CN114004430B - Wind speed forecasting method and system - Google Patents
Wind speed forecasting method and system Download PDFInfo
- Publication number
- CN114004430B CN114004430B CN202210002536.1A CN202210002536A CN114004430B CN 114004430 B CN114004430 B CN 114004430B CN 202210002536 A CN202210002536 A CN 202210002536A CN 114004430 B CN114004430 B CN 114004430B
- Authority
- CN
- China
- Prior art keywords
- wind speed
- field
- mode
- assimilation
- scheme
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013277 forecasting method Methods 0.000 title abstract description 17
- 238000013473 artificial intelligence Methods 0.000 claims abstract description 50
- 230000007246 mechanism Effects 0.000 claims abstract description 36
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 27
- 238000000034 method Methods 0.000 claims description 102
- 238000005457 optimization Methods 0.000 claims description 65
- 230000008569 process Effects 0.000 claims description 65
- 238000012549 training Methods 0.000 claims description 25
- 238000004364 calculation method Methods 0.000 claims description 24
- 238000004088 simulation Methods 0.000 claims description 21
- 238000013135 deep learning Methods 0.000 claims description 20
- 230000006870 function Effects 0.000 claims description 18
- 238000005259 measurement Methods 0.000 claims description 16
- 238000013527 convolutional neural network Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 230000000306 recurrent effect Effects 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 description 14
- 238000012545 processing Methods 0.000 description 4
- 238000002790 cross-validation Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000002068 genetic effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000739 chaotic effect Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Chemical compound O XLYOFNOQVPJJNP-UHFFFAOYSA-N 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
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- 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
-
- 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
- G06N3/08—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
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Resources & Organizations (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Tourism & Hospitality (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Environmental & Geological Engineering (AREA)
- Entrepreneurship & Innovation (AREA)
- Atmospheric Sciences (AREA)
- Environmental Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Ecology (AREA)
- Educational Administration (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a wind speed forecasting method and a system, comprising the following steps: performing data assimilation by adopting a pre-constructed global optimal assimilation scheme based on the mode background field and the latest meteorological element data to obtain a mode initial state; based on the mode initial state, adopting a pre-trained mode parameterization scheme to predict to obtain a wind speed forecasting field; aiming at the links of data assimilation and mode parameterization schemes, on the basis of a traditional numerical weather forecast mode based on physical mechanisms and known rule constraints, and in the case of lack of or no physical mechanism constraints, the invention trains a global optimal assimilation scheme and a mode parameterization scheme by adopting an artificial intelligence algorithm, so that the advantages of the artificial intelligence algorithm are fully exerted, and the accuracy of wind speed forecasting is obviously improved.
Description
Technical Field
The invention belongs to the technical field of electric power weather, and particularly relates to a wind speed forecasting method and system.
Background
The core of weather forecasting is a numerical weather forecasting mode (NWP), the numerical weather forecasting mode carries out certain simplification and approximation on an atmospheric motion basic equation set according to actual application requirements and scenes, and meanwhile, the processes of boundary layers, cloud accumulation convection, micro physics, radiation and the like are introduced, so that numerical simulation and forecasting of the weather elements with the required scale are realized. The weather forecasting process based on the numerical weather forecasting mode comprises links such as Data Assimilation (DA), a mode parameterization scheme, forecasting error calibration and the like, and the current links completely follow the constraints of a weather physical mechanism and a related rule. Meanwhile, the accuracy of weather forecast is mainly limited by the lack of accurate understanding of physical mechanisms in all links or the lack of accurate understanding of physical mechanisms, but the current atmospheric scientific theory and the development of related technologies cannot effectively solve the problem.
Data Assimilation (DA) is one of the effective technical means for improving the accuracy of numerical weather forecast, and the objective is to provide the best estimation (initial guess field) of the current state of the numerical forecast model by fusing the latest weather observation data with the model background field or short-term weather forecast. According to the technical scheme, a better estimation of the initial state of the mode is obtained by comparing the observation of the initial estimation of the numerical mode with the real observation according to the currently known initial state of the observation field and the numerical mode.
The premise of data assimilation is that an accurate and reliable meteorological model based on a physical mechanism is assumed, the initial state of the model is an unknown item, and various physical mechanism constraints such as atmospheric physical state, wind pressure balance and the like are added, so that the more accurate mode initial state than that obtained from observation is obtained. However, the actual atmospheric variation process is a highly nonlinear chaotic system, and at present, there are many links which are lack of or have no physical mechanism constraint, and the processes which are not scientifically and accurately recognized severely restrict the accuracy of weather forecast.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a wind speed forecasting method, which comprises the following steps:
performing data assimilation by adopting a pre-constructed global optimal assimilation scheme based on the mode background field and the latest meteorological element data to obtain a mode initial state;
based on the mode initial state, adopting a pre-trained mode parameterization scheme to predict to obtain a wind speed forecasting field;
the global optimal assimilation scheme is constructed by adopting a method of combining artificial intelligence and a physical mechanism based on numerical simulation data of meteorological elements and forecast field data corresponding to a numerical forecast mode; the mode parameterization scheme is obtained by training by adopting an artificial intelligence method based on the mode initial state and the wind speed forecasting field corresponding to the historical meteorological element data.
Preferably, the construction of the global optimal assimilation scheme includes:
acquiring the quality of meteorological element data in time and space ranges and forecasting precision requirements;
optimizing an artificial intelligence model for assimilation information attenuation and optimization in an assimilation time window by taking numerical simulation data of meteorological elements meeting quality and forecast accuracy requirements as input and forecasting field data of a corresponding numerical forecast mode as output to obtain an assimilation time window configuration scheme;
aiming at the spatial range of the data assimilation process, fused observation field data corresponding to the numerical simulation data are used as input, forecast field data of a corresponding numerical forecasting mode are used as output, an assimilation area and an assimilation quality self-adaptive artificial intelligence model are optimized, and an assimilation spatial range configuration scheme is obtained;
and constructing a global optimal assimilation scheme of meteorological element data and a mode background field based on the time window configuration scheme and the space range configuration scheme and according to the generalization capability of an artificial intelligence algorithm.
Preferably, the training of the schema parameterization scheme comprises:
acquiring a mode key physical process parameterization scheme and sensitive parameters thereof which influence the wind speed of the key area to be forecasted;
aiming at the mode key physical process parameterization scheme, optimizing the mode key physical process parameterization scheme by adopting a heuristic intelligent optimization algorithm and taking wind speed forecasting precision as an optimization target on the basis of a mode initial state and a wind speed forecasting field corresponding to historical meteorological element data;
designing an optimization objective function of sample set error feedback aiming at sensitive parameters in the optimized mode key physical process parameterization scheme based on the optimized mode key physical process parameterization scheme, constructing a nestable optimization objective function integrating mode integral calculation and error index calculation, and carrying out sensitive parameter optimization in the optimized mode key physical process parameterization scheme through repeated iterative calculation;
and taking an optimized mode-critical physical process parameterization scheme for completing sensitive parameter optimization as a mode parameterization scheme.
Preferably, after obtaining the wind speed forecasting field, the method further includes:
calibrating a wind speed forecasting field by adopting a pre-trained wind speed calibration model;
the wind speed calibration model is obtained by training a physical mechanism and artificial intelligence based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period.
Preferably, the training of the wind speed calibration model comprises:
based on the numerical weather forecast wind speed field and the actually measured wind speed field in the historical time period, learning the spatial information of the weather forecast multi-grid-point element by adopting a convolutional neural network or a graph convolutional neural network;
based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period, learning relevant information hidden in each meteorological element on a time sequence through a recurrent neural network or a long-short term memory network;
the spatial information and the time sequence related information are used as physical mechanisms, and the hyper-parameters of the deep learning network model are optimized by combining a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period to obtain a wind speed calibration model;
wherein the hyper-parameters of the deep learning network model comprise: network structure and learning rate; the deep learning network model is designed according to the application scene of wind speed forecasting.
Preferably, after the wind speed forecasting field is calibrated by using the pre-trained wind speed calibration model, the method further includes:
and taking the calibrated wind speed forecast field as a new numerical weather forecast wind speed field, and carrying out online optimization updating on the wind speed calibration model by combining with an actually measured wind speed field corresponding to the calibrated wind speed forecast field.
Based on the same inventive concept, the invention also provides a wind speed forecasting system, which comprises: data assimilation module and wind speed forecast module, wherein:
the data assimilation module is used for carrying out data assimilation by adopting a pre-constructed global optimal assimilation scheme based on the mode background field and the latest meteorological element data to obtain a mode initial state;
the wind speed forecasting module is used for forecasting by adopting a pre-trained mode parameterization scheme based on the mode initial state to obtain a wind speed forecasting field;
the global optimal assimilation scheme is constructed by adopting a method of combining artificial intelligence and a physical mechanism based on numerical simulation data of meteorological elements and forecast field data corresponding to a numerical forecast mode; the mode parameterization scheme is obtained by training by adopting an artificial intelligence method based on the mode initial state and the wind speed forecasting field corresponding to the historical meteorological element data.
Preferably, the construction of the global optimal assimilation scheme includes:
acquiring the quality of meteorological element data in time and space ranges and forecasting precision requirements;
optimizing an artificial intelligence model for assimilation information attenuation and optimization in an assimilation time window by taking numerical simulation data of meteorological elements meeting quality and forecast accuracy requirements as input and forecasting field data of a corresponding numerical forecast mode as output to obtain an assimilation time window configuration scheme;
aiming at the spatial range of the data assimilation process, fused observation field data corresponding to the numerical simulation data are used as input, forecast field data of a corresponding numerical forecasting mode are used as output, an assimilation area and an assimilation quality self-adaptive artificial intelligence model are optimized, and an assimilation spatial range configuration scheme is obtained;
and constructing a global optimal assimilation scheme of meteorological element data and a mode background field based on the time window configuration scheme and the space range configuration scheme and according to the generalization capability of an artificial intelligence algorithm.
Preferably, the training of the schema parameterization scheme comprises:
acquiring a mode key physical process parameterization scheme and sensitive parameters thereof which influence the wind speed of the key area to be forecasted;
aiming at the mode key physical process parameterization scheme, optimizing the mode key physical process parameterization scheme by adopting a heuristic intelligent optimization algorithm and taking wind speed forecasting precision as an optimization target on the basis of a mode initial state and a wind speed forecasting field corresponding to historical meteorological element data;
designing an optimization objective function of sample set error feedback aiming at sensitive parameters in the optimized mode key physical process parameterization scheme based on the optimized mode key physical process parameterization scheme, constructing a nestable optimization objective function integrating mode integral calculation and error index calculation, and carrying out sensitive parameter optimization in the optimized mode key physical process parameterization scheme through repeated iterative calculation;
and taking an optimized mode-critical physical process parameterization scheme for completing sensitive parameter optimization as a mode parameterization scheme.
Preferably, the system further comprises: a wind speed calibration module;
the wind speed calibration module is used for calibrating a wind speed forecasting field by adopting a pre-trained wind speed calibration model;
the wind speed calibration model is obtained by training a physical mechanism and artificial intelligence based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period.
Preferably, the training of the wind speed calibration model comprises:
based on the numerical weather forecast wind speed field and the actually measured wind speed field in the historical time period, learning the spatial information of the weather forecast multi-grid-point element by adopting a convolutional neural network or a graph convolutional neural network;
based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period, learning relevant information hidden in each meteorological element on a time sequence through a recurrent neural network or a long-short term memory network;
the spatial information and the time sequence related information are used as physical mechanisms, and the hyper-parameters of the deep learning network model are optimized by combining a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period to obtain a wind speed calibration model;
wherein the hyper-parameters of the deep learning network model comprise: network structure and learning rate; the deep learning network model is designed according to the application scene of wind speed forecasting.
Preferably, after the wind speed forecasting field is calibrated by using the pre-trained wind speed calibration model, the method further includes:
and taking the calibrated wind speed forecast field as a new numerical weather forecast wind speed field, and carrying out online optimization updating on the wind speed calibration model by combining with an actually measured wind speed field corresponding to the calibrated wind speed forecast field.
The present invention also provides a computer device comprising: one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the one or more processors, implement a wind speed forecasting method as previously described.
The invention also provides a computer-readable storage medium having stored thereon a computer program which, when executed, implements a wind speed forecasting method as described above.
Compared with the closest prior art, the invention has the following beneficial effects:
the invention provides a wind speed forecasting method and a system, comprising the following steps: performing data assimilation by adopting a pre-constructed global optimal assimilation scheme based on the mode background field and the latest meteorological element data to obtain a mode initial state; based on the mode initial state, adopting a pre-trained mode parameterization scheme to predict to obtain a wind speed forecasting field; the global optimal assimilation scheme is constructed by adopting a method of combining artificial intelligence and a physical mechanism based on numerical simulation data of meteorological elements and forecast field data corresponding to a numerical forecast mode; the mode parameterization scheme is obtained by training by adopting an artificial intelligence method based on the mode initial state and the wind speed forecasting field corresponding to the historical meteorological element data; aiming at the links of data assimilation and mode parameterization schemes, on the basis of a traditional numerical weather forecast mode based on physical mechanisms and known rule constraints, and in the case of lack of or no physical mechanism constraints, the invention trains a global optimal assimilation scheme and a mode parameterization scheme by adopting an artificial intelligence algorithm, so that the advantages of the artificial intelligence algorithm are fully exerted, and the accuracy of wind speed forecasting is obviously improved.
Drawings
FIG. 1 is a schematic flow chart of a wind speed forecasting method according to the present invention;
FIG. 2 is a schematic overall route diagram of an example of a wind speed forecasting method for wind power prediction according to the present invention;
FIG. 3 is a schematic diagram of a technical route of a data assimilation method in an example of a wind speed forecasting method for wind power prediction provided by the invention;
FIG. 4 is a schematic diagram of a technical route of a mode parameterization scheme optimization method in an example of a wind speed forecasting method for wind power prediction provided by the invention;
FIG. 5 is a schematic diagram of a technical route of forecast error calibration in an example of a wind speed forecasting method for wind power forecasting according to the present invention;
fig. 6 is a schematic structural diagram of a wind speed forecasting system according to the present invention.
Detailed Description
The following describes embodiments of the present invention in further detail with reference to the accompanying drawings.
Example 1:
the flow diagram of the wind speed forecasting method provided by the invention is shown in fig. 1, and the method comprises the following steps:
step 1: performing data assimilation by adopting a pre-constructed global optimal assimilation scheme based on the mode background field and the latest meteorological element data to obtain a mode initial state;
step 2: based on the mode initial state, adopting a pre-trained mode parameterization scheme to predict to obtain a wind speed forecasting field;
the global optimal assimilation scheme is constructed by adopting a method of combining artificial intelligence and a physical mechanism based on numerical simulation data of meteorological elements and forecast field data corresponding to a numerical forecast mode; the mode parameterization scheme is obtained by training by adopting an artificial intelligence method based on the mode initial state and the wind speed forecasting field corresponding to the historical meteorological element data.
In step 1, the construction of the global optimal assimilation scheme comprises the following steps:
acquiring the quality of meteorological element data in time and space ranges and forecasting precision requirements;
optimizing an artificial intelligence model for assimilation information attenuation and optimization in an assimilation time window by taking numerical simulation data of meteorological elements meeting quality and forecast accuracy requirements as input and forecasting field data of a corresponding numerical forecast mode as output to obtain an assimilation time window configuration scheme;
aiming at the spatial range of the data assimilation process, fused observation field data corresponding to numerical simulation data are used as input, forecast field data of a corresponding numerical forecasting mode are used as output, an assimilation area and an assimilation quality self-adaptive artificial intelligence model are optimized, and an assimilation spatial range configuration scheme is obtained;
and constructing a global optimal assimilation scheme of meteorological element data and a mode background field based on a time window configuration scheme and a space range configuration scheme according to the generalization capability of an artificial intelligence algorithm.
In step 2, the training of the mode parameterization scheme comprises the following steps:
acquiring a mode key physical process parameterization scheme and sensitive parameters thereof which influence the wind speed of the key area to be forecasted;
aiming at the mode key physical process parameterization scheme, optimizing the mode key physical process parameterization scheme by adopting a heuristic intelligent optimization algorithm and taking wind speed forecasting precision as an optimization target on the basis of a mode initial state and a wind speed forecasting field corresponding to historical meteorological element data;
designing an optimization objective function of sample set error feedback aiming at sensitive parameters in the optimized mode key physical process parameterization scheme based on the optimized mode key physical process parameterization scheme, constructing a nestable optimization objective function integrating mode integral calculation and error index calculation, and developing sensitive parameter optimization in the optimized mode key physical process parameterization scheme through repeated iterative calculation;
and taking an optimized mode-critical physical process parameterization scheme for completing sensitive parameter optimization as a mode parameterization scheme.
After step 2, the wind speed forecasting field can be further calibrated, including: and calibrating the wind speed forecasting field by adopting a pre-trained wind speed calibration model.
The training process of the wind speed calibration model comprises the following steps: based on the numerical weather forecast wind speed field and the actually measured wind speed field in the historical time period, learning the spatial information of the weather forecast multi-grid-point element by adopting a convolutional neural network or a graph convolutional neural network;
based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period, learning relevant information hidden in each meteorological element on a time sequence through a recurrent neural network or a long-short term memory network;
the spatial information and the time sequence related information are used as physical mechanisms, and the hyper-parameters of the deep learning network model are optimized by combining the numerical weather forecast wind speed field and the measured wind speed field in the historical time period to obtain a wind speed calibration model;
wherein, the hyper-parameters of the deep learning network model comprise: network structure and learning rate; the deep learning network model is designed according to the application scene of wind speed forecasting.
After the wind speed forecasting field is calibrated, the calibrated wind speed forecasting field can be further used as a new numerical weather forecasting wind speed field, and the wind speed calibration model is optimized and updated on line by combining with an actually measured wind speed field corresponding to the calibrated wind speed forecasting field.
The method further gives full play to the advantages of an artificial intelligence algorithm in the forecast error calibration link and on the basis of the traditional numerical weather forecast mode based on physical mechanism and known rule constraint for the condition of lack or no physical mechanism constraint, thereby obviously improving the accuracy of wind speed forecast.
Example 2:
the method provided by the invention is explained below with a wind speed forecast predicted for wind power as a specific example.
A wind speed forecasting method for wind power prediction has an overall technical route as shown in FIG. 2, and comprises the following 3 steps:
step S1: artificial intelligence and physical mechanism.
On the basis of the assimilation of the traditional numerical mode data, based on the time-space characteristics of the electric power meteorological fusion field data and specific forecasting requirements, aiming at the defects that the effective time window, the space range and the like of the assimilation process of the numerical mode data are uniformly set, the self-adaptive optimization is realized by adopting an artificial intelligence algorithm, the numerical mode calculation efficiency is improved, and meanwhile, the forecasting effect is improved.
Specifically, the method comprises the following 4 steps as shown in fig. 3:
step S1-1: and analyzing the space-time distribution characteristics of the electric power meteorological observation fusion data, and determining the quality and the forecasting precision of the corresponding meteorological element data in the time and space ranges according to the accurate wind speed forecasting requirement required by wind power forecasting.
Step S1-2: and training an artificial intelligence model for assimilation information attenuation and optimization in an assimilation time window by taking high-quality numerical simulation data as input and forecast field data of a corresponding numerical forecast mode as output, and realizing adaptive optimization of effective information attenuation in the assimilation time window based on the model during service forecast.
Step S1-3: aiming at the space range of the data assimilation process, fused observation field data are used as input, forecast field data of a corresponding numerical forecast mode are used as output, an assimilation area and an assimilation quality self-adaptive artificial intelligence model are trained, and self-adaptive optimization of the mode assimilation area is achieved based on the model during service forecast.
Step S1-4: based on the assimilation time window and space range configuration scheme, a global optimal assimilation scheme of an observation field (namely the latest meteorological element data) and a background field (namely a mode background field) is constructed according to the generalization capability of an artificial intelligence algorithm.
Step S2: a method for optimizing a mode parameterization scheme based on artificial intelligence.
Based on the physical mechanism of momentum, heat and water vapor interaction and mutual influence among the physical process parameterization schemes in the numerical weather forecasting mode, a heuristic intelligent optimization algorithm suitable for mode parameter optimization is constructed based on a heuristic artificial intelligent optimization algorithm framework by adopting optimization algorithms such as a genetic algorithm, a particle swarm algorithm and the like, and the evaluation and optimization of the numerical mode physical process parameterization scheme (namely the mode key physical process parameterization scheme) are realized by constructing a nestable optimization target integrating mode integral calculation and forecast error index calculation.
Specifically, the method comprises the following 3 steps as shown in fig. 4:
step S2-1: and analyzing a key physical process parameterization scheme and sensitive parameters of a mode influencing the wind speed of the key wind power plant by aiming at wind speed forecasting required by wind power prediction.
Step S2-2: aiming at the parameterization scheme of the mode key physical process, heuristic intelligent optimization algorithms such as a genetic algorithm, a particle swarm algorithm and the like are adopted, the forecasting precision is used as an optimization target, and the optimization of the parameterization scheme of the numerical mode physical process is developed.
Step S2-3: based on an optimized mode key physical process parameterization scheme, aiming at sensitive parameters in the parameterization scheme and semi-empirical parameters which are difficult to be explicitly analyzed and expressed in a certain sub-grid process and the like, an optimization objective function for sample set error feedback is designed, a nestable optimization objective function integrating mode integral calculation and error index calculation is constructed, and sensitive parameter optimization in the key physical process parameterization scheme is carried out through repeated iterative calculation.
Step S3: a forecast error calibration method combining physical mechanism and artificial intelligence.
In the forecast error calibration step, on the basis of a weather situation method based on a physical mechanism and a similar weather calibration technology, a convolution neural network and a circulation neural network are adopted to respectively construct a deep learning network model aiming at time sequence meteorological elements and event meteorological forecasts, and cross validation is adopted to optimize super parameters such as a model structure, a learning rate and the like, so that the meteorological forecast error calibration based on the combination of the physical mechanism and artificial intelligence is realized.
Specifically, the method comprises the following 4 steps as shown in fig. 5:
step S3-1: based on massive numerical weather forecast results (namely, numerical weather forecast wind speed fields in historical time periods) and weather live data (namely, actually measured wind speed fields in historical time periods), spatial information of weather forecast multi-grid-point elements is learned by adopting a convolutional neural network, a graph convolutional neural network and the like.
Step S3-2: based on massive numerical weather forecast results and weather live data, learning and hiding relevant information of each weather element on time sequence through a recurrent neural network, a long-time and short-time memory network and the like.
Step S3-3: aiming at the wind speed forecasting scene required by wind power forecasting, a convolution network or a circulation network can be independently adopted, and the convolution network and the circulation network can be fused for meteorological feature learning. Aiming at the calibration of the wind speed sequence, a deep learning network model suitable for the application scene is designed, and the super parameters such as the model structure, the learning rate and the like are optimized based on cross validation.
Step S3-4: and establishing an online optimization updating mechanism of a deep learning model (namely a wind speed calibration model), and performing online calibration on the prediction result by collecting the prediction error evaluation condition of the near stage.
Example 3:
based on the same inventive concept, the present invention further provides a wind speed forecasting system, the structure of which is shown in fig. 6, and the system comprises:
data assimilation module and wind speed forecast module, wherein:
the data assimilation module is used for carrying out data assimilation by adopting a pre-constructed global optimal assimilation scheme based on the mode background field and the latest meteorological element data to obtain a mode initial state;
the wind speed forecasting module is used for forecasting by adopting a pre-trained mode parameterization scheme based on the mode initial state to obtain a wind speed forecasting field;
the global optimal assimilation scheme is constructed by adopting a method of combining artificial intelligence and a physical mechanism based on numerical simulation data of meteorological elements and forecast field data corresponding to a numerical forecast mode; the mode parameterization scheme is obtained by training by adopting an artificial intelligence method based on the mode initial state and the wind speed forecasting field corresponding to the historical meteorological element data.
The construction of the global optimal assimilation scheme comprises the following steps:
acquiring the quality of meteorological element data in time and space ranges and forecasting precision requirements;
optimizing an artificial intelligence model for assimilation information attenuation and optimization in an assimilation time window by taking numerical simulation data of meteorological elements meeting quality and forecast accuracy requirements as input and forecasting field data of a corresponding numerical forecast mode as output to obtain an assimilation time window configuration scheme;
aiming at the spatial range of the data assimilation process, fused observation field data corresponding to numerical simulation data are used as input, forecast field data of a corresponding numerical forecasting mode are used as output, an assimilation area and an assimilation quality self-adaptive artificial intelligence model are optimized, and an assimilation spatial range configuration scheme is obtained;
and constructing a global optimal assimilation scheme of meteorological element data and a mode background field based on a time window configuration scheme and a space range configuration scheme according to the generalization capability of an artificial intelligence algorithm.
The training of the mode parameterization scheme comprises the following steps:
acquiring a mode key physical process parameterization scheme and sensitive parameters thereof which influence the wind speed of the key area to be forecasted;
aiming at the mode key physical process parameterization scheme, optimizing the mode key physical process parameterization scheme by adopting a heuristic intelligent optimization algorithm and taking wind speed forecasting precision as an optimization target on the basis of a mode initial state and a wind speed forecasting field corresponding to historical meteorological element data;
designing an optimization objective function of sample set error feedback aiming at sensitive parameters in the optimized mode key physical process parameterization scheme based on the optimized mode key physical process parameterization scheme, constructing a nestable optimization objective function integrating mode integral calculation and error index calculation, and developing sensitive parameter optimization in the optimized mode key physical process parameterization scheme through repeated iterative calculation;
and taking an optimized mode-critical physical process parameterization scheme for completing sensitive parameter optimization as a mode parameterization scheme.
Wherein, this system still includes: a wind speed calibration module;
the wind speed calibration module is used for calibrating a wind speed forecasting field by adopting a pre-trained wind speed calibration model;
the wind speed calibration model is obtained by training a physical mechanism and artificial intelligence based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period.
Wherein, the training of wind speed calibration model includes:
based on the numerical weather forecast wind speed field and the actually measured wind speed field in the historical time period, learning the spatial information of the weather forecast multi-grid-point element by adopting a convolutional neural network or a graph convolutional neural network;
based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period, learning relevant information hidden in each meteorological element on a time sequence through a recurrent neural network or a long-short term memory network;
the spatial information and the time sequence related information are used as physical mechanisms, and the hyper-parameters of the deep learning network model are optimized by combining the numerical weather forecast wind speed field and the measured wind speed field in the historical time period to obtain a wind speed calibration model;
wherein, the hyper-parameters of the deep learning network model comprise: network structure and learning rate; the deep learning network model is designed according to the application scene of wind speed forecasting.
After the wind speed forecasting field is calibrated by adopting the pre-trained wind speed calibration model, the method further comprises the following steps:
and taking the calibrated wind speed forecast field as a new numerical weather forecast wind speed field, and performing online optimization updating on the wind speed calibration model by combining with the actual measurement wind speed field corresponding to the calibrated wind speed forecast field.
Example 4:
the present invention also provides a computer apparatus comprising: one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the one or more processors, implement a wind speed forecasting method as previously described.
Example 5:
the invention also provides a computer-readable storage medium having stored thereon a computer program which, when executed, implements a wind speed forecasting method as described above.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It should be noted that the above-mentioned embodiments are only for illustrating the technical solutions of the present invention and not for limiting the protection scope thereof, and although the present invention is described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that after reading the present invention, they can make various changes, modifications or equivalents to the specific embodiments of the application, but these changes, modifications or equivalents are all within the protection scope of the claims of the application.
Claims (6)
1. A method of wind speed forecasting, comprising:
performing data assimilation by adopting a pre-constructed global optimal assimilation scheme based on the mode background field and the latest meteorological element data to obtain a mode initial state;
based on the mode initial state, adopting a pre-trained mode parameterization scheme to predict to obtain a wind speed forecasting field;
the global optimal assimilation scheme is constructed by adopting a method of combining artificial intelligence and a physical mechanism based on numerical simulation data of meteorological elements and forecast field data corresponding to a numerical forecast mode; the mode parameterization scheme is obtained by training by adopting an artificial intelligence method based on a mode initial state and a wind speed forecasting field corresponding to historical meteorological element data;
the construction of the global optimal assimilation scheme comprises the following steps:
acquiring the quality of meteorological element data in time and space ranges and forecasting precision requirements;
optimizing an artificial intelligence model for assimilation information attenuation and optimization in an assimilation time window by taking numerical simulation data of meteorological elements meeting quality and forecast accuracy requirements as input and forecasting field data of a corresponding numerical forecast mode as output to obtain an assimilation time window configuration scheme;
aiming at the spatial range of the data assimilation process, fused observation field data corresponding to the numerical simulation data are used as input, forecast field data of a corresponding numerical forecasting mode are used as output, an assimilation area and an assimilation quality self-adaptive artificial intelligence model are optimized, and an assimilation spatial range configuration scheme is obtained;
constructing a global optimal assimilation scheme of meteorological element data and a mode background field based on the time window configuration scheme and the space range configuration scheme and according to the generalization capability of an artificial intelligence algorithm;
after the wind speed forecasting field is obtained, the method further comprises the following steps:
calibrating a wind speed forecasting field by adopting a pre-trained wind speed calibration model;
the wind speed calibration model is obtained by training a physical mechanism and artificial intelligence based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period;
training of the wind speed calibration model comprises:
based on the numerical weather forecast wind speed field and the actually measured wind speed field in the historical time period, learning the spatial information of the weather forecast multi-grid-point element by adopting a convolutional neural network or a graph convolutional neural network;
based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period, learning relevant information hidden in each meteorological element on a time sequence through a recurrent neural network or a long-short term memory network;
the spatial information and the time sequence related information are used as physical mechanisms, and the hyper-parameters of the deep learning network model are optimized by combining a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period to obtain a wind speed calibration model;
wherein the hyper-parameters of the deep learning network model comprise: network structure and learning rate; the deep learning network model is designed according to the application scene of wind speed forecast;
training of the pattern parameterization scheme, comprising:
acquiring a mode key physical process parameterization scheme and sensitive parameters thereof which influence the wind speed of the key area to be forecasted;
aiming at the mode key physical process parameterization scheme, optimizing the mode key physical process parameterization scheme by adopting a heuristic intelligent optimization algorithm and taking wind speed forecasting precision as an optimization target on the basis of a mode initial state and a wind speed forecasting field corresponding to historical meteorological element data;
designing an optimization objective function of sample set error feedback aiming at sensitive parameters in the optimized mode key physical process parameterization scheme based on the optimized mode key physical process parameterization scheme, constructing a nestable optimization objective function integrating mode integral calculation and error index calculation, and carrying out sensitive parameter optimization in the optimized mode key physical process parameterization scheme through repeated iterative calculation;
and taking an optimized mode-critical physical process parameterization scheme for completing sensitive parameter optimization as a mode parameterization scheme.
2. The method of claim 1, wherein after the calibrating the wind speed forecasting field using the pre-trained wind speed calibration model, further comprising:
and taking the calibrated wind speed forecast field as a new numerical weather forecast wind speed field, and carrying out online optimization updating on the wind speed calibration model by combining with an actually measured wind speed field corresponding to the calibrated wind speed forecast field.
3. A wind speed forecasting system, comprising: data assimilation module and wind speed forecast module, wherein:
the data assimilation module is used for carrying out data assimilation by adopting a pre-constructed global optimal assimilation scheme based on the mode background field and the latest meteorological element data to obtain a mode initial state;
the wind speed forecasting module is used for forecasting by adopting a pre-trained mode parameterization scheme based on the mode initial state to obtain a wind speed forecasting field;
the global optimal assimilation scheme is constructed by adopting a method of combining artificial intelligence and a physical mechanism based on numerical simulation data of meteorological elements and forecast field data corresponding to a numerical forecast mode; the mode parameterization scheme is obtained by training by adopting an artificial intelligence method based on a mode initial state and a wind speed forecasting field corresponding to historical meteorological element data;
the construction of the global optimal assimilation scheme comprises the following steps:
acquiring the quality of meteorological element data in time and space ranges and forecasting precision requirements;
optimizing an artificial intelligence model for assimilation information attenuation and optimization in an assimilation time window by taking numerical simulation data of meteorological elements meeting quality and forecast accuracy requirements as input and forecasting field data of a corresponding numerical forecast mode as output to obtain an assimilation time window configuration scheme;
aiming at the spatial range of the data assimilation process, fused observation field data corresponding to the numerical simulation data are used as input, forecast field data of a corresponding numerical forecasting mode are used as output, an assimilation area and an assimilation quality self-adaptive artificial intelligence model are optimized, and an assimilation spatial range configuration scheme is obtained;
constructing a global optimal assimilation scheme of meteorological element data and a mode background field based on the time window configuration scheme and the space range configuration scheme and according to the generalization capability of an artificial intelligence algorithm;
further comprising: a wind speed calibration module;
the wind speed calibration module is used for calibrating a wind speed forecasting field by adopting a pre-trained wind speed calibration model;
the wind speed calibration model is obtained by training a physical mechanism and artificial intelligence based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period;
training of the wind speed calibration model comprises:
based on the numerical weather forecast wind speed field and the actually measured wind speed field in the historical time period, learning the spatial information of the weather forecast multi-grid-point element by adopting a convolutional neural network or a graph convolutional neural network;
based on a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period, learning relevant information hidden in each meteorological element on a time sequence through a recurrent neural network or a long-short term memory network;
the spatial information and the time sequence related information are used as physical mechanisms, and the hyper-parameters of the deep learning network model are optimized by combining a numerical weather forecast wind speed field and an actual measurement wind speed field in a historical time period to obtain a wind speed calibration model;
wherein the hyper-parameters of the deep learning network model comprise: network structure and learning rate; the deep learning network model is designed according to the application scene of wind speed forecast;
training of the pattern parameterization scheme, comprising:
acquiring a mode key physical process parameterization scheme and sensitive parameters thereof which influence the wind speed of the key area to be forecasted;
aiming at the mode key physical process parameterization scheme, optimizing the mode key physical process parameterization scheme by adopting a heuristic intelligent optimization algorithm and taking wind speed forecasting precision as an optimization target on the basis of a mode initial state and a wind speed forecasting field corresponding to historical meteorological element data;
designing an optimization objective function of sample set error feedback aiming at sensitive parameters in the optimized mode key physical process parameterization scheme based on the optimized mode key physical process parameterization scheme, constructing a nestable optimization objective function integrating mode integral calculation and error index calculation, and carrying out sensitive parameter optimization in the optimized mode key physical process parameterization scheme through repeated iterative calculation;
and taking an optimized mode-critical physical process parameterization scheme for completing sensitive parameter optimization as a mode parameterization scheme.
4. The system of claim 3, wherein after the wind speed forecasting field is calibrated using the pre-trained wind speed calibration model, the method further comprises:
and taking the calibrated wind speed forecast field as a new numerical weather forecast wind speed field, and carrying out online optimization updating on the wind speed calibration model by combining with an actually measured wind speed field corresponding to the calibrated wind speed forecast field.
5. A computer device, comprising: one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the one or more processors, implement the method of any of claims 1-2.
6. A computer-readable storage medium, having stored thereon a computer program which, when executed, implements the method of any of claims 1-2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210002536.1A CN114004430B (en) | 2022-01-05 | 2022-01-05 | Wind speed forecasting method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210002536.1A CN114004430B (en) | 2022-01-05 | 2022-01-05 | Wind speed forecasting method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114004430A CN114004430A (en) | 2022-02-01 |
| CN114004430B true CN114004430B (en) | 2022-04-29 |
Family
ID=79932576
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210002536.1A Active CN114004430B (en) | 2022-01-05 | 2022-01-05 | Wind speed forecasting method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114004430B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115293393B (en) * | 2022-04-11 | 2023-04-18 | 北京城市气象研究院 | Near-ground wind speed prediction method combining turbulence physical model and historical data optimization |
| CN115049024B (en) * | 2022-08-15 | 2022-11-01 | 国能日新科技股份有限公司 | Training method and device of wind speed prediction model, electronic equipment and storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102628876B (en) * | 2012-02-13 | 2013-07-31 | 甘肃省电力公司风电技术中心 | Ultra-short term prediction method comprising real-time upstream and downstream effect monitoring |
| US9952352B2 (en) * | 2015-09-25 | 2018-04-24 | The United States Of America As Represented By The Secretary Of The Army | Device and system for generating a weather forecast for a selected zone |
| CN109325633A (en) * | 2018-10-23 | 2019-02-12 | 中国电力科学研究院有限公司 | A kind of weather forecast set member choosing method and system |
| CN112149859B (en) * | 2019-06-28 | 2023-10-31 | 北京金风慧能技术有限公司 | Wind speed prediction method and wind speed prediction device |
-
2022
- 2022-01-05 CN CN202210002536.1A patent/CN114004430B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114004430A (en) | 2022-02-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114004430B (en) | Wind speed forecasting method and system | |
| CN110135630B (en) | Short-term load demand prediction method based on random forest regression and multi-step optimization | |
| CN109472321A (en) | A kind of prediction towards time series type surface water quality big data and assessment models construction method | |
| CN112734128A (en) | 7-day power load peak value prediction method based on optimized RBF | |
| CN109598052B (en) | Intelligent ammeter life cycle prediction method and device based on correlation coefficient analysis | |
| Li et al. | A novel multichannel long short-term memory method with time series for soil temperature modeling | |
| CN113325721B (en) | Model-free adaptive control method and system for industrial system | |
| CN114492922A (en) | Medium-and-long-term power generation capacity prediction method | |
| CN115271186B (en) | Reservoir water level prediction and early warning method based on delay factor and PSO RNN Attention model | |
| CN112735541A (en) | Sewage treatment water quality prediction method based on simple circulation unit neural network | |
| CN110097929A (en) | A kind of blast furnace molten iron silicon content on-line prediction method | |
| CN116384244A (en) | Electromagnetic field prediction method based on physical enhancement neural network | |
| Dong et al. | Short-term building cooling load prediction model based on DwdAdam-ILSTM algorithm: A case study of a commercial building | |
| Sankalp et al. | Deep learning models comparable assessment and uncertainty analysis for diurnal temperature range (DTR) predictions over Indian urban cities | |
| Zhu et al. | Time-varying interval prediction and decision-making for short-term wind power using convolutional gated recurrent unit and multi-objective elephant clan optimization | |
| Mystakidis et al. | One step ahead energy load forecasting: A multi-model approach utilizing machine and deep learning | |
| CN110457737B (en) | Method for quickly positioning water pollution source based on neural network | |
| CN117252085A (en) | Cable-stayed bridge cable force tolerance optimization method and system based on neural network and particle swarm | |
| Li et al. | Discovering Mathematical Formulas from Data via GPT-guided Monte Carlo Tree Search | |
| Zhang et al. | A deep reinforcement learning based human behavior prediction approach in smart home environments | |
| Liu et al. | Application of least square support vector machine based on particle swarm optimization to chaotic time series prediction | |
| Li et al. | The research of temperature compensation for thermopile sensor based on improved PSO-BP algorithm | |
| CN113987904A (en) | Method, device, equipment and storage medium for measuring and calculating repair cost of power transmission project | |
| KR20230064389A (en) | Method and system for optimizing parameter intervals of manufacturing processes based on prediction intervals | |
| CN116544931B (en) | Power load distribution prediction method based on integrated fragment transformation and time convolution 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 |