³í¹®¸í |
°èÀýº° ½Ç³» ȯ°æ µ¥ÀÌÅÍ ±â¹Ý Àç½ÇÀÚ Çൿ ºÐ·ù ¸ðµ¨ °³¹ß / A Model for Classification of Occupant Behavior based on Building Environmental Data by Seasons |
ÀúÀÚ¸í |
ÀÌ¿¹¸°(Ye Rin Lee) ; À±¿µ¶õ(Young Ran Yoon) ; ¹®ÇöÁØ(Hyeun Jun Moon) |
¼ö·Ï»çÇ× |
´ëÇÑ°ÇÃàÇÐȸ³í¹®Áý, Vol.36 No.11 (2020-11) |
ÆäÀÌÁö |
½ÃÀÛÆäÀÌÁö(239) ÃÑÆäÀÌÁö(7) |
ÁÖÁ¦¾î |
Àç½ÇÀÚ ÇàÅÂ; Àç½ÇÀÚ Çൿ; °Ç¹° ȯ°æ µ¥ÀÌÅÍ; ºÐ·ù¾Ë°í¸®Áò; ±â°èÇнÀ ; Occupant status; Occupant activity; Environmental data; Classification; Machine Learning |
¿ä¾à1 |
HVAC ½Ã½ºÅÛÀÇ ÀûÀýÇÑ °Ç¹° ¿î¿µ ¹× Á¦¾î¸¦ À§ÇØ °ÅÁÖÀÚÀÇ ¼ö¿Í ±× È°µ¿¿¡ ´ëÇÑ ÀÚ¼¼ÇÑ Á¤º¸¸¦ °®´Â °ÍÀÌ Áß¿äÇÏ´Ù. ½Ç³» ȯ°æÀº ±â±â »ç¿ë°ú °ÅÁÖÀÚÀÇ È°µ¿¿¡ ¿µÇâÀ» ¹Þ´Â´Ù. µû¶ó¼ º» ¿¬±¸´Â ¸Ó½Å·¯´× ¾Ë°í¸®ÁòÀ» ÀÌ¿ëÇÑ °ÅÁÖÀÚÀÇ È°µ¿ ºÐ·ù¸¦ ¸ñÀûÀ¸·Î ÇÑ´Ù. ºÐ·ù ¾Ë°í¸®ÁòÀ» ÀÌ¿ëÇÏ¿© °èÀýº°(¿©¸§, °Ü¿ï, ¿©¸§+°Ü¿ï)º° Àç½ÇÀÚ Çൿ ÀÎÁö ¸ðµ¨À» °³¹ßÇÏ¿´´Ù. µ¥ÀÌÅÍ ¼öÁýÀº ½º¸¶Æ® ¸®ºù Å×½ºÆ®º£µå¿¡¼ ¼öÇàµÇ¾úÀ¸¸ç, Àç½ÇÀÚÀÇ »óŸ¦ ¼ö¸é, ÈÞ½Ä, ÀÛ¾÷, ¿ä¸®, ½Ä»ç, ¿îµ¿, ¿ÜÃâÀÇ 7°¡Áö È°µ¿À¸·Î ºÐ·ùÇÏ¿´´Ù. Àç½ÇÀÚÀÇ Çൿ ºÐ·ù ¸ðµ¨ °³¹ßÀ» À§ÇØ µÎ °¡Áö ºÐ·ù ¾Ë°í¸®Áò(KNN, Random Forest)À» »ç¿ëÇÏ¿´´Ù. ¿©¸§Ã¶ µ¥ÀÌÅ͸¦ ÀÌ¿ëÇÑ ·£´ý Æ÷·¹½ºÆ® ¸ðµ¨ÀÇ °æ¿ì, ¸ðµ¨ÀÇ Á¤È®µµ´Â 95.96%À̸ç KNNÀÇ °æ¿ì, Á¤È®µµ°¡ 94.75%·Î ³ªÅ¸³µ´Ù. °Ü¿ïö µ¥ÀÌÅ͸¦ ÀÌ¿ëÇÑ ¸ðµ¨ÀÇ °æ¿ì ·£´ý Æ÷·¹½ºÆ® ¸ðµ¨ÀÇ Á¤È®µµ´Â 98.91%, KNNÀº 98.90%·Î ³ªÅ¸³µ´Ù. ¿©¸§Ã¶°ú °Ü¿ïö µ¥ÀÌÅ͸¦ ÇÔ²² »ç¿ëÇßÀ» ¶§, µÎ ¸ðµ¨ÀÇ Á¤È®µµ´Â °¢°¢ ·£´ý Æ÷·¹½ºÆ® 97.82%, KNN 97.16%·Î ³ªÅ¸³µ´Ù. ±×·¯³ª ¿ä¸®¿Í ÈÞ½ÄÀº ´Ù¸¥ È°µ¿¿¡ ºñÇØ Á¤È®µµ°¡ ³·¾Ò´Ù. |
¿ä¾à2 |
It is important to have detailed information on the number of occupants and their activities for appropriate building operation and control of HVAC systems. Indoor environment is affected by using thermal environmental devices, and the occupant¡¯s activities as well. Thus, this study focuses on the classification of occupant¡¯s activities using machine learning algorithms with indoor environmental data. We developed an occupant¡¯s status detection model by seasons(summer, winter, summer and winter) using classification algorithms. Data collection was performed in a Smart Living Testbed. This study categorized occupant¡¯s status into 7 activities; sleeping, resting, working, cooking, eating, exercising, or away. Two classification algorithms(KNN, Random Forest) were evaluated for the development of an occupant¡¯s behavior classification model. For Random Forest model using summer data, the accuracy of the occupant behavior detection model was 95.96% and for KNN, the accuracy was 94.75%. For models using winter data, the accuracy of Random Forest model was 98.91% and KNN was 98.90%. When we used summer and winter data together for the classification models, the accuracies of both models were 97.82% for Random Forest and 97.16% for KNN, respectively. However, cooking and rest showed lower accuracies compared to other activities. |