We developed a prediction model based on machine learning to provide a higher precision baseline in the energy auditing process.
Initialized with this objective, we have applied this model to the residential field, considering the high energy consumption in the sector. The project's further aspiration is to implement a prototype of a civilian Smart Energy Consumption Management System (SECMS). It provides functions including precise baseline prediction, energy-saving level indication, energy monitoring, and household appliance adoption suggestions. Eventually, it is expected to encourage residents to optimize their energy consumption behavior.
The SECMS is based on machine learning and realized through a prototype program. Sensors automatically collect appliance operation data to feed into the system and train corresponding models. SECMS will learn users' energy consumption behavior and make predictions. Users can operate the system through a graphical user interface (GUI) and access five main functions: tariff summary, usage pattern statistics, subsystem consumption profile, Energy Consumption Management (ECM) economic analysis, and model fit summary. The system features accuracy to exclude effects from external factors, flexibility to adapt to various project scopes, user-friendliness by minimizing manual work, and timely self-upgrades.
The SECMS is mainly applied to residential buildings. It requires a minimum of one sensor with the lowest initial investment. Residents can utilize it to perform energy consumption level calculations, detect abnormal electricity consumption in appliances, and obtain furniture adjustment suggestions. They can fully optimize their household appliances and spontaneously optimize their energy consumption behavior.
For future developments, there are two promising functions that can be further investigated. After defining a clear scope and being fed with big data, SECMS can function in the M&V (Measurement and Verification) process. Moreover, applying it to IoT (Internet of Things) is another potential direction.
The contents provided herein are from the winning entries. The Electrical and Mechanical Services Department does not endorse or assume any liability for the accuracy, completeness, or reliability of the contents presented.
我們開發了一個基於機器學習，可以實現精確的民用智能能源消耗管理的系統（SECMS）。SECMS的原理是利用傳感器自動收集電器的使用數據，供系統訓練、檢測和生成模型。用戶可以通過圖形使用者介面（GUI）操作系統，並允許存取五個主要功能：費率摘要、使用模式統計、子系統消耗概況、能源消費經濟分析和模型分析。該系統具有準確性、靈活性、簡便性和及時自我升級的特點。
SECMS主要應用於住宅建築。最低的初始投資僅為一個傳感器。居民可以利用它進行能源消耗水平計算，檢測電器的異常耗電量，並獲取家電購買建議。他們可以充份了解家用電器的耗電情況，不斷優化其能源消費行為。對於未來的發展，在更大的數據範本上，SECMS可以在量測和驗證（M&V）過程中發揮作用，或是應用於物聯網（IoT）。
本頁內容來自得獎者參賽作品。機電工程署概不對內容的準確性、完整性或可靠性作出任何保證或承擔任何責任。