HOME > 상세정보

상세정보

Distributed energy resource management in microgrid environments using artificial intelligence

Distributed energy resource management in microgrid environments using artificial intelligence

자료유형
학위논문
개인저자
손준서, 孫焌瑞
서명 / 저자사항
Distributed energy resource management in microgrid environments using artificial intelligence / Junseo Son
발행사항
Seoul :   Graduate School, Korea University,   2020  
형태사항
x, 123장 : 삽화, 도표 ; 26 cm
기타형태 저록
Distributed Energy Resource Management in Microgrid Environments Using Artificial Intelligence   (DCOLL211009)000000127357  
학위논문주기
학위논문(박사)-- 고려대학교 대학원: 컴퓨터학과(정보통신대학), 2020. 2
학과코드
0510   6YD36   370  
일반주기
지도교수: 김효곤  
서지주기
참고문헌: 장 104-119
이용가능한 다른형태자료
PDF 파일로도 이용가능;   Requires PDF file reader(application/pdf)  
비통제주제어
distributed energy resource , artificial intelligence , deep neural network , deep reinforcement learning , microgrid,,
000 00000nam c2200205 c 4500
001 000046025845
005 20200428100754
007 ta
008 191226s2020 ulkad bmAC 000c eng
040 ▼a 211009 ▼c 211009 ▼d 211009
041 0 ▼a eng ▼b kor
085 0 ▼a 0510 ▼2 KDCP
090 ▼a 0510 ▼b 6YD36 ▼c 370
100 1 ▼a 손준서, ▼g 孫焌瑞
245 1 0 ▼a Distributed energy resource management in microgrid environments using artificial intelligence / ▼d Junseo Son
260 ▼a Seoul : ▼b Graduate School, Korea University, ▼c 2020
300 ▼a x, 123장 : ▼b 삽화, 도표 ; ▼c 26 cm
500 ▼a 지도교수: 김효곤
502 1 ▼a 학위논문(박사)-- ▼b 고려대학교 대학원: ▼c 컴퓨터학과(정보통신대학), ▼d 2020. 2
504 ▼a 참고문헌: 장 104-119
530 ▼a PDF 파일로도 이용가능; ▼c Requires PDF file reader(application/pdf)
653 ▼a distributed energy resource ▼a artificial intelligence ▼a deep neural network ▼a deep reinforcement learning ▼a microgrid
776 0 ▼t Distributed Energy Resource Management in Microgrid Environments Using Artificial Intelligence ▼w (DCOLL211009)000000127357
900 1 0 ▼a Son, Jun-seo, ▼e
900 1 0 ▼a 김효곤, ▼g 金孝坤, ▼e 지도교수
945 ▼a KLPA

전자정보

No. 원문명 서비스
1
Distributed energy resource management in microgrid environments using artificial intelligence (22회 열람)
PDF 초록 목차

소장정보

No. 소장처 청구기호 등록번호 도서상태 반납예정일 예약 서비스
No. 1 소장처 과학도서관/학위논문서고/ 청구기호 0510 6YD36 370 등록번호 123063627 도서상태 대출가능 반납예정일 예약 서비스 B M
No. 2 소장처 과학도서관/학위논문서고/ 청구기호 0510 6YD36 370 등록번호 123063628 도서상태 대출가능 반납예정일 예약 서비스 B M

컨텐츠정보

초록

Conventional power systems are unidirectional configurations that deliver electricity produced at the power plant to the consumer, which generates a large amount of power and supplies it to electrical users through the grid. This centralized power system is quite inefficient in power generation, delivery and supply. In case of power shortages, power generation facilities should be operated in excess of the required power, and the transmission network should be operated to deliver the generated power to long-distance consumers. Therefore, this system has poor power usage efficiency and wastes a lot of maintenance costs. More important than efficiency is a power outage due to the shutdown of a power plant in a disaster situation. Most countries are converting their grids into decentralized power plants in response to these disasters. It is a way of operating as a single power plant by integrating small power generation facilities such as solar power distributed at each site and power management system. The distributed generation system provides excellent resilience and efficient power operation in disasters such as earthquakes.

The microgrid is power systems consisting of these small distributed sources and energy storage devices. The microgrid has made the grid smaller and more power efficient. Small-scale power grid can be operated efficiently through power generation through renewable energy sources and charging / discharging of power using power storage devices. In this paper, we describe an effective distributed power management technique using artificial intelligence in microgrid environment composed of photovoltaic sources and ESS (Energy Storage System).

Forecasting the photovoltaic (PV) power output in the next day is an important factor to plan for the cost-effective power supply in grid-connected smart buildings. Having the information on the expected shortage or surplus of the next day's solar power supply by hours, one can perform a cost-effective charge / discharge control of the associated energy storage system (ESS). The solar power generation is heavily influenced by weather parameters such as the solar radiation at the site the solar panels are installed. So, the conventional procedure predicts weather parameters at the site, and calculates the expected amount of PV power output based on the predicted parameters using a PV model.

Existing works in photovoltaic (PV) power generation focus on accurately predicting the PV power output on a forecast horizon.
If done successfully, it allows for the planning various aspects of future energy supply, for instance supplementing it by other more traditional means at a more cost-effective way. As the solar power generation is heavily influenced by meteorological conditions such as solar radiation, the weather forecast is a critical input in the prediction performance. But the weather forecast is traditionally considered to have coarse granularity, so many are compelled to use on-site meteorological sensors to complement it. Typically, they first model the relation between the regional weather forecast and the precise on-site measurement at the forecasted time. Then, using the precise historical on-site measurement values inferred from the relation and the weather forecast input, they use a PV model (typically implemented in commercial software) to forecast the PV power output. However, the approach involving on-site sensors has several issues. First, it incurs the cost in the installation, operation, and management of the sensors. Second, the physical model of the sensor dynamics itself can be a source of forecast errors. Third, it requires an accumulation of sensory data that represent all seasonal variations, which takes time to collect. In this paper, we take an alternative approach to use a relatively large deep neural network (DNN) instead of the on-site sensors to cope with the coarse-grained weather forecast. With historical PV output power data from our grid-connected building with a rooftop PV power generation facility and the publicly available weather forecast history data, we demonstrate that we can train a 6-layer feed-forward DNN for the day-ahead forecast. It achieves the average mean absolute error (MAE) of 2.9%, comparable to that of the conventional model, but without involving the on-site sensors.

Based on the estimated photovoltaic power generation, stable operation of the microgrid constructed through efficient operation of storage devices is possible. In the existing microgrid operation, the ESS schedule was operated by finding the optimal solution for such operation. However, due to the forecasting error of power generation and demand, it is difficult to maximize the efficiency of the actual operation of the ESS with a simple optimal solution. In this paper, we introduce artificial intelligence-based ESS optimization planning along with artificial intelligence-based photovoltaic forecasting. Reinforcement learning was used to learn about prediction failures, and the optimal operation of ESS considering the prediction error was possible. If the predicted error rate is 29.73%, the optimal operation of the ESS can be performed within the 5.27% error range. This result is 7.5% better than the existing optimal algorithm. As a result of this experiment, we learned that if we learn more data with various prediction errors, we can cope with all days. The proposed method also confirmed the possibility of establishing an optimal power generation plan for the climate of the site where the microgrid is installed.

We propose solar power generation forecasting based on DNN and the optimal generation control using  reinforcement learning. And also we confirm that microgrid operation can be performed more efficiently in the unpredictable environment using artificial intelligence. It is expected that the integration of artificial intelligence technology will be an effective technique for efficient and stable power operation. 

목차

1 Introduction  1
2 Related work  8
 2.1 PV Forecasting 8
 2.2 Optimization Techniques 17
3 Material and Methods 20
 3.1 PV Forecasting  20
  3.1.1 Historical data  20
  3.1.2 Comparison of the existing and the proposed approaches 34
  3.1.3 Architecturing the DNN  44
 3.2 Optimal Generation Plan   59
  3.2.1 Harmony search  64
  3.2.2 Deep Q-learning  68
4 Results   73
 4.1 PV Foresting 73
  4.1.1 Changes in input and output   75
  4.1.2 Visual comparison  75
  4.1.3 Numeric comparison  80
 4.2 Optimal Generation Plan 86
  4.2.1 Environment  86
  4.2.2 Performance of optimal generation plan  88
5 Conclusion   100
Bibliography   104
Summary (in Korean)   120