KR101749427B1 - Method for forecasting wind speed based on artificial neural networks having different features - Google Patents

Method for forecasting wind speed based on artificial neural networks having different features Download PDF

Info

Publication number
KR101749427B1
KR101749427B1 KR1020150166316A KR20150166316A KR101749427B1 KR 101749427 B1 KR101749427 B1 KR 101749427B1 KR 1020150166316 A KR1020150166316 A KR 1020150166316A KR 20150166316 A KR20150166316 A KR 20150166316A KR 101749427 B1 KR101749427 B1 KR 101749427B1
Authority
KR
South Korea
Prior art keywords
wind speed
data
initial
prediction
learning
Prior art date
Application number
KR1020150166316A
Other languages
Korean (ko)
Other versions
KR20170061377A (en
Inventor
이정훈
박경린
현예빈
이윤지
윤영미
Original Assignee
제주대학교 산학협력단
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 제주대학교 산학협력단 filed Critical 제주대학교 산학협력단
Priority to KR1020150166316A priority Critical patent/KR101749427B1/en
Publication of KR20170061377A publication Critical patent/KR20170061377A/en
Application granted granted Critical
Publication of KR101749427B1 publication Critical patent/KR101749427B1/en

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01W—METEOROLOGY
    • G01W1/00—Meteorology
    • G01W1/10—Devices for predicting weather conditions
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
    • F03D—WIND MOTORS
    • F03D80/00—Details, components or accessories not provided for in groups F03D1/00 - F03D17/00
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01P—MEASURING LINEAR OR ANGULAR SPEED, ACCELERATION, DECELERATION, OR SHOCK; INDICATING PRESENCE, ABSENCE, OR DIRECTION, OF MOVEMENT
    • G01P5/00—Measuring speed of fluids, e.g. of air stream; Measuring speed of bodies relative to fluids, e.g. of ship, of aircraft
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
    • G05B13/027—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion using neural networks only

Landscapes

  • Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Environmental & Geological Engineering (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Mechanical Engineering (AREA)
  • Automation & Control Theory (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Sustainable Development (AREA)
  • Sustainable Energy (AREA)
  • Chemical & Material Sciences (AREA)
  • Combustion & Propulsion (AREA)
  • Medical Informatics (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Atmospheric Sciences (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Ecology (AREA)
  • Environmental Sciences (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Air Conditioning Control Device (AREA)

Abstract

A method for predicting wind speed based on an artificial neural network having various characteristics and an apparatus using the method are disclosed. A method according to one aspect of the present invention includes receiving wind speed data for a target time point for predicting the wind speed and a predetermined time period before the target time point and using a plurality of artificial neural networks (ANN) Generating initial prediction data on the wind speed of the target point and generating final prediction data using at least a part of the initial prediction data.

Description

TECHNICAL FIELD [0001] The present invention relates to an artificial neural network based on artificial neural networks and a method for predicting wind speed based on artificial neural networks,

The following embodiments relate to a wind speed prediction technique, and more particularly to an artificial neural network-based wind speed prediction technique having various characteristics.

Renewable energy is attracting attention in recent power systems that pursue energy efficiency and environment. Renewable energy can reduce dependence on fossil fuels or power plants powered by nuclear fuels. Many regions or countries may have potential due to sources of renewable energy such as wind, sunlight, and waves.

However, these energies are difficult to integrate into the main grid due to their intermittent nature. For example, if the wind is heavy, renewable energy can be overproduced and wasted. In addition, it may be impossible to produce renewable energy in a situation where power demand is high. This problem is due to the fact that the power can not be economically stored for later use. For efficient operation, the main grid must consider how much energy it will produce at the plant, taking into account the power available from renewable energy.

For this purpose, it is required to accurately predict the availability of renewable energy. For example, in the case of wind energy, it is required to predict the wind speed. Prediction of wind speed is an old task for many researchers and engineers in various countries. However, there is no universal way to predict wind speeds, and each region predicts wind speed according to their model.

Furthermore, such predictions of wind speed can be combined with technologies such as ANNs (Artificial Neural Networks), fuzzy logic, and ARIMA (Auto Regressive Integral Moving Averate). ANNs can be used to predict the nonlinear motion of the target object. Thus, the dependency of input variables and output variables can be found in the learning stage through ANNs, and learning data for learning can be obtained through observation and measurement.

The following embodiments are intended to provide a technique capable of accurately predicting wind speed based on various artificial neural networks.

The method of predicting wind speed according to one side includes receiving a wind speed data for a target time point for predicting the wind speed and a predetermined time period before the target time point; Generating initial prediction data on the wind speed of the target point of time using a plurality of artificial neural networks (ANN) having different characteristics; And generating final prediction data using at least a part of the initial prediction data.

The plurality of ANNs are preliminarily learned with learning data of a period classified by different criteria, and the different criteria may include a season and a month.

The plurality of ANNs may include a plurality of input nodes corresponding to the wind speed data and an output node for outputting the initial predicted data.

The initial predicted data may include initial predicted data on seasonal variation characteristics of wind speed and initial predicted data on monthly variation characteristics of wind speed. ANN1 for generating first initial predictive data, ANN2 for generating second initial predictive data based on seasonal characteristics of the whole period, and ANN3 for generating third initial predictive data based on monthly characteristics of the whole period .

The generating of the initial prediction data may include: selecting subnetworks corresponding to the target time point in the ANN2 and the ANN3; And generating the initial prediction data using the subnetworks.

The ANN2 may include four subnetworks corresponding to spring, summer, autumn and winter, and the ANN3 may include 12 subnetworks corresponding to January to December.

delete

The plurality of ANNs includes ANN1 learned with learning data of the entire period, ANN2 learned with learning data classified by season as a whole, and ANN3 learned with learning data obtained by classifying the whole period as a monthly can do.

An apparatus for predicting the wind speed according to one side includes a plurality of ANNs for receiving wind speed data for a target time point for predicting the wind speed and a predetermined time period before the target time point and generating initial predicted data on the wind speed of the target time point (Artificial Neural Network); And a selection processing unit for generating final prediction data using at least a part of the initial prediction data, wherein each of the plurality of ANNs has different characteristics.

The plurality of ANNs may be previously learned with learning data of a period classified by different criteria, and the different criteria may include a season and a month.

The plurality of ANNs may include a plurality of input nodes corresponding to the wind speed data and an output node for outputting the initial predicted data.

The initial predicted data may include initial predicted data on seasonal variation characteristics of wind speed and initial predicted data on monthly variation characteristics of wind speed.

Wherein the plurality of ANNs comprises ANN1 for generating first initial predictive data based on characteristics of the entire period, ANN2 for generating second initial predictive data based on seasonal characteristics of the entire period, Lt; RTI ID = 0.0 > ANN3 < / RTI >

The method may further comprise a switch for selecting the subnetworks corresponding to the target time point in the ANN2 and the ANN3 and transmitting the wind speed data to the subnetworks, Data can be generated.

The ANN2 may include four subnetworks corresponding to spring, summer, autumn and winter, and the ANN3 may include 12 subnetworks corresponding to January to December.

delete

The plurality of ANNs includes ANN1 learned with learning data of the entire period, ANN2 learned with learning data classified by season as a whole, and ANN3 learned with learning data obtained by classifying the whole period as a monthly can do.

According to the embodiments described below, it is possible to accurately predict the wind speed using artificial neural networks of various standards.

Also, according to the embodiments described below, the utilization of the regenerated energy can be improved by using the accurately predicted wind speed.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing an apparatus for predicting wind speed according to an embodiment; FIG.
2 is a block diagram showing a detailed structure of an ANN according to an embodiment;
3 is a block diagram illustrating a data processing procedure of an ANN according to an exemplary embodiment;
4 is a block diagram for explaining a learning process of an ANN according to an embodiment;
5 is a block diagram illustrating a process of generating learning data according to an embodiment;
6 is an operational flowchart illustrating a wind speed prediction method according to an embodiment.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. Like reference symbols in the drawings denote like elements.

1 is a block diagram showing a wind speed predicting apparatus according to an embodiment. Referring to FIG. 1, an apparatus for predicting wind speed according to an embodiment may include an ANN (Artificial Neural Network) 100 and a selection processing unit 200. The wind speed predicting device can generate the predicted data based on the wind speed data and the target point of view. The wind speed predicting device can predict the wind speed at a specific point on the basis of the change of the wind speed during the previous predetermined period. The wind speed predicting device can learn the ANN 100 to predict the wind speed at the next point of time based on the wind speed at the previous point of time. The wind speed predicting device can generate the final predicted data based on the calculation on the initial predicted data after generating the initial predicted data first. The initial predictive data may include initial predictive data on seasonal variation characteristics of wind speed and initial predictive data on monthly variation characteristics of wind speed.

The wind speed prediction device can receive the wind speed data from the weather data server. The weather data server can provide wind speed records along with meteorological data such as sunlight and precipitation over time. The wind speed predicting device can extract necessary data from the data received from the weather data server. Thereafter, the extracted data may be converted into a series of Structured Query Language (SQL) to be inserted into a database table. The table can be defined as a MySQL database running on a Linux machine. Data analysis modules can find the information they need through the MySQL communications section of their platform. Furthermore, Open DataBase Connectivity (ODBC) allows a generic application to find a specific database field and process it according to its workspace analysis strategy. Embodiments may include a Fast ANN (FANN) library that provides sufficient ANN APIs for C language applications.

The ANN 100 receives the target time and wind speed data. The target time point means a specific time point to which the wind speed is predicted. For example, if you want to know the wind speed of tomorrow, the target time can be tomorrow. As will be described in greater detail below, the ANN 100 may include subnetworks having different characteristics, generated on various criteria, and the target time point may be used to select a particular network among the subnetworks. The selected specific network can generate initial predicted data based on wind speed data.

The wind speed data may include wind speed data for a predetermined time interval before the target point. For example, the wind speed data may include wind speed data for five days before the target point. The time interval of the wind speed data included in the wind speed data may correspond to the input node of the ANN 100. [ For example, if the ANN 100 has five input nodes, the wind speed data may include wind speed data for a time interval that fits the five input nodes. For example, if the ANN 100 has five input nodes, the wind speed data may include the average wind speed for one day of each day for five days prior to the target time point. The ANN 100 can predict the wind speed data at the target time point based on the wind speed data for the previous five days. Hereinafter, the wind speed prediction process will be described in units of days, but the embodiments described below can be applied to units of time and smaller units.

The ANN 100 can generate initial prediction data for the target time point based on the wind speed data. The ANN 100 may include a plurality of input nodes and an output node. The wind speed data can be input to the input node of the ANN 100 and the initial predicted data can be output from the output node of the ANN 100. [ As will be described in detail later, the ANN 100 may include a plurality of sub-networks having different characteristics learned according to various criteria. Accordingly, the initial prediction data may include a plurality of sub data by a plurality of sub-networks.

The selection processing unit (200) generates final prediction data based on the initial prediction data. The selection processing unit 200 can select some data among a plurality of sub data included in the initial prediction data. The selection processing unit 200 can generate final prediction data using the selected data. For example, the selection processing unit 200 may generate the final prediction data based on the average of the remaining data after removing the data with the largest difference among the plurality of sub data included in the initial prediction data. Specifically, the initial prediction data may include first data, second data, and third data. The selection processing unit 200 may remove one of the first data, the second data, and the third data, which is the largest difference from the other two. The selection processing unit 200 can determine the average of the remaining two data as the final prediction data. In other words, the selection processing unit 200 can generate the final prediction data based on the average of the remaining data excluding the data having the largest difference from the remaining data among the initial prediction data.

2 is a block diagram showing a detailed structure of an ANN according to an embodiment. Referring to FIG. 2, an ANN 100 according to an embodiment includes ANN1 110, ANN2 120, and ANN3 130. ANN1 110, ANN2 120 and ANN3 130 are independent of each other. Also, the first data, the second data, and the third data output by ANN1 110, ANN2 120, and ANN3 130 are independent of each other. ANN1 110, ANN2 120 and ANN3 130 generate initial prediction data based on wind speed data, respectively. ANN1 110 generates the first data based on the wind speed data. ANN2 120 generates second data based on the wind speed data. ANN3 130 generates third data based on the wind speed data.

ANN1 110, ANN2 120 and ANN3 130 may be networks learning different learning data. For example, the ANN1 110 can learn the learning data as a whole, the ANN2 120 can learn the data classified according to the first criterion, and the ANN3 130 can learn the learning data as the second criterion Can be learned. For example, the ANN1 110 may learn the entire wind speed data for the past year, the ANN2 120 may learn the wind speed data for the past year by season, and the ANN3 (130) The wind speed data during the day can be learned monthly. Accordingly, the ANN1 110, the ANN2 120, and the ANN3 130 can predict the wind speed at the target time point according to different standards, respectively. ANN1 generates first initial prediction data based on the characteristics of the entire period, ANN2 generates second initial prediction data based on the seasonal characteristics of the entire period, and ANN3 is based on the monthly characteristic of the entire period It is possible to generate the third initial prediction data. That is, diversity related to wind speed prediction can be ensured through a plurality of sub-networks having different characteristics learned by different criteria.

ANN2 120 may include subnetworks corresponding to a plurality of seasons. For example, ANN2 120 may include four subnetworks corresponding to spring, summer, autumn, and winter. The subnetworks can learn the wind speed data corresponding to spring, summer, autumn and winter in advance. ANN3 130 may include subnetworks corresponding to a plurality of months. For example, ANN2 120 may include 12 subnetworks corresponding to January through December. The subnetworks can learn the wind speed data corresponding to January to December respectively. The following examples illustrate learned networks based on season and month, but networks can be learned according to various additional criteria that can distinguish characteristics of wind speed. The subnetworks of ANN3 130 and ANN3 130 are independent of each other.

3 is a block diagram for explaining a data processing procedure of the ANN according to an embodiment. Referring to FIG. 3, the first switch 310 receives the target time and wind speed data. The first switch 310 transmits the wind speed data to the subnetwork corresponding to the target time point among the ANN2 120. The first switch 310 can select a subnetwork corresponding to the target time point based on the target time point. Similarly, the second switch 320 receives the target point of view and the wind speed data, and transmits the wind speed data to the subnetwork corresponding to the target point in the ANN3 130.

The subnetworks ANN2 1 through ANN2 4 of ANN2 120 may correspond to seasons, respectively. For example, the subnetwork ANN2 1 corresponds to the spring, the subnetwork ANN2 2 corresponds to the summer, the subnetwork ANN2 3 corresponds to the fall, and the subnetwork ANN2 4 corresponds to the winter . Each season can be defined by a certain standard. For example, in Korea, March to May may be defined as spring, June to August as summer, September to November as autumn, and December to February as winter. As will be described later in detail, the subnetworks ANN2 1 to ANN2 4 of ANN2 120 can learn learning data corresponding to each season in advance.

The subnetworks ANN3 1 through ANN3 12 of ANN3 130 may correspond to months, respectively. For example, the subnetwork ANN3 1 corresponds to January, the subnetwork ANN3 2 corresponds to February, the subnetwork ANN3 3 corresponds to March, and the subnetwork ANN3 12 corresponds to December, . As will be described in detail later, the subnetworks ANN3 1 to ANN3 12 of ANN3 130 may learn learning data corresponding to each month in advance.

As described above, the first switch 310 transmits the wind speed data to the subnetwork corresponding to the target time point of the ANN2 120. [ The sub-network receiving the wind speed data from the first switch 310 generates the second data. For example, when the target time is May 5, the first switch 310 may transmit the wind speed data to the subnetwork ANN2 1 corresponding to the spring. The subnetwork (ANN2 1 ) can generate the second data based on the wind speed data.

Also, the second switch 320 transmits the wind speed data to the subnetwork corresponding to the target time point among the ANN3 130. The sub-network receiving the wind speed data from the second switch 320 generates the third data. For example, when the target time is May 5, the second switch 320 may transmit the wind speed data to the subnetwork (ANN3 5 ) corresponding to May . The subnetwork (ANN3 5 ) can generate the third data based on the wind speed data.

FIG. 4 is a block diagram for explaining a learning process of the ANN according to an embodiment. The learning device 400 according to an embodiment can learn the ANN 100 based on the learning data. According to one aspect, the training data may be wind speed data of the previous year. The learning data may include wind speed data for at least one year. For example, in the case of predicting the wind speed in 2015, the learning data may be data on the wind speed in 2014, or data on wind speed from January 1, 2013 to December 31, 2014. Alternatively, the wind speed data may include data for the previous 15 years.

As described above, the ANN 100 may include a plurality of independent subnetworks. The learning device 400 may learn sub-networks according to various criteria. Thus, the subnetworks have different characteristics. For example, the learning device 400 can learn a specific sub-network seasonally and can learn another sub-network according to a month. The learning device 400 may learn the ANN 100 according to various machine learning techniques. Learning ANN 100 can be understood as learning ANN 100 parameters.

The ANN 100 may include a plurality of input nodes and an output node. In addition, the ANN 100 may include a plurality of layers between input nodes and output nodes. The number of hidden nodes in the plurality of layers may be selected by trial and error. The learning device 400 may adjust the parameters between the input nodes, the plurality of layers, and the output nodes based on the training data, so as to learn the ANN 100. [ The plurality of input nodes may correspond to the time points before the target time point. For example, the number of input nodes may be five, and each input node may correspond to a period of five days before the target time point. That is, the ANN 100 can learn the wind speed at the target time point based on the wind speed data for the five days prior to the target time point. More specifically, the ANN 100 may receive five wind speed data through the input node from April 30 to May 4 to learn the wind speed of May 5. In this case, the parameter of the ANN 100 can be adjusted by comparing the predicted data output through the output node with the wind speed of May 5th.

The learning device 400 can repeat the above-mentioned ANN1 by repeating this process from January 1 to December 31. Also, the learning device 400 may repeat the above-mentioned ANN2 1 by repeating this process from January 1 to February 28. Further, the learning machine (400) is capable of learning the aforementioned ANN3 1 by performing the above process repeatedly from 1 January 31 January. The remaining subnetworks may be learned in a similar manner. Because the change of wind speed is seasonal and monthly specific, subnetworks based on various criteria can improve the accuracy of wind speed prediction.

5 is a block diagram illustrating a process of generating learning data according to an embodiment of the present invention. Referring to FIG. 5, the learning data includes first learning data 510, second learning data 520, and third learning data 530. The second learning data 520 and the third learning data 530 may be generated based on the first learning data 510. [ The first learning data 510 includes learning data for learning AAN1 described above and the second learning data 520 includes learning data for learning AAN2 described above and the third learning data 530 ) Contains learning data for learning AAN3 described above.

The first learning data DB1 may include total wind speed data for at least one year before the target time point for predicting the wind speed. For example, when the target time is 2015, the first learning data DB1 may be data on the wind speed of the whole of 2014, or data on the wind speed of the whole two years from January 1, 2013 to December 31, have.

The second learning data DB2 and the third learning data DB3 can be generated by classifying the first learning data DB1 on a constant basis. The second learning data DB2 includes the sub learning data DB2 1 to DB2 4 . The second learning data (DB2) can be generated by classifying the first learning data (DB1) by season. For example, wind speed data from January 1 to February 28 corresponding to spring in the first learning data DB 1 can be classified into sub learning data DB2 1 . The third learning data DB3 includes sub learning data DB3 1 to DB3 12 . The third learning data DB3 can be generated by classifying the first learning data DB1 monthly. For example, wind speed data from January 1 to January 31 corresponding to January in the first learning data DB1 can be classified into sub learning data DB3 1 .

The above-described learning device 400 uses the first learning data DB1, the second learning data DB2 and the third learning data DB3 to generate the above-mentioned ANN1 110, ANN2 120 and ANN3 130 Respectively. Also, the learning device 400 can learn the subnetworks ANN2 1 to ANN2 4 mentioned above by using the sub learning data DB2 1 to DB2 4 , respectively, and the sub learning data DB3 1 to DB3 12 Can learn each of the above-mentioned subnetworks (ANN3 1 to ANN3 12 ).

The embodiments described above may be implemented in hardware components, software components, and / or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented within a computer system, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, such as an array, a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For ease of understanding, the processing apparatus may be described as being used singly, but those skilled in the art will recognize that the processing apparatus may have a plurality of processing elements and / As shown in FIG. For example, the processing unit may comprise a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as a parallel processor.

The software may include a computer program, code, instructions, or a combination of one or more of the foregoing, and may be configured to configure the processing device to operate as desired or to process it collectively or collectively Device can be commanded. The software and / or data may be in the form of any type of machine, component, physical device, virtual equipment, computer storage media, or device , Or may be permanently or temporarily embodied in a transmitted signal wave. The software may be distributed over a networked computer system and stored or executed in a distributed manner. The software and data may be stored on one or more computer readable recording media.

The method according to an embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions to be recorded on the medium may be those specially designed and configured for the embodiments or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Although the embodiments have been described with reference to the drawings, various technical modifications and variations may be applied to those skilled in the art. For example, it is to be understood that the techniques described may be performed in a different order than the described methods, and / or that components of the described systems, structures, devices, circuits, Lt; / RTI > or equivalents, even if it is replaced or replaced.

Claims (19)

Receiving wind speed data for a target time point for predicting the wind speed and a predetermined time period before the target time point;
Generating initial prediction data on the wind speed of the target point of time using a plurality of artificial neural networks (ANN) having different characteristics; And
Generating final prediction data using at least a part of the initial prediction data;
And estimating the wind speed. The method according to claim 1,
Wherein the plurality of ANNs comprises:
Learning is preliminarily learned with learning data of periods classified on different criteria,
The different criteria may include season and month,
Wind speed prediction method. The method according to claim 1,
Wherein the plurality of ANNs comprises:
A plurality of input nodes corresponding to the wind speed data and an output node for outputting the initial predicted data;
Wind speed prediction method. The method according to claim 1,
Wherein the initial prediction data includes:
Including initial predictive data on seasonal variation characteristics of wind speed and initial predictive data on monthly variation characteristics of wind speed,
Wind speed prediction method. The method according to claim 1,
Wherein the plurality of ANNs comprises:
ANN1 for generating second initial predictive data based on the seasonal characteristic of the whole period, and ANN2 for generating second initial predictive data based on the monthly characteristic of the whole period, RTI ID = 0.0 > ANN3 < / RTI >
Wind speed prediction method. 6. The method of claim 5,
Wherein the generating the initial prediction data comprises:
Selecting subnetworks corresponding to the target time point in the ANN2 and the ANN3; And
Generating the initial prediction data using the subnetworks
And estimating the wind speed. 6. The method of claim 5,
Wherein the ANN2 comprises four subnetworks corresponding to spring, summer, autumn and winter, and the ANN3 comprises twelve subnetworks corresponding to January to December,
Wind speed prediction method. delete The method according to claim 1,
Wherein the plurality of ANNs comprises:
An ANN1 that is learned by learning data of the entire period, ANN2 that is learned by learning data that classifies the whole period by season, and ANN3 that is learned by learning data that classifies the whole period by a monthly,
Wind speed prediction method. 9. A computer program stored on a medium for executing the method of any one of claims 1 to 7 and 9 in combination with hardware. A plurality of ANNs (Artificial Neural Networks) for receiving wind speed data for a target time point for predicting the wind speed and a predetermined time interval before the target time point and generating initial predicted data on the wind speed of the target time point; And
A selection processing unit for generating final prediction data using at least a part of the initial prediction data;
Lt; / RTI >
Each of the plurality of ANNs having different characteristics,
Wind speed prediction device. 12. The method of claim 11,
Wherein the plurality of ANNs comprises:
Learning is preliminarily learned with learning data of periods classified on different criteria,
The different criteria may include season and month,
Wind speed prediction device. 12. The method of claim 11,
Wherein the plurality of ANNs comprises:
A plurality of input nodes corresponding to the wind speed data and an output node for outputting the initial predicted data;
Wind speed prediction device. 12. The method of claim 11,
Wherein the initial prediction data includes:
Including initial predictive data on seasonal variation characteristics of wind speed and initial predictive data on monthly variation characteristics of wind speed,
Wind speed prediction device. 12. The method of claim 11,
Wherein the plurality of ANNs comprises:
ANN1 for generating second initial predictive data based on the seasonal characteristic of the whole period, and ANN2 for generating second initial predictive data based on the monthly characteristic of the whole period, RTI ID = 0.0 > ANN3 < / RTI >
Wind speed prediction device. 16. The method of claim 15,
Further comprising a switch for selecting the subnetworks corresponding to the target time point in the ANN2 and the ANN3 and transmitting the wind speed data to the subnetworks,
Wherein the subnetworks generate the initial prediction data based on the wind speed data,
Wind speed prediction device. 16. The method of claim 15,
Wherein the ANN2 comprises four subnetworks corresponding to spring, summer, autumn and winter, and the ANN3 comprises twelve subnetworks corresponding to January to December,
Wind speed prediction device. delete 12. The method of claim 11,
Wherein the plurality of ANNs comprises:
An ANN1 that is learned by learning data of the entire period, ANN2 that is learned by learning data that classifies the whole period by season, and ANN3 that is learned by learning data that classifies the whole period by a monthly,
Wind speed prediction device.
KR1020150166316A 2015-11-26 2015-11-26 Method for forecasting wind speed based on artificial neural networks having different features KR101749427B1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1020150166316A KR101749427B1 (en) 2015-11-26 2015-11-26 Method for forecasting wind speed based on artificial neural networks having different features

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1020150166316A KR101749427B1 (en) 2015-11-26 2015-11-26 Method for forecasting wind speed based on artificial neural networks having different features

Publications (2)

Publication Number Publication Date
KR20170061377A KR20170061377A (en) 2017-06-05
KR101749427B1 true KR101749427B1 (en) 2017-07-03

Family

ID=59222935

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1020150166316A KR101749427B1 (en) 2015-11-26 2015-11-26 Method for forecasting wind speed based on artificial neural networks having different features

Country Status (1)

Country Link
KR (1) KR101749427B1 (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101956717B1 (en) * 2017-09-01 2019-03-11 군산대학교산학협력단 Method and apparatus for prediction of wind direction, and method for yaw control of wind turbines using the same
KR101956715B1 (en) * 2017-09-01 2019-03-11 군산대학교산학협력단 Wind direction prediction method and apparatus for yaw control of wind turbines
KR20210154581A (en) 2020-06-12 2021-12-21 부산대학교 산학협력단 Method for predicting the formulation of structural epoxy adhesives using machine learning
KR20230083860A (en) 2021-12-03 2023-06-12 전기은 Generator control system and method based on climate prediction through artificial intelligence

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102481885B1 (en) * 2017-09-08 2022-12-28 삼성전자주식회사 Method and device for learning neural network for recognizing class
CN108022025B (en) * 2017-12-28 2020-08-18 华中科技大学 Wind speed interval prediction method and system based on artificial neural network
KR102396290B1 (en) * 2019-12-23 2022-05-11 주식회사 포디솔루션 Method for providing ultra low altitude wind prediction information
CN111427101B (en) * 2020-04-07 2022-04-26 南京气象科技创新研究院 Thunderstorm strong wind grading early warning method, system and storage medium
CN112990354B (en) * 2021-04-15 2021-08-13 中国气象局公共气象服务中心(国家预警信息发布中心) Method and device for constructing deep convolution regression network for wind speed prediction

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
이동욱, '하둡 기반 제주 풍속 변화의 분석 플렛폼 구축', 제주대학교 대학원, 2015.02., 석사학위논문.

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101956717B1 (en) * 2017-09-01 2019-03-11 군산대학교산학협력단 Method and apparatus for prediction of wind direction, and method for yaw control of wind turbines using the same
KR101956715B1 (en) * 2017-09-01 2019-03-11 군산대학교산학협력단 Wind direction prediction method and apparatus for yaw control of wind turbines
KR20210154581A (en) 2020-06-12 2021-12-21 부산대학교 산학협력단 Method for predicting the formulation of structural epoxy adhesives using machine learning
KR20230083860A (en) 2021-12-03 2023-06-12 전기은 Generator control system and method based on climate prediction through artificial intelligence

Also Published As

Publication number Publication date
KR20170061377A (en) 2017-06-05

Similar Documents

Publication Publication Date Title
KR101749427B1 (en) Method for forecasting wind speed based on artificial neural networks having different features
Ibrahim et al. A novel hybrid model for hourly global solar radiation prediction using random forests technique and firefly algorithm
Capizzi et al. Advanced and adaptive dispatch for smart grids by means of predictive models
Carneiro et al. Review on photovoltaic power and solar resource forecasting: current status and trends
JP6980521B2 (en) Data metascaling device and method for continuous learning
CN105117593A (en) Wavelet transform and particle swarm optimized grey model-based short-term wind speed forecasting method
KR20220040968A (en) Short-term load forecasting system and method using temporal convolutional neural network
CN109412161A (en) A kind of Probabilistic Load calculation method and system
CN112925824A (en) Photovoltaic power prediction method and system for extreme weather type
CN115310536A (en) Reservoir water level prediction early warning method based on neural network and GCN deep learning model
CN116843083A (en) Carbon emission prediction system and method based on hybrid neural network model
Qadrdan et al. Probabilistic wind power forecasting using a single forecast
Bandong et al. Performance ratio estimation and prediction of solar power plants using machine learning to improve energy reliability
Shabani et al. Predicting building energy consumption using engineering and data driven approaches: A review
Li et al. Temporal attention based tcn-bigru model for energy time series forecasting
Dwivedi et al. Climate change: Prediction of solar radiation using advanced machine learning techniques
El Shahat et al. An artificial neural network model for wind energy estimation
Villa et al. A stochastic model of future extreme temperature events for infrastructure analysis
CN113240219A (en) Land utilization simulation and prediction method
CN112668650A (en) Industrial data model generation method, system, device and medium
Osório et al. Hybrid evolutionary-adaptive approach to predict electricity prices and wind power in the short-term
Dokuz et al. Year-ahead wind speed forecasting using a clustering-statistical hybrid method
CN113962357A (en) GWO-WNN-based distributed photovoltaic power data virtual acquisition method
Huang et al. Probabilistic prediction intervals of wind speed based on explainable neural network
Sundaram et al. Solution to profit based unit commitment using swarm intelligence technique

Legal Events

Date Code Title Description
E701 Decision to grant or registration of patent right