CN115099153A - Wind power prediction model training method, wind power prediction device and medium - Google Patents
Wind power prediction model training method, wind power prediction device and medium Download PDFInfo
- Publication number
- CN115099153A CN115099153A CN202210780732.1A CN202210780732A CN115099153A CN 115099153 A CN115099153 A CN 115099153A CN 202210780732 A CN202210780732 A CN 202210780732A CN 115099153 A CN115099153 A CN 115099153A
- Authority
- CN
- China
- Prior art keywords
- data
- fan
- wind power
- power prediction
- time sequence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Abstract
The invention provides a wind power prediction model training method, a wind power prediction device and a medium, wherein the training method comprises the following steps: acquiring time sequence fan data of a fan and meteorological numerical data of a wind field where the fan is located, wherein the meteorological numerical data represent meteorological data of the wind field, and the time sequence fan data represent fan data which are sequenced according to time; performing feature extraction on the time sequence fan data based on a self-attention mechanism to obtain first feature data, and performing feature extraction on the meteorological numerical data based on a cross-attention mechanism to obtain second feature data; fusing the first characteristic data and the second characteristic data to obtain training data; and training a pre-constructed deep neural network by adopting the training data to obtain a wind power prediction model. The technical scheme of the invention improves the accuracy of wind power prediction.
Description
Technical Field
The invention relates to the technical field of wind power generation, in particular to a wind power prediction model training method, a wind power prediction device and a medium.
Background
With the annual increase of the electricity consumption and the fossil fuel cost of society, the demand of society for new energy power generation is higher and higher. Among them, wind power generation has received more and more attention as a main new energy power generation method because of its characteristics of cleanness, no public nuisance, renewability and large energy storage. However, the wind power generation is influenced by the fluctuation of the wind power, so that the output power is unstable, the price pricing of the wind power generation is difficult, and the wind power refers to the wind power generation power of a fan. In addition, if the wind power suddenly changes, the load of the fan is likely to suddenly increase due to hysteresis of the fan control system, and the fan may be damaged.
In order to optimize a wind power pricing strategy and protect the safety of a fan, a deep learning model is usually trained through collected fan data at present, the trained deep learning model is used for predicting wind power, and wind power pricing is optimized in time and self-adaptive control is carried out on the fan according to the predicted wind power. However, the method only considers the data of the fan, does not consider the influence of other factors on the wind power, and is poor in accuracy.
Disclosure of Invention
The invention solves the problem of how to improve the accuracy of wind power prediction.
In order to solve the above problems, the present invention provides a wind power prediction model training method, a wind power prediction device, and a medium.
In a first aspect, the present invention provides a method for training a wind power prediction model, including:
acquiring time sequence fan data of a fan and meteorological numerical data of a wind field where the fan is located, wherein the meteorological numerical data represent meteorological data of the wind field, and the time sequence fan data represent fan data which are sequenced according to time;
performing feature extraction on the time sequence fan data based on a self-attention mechanism to obtain first feature data, and performing feature extraction on the meteorological numerical data based on a cross-attention mechanism to obtain second feature data;
fusing the first characteristic data and the second characteristic data to obtain training data;
and training a pre-constructed deep neural network by adopting the training data to obtain a wind power prediction model.
Optionally, the wind turbine data comprises at least one of average wind speed, rotor speed and wind turbine condition, and the meteorological data comprises at least one of cross-longitudinal wind speed, temperature and humidity;
the acquiring of the time sequence fan data of the fan and the meteorological numerical data of the wind field where the fan is located comprises the following steps: acquiring labeled fan data of the fan at different times, acquiring the meteorological numerical data in an N-N grid with the wind field as the center, and generating the time sequence fan data with a label according to the labeled fan data at each time, wherein the label is wind power, and N is greater than or equal to 1.
Optionally, before the feature extraction is performed on the time-series fan data based on the self-attention mechanism, the method further includes:
sequentially dividing the time sequence fan data according to a preset time span to obtain a plurality of fan data sequences;
and determining the average value of the fan data in each fan data sequence, and sequentially combining the average values corresponding to each fan data sequence to obtain the processed time sequence fan data.
Optionally, the feature extraction of the time-series fan data based on the self-attention mechanism includes:
acquiring the relative position of each fan data in the time sequence fan data, the time information of each fan data and the fan state information at different times;
determining the position code of each fan data according to the relative position, determining the time sequence code of each fan data according to the time information, and determining the state code corresponding to each fan data according to the fan state information;
converting the fan data, the position codes, the time sequence codes and the state codes into the same dimension, and accumulating to obtain an input data sequence;
and performing feature extraction on the input data sequence based on a multi-head self-attention mechanism to obtain the first feature data.
Optionally, the input data sequence includes a plurality of input data, and the feature extraction on the input data sequence based on the multi-head self-attention mechanism includes:
for one input data, respectively extracting features according to preset multiple groups of weights to obtain multiple sub-feature data;
splicing all the sub-feature data of the input data to obtain a feature vector;
and performing linear conversion on the feature vector to obtain the first feature data corresponding to the input data.
In a second aspect, the present invention provides a wind power prediction method, including:
acquiring time sequence fan data of a target fan and meteorological numerical data of a wind field where the target fan is located;
performing feature extraction on the time sequence fan data based on a self-attention mechanism to obtain first feature data, and performing feature extraction on the meteorological numerical data based on a cross-attention mechanism to obtain second feature data;
fusing the first characteristic data and the second characteristic data to obtain input data;
inputting the input data into a wind power prediction model to obtain the wind power corresponding to the target fan;
wherein the wind power prediction model adopts the wind power prediction model training method according to any one of the first aspect.
In a third aspect, the present invention provides a wind power prediction model training apparatus, including:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring time sequence fan data of a fan and meteorological numerical data of a wind field where the fan is located, the meteorological numerical data represent meteorological data of the wind field, and the time sequence fan data represent fan data which are sequenced according to time;
the first extraction module is used for extracting the characteristics of the time sequence fan data based on a self-attention mechanism to obtain first characteristic data, and extracting the characteristics of the meteorological numerical data based on a cross-attention mechanism to obtain second characteristic data;
the first fusion module is used for fusing the first characteristic data and the second characteristic data to obtain training data;
and the training module is used for training a pre-constructed deep neural network by adopting the training data to obtain a wind power prediction model.
In a fourth aspect, the present invention provides a wind power prediction apparatus, comprising:
the second acquisition module is used for acquiring time sequence fan data of a target fan and meteorological numerical data of a wind field where the target fan is located;
the second extraction module is used for extracting the characteristics of the time sequence fan data based on a self-attention mechanism to obtain first characteristic data, and extracting the characteristics of the meteorological numerical data based on a cross-attention mechanism to obtain second characteristic data;
the second fusion module is used for fusing the first characteristic data and the second characteristic data to obtain input data;
the prediction module is used for inputting the input data into a wind power prediction model to obtain the wind power corresponding to the target fan;
wherein the wind power prediction model adopts the wind power prediction model training method according to any one of the first aspect.
In a fifth aspect, the present invention provides an electronic device comprising a memory and a processor;
the memory for storing a computer program;
the processor is configured to, when executing the computer program, implement the wind power prediction model training method according to any one of the first aspect or the wind power prediction method according to the second aspect.
In a sixth aspect, the invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the wind power prediction model training method according to any one of the first aspect or the wind power prediction method according to the second aspect.
The wind power prediction model training method, the wind power prediction device and the medium have the beneficial effects that: the method comprises the steps of obtaining time sequence fan data of a fan and meteorological numerical data of a wind field where the fan is located, extracting first characteristic data from the time sequence fan data based on a self-attention mechanism, fully learning historical time sequence fan data, filtering noise, improving accuracy of the extracted first characteristic data, and further improving accuracy of a wind power prediction model obtained through follow-up training. The second characteristic data is extracted from the meteorological numerical data based on a cross attention mechanism, data overflow can be prevented through a data scaling method, calculation efficiency is improved, and the meteorological data with larger dimensionality can be effectively processed. The method comprises the steps of fusing first characteristic data and second characteristic data to obtain training data, training a pre-constructed deep neural network by adopting the training data to obtain a wind power prediction model, considering fan data and fusing meteorological numerical data influencing wind power, improving comprehensiveness of the training data, improving accuracy of the wind power prediction model obtained by training, and enabling the wind power prediction model to accurately predict wind power in a certain period of time in the future.
Drawings
FIG. 1 is a schematic flow chart of a wind power prediction model training method according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart illustrating feature extraction based on the self-attention mechanism according to an embodiment of the present invention;
FIG. 3 is a flow chart illustrating a wind power prediction method according to another embodiment of the present invention;
FIG. 4 is a schematic structural diagram of a wind power prediction model training apparatus according to another embodiment of the present invention;
fig. 5 is a schematic structural diagram of a wind power prediction apparatus according to another embodiment of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below. While certain embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be construed as limited to the embodiments set forth herein, but rather are provided for a more complete and thorough understanding of the present invention. It should be understood that the drawings and the embodiments of the present invention are illustrative only and are not intended to limit the scope of the present invention.
It should be understood that the various steps recited in method embodiments of the present invention may be performed in a different order, and/or performed in parallel. Moreover, method embodiments may include additional steps and/or omit performing the illustrated steps. The scope of the invention is not limited in this respect.
The term "include" and variations thereof as used herein are open-ended, i.e., "including but not limited to". The term "based on" is "based, at least in part, on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "alternative embodiments". Relevant definitions for other terms will be given in the following description. It should be noted that the terms "first", "second", and the like in the present invention are only used for distinguishing different devices, modules or units, and are not used for limiting the order or interdependence of the functions performed by the devices, modules or units.
It is noted that references to "a", "an", and "the" modifications in the present invention are intended to be illustrative rather than limiting, and that those skilled in the art will recognize that reference to "one or more" unless the context clearly dictates otherwise.
The names of messages or information exchanged between devices in the embodiments of the present invention are for illustrative purposes only, and are not intended to limit the scope of the messages or information.
As shown in fig. 1, a method for training a wind power prediction model according to an embodiment of the present invention includes:
step S110, acquiring time sequence fan data of a fan and meteorological numerical data of a wind field where the fan is located, wherein the meteorological numerical data represent meteorological data of the wind field, and the time sequence fan data represent fan data which are sorted according to time.
Specifically, the wind turbine data includes at least one of an average wind speed, a rotor speed, and a wind turbine state, and the meteorological data includes at least one of a cross-longitudinal wind speed, a temperature, and a humidity. The data acquisition and monitoring system can acquire the fan data of the fan at different times, the time sequence fan data can be obtained according to time sequence, and numerical meteorological data can be acquired from a meteorological center. The time sequence fan data may be fan data with a tag for the corresponding wind power.
And step S120, performing feature extraction on the time sequence fan data based on a self-attention mechanism to obtain first feature data, and performing feature extraction on the meteorological numerical data based on a cross-attention mechanism to obtain second feature data.
Specifically, the time-series fan data is W × F data, where W represents the number of time points and F represents the data amount at each time point, and the time-series fan data may be converted into 1 × D1-dimensional input data by a flat layer and a linear layer, and feature extraction may be performed by an auto-attention mechanism. The meteorological numerical data are N × G data, wherein N is the size of a grid with a fan as the center and is larger than or equal to 1, G is the characteristic dimension of the meteorological data, the meteorological numerical data can be convoluted into input data with 1 × D2 dimensions through series-connected 3 × 3 convolution kernels, feature extraction is carried out through a cross attention mechanism, and compared with the large convolution kernel, the nonlinear features of the meteorological data can be extracted more accurately by adopting the small convolution kernel of 3 × 3. The cross attention mechanism can adopt a method of scaling dot multiplication, so that the calculation efficiency is improved, data overflow is prevented, and the calculation effect on a larger dimension vector is effectively improved.
And S130, fusing the first characteristic data and the second characteristic data to obtain training data.
In particular, the spatial properties of the wind are applied, i.e. the wind speed at the next time step is related to the wind speed at the current neighbouring grid point, e.g. the wind blows from the current grid point to the neighbouring grid point, the current wind speed is related to the current grid point and the wind speed at the next time step is related to the neighbouring grid point. And performing space-time fusion on the first characteristic data of the time sequence fan data and the second characteristic data of the meteorological numerical data, and obtaining training data by adopting a method of changing space into time.
And S140, training a pre-constructed deep neural network by using the training data to obtain a wind power prediction model.
Specifically, the training data can be divided into a training set and a test set, and the deep neural network of the construction number is trained by the training set to obtain a trained deep neural network, namely, a wind power prediction model. The test set may then be employed to evaluate the prediction accuracy of the wind power prediction model.
In the embodiment, the time sequence fan data of the fan and the meteorological numerical data of the wind field where the fan is located are obtained, the first characteristic data are extracted from the time sequence fan data based on the self-attention mechanism, the historical time sequence fan data can be fully learned, noise is filtered, the accuracy of the extracted first characteristic data is improved, and the accuracy of the wind power prediction model obtained through follow-up training is further improved. The second characteristic data is extracted from the meteorological numerical data based on a cross attention mechanism, data overflow can be prevented through a data scaling method, calculation efficiency is improved, and the meteorological data with larger dimensionality can be effectively processed. The method comprises the steps of fusing first characteristic data and second characteristic data to obtain training data, training a pre-constructed deep neural network by adopting the training data to obtain a wind power prediction model, considering fan data and meteorological numerical data influencing wind power, improving comprehensiveness of the training data, improving accuracy of the wind power prediction model obtained by training, and predicting the wind power in an ultra-short period, such as 4 hours after the starting point.
Optionally, the acquiring time sequence fan data of the fan and meteorological numerical data of a wind field where the fan is located includes: acquiring the labeled fan data of the fan at different times, acquiring the meteorological numerical data in an N-x-N grid with the wind field as the center, and generating the labeled time sequence fan data according to the labeled fan data at each time, wherein the label is wind power, and N is greater than or equal to 1.
Specifically, the fan data with the labels of the fans at different times are obtained, and the fan data with the labels can be sequenced according to the sequence of time to generate the time sequence fan data with the labels. For the meteorological numerical data, the meteorological numerical data downloaded from the meteorological center are nc format data, and can be read in a coordinate classification mode by using a python library function provided by the meteorological center, then the read data are re-integrated into form data sorted according to time, and further are arranged into three-dimensional grid data of N G, wherein N is the grid size, N can be preferably 7, the width size of each grid can be 0.125 degrees, and G is a characteristic number.
It should be noted that, in the prior art, an interpolation method is usually adopted for fusing two types of data, for example, time granularity refinement is performed on meteorological data through interpolation, so that the meteorological data and the wind turbine data are aligned in a time step, and then the aligned meteorological data and the wind turbine data are spliced and fused. The method is simple and easy to implement, but the meteorological numerical data refined by the interpolation method loses the volatility of the meteorological numerical data, the difference between the meteorological numerical data such as wind speed and the like generated by interpolation and the real data is very large, namely noise data is introduced, and adverse effects are brought to the prediction effect of the wind power prediction model obtained by training.
In the optional embodiment, compared with the prior art in which data refinement is performed by an interpolation method, the spatial attribute of wind is applied, that is, the next time step is related to the adjacent grid point of the current grid point, so that a method of changing space time is adopted to refine the meteorological numerical data in spatial granularity, and then perform space-time fusion on the meteorological numerical data and time sequence fan data, thereby solving the problems of data loss volatility and introduction of noise data caused by data refinement by the interpolation method, improving the training effect of a depth model, and improving the prediction accuracy of a wind power prediction model obtained by training.
Optionally, before the feature extraction is performed on the time-series fan data based on the self-attention mechanism, the method further includes:
and dividing the time sequence fan data in sequence according to a preset time span to obtain a plurality of fan data sequences.
For example, assuming that the time span is 10 minutes, the fan data in the time series fan data is sequentially intercepted, and a plurality of fan data sequences are obtained, where each fan data sequence includes 10 minutes of fan data.
And determining the average value of the fan data in each fan data sequence, and sequentially combining the average values corresponding to each fan data sequence to obtain the processed time sequence fan data.
Specifically, fan data in each fan data sequence are averaged to obtain an average value corresponding to each fan data sequence, and all average values are combined according to a time sequence to obtain processed time sequence fan data which are used as input data of feature extraction.
In this optional embodiment, the average value can represent the fan speed condition in a time period, and compared with the case of selecting fan data at a specific time point as input data for feature extraction, the influence of wind fluctuation on the data can be avoided, so that the accuracy of extracted features is improved, and meanwhile, for the data in a missing time period, filling is not needed, so that data noise is prevented from being introduced, and the accuracy of feature extraction is prevented from being influenced.
Optionally, the feature extraction of the time-series fan data based on the self-attention mechanism includes:
and acquiring the relative position of each fan data in the time sequence fan data, the time information of each fan data and the fan state information at different times.
And determining the position code of each fan data according to the relative position, determining the time sequence code of each fan data according to the time information, and determining the state code corresponding to each fan data according to the fan state information.
Specifically, the fan data can also be encoded to obtain the feature code. The time sequence code can comprise month code, date code and hour code, and the state code can comprise limited power operation code and fan state code.
And converting the fan data, the position codes, the time sequence codes and the state codes into the same dimension, and accumulating to obtain an input data sequence.
And performing feature extraction on the input data sequence based on a multi-head self-attention mechanism to obtain the first feature data.
It should be noted that, two segments of time sequence fan data in different seasons and times or in different power-limited operation states of the fans may not be distinguished from each other by feature extraction based on the self-attention mechanism, so that feature data extracted from the two segments of time sequence fan data are the same, and prediction deviation of wind power is caused.
In this optional embodiment, position information, state information, and time information are integrated into the wind turbine data, so that input data obtained after information is integrated can be distinguished from one another, the problem that similar time sequence wind turbine data at different times and states are difficult to distinguish when time sequence features are extracted based on a self-attention mechanism is avoided, and the prediction accuracy of the wind power is improved.
Optionally, the input data sequence includes a plurality of input data, and the feature extraction on the input data sequence based on the multi-head self-attention mechanism includes:
for one input data, respectively extracting features according to preset multiple groups of weights to obtain multiple sub-feature data;
splicing all the sub-feature data of the input data to obtain a feature vector;
and performing linear conversion on the feature vector to obtain the first feature data corresponding to the input data.
Specifically, as shown in fig. 2, a set of weight pairs (Wq, Wk, Wv) is used to perform feature extraction on the input data sequence Xt through a self-attention mechanism, so as to obtain a sub-feature sequence Xt ', where the sub-feature sequence Xt' includes multiple sub-feature data, and the sub-feature data corresponds to the input data. Respectively extracting the features of the input data sequence through a plurality of groups of weight pairs to obtain a plurality of sub-feature sequences, respectively splicing the corresponding sub-feature data in each sub-feature sequence to obtain a feature vector corresponding to each sub-feature data, for example: and splicing the first sub-feature data in the first sub-feature sequence with the first sub-feature data in the second sub-feature sequence until the first sub-feature data in the Nth sub-feature sequence to obtain a feature vector corresponding to the sub-feature data. The feature vectors can be converted into first feature data through linear conversion, and the first feature data corresponding to each sub-feature data are combined to obtain a first feature data sequence. The specific process of feature extraction by the self-attention mechanism is the prior art, and is not described herein again.
In this optional embodiment, the feature extraction is performed through a multi-head self-attention mechanism, and compared with the feature extraction performed by a common self-attention mechanism, the data can be learned from multiple angles, so that the occurrence of deviation is avoided, the accuracy of the extracted features can be improved, the effect of model training is further improved, and the prediction accuracy of the wind power is improved.
As shown in fig. 3, another embodiment of the present invention provides a wind power prediction method, including:
step S210, acquiring time sequence fan data of a target fan and meteorological numerical data of a wind field where the target fan is located;
step S220, performing feature extraction on the time sequence fan data based on a self-attention mechanism to obtain first feature data, and performing feature extraction on the meteorological numerical data based on a cross-attention mechanism to obtain second feature data;
step S230, fusing the first feature data and the second feature data to obtain input data;
step S240, inputting the input data into a wind power prediction model to obtain the wind power corresponding to the target fan;
the wind power prediction model adopts the wind power prediction model training method.
In the embodiment, the time sequence fan data of the target fan and the meteorological numerical data of the wind field where the target fan is located are obtained, the first characteristic data can be accurately extracted from the time sequence fan data based on the self-attention mechanism, and the accuracy of wind power prediction is improved. The second characteristic data can be effectively extracted from the meteorological numerical data based on the cross attention mechanism, the data overflow can be prevented, and the calculation efficiency is improved. The first characteristic data and the second characteristic data are fused, so that the time sequence fan data and meteorological numerical data influencing wind power are fused into the input data, and comprehensiveness of factors influencing the wind power is improved. The input data are input into the wind power prediction model obtained by training through the wind power prediction model training method, the wind power of the target fan can be predicted in an ultra-short period, for example, the wind power within 4 hours after the starting point is predicted, and the accuracy is high.
As shown in fig. 4, a wind power prediction model training apparatus according to another embodiment of the present invention includes:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring time sequence fan data of a fan and meteorological numerical data of a wind field where the fan is located, the meteorological numerical data represent meteorological data of the wind field, and the time sequence fan data represent fan data which are sequenced according to time;
the first extraction module is used for extracting the characteristics of the time sequence fan data based on a self-attention mechanism to obtain first characteristic data, and extracting the characteristics of the meteorological numerical data based on a cross-attention mechanism to obtain second characteristic data;
the first fusion module is used for fusing the first characteristic data and the second characteristic data to obtain training data;
and the training module is used for training a pre-constructed deep neural network by adopting the training data to obtain a wind power prediction model.
The wind power prediction model training device is used for realizing the wind power prediction model training method so as to obtain corresponding beneficial effects.
As shown in fig. 5, a wind power prediction apparatus according to another embodiment of the present invention includes:
the second acquisition module is used for acquiring time sequence fan data of a target fan and meteorological numerical data of a wind field where the target fan is located;
the second extraction module is used for extracting the characteristics of the time sequence fan data based on a self-attention mechanism to obtain first characteristic data, and extracting the characteristics of the meteorological numerical data based on a cross-attention mechanism to obtain second characteristic data;
the second fusion module is used for fusing the first characteristic data and the second characteristic data to obtain input data;
the prediction module is used for inputting the input data into a wind power prediction model to obtain the wind power corresponding to the target fan;
the wind power prediction model adopts the wind power prediction model training method.
The wind power prediction device is used for realizing the wind power prediction method so as to obtain corresponding beneficial effects.
Another embodiment of the present invention provides an electronic device including a memory and a processor; the memory for storing a computer program; the processor, when executing the computer program, is configured to implement the wind power prediction model training method as described above or the wind power prediction method as described above.
A further embodiment of the invention provides a computer readable storage medium having stored thereon a computer program which, when being executed by a processor, carries out the wind power prediction model training method as described above or the wind power prediction method as described above.
An electronic device that may be a server or a client of the present invention, which is an example of a hardware device that may be applied to aspects of the present invention, will now be described. Electronic device is intended to represent various forms of digital electronic computer devices, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed herein.
The electronic device includes a computing unit that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) or a computer program loaded from a storage unit into a Random Access Memory (RAM). In the RAM, various programs and data required for the operation of the device can also be stored. The computing unit, the ROM, and the RAM are connected to each other by a bus. An input/output (I/O) interface is also connected to the bus.
The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like. In this application, the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment of the present invention. In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
Although the present disclosure has been described with reference to the above embodiments, the scope of the present disclosure is not limited thereto. Various changes and modifications may be effected therein by one of ordinary skill in the pertinent art without departing from the spirit and scope of the present disclosure, and these changes and modifications are intended to be within the scope of the present disclosure.
Claims (10)
1. A wind power prediction model training method is characterized by comprising the following steps:
acquiring time sequence fan data of a fan and meteorological numerical data of a wind field where the fan is located, wherein the meteorological numerical data represent meteorological data of the wind field, and the time sequence fan data represent fan data which are sequenced according to time;
performing feature extraction on the time sequence fan data based on a self-attention mechanism to obtain first feature data, and performing feature extraction on the meteorological numerical data based on a cross-attention mechanism to obtain second feature data;
fusing the first characteristic data and the second characteristic data to obtain training data;
and training a pre-constructed deep neural network by adopting the training data to obtain a wind power prediction model.
2. The method of claim 1, wherein the wind turbine data includes at least one of average wind speed, rotor speed, and wind turbine state, and the meteorological data includes at least one of cross-longitudinal wind speed, temperature, and humidity;
the acquiring of the time sequence fan data of the fan and the meteorological numerical data of the wind field where the fan is located comprises the following steps: acquiring labeled fan data of the fan at different times, acquiring the meteorological numerical data in an N-N grid with the wind field as the center, and generating the time sequence fan data with a label according to the labeled fan data at each time, wherein the label is wind power, and N is greater than or equal to 1.
3. The method for training the wind power prediction model according to claim 2, wherein before the feature extraction of the time-series wind turbine data based on the self-attention mechanism, the method further comprises:
sequentially dividing the time sequence fan data according to a preset time span to obtain a plurality of fan data sequences;
and determining the average value of the fan data in each fan data sequence, and sequentially combining the average values corresponding to each fan data sequence to obtain the processed time sequence fan data.
4. The training method of the wind power prediction model according to any one of claims 1 to 3, wherein the feature extraction of the time series wind turbine data based on the self-attention mechanism comprises:
acquiring the relative position of each fan data in the time sequence fan data, the time information of each fan data and the fan state information at different times;
determining the position code of each fan data according to the relative position, determining the time sequence code of each fan data according to the time information, and determining the state code corresponding to each fan data according to the fan state information;
converting the fan data, the position codes, the time sequence codes and the state codes into the same dimension, and accumulating to obtain an input data sequence;
and performing feature extraction on the input data sequence based on a multi-head self-attention mechanism to obtain the first feature data.
5. The method of claim 4, wherein the input data sequence comprises a plurality of input data, and wherein the feature extracting the input data sequence based on the multi-head self-attention mechanism comprises:
respectively extracting features of the input data according to a plurality of preset groups of weights to obtain a plurality of sub-feature data;
splicing all the sub-feature data of the input data to obtain a feature vector;
and performing linear conversion on the feature vector to obtain the first feature data corresponding to the input data.
6. A method of wind power prediction, comprising:
acquiring time sequence fan data of a target fan and meteorological numerical data of a wind field where the target fan is located;
performing feature extraction on the time sequence fan data based on a self-attention mechanism to obtain first feature data, and performing feature extraction on the meteorological numerical data based on a cross-attention mechanism to obtain second feature data;
fusing the first characteristic data and the second characteristic data to obtain input data;
inputting the input data into a wind power prediction model to obtain the wind power corresponding to the target fan;
wherein the wind power prediction model adopts a wind power prediction model training method according to any one of claims 1 to 5.
7. A wind power prediction model training device, comprising:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring time sequence fan data of a fan and meteorological numerical data of a wind field where the fan is located, the meteorological numerical data represent meteorological data of the wind field, and the time sequence fan data represent fan data which are sequenced according to time;
the first extraction module is used for performing feature extraction on the time sequence fan data based on a self-attention mechanism to obtain first feature data, and performing feature extraction on the meteorological numerical data based on a cross-attention mechanism to obtain second feature data;
the first fusion module is used for fusing the first characteristic data and the second characteristic data to obtain training data;
and the training module is used for training a pre-constructed deep neural network by adopting the training data to obtain a wind power prediction model.
8. A wind power prediction device, comprising:
the second acquisition module is used for acquiring time sequence fan data of a target fan and meteorological numerical data of a wind field where the target fan is located;
the second extraction module is used for performing feature extraction on the time sequence fan data based on a self-attention mechanism to obtain first feature data, and performing feature extraction on the meteorological numerical data based on a cross-attention mechanism to obtain second feature data;
the second fusion module is used for fusing the first characteristic data and the second characteristic data to obtain input data;
the prediction module is used for inputting the input data into a wind power prediction model to obtain the wind power corresponding to the target fan;
wherein the wind power prediction model adopts a wind power prediction model training method according to any one of claims 1 to 5.
9. An electronic device comprising a memory and a processor;
the memory for storing a computer program;
the processor, when executing the computer program, for implementing the wind power prediction model training method according to any of claims 1 to 5 or the wind power prediction method according to claim 6.
10. A computer-readable storage medium, characterized in that the storage medium has stored thereon a computer program which, when being executed by a processor, carries out a wind power prediction model training method according to any one of claims 1 to 5 or a wind power prediction method according to claim 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210780732.1A CN115099153A (en) | 2022-07-04 | 2022-07-04 | Wind power prediction model training method, wind power prediction device and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210780732.1A CN115099153A (en) | 2022-07-04 | 2022-07-04 | Wind power prediction model training method, wind power prediction device and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115099153A true CN115099153A (en) | 2022-09-23 |
Family
ID=83296092
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210780732.1A Pending CN115099153A (en) | 2022-07-04 | 2022-07-04 | Wind power prediction model training method, wind power prediction device and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115099153A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116826727A (en) * | 2023-06-28 | 2023-09-29 | 河海大学 | Ultra-short-term wind power prediction method and prediction system based on time sequence representation and multistage attention |
| CN117094452A (en) * | 2023-10-20 | 2023-11-21 | 浙江天演维真网络科技股份有限公司 | Drought state prediction method, and training method and device of drought state prediction model |
-
2022
- 2022-07-04 CN CN202210780732.1A patent/CN115099153A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116826727A (en) * | 2023-06-28 | 2023-09-29 | 河海大学 | Ultra-short-term wind power prediction method and prediction system based on time sequence representation and multistage attention |
| CN116826727B (en) * | 2023-06-28 | 2024-02-23 | 河海大学 | Ultra-short-term wind power prediction method and prediction system based on time sequence representation and multistage attention |
| CN117094452A (en) * | 2023-10-20 | 2023-11-21 | 浙江天演维真网络科技股份有限公司 | Drought state prediction method, and training method and device of drought state prediction model |
| CN117094452B (en) * | 2023-10-20 | 2024-02-06 | 浙江天演维真网络科技股份有限公司 | Drought state prediction method, and training method and device of drought state prediction model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fan et al. | Forecasting short-term electricity load using hybrid support vector regression with grey catastrophe and random forest modeling | |
| CN115099153A (en) | Wind power prediction model training method, wind power prediction device and medium | |
| Zareipour et al. | Classification of future electricity market prices | |
| Kang et al. | Big data analytics in China's electric power industry: modern information, communication technologies, and millions of smart meters | |
| Beldiceanu et al. | Describing and generating solutions for the EDF unit commitment problem with the ModelSeeker | |
| Jiang et al. | Wind speed forecasting using autoregressive moving average/generalized autoregressive conditional heteroscedasticity model | |
| CN111898247A (en) | Landslide displacement prediction method, equipment and storage medium | |
| Bouguera et al. | A novel solar energy predictor for communicating sensors | |
| CN110619407B (en) | Object sales prediction method and system, electronic device and storage medium | |
| CN112990583B (en) | Method and equipment for determining model entering characteristics of data prediction model | |
| CN117234844A (en) | Cloud server abnormality management method and device, computer equipment and storage medium | |
| CN115879956A (en) | E-commerce customer maintenance pushing method, device, equipment and storage medium | |
| CN116304699A (en) | Critical sample set generation method and system based on new energy multi-station short circuit ratio | |
| CN115563507A (en) | Generation method, device and equipment for renewable energy power generation scene | |
| CN114443948A (en) | Ranking model training method, ranking method and device based on multi-scene data | |
| Liu et al. | Quantized reservoir computing on edge devices for communication applications | |
| Deng et al. | Intelligent prediction of rolling bearing remaining useful life based on probabilistic DeepAR-Transformer model | |
| CN117748500B (en) | Photovoltaic power prediction method, device, equipment and medium | |
| CN116070138B (en) | State monitoring method, device, equipment and medium for pumped storage unit | |
| Bu et al. | Distributed unit commitment scheduling in the future smart grid with intermittent renewable energy resources and stochastic power demands | |
| CN116541766B (en) | Training method of electroencephalogram data restoration model, electroencephalogram data restoration method and device | |
| CN115640896B (en) | Household user power load prediction method under multi-user scene and related equipment | |
| CN114237915B (en) | Data restoration method, device and equipment based on machine learning in distributed environment | |
| CN117039870A (en) | Power load prediction method, device, equipment and storage medium | |
| Wu et al. | Data quality aware chance‐constrained DC‐OPF: a variational Bayesian Gaussian mixture approach |
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 |