daily temperature, TdewT represents dew point temperature to be obtainedmeanAnd TdewRespectively corresponding to the first substitution to the atmospheric saturated vapor pressure esThe relative humidity RH can be solved.
Step 3.2, performing statistical analysis on the daily runoff process simulated in the step 2 in the scarce data area and the actually measured daily runoff process to determine the time lag influencing the actually measured daily runoff;
as shown in fig. 3, a schematic diagram of the change of correlation coefficients of daily measured runoff and simulated runoff at different time lags is given; the correlation coefficient of the simulated runoff and the actually measured runoff generally decreases along with the prolonging of the time lag; further, selecting a correlation threshold value which accords with the characteristics of the underlying surface of the research basin to determine the simulation runoff duration of the machine learning model established with the actual measurement runoff; for example, 0.5 may be desirable.
Step 3.3, constructing a long-short term memory neural network with a three-layer neural network architecture according to the data obtained in the steps, adopting a long-short term memory neural network model, taking the meteorological data in the step S1, the simulated runoff series and the actual measured runoff series in the step S2 as input, calibrating the long-short term memory neural network model, and correcting the simulated runoff series through the calibrated long-short term memory neural network model to obtain a corrected simulated runoff series model;
the method constructs a long-short term memory neural network (LSTM) model with a three-layer neural network architecture, and is used for generalizing the regulation and storage effects of dams, reservoirs or water transfer projects on watersheds and improving the hydrological simulation precision; in the embodiment, a neural network interval simulation averaging method is used, a neural network model is independently operated for multiple times, and an average value is taken as a final simulation result, so that uncertainty is reduced.
In order to solve the problems of gradient explosion and gradient disappearance caused by a nonlinear autoregressive exogenous input mode (NARX) dynamic neural network in a deep learning process (the number of hidden layers is more than or equal to 2), the LSTM long-term and short-term memory neural network selectively memorizes current information or forgets past memory information (such as rainfall-runoff mapping relation) by introducing a storage unit, namely an input gate, a forgetting gate, an internal feedback connection and an output gate into a hidden layer of the NARX neural network so as to enhance the long-term memory capability of the NARX neural network. In short, the LSTM long-short term memory neural network is formed by replacing each hidden layer in the NARX dynamic neural network with a memory unit with a memory function, namely an LSTM unit, and the input layer and the output layer of the LSTM unit are the same as those of the NARX dynamic neural network.
As shown in FIG. 4, a schematic diagram of a long short term memory neural network (LSTM) model memory unit employed in the present embodiment is shown. Taking the meteorological data acquired in the step S1, the simulated runoff series and the actual measurement series in the step S2 as input, calibrating the LSTM model, and then correcting the simulated runoff series through the calibrated long-short term memory neural network (LSTM) model, wherein the corrected simulated runoff series equation can be expressed as:
Qcor(t)＝FLSTM[QM(t),QM(t-1),QM(t-2),…,QM(t-N)]] (2)
in the formula: qcor(t) represents runoff corrected at time t, QM (t) represents input variables of a calibration LSTM model, and the input variables comprise daily runoff series simulated by a GR4J model, precipitation series of GPM satellite inversion products, daily highest and lowest air temperature series of ERA5, daily relative humidity series obtained in step 3.1 and land water storage series of GLDAS second generation products; QM (t-1) represents a simulated runoff and meteorological series at the t-1 moment, and N represents the time lag determined by the LSTM model; fLSTMThe long-short term memory neural network (LSTM) model after calibration is represented. To optimize the parameters of the model, the LSTM model was trained using the least-batch gradient descent method, which is a technique conventional in the art.
And 4, inputting the long series meteorological data acquired in the step S1 into the hydrological model established in the step S2 and the simulated runoff system model corrected in the step S3, and simulating the long series warehousing runoff process of the reservoir.
The reservoir of the embodiment has a shorter series of water level and delivery flow, and the series of warehousing flow calculated by a water level-reservoir capacity curve and a water quantity balance principle is shorter; in order to prolong the warehousing runoff series, a GR4J-9 hydrological model and an LSTM model are respectively calibrated through a step 2 and a step 3 based on short-duration data; and then inputting long-term meteorological data into the calibrated GR4J-9 hydrological model and the LSTM model, thereby realizing long-series warehousing runoff simulation.
And 5, constructing a multi-objective optimization scheduling model based on the long series runoff data in the step 4, and solving by adopting a second generation non-dominated sorting genetic algorithm (NSGA-II) to obtain an optimized scheduling rule.
In the step, a multi-objective optimization scheduling model is constructed according to the specific task characteristics of the reservoir, and the multi-objective optimization scheduling model is generally constructed by considering the flood control, water supply, power generation and ecological functions of the reservoir. For example, in the reservoir considered in this embodiment, the engineering task requirement of the reservoir is that the reservoir generates as much water supply benefit and power generation benefit as possible on the premise of meeting the shipping requirement and the ecological water in the river as much as possible, and the scheduling is performed according to the criterion that the water supply benefit is the maximum and the power generation is the maximum, and the multi-objective optimization scheduling model of the reservoir is as follows:
in the formula: w*(T) and E*(T) water supply (m) in the planned period T3) And generating electricityAmount (kW. h);
and PtAverage water supply flow rate (m) of t-th time period3/s) and average output (kW); k is the comprehensive output coefficient of the power station;
generating flow (m) for t-th period3/s)；HtAveraging the power generation water purification heads (m) for the t-th period; mtThe number of hours in the t period; t is the length of the scheduling period;
the upper water level (m) of the reservoir dam in the t-th time period;
the water level (m) under the reservoir dam in the t-th time period; qtThe lower discharge quantity of the reservoir in the t-th time period; Δ HtHydropower station head loss (m) for a t-th period; f. ofZQ(x) is a downstream water level flow relation function; f. ofΔH(. x) is a hydropower station head loss function.
As shown in FIG. 5, a schematic diagram of a second generation non-dominated sorting genetic algorithm flow chart is given; the algorithm is widely applied to multi-objective optimization calculation at present due to the rapid operation effect and good convergence, and meanwhile, the complexity of the traditional non-inferior ranking genetic algorithm is reduced, so that the algorithm becomes the basis of the performance of other multi-objective optimization algorithms. The NSGA-II prevents the loss of the optimal individual by introducing an elite retention strategy, and improves the operation speed and the robustness of the algorithm. The algorithm has been applied to reservoir dispatching research by Chinese scholars in recent years, and a good effect is achieved. In the optimizing process of the dispatching model, water balance constraint, reservoir water storage capacity constraint, lower discharge flow constraint, power station output constraint, ecological flow constraint and the like are met.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention.
Claims (7)
1. A reservoir dispatching rule optimization method based on machine learning and multi-source remote sensing data fusion is characterized by comprising the following steps:
s1, collecting short series runoff observation data of a reservoir, collecting engineering characteristics and scheduling operation related data of the reservoir, extracting precipitation series of a basin where the reservoir is located, and analyzing data set meteorological data and land water storage quantity series;
s2, establishing a hydrological model of the basin where the reservoir is located according to the short series runoff observation data and the meteorological data obtained in the step S1, and simulating the primary runoff through the hydrological model;
s3, constructing a long-short term memory neural network model and correcting the simulated runoff in the step S2 by using the long-short term memory neural network model to obtain a corrected simulated runoff series model;
s4, inputting the long series meteorological data collected in the step S1 into the hydrological model established in the step S2 and the simulated runoff system model corrected in the step S3, and simulating the long series warehousing runoff process of the reservoir;
and S5, constructing a multi-objective optimization scheduling model according to the long series of warehousing runoff data of the reservoir obtained in the step S4, and solving by adopting a genetic algorithm to obtain an optimized scheduling rule.
2. The method for optimizing the reservoir dispatching rule based on machine learning and fusion of the multi-source remote sensing data according to claim 1, wherein the short series runoff observation data in the step S1 comprise sunrise reservoir flow and reservoir area daily water level data of the reservoir, and the reanalysis data set meteorological data comprise hour-scale near-earth air temperature, dew point temperature and horizontal wind speed.
3. The reservoir dispatching rule optimization method based on machine learning and multi-source remote sensing data fusion of claim 1, wherein the step S2 further comprises the following substeps:
step 2.1, extracting a daily maximum air temperature and a daily minimum air temperature series according to the meteorological data acquired in the step S1, and calculating a daily average dew point temperature, a daily average air temperature and a daily average wind speed;
step 2.2, acquiring short series runoff observation data of the reservoir according to the step S1, and calculating a warehousing daily runoff series according to a water level-reservoir capacity curve and a water quantity balance principle;
step 2.3, constructing a hydrological model for calibrating and considering the snow melting module according to the warehousing daily runoff series of the reservoir and precipitation data, daily maximum air temperature and daily minimum air temperature data of the same period;
and 2.4, simulating to obtain a long-series warehousing runoff process by inputting the long-series precipitation and the temperature data in the step 2.1 by adopting the calibrated hydrological model.
4. The reservoir dispatching rule optimization method based on machine learning and multi-source remote sensing data fusion of claim 1, wherein the step S3 further comprises the following substeps:
step 3.1, extracting the daily average air temperature and the dew point temperature according to the meteorological data obtained in the step S1, and then calculating a relative humidity series according to the daily average air temperature and the dew point temperature;
step 3.2, determining the time lag influencing the daily measured runoff by carrying out statistical analysis on the daily runoff process simulated in the step S2 and the actually measured daily runoff process;
and 3.3, constructing a long-short term memory neural network with a three-layer neural network architecture according to the data obtained in the steps, calibrating the long-short term memory neural network model by using meteorological data, the simulated runoff series in the step S2 and the actually-measured runoff series as input through a calibrated long-short term memory neural network model, and correcting the simulated runoff series in the step S2 through the calibrated long-short term memory neural network model to obtain a corrected simulated runoff series model.
5. The reservoir dispatching rule optimization method based on machine learning and multi-source remote sensing data fusion as claimed in claim 4, wherein the method for solving the relative humidity in step 3.1The method comprises the following steps: solving through a Clausius-Clappe-long equation and a given air temperature T to obtain the atmospheric saturated vapor pressure es：
In the formula: t is0And es0Are integration constants 273.16K and 611Pa, L respectivelyvIs latent heat of vaporization (2.5X 10)6J kg-1)，RvThen is the vapor gas constant (461J kg)-1K-1)；
Relative humidity RH ═ es(Tdew)/es(Tmean) Wherein, TmeanIndicating the average daily temperature, TdewT represents dew point temperature to be obtainedmeanAnd TdewRespectively correspondingly substituting into atmospheric saturated vapor pressure esThe relative humidity RH can be solved.
6. The reservoir dispatching rule optimization method based on machine learning and multi-source remote sensing data fusion of claim 4, wherein the model of the simulated runoff series after correction is as follows: qcor(t)＝FLSTM[QM(t),QM(t-1),QM(t-2),…,QM(t-N)]；
In the formula: qcor(t) represents runoff corrected at time t; QM (t) represents input variables of the long-term and short-term memory neural network model after calibration, including a runoff series, a precipitation series, a highest daily temperature series, a lowest daily temperature series, a relative humidity series of days obtained in the step 3.1 and a land water reserve series simulated by a hydrological model; QM (t-1) represents a simulated runoff and meteorological series at the t-1 moment; n represents the time lag determined by the long-term and short-term memory neural network model; fLSTMAnd representing the long-term and short-term memory neural network model after calibration.
7. The reservoir dispatching rule optimization method based on machine learning and multi-source remote sensing data fusion of claim 1, wherein in step S5, a multi-objective optimization dispatching model is established with the maximum water supply benefit and the maximum power generation capacity as dispatching targets, and the multi-objective optimization dispatching model is as follows:
in the formula: w*(T) and E*(T) water supply (m) in the planned period T3) And power generation (kW · h);
and PtAverage water supply flow rate (m) of t-th time period3/s) and average output (kW); k is the comprehensive output coefficient of the power station;
generating flow (m) for t-th period3/s)；HtAveraging the power generation water purification heads (m) for the t-th period; mtThe number of hours in the t period; t is the length of the scheduling period;
the upper water level (m) of the reservoir dam in the t-th time period;
the water level (m) under the reservoir dam in the t-th time period; qtThe lower discharge quantity of the reservoir in the t-th time period; Δ HtHydropower station head loss (m) for a t-th period; f. ofZQ(x) is a downstream water level flow relation function; f. ofΔH(. x) is a hydropower station head loss function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110053303.XA CN112700068B (en) | 2021-01-15 | 2021-01-15 | Reservoir dispatching rule optimization method based on machine learning fusion of multi-source remote sensing data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110053303.XA CN112700068B (en) | 2021-01-15 | 2021-01-15 | Reservoir dispatching rule optimization method based on machine learning fusion of multi-source remote sensing data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112700068A true CN112700068A (en) | 2021-04-23 |
| CN112700068B CN112700068B (en) | 2022-04-29 |
Family
ID=75515215
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110053303.XA Active CN112700068B (en) | 2021-01-15 | 2021-01-15 | Reservoir dispatching rule optimization method based on machine learning fusion of multi-source remote sensing data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112700068B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113205368A (en) * | 2021-05-25 | 2021-08-03 | 合肥供水集团有限公司 | Industrial and commercial customer clustering method based on time sequence water consumption data |
| CN115544785A (en) * | 2022-10-19 | 2022-12-30 | 中国科学院地理科学与资源研究所 | Data-free cascade reservoir basin hydrological simulation method and system |
| CN117332908A (en) * | 2023-12-01 | 2024-01-02 | 长江水利委员会水文局 | Multi-objective optimization scheduling method and system for cascade reservoir of coupling set forecast |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049671A (en) * | 2013-01-21 | 2013-04-17 | 武汉大学 | Method for drawing up multi-goal reservoir optimization scheduling graph capable of being self-adaptive to climate change |
| CN107563538A (en) * | 2017-07-13 | 2018-01-09 | 大连理工大学 | Multiple-use reservoir group's method for optimizing scheduling towards crucial water level control under bulk power grid platform |
| CN110442937A (en) * | 2019-07-24 | 2019-11-12 | 武汉大学 | A kind of valley water security management method merging satellite remote sensing and machine learning techniques |
| CN110851977A (en) * | 2019-11-06 | 2020-02-28 | 武汉大学 | Water supply-power generation-ecological multi-target scheduling graph optimization method based on ecological flow |
| CN111898660A (en) * | 2020-07-17 | 2020-11-06 | 武汉大学 | Hydrological simulation method for averagely fusing multi-source data based on Bayesian mode |
-
2021
- 2021-01-15 CN CN202110053303.XA patent/CN112700068B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049671A (en) * | 2013-01-21 | 2013-04-17 | 武汉大学 | Method for drawing up multi-goal reservoir optimization scheduling graph capable of being self-adaptive to climate change |
| CN107563538A (en) * | 2017-07-13 | 2018-01-09 | 大连理工大学 | Multiple-use reservoir group's method for optimizing scheduling towards crucial water level control under bulk power grid platform |
| CN110442937A (en) * | 2019-07-24 | 2019-11-12 | 武汉大学 | A kind of valley water security management method merging satellite remote sensing and machine learning techniques |
| CN110851977A (en) * | 2019-11-06 | 2020-02-28 | 武汉大学 | Water supply-power generation-ecological multi-target scheduling graph optimization method based on ecological flow |
| CN111898660A (en) * | 2020-07-17 | 2020-11-06 | 武汉大学 | Hydrological simulation method for averagely fusing multi-source data based on Bayesian mode |
Non-Patent Citations (4)
| Title |
|---|
| AUDREY VALÉRY,VAZKEN ANDRÉASSIAN,CHARLES PERRIN.: "As simple as possible but not simpler’: What is useful in a temperature-based snow-accounting routine? Part 2 – Sensitivity analysis of the Cemaneige snow accounting routine on 380 catchments", 《JOURNAL OF HYDROLOGY》 * |
| 杨光: "考虑未来径流变化的丹江口水库多目标调度规则研究", 《水力发电学报》 * |
| 胡彩虹: "土地利用变化对汾河流域典型区域径流影响模拟研究", 《人民珠江》 * |
| 陈晓宏: "气候变化对澜沧江下游梯级电站发电及生态调度的影响", 《水科学进展》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113205368A (en) * | 2021-05-25 | 2021-08-03 | 合肥供水集团有限公司 | Industrial and commercial customer clustering method based on time sequence water consumption data |
| CN113205368B (en) * | 2021-05-25 | 2022-11-29 | 合肥供水集团有限公司 | Industrial and commercial customer clustering method based on time sequence water consumption data |
| CN115544785A (en) * | 2022-10-19 | 2022-12-30 | 中国科学院地理科学与资源研究所 | Data-free cascade reservoir basin hydrological simulation method and system |
| CN117332908A (en) * | 2023-12-01 | 2024-01-02 | 长江水利委员会水文局 | Multi-objective optimization scheduling method and system for cascade reservoir of coupling set forecast |
| CN117332908B (en) * | 2023-12-01 | 2024-02-13 | 长江水利委员会水文局 | Multi-objective optimization scheduling method and system for cascade reservoir of coupling set forecast |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112700068B (en) | 2022-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112700068B (en) | Reservoir dispatching rule optimization method based on machine learning fusion of multi-source remote sensing data | |
| CN111626506B (en) | Regional photovoltaic power probability prediction method based on federal learning and cooperative regulation and control system thereof | |
| Talebmorad et al. | Evaluation of the impact of climate change on reference crop evapotranspiration in Hamedan-Bahar plain | |
| CN113158542B (en) | Multivariable design flood estimation method suitable for data-lacking area | |
| CN110442937B (en) | Drainage basin hydrological simulation method integrating satellite remote sensing and machine learning technology | |
| CN108983320B (en) | Numerical weather forecast-artificial intelligence coupling prediction method for coastal typhoon extreme wind speed | |
| CN107316095B (en) | Regional weather drought level prediction method coupled with multi-source data | |
| CN111898660B (en) | Hydrological simulation method for average fusion of multi-source data based on Bayesian mode | |
| CN105184423B (en) | A kind of wind power plant cluster wind speed forecasting method | |
| CN115507822A (en) | Flood risk prediction method under hydrographic cyclic variation drive | |
| CN114265836A (en) | All-weather assimilation method and device of satellite microwave hygrothermograph based on cloud area temperature and humidity profile inversion | |
| CN114019579B (en) | High space-time resolution near-surface air temperature reconstruction method, system and equipment | |
| Ma et al. | Prediction of outdoor air temperature and humidity using Xgboost | |
| CN107609713B (en) | Land-air coupling real-time forecasting method through rainfall and runoff double-factor correction | |
| US11944048B2 (en) | Decision-making method for variable rate irrigation management | |
| CN113553782B (en) | Downscaling method for forecasting wind speed | |
| CN114186723A (en) | Distributed photovoltaic power grid virtual prediction system based on space-time correlation | |
| CN108808671A (en) | A kind of short-term wind speed DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method of wind power plant | |
| CN114757416A (en) | Water demand prediction method and system based on rating method | |
| CN111859303B (en) | Soil humidity fusion method and system based on dynamic Bayesian average | |
| CN117493476A (en) | Runoff backtracking simulation method and system integrating physical mechanism and artificial intelligence | |
| CN117273200A (en) | Runoff interval forecasting method based on convolution optimization algorithm and Pyraformer neural network | |
| CN115912334A (en) | Method for establishing prediction model of output guarantee rate of wind power plant and prediction method | |
| CN106355264A (en) | Combined prediction method of reference crop evapotranspiration | |
| CN113435630B (en) | Basin hydrological forecasting method and system with self-adaptive runoff yield mode |
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 |