實時空調能耗預測模擬與預測控制

Real-time air conditioning energy consumption simulation and model predictive control

蔣星宇(Xing-Yu Jiang)

指導教授 ： 詹瀅潔

國立臺灣大學/工學院/土木工程學研究所/碩士(2018年)

https://doi.org/10.6342/NTU201801962

摘要

電力供給與需求之間的矛盾是個嚴峻的問題。在熱帶與亞熱帶地域，夏季高溫天氣極有可能引發電力供給緊缺。空調被廣泛運用耗電量高且使用時間段集中是導致用電需求量上漲的因素之一。若能從中節能可有效降低電力消耗。
建築室內溫度波動具有非瞬時性，最佳控制時機往往需要被預測。模型預測控制(MPC, Model predictive control)能夠將未來狀況納入控制策略的考量範圍，並可以採取相應措施。過去之文獻已證明模型預測控制適用於建築能耗實時預測並且具有良好的控制效果，但鮮有文獻以實際天氣數據進行實作與驗證其成效。本研究選擇結合建築模擬程式來實現模型預測控制並驗證其效果。
本研究運用商業數學軟體MATLAB和建築模擬軟體EnergyPlus開發了一套自動化模型預測控制系統，旨在對現地建物進行短期實時空調能耗預測模擬與預測控制，反饋基於預測天氣的最佳控制策略。系統以降低冷卻負載和轉移尖峰用電為切入點採用三種控制策略，分別是自然通風、空調預冷和儲冰槽儲冰。系統會自動從網絡獲取並且處理天氣數據，通過基因演算法結合EnergyPlus模擬以預測電費為目標求解控制策略的最佳化。本研究預測控制之電費節約幅度大約為29%~33%，但會受到預測天氣與實際天氣之間的誤差而產生不確定性。

關鍵字

實時 ； 空調能耗模擬 ； 模型預測控制 ； EnergyPlus ； 基因演算法

並列摘要

The cooling demand is one of the significant factors that lead to increased electricity consumption in summer. Researchers tried to find solutions to reduce energy usage while providing a comfortable indoor environment to the occupants.
Due to the effect of thermal mass, the impact of weather condition and the changes in indoor temperature are not instantaneous. Therefore, a suitable control strategy needs to include the considerations of what would happen shortly. The concept, model predictive control, is to anticipate future events and takes actions accordingly. Previous researchers have proven that model predictive control is suitable for real-time prediction of building energy consumption. However, most of the studies stayed at simulation phase and used the off-line data in their studies. Only a few studies use real-time weather prediction to test and verify the effectiveness of the developed control strategies. This study combines few technology including weather prediction, building simulation program, and optimization algorithms, to achieve model predictive control and to verify its effectiveness.
In this study, an automated model predictive control system has been developed in MATLAB and EnergyPlus was used as simulation engine. The system is designed to simulate and predict short-term and real-time air conditioning energy consumption for existing buildings. The system adopts three control strategies to reduce cooling load and transfer electricity consumption at daytime peak hours, including natural ventilation, pre-cooling, and thermal energy storage.
The system automatically gets and processes weather data from the Wunderground website. The system uses a genetic algorithm combined with EnergyPlus simulation to optimize an objective function. The objective function of the optimization is the predicted total cost of the electric energy used by air conditioning. Results indicate that the system can save the cost of electricity by about 29%~33%. However, the result is also sensible to the uncertainties of weather prediction.

並列關鍵字

real-time ； air conditioning energy consumption simulation ； model predictive control ； EnergyPlus ； Genetic Algorithm

參考文獻

[1] 105年能源供需概況. 經濟部能源局,2016.

[2] 辦公大樓節約能源手冊. 經濟部能源局，2009.

[3] 家庭節能應用技術手冊，經濟部能源局，2017

[4] ZHANG, H. 2002, “development of typical year weather data for chinese locations,” ashrae transactions 2002, v. 108. pt. 2.

[5] SEO, D. 2010, development of a universal model for predicting hourly solar radiation – application: evaluation of an optimal daylighting controller, Architectural Engineering, Sungkyunkwan University, Korea.