최근 건물에서 에너지 시스템을 최적 운전하기 위해 모델을 활용한 예측 제어(MPC)와 관련한 연구가 지속해서 증가하고 있다. 다수의 연구 논문과 실증 연구에도 불구하고 MPC의 일반건물에서...

http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
최근 건물에서 에너지 시스템을 최적 운전하기 위해 모델을 활용한 예측 제어(MPC)와 관련한 연구가 지속해서 증가하고 있다. 다수의 연구 논문과 실증 연구에도 불구하고 MPC의 일반건물에서...
최근 건물에서 에너지 시스템을 최적 운전하기 위해 모델을 활용한 예측 제어(MPC)와 관련한 연구가 지속해서 증가하고 있다. 다수의 연구 논문과 실증 연구에도 불구하고 MPC의 일반건물에서의 적용은 아직 초기 단계에 머무른다. MPC의 성공적인 적용은 건물의 에너지 수요 공급량을 예측하는 모델에 큰 영향을 받는데 현재의 MPC용 모델은 현장의 제한사항을 반영하지 못하기 때문에 실제 적용에 활용되기에는 정확도와 사용성이 떨어진다. 따라서 본 연구에서는 MPC용 시뮬레이션 모델이 현장에서 활용되기 어려운 이유를 분석하고 실제 활용단계에서 적용할 수 있는 에너지 수요, 공급 모델을 개발한다. 개발된 모델은 실제 우리나라에서 시행중인 변동요금제도(TOU, Time of Use)에서 최적제어를 구현해 모델과 MPC 효과를 모두 검증한다.
MPC용 시뮬레이션 모델은 에너지 공급모델과 수요모델로 구성된다. 공급모델은 수평면전일사량과 이를 입력값으로 하는 태양광 발전량을 예측하는 것을 목표로 하며 일반적인 딥러닝(LSTM) 학습알고리즘을 적용한다. 기존의 딥러닝 기반 일사예측 및 발전량 모델은 현장에서 오랫동안 측정한 기상 및 일사량 데이터를 입력값으로 하므로 과거 지속적으로 측정데이터를 누적하지 않은 건물에서는 MPC를 적용하고자 할 경우 활용하기 어려움이 따른다. 이에 본 연구에서는 Global 지역의 누적된 일사 데이터를 정규화해 일사 발생패턴을 분석하여 local 대상 건물에서 단 하루의 측정데이터만으로도 다음날의 일사량 및 태양광발전량을 예측할 수 있는 공급모델을 개발하였다. 제안된 모델의 오차는 30W/m2 수준으로 현장에서 1년 이상의 일사량 자료를 수집한 기존 선행연구와 유사한 성능을 보였다.
수요 예측 모델은 건물의 열적 거동을 물리적 관계를 수식으로 계산하는 TRNSYS를 통해 개발한다. 블랙박스 모델로 수요 예측 모델을 개발할 경우 모델의 구축이 쉽고 건물에서 측정된 데이터를 학습하기 때문에 실제를 묘사할 가능성이 더 크나 다양한 제어 시나리오를 시뮬레이션해 최적값을 도출하는 MPC 시뮬레이션에는 실제 건물에서 원하는 수준의 에너지 자료를 수집할 수 없으므로 사용에 제한이 있을 수 있다. 본 연구에서는 개발된 TRNSYS 모델의 결과와 실제 에너지 사용량의 차이를 좁히기 위해 보상을 최대화하기 위한 행동을 스스로 학습하는 강화학습 알고리즘을 적용해 주요 파라미터를 수정하는 캘리브레이션 방법을 제안하였다. 그 결과 기존 CVRMSE 58% 수준의 수요 모델의 오차를 20%로 개선하였으며 이는 ASHRAE 가이드 라인인 30% 오차수준을 만족하는 결과이다.
개발된 수요, 공급예측 모델은 모델 구축을 위해 건물에서 쉽게 확보할 수 있는 단기간의 측정데이터, 기상예보 등 가용한 데이터만을 사용하기 때문에 MPC 시뮬레이션 모델로 활용할 수 있다. 실제로 다목적 유전자 알고리즘을 적용하여 우리나라에서 시행 중인 일반건물 TOU에서 MPC 최적화를 통한 운전계획을 수립했을 때 실내 열 쾌적을 유지하면서 다음 날의 지출 전기요금을 감소시킬 수 있음을 확인하였다.
다국어 초록 (Multilingual Abstract)
Recently, there has been a continuous increase of the research related to model-based predictive control (MPC) in the field of building energy systems, aiming to optimize the operation of energy systems in buildings. While numerous research papers and...
Recently, there has been a continuous increase of the research related to model-based predictive control (MPC) in the field of building energy systems, aiming to optimize the operation of energy systems in buildings. While numerous research papers and some empirical studies exist, the application of MPC in real buildings is still in its early stages. The successful implementation of MPC heavily relies on models that must accurately predict the energy demand supply of a building. However, the current models used for MPC fail to adequately reflect the constraints present in the field, resulting in lower accuracy and usability for practical applications. Therefore, this study, aims to analyze the reasons behind the challenges in utilizing simulation models for MPC in real-world applications. Additionally, energy demand and supply models is developed that can be applied during the practical implementation phase. The proposed MPC models will be ued to investigate optimal operating strateges under the Time of Use (TOU) tariff system, which is currently being implemented in South Korea for commercial buildings. This will demonstrate the effectiveness of both the model and the MPC approach. For the MPC simulation model, separate energy supply and demand models will be developed. The supply model aims to predict the solar power generation based on horizontal solar irradiance and other relevant inputs. A commonly used deep learning algorithm, such as Long Short-Term Memory (LSTM), will be applied for training the supply model. One of the challenges in applying MPC to buildings where a long history of measured weather and solar irradiance data are scarce is that existing deep learning-based solar prediction models rely on historical data. Therefore, it becomes difficult to utilize those models for MPC in buildings where continuous measurng data has not been accumulated. In this study, a supply model is developed that can predict the next day's solar irradiance and solar power generation using only a single day's measurement data from the target building. To achieve this, accumulated solar irradiance data from global regions are normalized and the generalized patterns of solar occurrence is used for a target local. This approach enables us to develop a supply model that can make accurate predictions even with limited measurement data from the local building. The proposed model exhibited an error level of around 30W/m2, which is comparable to the performance of previous studies that collected over one year of solar irradiance data in the field.
The demand prediction model is developed using TRNSYS models, which calculates the thermal behavior of the building based on physical relationships represented by a set of equations. This approach enables us to develop an accurate demand prediction model for a target building.
If a black-box model is used for the demand prediction, it may be simple for model construction by learning the measuring data in the building, which increases the likelihood of accurately representing the reality. However, there may be limitations in using actual building energy data at the desired level, as it is challenging to collect such data. This study, proposes a calibration method that utilizes a reinforcement learning algorithm to adjust main parameters, aiming to minimize the discrepancy between the results obtained from the developed TRNSYS model and actual energy consumption. This calibration method involves self-learning behaviors to maximize rewards, ultimately improving the accuracy of the model.
As a result, errors found in the TRNSYS model was reduced from 58% to 20% in CVRMSE. This improvement meets the guideline of 30% set by ASHRAE.
The proposed demand and supply prediction models can be utilized as MPC simulation models since they are constructed using only readily available short-term measurement data from the building and other available data such as weather forecasts. This makes them suitable for a practical application of MPC based operating scenarios. A test MPC is achieved using a multi-objective genetic algorithm under the Time of Use (TOU) tariff system, which has been implemented in general buildings in South Korea.
The results showed the possibility of reduced electricity expenses while maintaining indoor thermal comfort for the following day, by means of an MPC scheme. This optimization approach provides a promising solution for energy cost reduction in buildings.
목차 (Table of Contents)
참고문헌 (Reference)
1. Long short-term memory, Hochreiter, S. & Schmidhuber, J, 9(8), 1735-1780, , 1997
2. PVWatts version 5 manual, Dobos, A. P., No. NREL/TP-6A20-62641, , 2014
3. Review of modeling methods for HVAC systems, Afram, A., Janabi-Sharifi, F., 67(1), 507–519. https://doi. org/10.1016/j, , 2014
4. HVAC system optimization— in-building section, Soh, Y. C., Lu, L., Xie, L., Li, S., Cai, W., 37(1), 11-22, , 2005
5. Control with building mass-part I: Thermal response model, Armstrong, P. R., Norford, L. K., Leeb, S. B., 112(1), 449, , 2006
6. Gaussian process model for control of an existing building, Ahn, K.-U., Kim, I.-H., Kim, Y.-J., Kim, D.-W., Park, C.-S., 78, 2136–2141. https:// doi. org/10.1016/j. egypro., , 2015
7. Review of building energy modeling for control and operation, Li, X., Wen, J, 37, 517-537, , 2014
8. Applying model predictive control to a LEED silver-certified building, Swan, L., Qin, Z., Kavgic, M., Hilliard, T., 78, 1817-1822, , 2015
9. Model predictive control for the operation of building cooling systems, Haves, P., Hencey, B., Borrelli, F., Ma, Y., Coffey, B., Bengea, S., 20(3), 796-803, , 2011
10. Model predictive control of a building heating system: The first experience, Privara, S., Cigler, J., Široký, J., Ferkl, L., 43(2-3), 564-572, , 2011
1. Long short-term memory, Hochreiter, S. & Schmidhuber, J, 9(8), 1735-1780, , 1997
2. PVWatts version 5 manual, Dobos, A. P., No. NREL/TP-6A20-62641, , 2014
3. Review of modeling methods for HVAC systems, Afram, A., Janabi-Sharifi, F., 67(1), 507–519. https://doi. org/10.1016/j, , 2014
4. HVAC system optimization— in-building section, Soh, Y. C., Lu, L., Xie, L., Li, S., Cai, W., 37(1), 11-22, , 2005
5. Control with building mass-part I: Thermal response model, Armstrong, P. R., Norford, L. K., Leeb, S. B., 112(1), 449, , 2006
6. Gaussian process model for control of an existing building, Ahn, K.-U., Kim, I.-H., Kim, Y.-J., Kim, D.-W., Park, C.-S., 78, 2136–2141. https:// doi. org/10.1016/j. egypro., , 2015
7. Review of building energy modeling for control and operation, Li, X., Wen, J, 37, 517-537, , 2014
8. Applying model predictive control to a LEED silver-certified building, Swan, L., Qin, Z., Kavgic, M., Hilliard, T., 78, 1817-1822, , 2015
9. Model predictive control for the operation of building cooling systems, Haves, P., Hencey, B., Borrelli, F., Ma, Y., Coffey, B., Bengea, S., 20(3), 796-803, , 2011
10. Model predictive control of a building heating system: The first experience, Privara, S., Cigler, J., Široký, J., Ferkl, L., 43(2-3), 564-572, , 2011
11. Predictive control of a building hybrid heating system for energy cost reduction, Esmaeilzadeh, A., Markazi, A. H. D., Khanmirza, E., 46, 407-423, , 2016
12. Building modeling: Selection of the most appropriate model for predictive control, Váňa, Z., Cigler, J., Žáčeková, E., Privara, S., 55, 341-350, , 2012
13. Review and classification of recent observers applied in chemical process systems, Ali, J. M., Hussain, M. A., Hoang, N. H., Dochain, D., 76, 27–41, , 2015
14. Using particle swarm optimization to accurately identify syntactic phrases in free text, Tambouratzis, G., 8(1), 63-77, , 2018
15. Electricity price forecasting: A review of the state-of-the-art with a look into the future, Weron, R., 30(4), 1030-1081, , 2014
16. Theory and applications of HVAC control systems–A review of model predictive control (MPC), Afram A, Janabi-Sharifi F., 72, 343-355, , 2014
17. Robustness analysis of a hybrid ground coupled heat pump system with model predictive control, Helsen, L., Antonov, S., 47, 191–200. https://doi. org/10.1016/j. jprocont.2016.08.009, , 2016
18. Building automation and control systems and performance optimization: A framework for analysis, Aste, N., Manfren, M., Marenzi, G., 75, 313–330. https://doi. org/10.1016/j. rser.2016.10.072, , 2017
19. A study of energy management in domestic micro-grids based on model predictive control strategies, Bruni, G., Cordiner, S., Spagnolo, F., Rocco, V., Mulone, V., 61, 1012-1016, , 2014
20. Minimization of energy consumption in HVAC systems with data-driven models and an interior-point method, Xu, G., Zhang, Z., Kusiak, A., 85, 146-153, , 2014
21. Advanced control systems engineering for energy and comfort management in a building environment A review, Dounis, A. I., Caraiscos, C., 13(6-7), 1246-1261, , 2009
22. A new simplified modeling method for model predictive control in a medium-sized commercial building: A case study, Chen, Y., Zhuang, J., Chen, X., 127, 1-12, , 2018
23. Next-Day Prediction of Hourly Solar Irradiance Using Local Weather Forecasts and LSTM Trained with Non-Local Data, Jeon, B. K., Kim, E. J., 13(20), 5258, , 2020
24. Simulation-based optimization of an integrated daylighting and HVAC system using the design of experiments method, Jeon, Y., Kim, Y., Kim, W., 162, 666-674, , 2016
25. Learning-based predictive building energy model using weather forecasts for optimal control of domestic energy systems, Jeon, B. K., Lee, K. H., Shin, Y., Kim, E. J., 11(1), 147, , 2019
26. Optimal Cooling Operation of a Single Family House Model Equipped with Renewable Energy Facility by Linear Programming, Kim, E., Shin, Y., Lee, K., 29(12), 638-644, , 2017
27. Simulation-based model predictive control by the multi-objective optimization of building energy performance and thermal comfort, De Stasio, C., Mauro, G. M., Ascione, F., Bianco, N., Vanoli, G. P., 111, 131–144. https://doi. org/10.1016/j. enbuild.2015.11.033., , 2016
28. A new model predictive control scheme for energy and cost savings in commercial buildings: An airport terminal building case study, Hu, E., Chen, L., Huang, H., 89, 203-216, , 2015
29. Weather-data-based control of space heating operation via multi-objective optimization: Application to italian residential buildings, Napolitano, D. F., Ascione, F., Mauro, G. M., Bianco, N., Vanoli, G. P., 163, 114384. https://doi. org/10.1016/j. applthermaleng.2019.114384., , 2019
30. Supervisory model predictive controller (MPC) for residential HVAC systems: Implementation and experimentation on archetype sustainable house in Toronto, Afram, A., Janabi-Sharifi, F., 154, 268–282. https://doi. org/ 10.1016/j. enbuild.2017.08.060, , 2017
31. Artificial neural network (ANN) based model predictive control (MPC) and optimization of HVAC systems: A state of the art review and case study of a residential HVAC system, Janabi-Sharifi, F., Afram, A., Raahemifar, K., Fung, A. S., 141, 96-113, , 2017
32. Real-time electricity pricing: TOU-MPC based energy management for commercial buildings Empirical Modeling of Direct Expansion (DX) Cooling System for Multiple Research Use Cases. Sustainability, 12(20), 8738., Naidoo, R. M., Joe, J., Dong, J., Im, P., Mbungu, N. T., Bansal, R. C., 105, 3419-3424, , 2017