KR20210132503A - National power demand forecasting using cyclic neural network model - Google Patents

National power demand forecasting using cyclic neural network model Download PDF

Info

Publication number
KR20210132503A
KR20210132503A KR1020200050984A KR20200050984A KR20210132503A KR 20210132503 A KR20210132503 A KR 20210132503A KR 1020200050984 A KR1020200050984 A KR 1020200050984A KR 20200050984 A KR20200050984 A KR 20200050984A KR 20210132503 A KR20210132503 A KR 20210132503A
Authority
KR
South Korea
Prior art keywords
data
power demand
neural network
reinforcement learning
network model
Prior art date
Application number
KR1020200050984A
Other languages
Korean (ko)
Inventor
김영진
신범수
변광규
Original Assignee
주식회사 와이즈테크놀로지
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 주식회사 와이즈테크놀로지 filed Critical 주식회사 와이즈테크놀로지
Priority to KR1020200050984A priority Critical patent/KR20210132503A/en
Publication of KR20210132503A publication Critical patent/KR20210132503A/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S40/00—Systems for electrical power generation, transmission, distribution or end-user application management characterised by the use of communication or information technologies, or communication or information technology specific aspects supporting them
    • Y04S40/20—Information technology specific aspects, e.g. CAD, simulation, modelling, system security

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • General Business, Economics & Management (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Primary Health Care (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The present invention relates to a power demand forecasting apparatus and method. For this purpose, the power demand forecasting device of the present invention collects weather data, power usage data, and date data for forecasting power demand, generates a training data set based on the collected data, and generates a power demand model based on the training data set to perform reinforcement learning, and forecasts power demand using a micro service.

Description

순환형 신경망 모델(CuDNNLSTM)을 이용한 전국 단위 전력 수요예측{National power demand forecasting using cyclic neural network model}National power demand forecasting using cyclic neural network model

본 발명은 익일의 24시간에 대한 전력 수요예측 장치 및 전력 수요예측 방법에 관한 것으로서, 시계열 예측에 성능이 우수한 RNN(Recurrent Neural Network) 모델의 변형인 LSTM(Long Short-Term Memory) 모델을 GPU 환경에서 최적의 성능을 발휘할 수 있도록 구현한 CuDNNLSTM 모델을 이용하여, 익일에 대해 예측된 시간별 기상정보를 참고하여 학습 데이터를 생성하고 강화 학습을 통해 전력 수요를 예측할 수 있는 장치 및 방법에 관한 것이다.The present invention relates to a power demand prediction device and a power demand prediction method for 24 hours of the next day, and a Long Short-Term Memory (LSTM) model, which is a variation of a Recurrent Neural Network (RNN) model, which has excellent performance in time series prediction, is applied to a GPU environment. It relates to an apparatus and method that can generate training data by referring to the hourly weather information predicted for the next day using the CuDNNLSTM model implemented so that it can exhibit optimal performance and predict power demand through reinforcement learning.

전력 수요예측은 안정적이고 효율적인 전력 수급계획을 수립하는데 중요한 요소이다. 전력 수요예측 결과는 전력 가격을 결정하거나, 전력 계통운영을 위해 사용되므로, 전력 수요예측의 오차는 안정적인 전력 계통운영을 방해하고 경제적 손실을 야기하는 원인이 될 수 있다. 따라서, 전력 수요예측의 오차를 줄이기 위하여, 시계열 분석법, 회귀 분석법, 인공신경망, 지식기반 전문가 시스템 등 다양한 전력 수요예측 기법들이 제시되어 왔다.Electricity demand forecasting is an important factor in establishing a stable and efficient electricity supply and demand plan. Since the power demand forecast results are used to determine the power price or operate the power system, errors in the power demand forecast may interfere with stable power system operation and cause economic loss. Accordingly, in order to reduce the error in the power demand forecasting, various power demand forecasting techniques such as time series analysis, regression analysis, artificial neural networks, and knowledge-based expert systems have been proposed.

그러나, 종래의 전력 수요예측 기법들은 일자별 기온 정보만을 사용하여 예측하는 것이 대부분으로, 단순한 유사도 측정 값에 의존하여 목표일의 전력 수요패턴을 예측한 바 전력 수요실적 값과 전력 수요예측 결과의 오차가 크게 발생하는 등의 문제점이 있었다.However, most of the conventional power demand forecasting techniques predict using only the daily temperature information. As a result of predicting the power demand pattern of the target day based on a simple similarity measurement value, the error between the power demand performance value and the power demand forecast result is There were problems, such as a large occurrence.

또한, 다양한 수요예측 방법을 크게 시계열 분석을 통한 예측과 인공신경망에 기반한 예측 방법 등으로 구분 가능한데, 이들 방법 중 시계열 분석 기법은 주기별로 반복되는 전력 패턴의 특징을 반영할 수 있으나 과거 전력 사용내역 외에 영향을 끼치는 외부 요인들데 대한 분석 및 적용이 상대적으로 어려운 단점이 존재한다.In addition, various demand forecasting methods can be broadly divided into prediction through time series analysis and prediction methods based on artificial neural networks. There are disadvantages in that it is relatively difficult to analyze and apply external factors that influence them.

또한, 인공신경망 기반의 수요예측 방법은 주로 통계적 모형에 기반을 두기보다는 학습 모형을 통하여 알고리즘을 형성하는데, 중간에서 처리하는 노드의 개수가 얼마나 복잡하고 정교한지에 따라 그 예측 능력이 좌우되기 때문에 정확한 전력 수요예측을 위해서 노드의 개수가 매우 복잡하며 진화된 연산처리 기법이 필요하다는 문제점이 있다.In addition, the artificial neural network-based demand prediction method mainly forms an algorithm through a learning model rather than based on a statistical model. There is a problem in that the number of nodes is very complicated for demand prediction and an advanced computational processing technique is required.

본 발명이 이루고자 하는 기술적 과제는 데이터 전처리, 딥러닝 기법을 활용한 모델을 이용하여 전력 사용량을 정확하게 예측할 수 있는 전력 수요예측 장치 및 방법을 제공하는데 있다.An object of the present invention is to provide an apparatus and method for predicting power demand that can accurately predict power usage by using a model using data preprocessing and deep learning techniques.

본 발명의 실시예에 따르면, 수요예측 활용 데이터 수집 및 전처리 단계; 딥러닝 활용 수요예측 모델 개발 단계; 하이퍼 파라미터 탐색을 통한 최적 모델 도출 단계; 수요예측 모델 강화 학습 단계; 마이크로 서비스를 활용한 수요예측 단계; 를 포함한다.According to an embodiment of the present invention, the demand forecast utilization data collection and pre-processing steps; Development of a demand forecasting model using deep learning; deriving an optimal model through hyperparameter search; Demand forecasting model reinforcement learning stage; Demand forecasting stage using microservices; includes

본 발명의 실시예에 따른 전력 수요예측 장치 및 방법에 따르면 동적으로 변화하는 지역 단위 전력 수요를 정확히 예측하여, 안정적인 전력 수급이 가능하고 계획적인 전력 계통운영을 통하여 에너지 비용 절감이 가능한 효과가 있다.According to the apparatus and method for predicting power demand according to an embodiment of the present invention, it is possible to accurately predict dynamically changing regional power demand, thereby enabling stable power supply and demand and reducing energy costs through planned power system operation.

도 1은 본 발명의 실시예에 따른 전력 수요예측 방법에 대한 흐름도이다.
도 2는 본 발명의 실시예에 따른 데이터 중간 누락을 보여주는 도면이다.
도 3은 본 발명의 실시예에 따른 데이터 마지막 부분 누락을 보여주는 도면이다.
도 4는 본 발명의 실시예에 따른 기상 데이터 보정을 보여주는 도면이다.
도 5는 본 발명의 실시예에 따른 전력 수요예측 방법에 대한 개념도이다.
도 6은 본 발명의 실시예에 따른 수요예측 강화학습에 대한 개념도이다.
도 7은 본 발명의 실시예에 따른 수요예측 강화학습의 시간 구성에 대한 개념도이다.
도 8은 본 발명의 실시예에 따른 수요예측 모델의 강화학습에 대한 개념도이다.1 is a flowchart of a method for predicting power demand according to an embodiment of the present invention.
2 is a diagram illustrating data omission according to an embodiment of the present invention.
3 is a diagram illustrating an omission of a last part of data according to an embodiment of the present invention.
4 is a view showing meteorological data correction according to an embodiment of the present invention.
5 is a conceptual diagram of a method for predicting power demand according to an embodiment of the present invention.
6 is a conceptual diagram of demand prediction reinforcement learning according to an embodiment of the present invention.
7 is a conceptual diagram of a time configuration of demand prediction reinforcement learning according to an embodiment of the present invention.
8 is a conceptual diagram of reinforcement learning of a demand prediction model according to an embodiment of the present invention.

본 발명은 첨부된 도면을 참조하여 상세히 설명하면 다음과 같다. 본 발명을 설명하는데 참조하는 도면에 도시된 구성요소의 크기, 선의 두께 등은 이해의 편의상 다소 과장되게 표현되어 있을 수 있다. 또, 본 발명의 설명에 사용되는 용어들은 본 발명에서의 기능을 고려하여 당 업계에서 평균적인 지식을 가진 자에게 본 발명을 보다 완전하게 설명하기 위해서 제공되는 것으로, 본 발명의 요지를 불필요하게 흐릴 수 있는 일반적인 기능 및 구성에 대한 상세한 설명은 생략될 수 있다.The present invention will be described in detail with reference to the accompanying drawings as follows. Sizes of components, thicknesses of lines, etc. shown in the drawings referenced to explain the present invention may be expressed somewhat exaggeratedly for convenience of understanding. In addition, the terms used in the description of the present invention are provided to more completely explain the present invention to those of ordinary skill in the art in consideration of the functions in the present invention, and may not unnecessarily obscure the gist of the present invention. Detailed descriptions of possible general functions and configurations may be omitted.

도 1은 본 발명의 실시예에 따른 지역 단위 전력 수요예측 방법에 대한 흐름도이다.1 is a flowchart of a method for predicting power demand for each region according to an embodiment of the present invention.

S110 단계는 전력 수요예측을 위한 데이터를 수집하는 단계로서, 수집 하는 데이터는 발전단 부하, 통계지표, 경제통계지표, 시간별 기상관측, 일별 기상관측, AWS 기상관측, 동네예보, 중기예보, 주간예보 등이 될 수 있다.Step S110 is a step of collecting data for power demand forecasting, and the data collected are power generation stage load, statistical indicators, economic statistical indicators, hourly weather observation, daily weather observation, AWS weather observation, neighborhood forecast, medium-term forecast, weekly forecast etc. can be

또한, S110 단계는 이렇게 수집한 데이터들을 데이터 저장이 가능한 저장소에 저장하는 단계를 포함할 수 있다.In addition, step S110 may include a step of storing the collected data in a storage capable of data storage.

S120 단계는 데이터 전처리 단계로서 수집된 데이터에서 누락된 부분을 보정하고, 모델 개발을 위한 시계열 입력 데이터에 적합하도록 전처리하고 변수화하는 단계이다. Step S120 is a data pre-processing step, which is a step of correcting missing parts in the collected data, pre-processing and variableizing it to fit the time series input data for model development.

도 2는 데이터의 중간 부분이 누락된 예로 12월 20일에 1차 누락과 12월 21일에 2차 누락이 발생하였다. 12월 19일 이전 데이터와 12월 22일 이후 데이터를 이용하여 인공신경망 알고리즘을 통해 12월 20일의 1차 누락을 보정하고, 12월 19일 이전 데이터와 12월 20일의 1차 누락 보정값, 12월 21일 이후 데이터를 이용하여 인공신경말 알고리즘을 통해 12월 21일의 2차 누락을 보정한다. 다시, 위의 1, 2차의 과정을 n번 반복 수행하여 데이터를 보정할 수 있다.2 shows an example in which the middle part of the data is omitted, with the primary omission on December 20 and the secondary omission on December 21. Using the data before December 19 and after December 22, the first omission of December 20 is corrected through an artificial neural network algorithm, and the data before December 19 and the correction of the first omission of December 20 are corrected. , the secondary omission of December 21 is corrected through artificial neural algorithm using data after December 21. Again, the data can be corrected by repeating the above 1st and 2nd steps n times.

도 3은 데이터의 마지막 부분이 누락된 예로 12월 25일에 1차 누락과 12월 26일에 2차 누락이 발생하였다. 12월 24일 이전 데이터를 이용하여 인공신경망 알고리즘을 통해 1차 누락을 보정하고, 12월 24일 이전 데이터와 12월 25일 1차 누락 보정값 데이터를 이용하여 인공신경망 알고리즘을 통해 2차 누락을 보정한다. 다시, 위의 1, 2차 과정을 n번 반복 수행하여 데이터를 보정할 수 있다.3 shows an example in which the last part of the data is omitted, the primary omission occurred on December 25th and the secondary omission occurred on December 26th. The primary omission is corrected through the artificial neural network algorithm using the data before December 24, and the secondary omission is corrected through the artificial neural network algorithm using the data before December 24 and the primary omission correction value data on December 25. Correct. Again, the data can be corrected by repeating the above 1st and 2nd processes n times.

도 4는 기상 예측 데이터를 보정하는 실시예이며, S123은 동네예보 보정 가중치, S124는 동네예보 데이터, S125는 동네예보 보정결과의 일부를 표로 나타낸 것이다. 실시예에서는 3시간 간격으로 수집된 동네예보 데이터를 동네예보 보정 가중치를 이용하여 1시간 간격을 동네예보 보정결과(S126, S127)를 도출해 냈으며, 1시간 간격 동네예보 보정결과는 아래의 수학식 1로 표현될 수 있다.4 is an embodiment of correcting weather forecast data, S123 is a neighborhood forecast correction weight, S124 is neighborhood forecast data, and S125 is a table showing a part of the neighborhood forecast correction result. In the embodiment, the neighborhood forecast correction results (S126, S127) for 1-hour intervals were derived using the neighborhood forecast correction weights for the neighborhood forecast data collected at 3-hour intervals, and the 1-hour interval neighborhood forecast correction results are expressed in Equation 1 below. can be expressed as

Figure pat00001
Figure pat00001

수학식 1에서 hT는 예측되야 하는 1시간 평균 기온이고, h₁T 는 참조해야 하는 첫번째 예측 기온, h₁w는 참조가 되야 하는 첫번째 가중치, h₂T는 참조해야 하는 두번째 예측 기온, h₂w는 참조가 되야 하는 두번째 가중치이다.In Equation 1, hT is the hourly average temperature to be predicted, h₁T is the first predicted temperature to be referenced, h₁w is the first weight to be referenced, h₂T is the second predicted temperature to be referenced, and h₂w is the second weight to be referenced am.

도 5는 수요예측에 대한 실시예로서, 예측 수행일(D) 12시까지의 데이터를 이용하여 예측 수행일(D) 13시부터 예측 대상일(D+1)의 1시부터 24시까지 36시간의 시간대별 수요예측을 수행하고, 36시간의 예측 수요 중 예측 대상일(D+1)의 12시간 데이터를 제거하여 예측 대상일(D+1)의 24시간 수요 데이터를 예측한다.5 is an embodiment of demand forecasting, 36 hours from 1:00 to 24:00 on the date of prediction (D+1) from 13:00 on the day of prediction (D) using data up to 12:00 on the day of prediction (D) , and predicts 24-hour demand data of the forecast target date (D+1) by removing 12-hour data of the forecast target date (D+1) from the 36-hour forecast demand.

도 6은 수요예측 모델의 강화학습 개념도로서, 어떤 환경(Environment)을 탐색하는 에이전트(Agent) 현재의 상태(State)를 인식하여 행동(Action)을 취하고 행동에 대한 보상(Reward)을 최대화하는 일련의 행동으로 정의 되는 정책을 찾는 기계학습의 방법을 적용하여 강화학습을 수행한다.6 is a conceptual diagram of reinforcement learning of a demand prediction model, an agent searching for a certain environment, recognizing the current state, taking an action, and maximizing a reward for the action. Reinforcement learning is performed by applying the method of machine learning to find a policy defined by the behavior of

도 7은 본 발명의 실시예에 따른 수요예측 모델의 강화학습 시에 적용되어지는 시간 구성을 나타내는 개념도로서, 강화학습 시 딥러닝 모델을 생성하는 기간, 강화학습 기간, 강화학습 검증기간, 강화학습 모델 적용기간으로 구분된다.7 is a conceptual diagram illustrating a time configuration applied during reinforcement learning of a demand prediction model according to an embodiment of the present invention. It is divided by the model application period.

도 8은 본 발명의 실시예에 따른 수요예측 모델의 강화학습 개념도로서, 대상 기간동안 학습을 계속적으로 수행하면서 반복하여 최적의 행동을 학습하여 예측정확도를 향상 시키게 된다. 8 is a conceptual diagram of reinforcement learning of a demand prediction model according to an embodiment of the present invention, in which the optimal behavior is learned repeatedly while continuously performing learning during the target period to improve prediction accuracy.

본 발명의 실시예에 따른 도면을 참조하여 설명하였지만, 본 발명이 속한 기술분야에서 통상의 지식을 가진 자라면 상기 내용을 바탕으로 본 발명의 범주 내에서 다양한 응용, 변형 및 개작을 행하는 것이 가능할 것이다. 이에, 본 발명의 진정한 보호 범위는 청구된 청구 범위에 의해서만 정해져야 할 것이다.Although described with reference to the drawings according to the embodiment of the present invention, those skilled in the art to which the present invention pertains will be able to make various applications, modifications and adaptations within the scope of the present invention based on the above contents . Accordingly, the true protection scope of the present invention should be defined only by the claims.

S110 : 데이터 수집부
S120 : 데이터 전처리부
S130 : 수요예측 모델 개발부
S140 : 최적 모델 도출부
S150 : 모델 강화학습부
S160 : 전력 수요예측부S110: data collection unit
S120: data preprocessor
S130: Demand forecasting model development department
S140: Optimal model derivation unit
S150: model reinforcement learning unit
S160: Power demand forecasting unit

Claims (3)

전력 수요예측을 위해 기상 데이터와 전력 사용량 데이터를 수집하는 데이터 수집부; 상기 기상 데이터, 상기 전력 사용량 데이터의 누락 데이터를 보정하고 시계열 입력 변수화하는 전처리부; 학습 데이터를 이용하여 수요예측 모델을 학습 하는 강화 학습부; 및 순환형 신경망 모델을 이용한 지역 단위 전력 수요를 예측하는 수요예측부를 포함하는 것을 특징으로 하는 순환형 신경망 모델을 이용한 전력 수요예측 장치.a data collection unit that collects weather data and power usage data for power demand prediction; a pre-processing unit that corrects missing data of the weather data and the power usage data and converts it into a time series input variable; a reinforcement learning unit that learns a demand forecasting model using learning data; and a demand prediction unit for predicting regional power demand using a recurrent neural network model. 제 1항에 있어서,
상기 순환형 신경망 모델을 이용한 전력 수요예측을 누락 데이터 보정을 통해 생성되는 전처리 데이터를 이용하여 학습 데이터를 생성하는 것을 특징으로 하는 순환형 신경망 모델을 이용한 전력 수요예측 장치.The method of claim 1,
Power demand prediction apparatus using a cyclic neural network model, characterized in that the training data is generated by using pre-processing data generated by correcting missing data for power demand prediction using the cyclic neural network model. 제 1항에 있어서,
수요예측 강화학습의 시간 구성을 딥러닝 모델 생성기간, 강화학습 기간, 강화학습 검증기간, 모델적용 기간으로 구분하여 반복적인 강화 학습을 특징으로 하는 순환형 신경망 모델을 이용한 전력 수요예측 장치.
The method of claim 1,
Power demand prediction device using a cyclical neural network model characterized by repetitive reinforcement learning by dividing the time composition of demand prediction reinforcement learning into deep learning model generation period, reinforcement learning period, reinforcement learning verification period, and model application period.
KR1020200050984A 2020-04-27 2020-04-27 National power demand forecasting using cyclic neural network model KR20210132503A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1020200050984A KR20210132503A (en) 2020-04-27 2020-04-27 National power demand forecasting using cyclic neural network model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1020200050984A KR20210132503A (en) 2020-04-27 2020-04-27 National power demand forecasting using cyclic neural network model

Publications (1)

Publication Number Publication Date
KR20210132503A true KR20210132503A (en) 2021-11-04

Family

ID=78521860

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1020200050984A KR20210132503A (en) 2020-04-27 2020-04-27 National power demand forecasting using cyclic neural network model

Country Status (1)

Country Link
KR (1) KR20210132503A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102514996B1 (en) * 2022-09-13 2023-03-29 성한 주식회사 Power demand forecasting system and method using unsupervised learning and supervised learning in machine learning

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102514996B1 (en) * 2022-09-13 2023-03-29 성한 주식회사 Power demand forecasting system and method using unsupervised learning and supervised learning in machine learning

Similar Documents

Publication Publication Date Title
US11243262B2 (en) Degradation estimation apparatus, computer program, and degradation estimation method
Nguyen et al. A new method for forecasting energy output of a large-scale solar power plant based on long short-term memory networks a case study in Vietnam
JP2019204458A (en) Electric power demand prediction system, learning device, and electric power demand prediction method
Colak et al. Multi-time series and-time scale modeling for wind speed and wind power forecasting part I: Statistical methods, very short-term and short-term applications
Ziyabari et al. A spatio-temporal hybrid deep learning architecture for short-term solar irradiance forecasting
CN110705768B (en) Wind power generation power prediction optimization method and device for wind power plant
US11830090B2 (en) Methods and systems for an enhanced energy grid system
KR20210132503A (en) National power demand forecasting using cyclic neural network model
CN112612781A (en) Data correction method, device, equipment and medium
El Shahat et al. An artificial neural network model for wind energy estimation
CN116205382B (en) Prediction method and prediction device for electricity consumption, electronic device and electronic equipment
Islam et al. Location agnostic source-free domain adaptive learning to predict solar power generation
CN117220283A (en) Photovoltaic power generation power prediction method, device and equipment
Chen et al. Machine learning forecasts of Scandinavian numerical weather prediction wind model residuals with control theory for wind energy
Romero et al. Short-term active power forecasting of a photovoltaic power plant using an artificial neural network
Ehsan et al. Artificial neural network predictor for grid-connected solar photovoltaic installations at atmospheric temperature
CN111476437A (en) Method and device for predicting ultra-short term interval of photovoltaic power generation power
Lin et al. Hpt-rl: Calibrating power system models based on hierarchical parameter tuning and reinforcement learning
CN115618286A (en) Transformer partial discharge type identification method, system, equipment, terminal and application
Belmahdi et al. Forecasting solar radiation using machine learning method: new optimization algorithm
Zjavka Photo-voltaic power intra-day and daily statistical predictions using sum models composed from L-transformed PDE components in nodes of step by step developed polynomial neural networks
Jabari et al. Long-term solar irradiance forecasting using feed-forward back-propagation neural network
Yagli et al. Using combinational methods for forecast improvement in PV power plants
Chen et al. Short-term wind power forecasting based on spatial correlation and artificial neural network
US20190006850A1 (en) Method for forecasting the power daily generable by a solar inverter