• Àüü
  • ÀüÀÚ/Àü±â
  • Åë½Å
  • ÄÄÇ»ÅÍ
´Ý±â

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ÇÐȸÁö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ÇÐȸÁö > µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) C-LSTM ¸ðµ¨À» ÀÌ¿ëÇÑ ´Ü±â Àü·Â ¿¹Ãø
¿µ¹®Á¦¸ñ(English Title) Short-term Electricity Forecasting Using Convolutional Long Short Term Memory(C-LSTM)
ÀúÀÚ(Author) ¼­ÈÖ   Á¶¹®Áõ   Hwi Seo   Moon-Jeung Joe  
¿ø¹®¼ö·Ïó(Citation) VOL 39 NO. 01 PP. 0006 ~ 0017 (2023. 04)
Çѱ۳»¿ë
(Korean Abstract)
È¿À²ÀûÀÎ ¹ßÀü ¼³ºñ ¿î¿ë°ú ¿¡³ÊÁö °ø±ÞÀ» À§Çؼ­´Â Á¤È®ÇÑ Àü·Â ¿¹ÃøÀÌ ÇʼöÀûÀÌ´Ù. Àü·Â ¿¹ÃøÀº ¼öÀÏ ³»ÀÇ Àü·Â »ç¿ë·®ÀÇ º¯È­¸¦ ¿¹ÃøÇÏ´Â ´Ü±â ¿¹Ãø°ú ¼ö°³¿ù °£ÀÇ Àü·Â »ç¿ë·®ÀÇ È帧À» ¿¹ÃøÀ» ÇÏ´Â Àå±â ¿¹Ãø À¸·Î ³ª´©¾îÁø´Ù. º» ³í¹®Àº ½Å°æ¸ÁÀ» È°¿ëÇÑ ´Ü±â Àü·Â ¿¹Ãø¿¡ ´ëÇØ ºÐ¼®ÇÏ°í ¼º´ÉÀ» °³¼±ÇÏ´Â ¹æ¾ÈÀ» Á¦¾ÈÇÑ´Ù. ½Ã°è¿­ µ¥ÀÌÅ͸¦ È¿°úÀûÀ¸·Î ó¸®Çϱâ À§ÇØ ÇÕ¼º°ö ½Å°æ¸Á(CNN: Convolutional Neural Network)°ú ¼øȯ ½Å°æ¸Á(RNN: Recurrent Neural Network)ÀÇ ÀÏÁ¾ÀÎ LSTM(Long short-Term Memory)À» °áÇÕÇÑ C-LSTM µö·¯´× ¸ðµ¨À» Á¦¾ÈÇÑ´Ù. º» ¿¬±¸¿¡¼­´Â ½Å°æ¸ÁÀ» È°¿ëÇÏ¿© Àü·Â·® °°Àº °ªÀÇ º¯È­ ¹üÀ§°¡ ³Ð°í ¿¹ÃøÀÌ ¾î·Á¿î ½Ã°è¿­ µ¥ÀÌÅÍ ¿¹Ãø¿¡ ÀÖ¾î Á¤È®µµ¿¡ ¹ÌÄ¡´Â ¿äÀÎÀ» ºÐ¼®ÇÏ¿´´Ù. ±× °á°ú 1) ½Å°æ¸Á ÀÔ·ÂÀÇ ¹üÀ§¸¦ °áÁ¤ÇÏ´Â µ¥ÀÌÅÍÀÇ Á¤±ÔÈ­, 2) ½Å°æ¸Á Ãâ·Â¿¡ ¿µÇâÀ» ÁÖ´Â È°¼ºÈ­ ÇÔ¼ö°¡ Á¤È®µµ¿¡ ¿µÇâÀ» ¹ÌÄ¡´Â °Í À» È®ÀÎÇÏ¿´´Ù. UK-DALE µ¥ÀÌÅ͸¦ »ç¿ëÇÑ ´Ü±â Àü·Â ¿¹Ãø ½ÇÇè¿¡¼­ ±âÁ¸ÀÇ Á¤±ÔÈ­ ¹æ¹ý Áß¿¡¼­ MinMax Á¤±ÔÈ­ ¹æ¹ý°ú ½Å°æ¸ÁÀÇ È°¼ºÈ­ ÇÔ¼ö Leaky-ReLU¸¦ °áÇÕÇÏ¿´À» ¶§ MAE ±âÁØ 98%ÀÇ °¡Àå ³ôÀº Á¤È®µµ¸¦ º¸ÀÌ´Â °ÍÀ» È®ÀÎÇÏ¿´´Ù. ¶ÇÇÑ, Àü·Â·®ÀÇ Á¤È®µµ¸¦ ³ôÀ̱â À§ÇØ ¿ÜºÎ ¿äÀÎÀ¸·Î ³¯¾¾¿Í ³¯Â¥¸¦ Ãß°¡ÀûÀÎ ÀÔ·Â À¸·Î »ç¿ëÇÑ ½ÇÇèÀ» ¼öÇàÇÏ¿´À¸³ª Á¤È®µµ Çâ»ó¿¡´Â Å©°Ô ¿µÇâÀÌ ¾ø¾ú´Ù. ÀÌ´Â Àü·Â·® ´Ü±â ¿¹Ãø¿¡´Â ¿ÜºÎ ¿äÀÎÀÇ ¿µÇâ·ÂÀÌ ÀûÀ¸¸ç, °èÀýÀÇ º¯È­¿¡ µû¸¥ Àå±â ¿¹Ãø °á°ú¿¡ Á» ´õ ¿µÇâÀÌ Å« °ÍÀ¸·Î ¿¹ÃøµÈ´Ù.
¿µ¹®³»¿ë
(English Abstract)
Electricity forecasting is essential for efficient operation of power generation facilities and energy supply. Electricity forecasts are divided into short-term forecasts that predict changes in electricity usage within days and long-term forecasts that predict the flow of electricity usage over several months. This paper proposes a short-term electricity forecasting method using neural networks. To handle time series data effectively, we propose a C-LSTM deep learning model that combines convolutional neural networks (CNN) and long short-term memory (LSTM), a type of recurrent neural network (RNN). In this study, a neural network was used to analyze the factors affecting accuracy in predicting time series data with no fixed range of values such as electricity. As a result, we verified 1) normalization of data that determines the range of neural network input, and 2) activation functions that affect neural network output affect prediction accuracy. Short-term electricity forecasting experiments using UK-DALE data showed the highest accuracy of 98% based on MAE when combining the MinMax normalization method with the neural network activation function Leaky-ReLU. In addition, an experiment using weather and date as additional inputs was conducted to increase the accuracy of electricity, but there was no significant effect on the prediction accuracy improvement. We assumed that the weather and date information does not affect short-term electricity prediction, but is a useful factor for long-term forecasting that reflects the season.
Å°¿öµå(Keyword) µö·¯´×   C-LSTM   Àü·Â ¿¹Ãø   Deep Learning   C-LSTM   Electricity Forecasting  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå