بررسي تاثير ادغام اطلاعات مناطق الكتريكي مختلف در پيش‌بيني بار شبكه قدرت با ارائه يك روش نوين پيش‌بيني سلسله‌مراتبي

Effect of Information Ensemble on Electricity Load Forecasting by Proposing a Novel Hierarchical Forecasting Method

ﭼﮑﯿﺪه: ﺑﻪ ﻣﻨﻈﻮر ﭘﯿﺶﺑﯿﻨﯽ ﺗﻘﺎﺿﺎي ﻣﺼﺮف اﻧﺮژي اﻟﮑﺘﺮﯾﮑﯽ ﯾﮏ ﺷﻬﺮ ﯾﺎ ﮐﺸﻮر، ﻣﺘﺪاول اﺳﺖ ﮐﻪ دادهﻫﺎي ﺗﺎرﯾﺨﯽ ﺑﺎر ﻣﻨﺎﻃﻖ ﻣﺨﺘﻠﻒ آن ﺷﻬﺮ و ﯾﺎ ﺷﻬﺮﻫﺎي ﻣﺨﺘﻠﻒ آن ﮐﺸﻮر ﺟﻤﻊآوري و ﺑﺮ اﺳﺎس آن ﺗﺼﻤﯿﻤﺎت راﻫﺒﺮي و ﯾﺎ راﻫﺒﺮدي اﺗﺨﺎذ ﺷﻮد. ﺑﺎ اﯾﻦ ﺣﺎل، ﻣﻤﮑﻦ اﺳﺖ اﺳﺘﻔﺎده از داده ﻫﺎي ﺗﺎرﯾﺨﯽ ﺗﻤﺎم ﻣﻨﺎﻃﻖ و ﯾﺎ اﻧﻮاع ﻣﺨﺘﻠﻒ ﺑﺎرﻫﺎي ﻣﺼﺮﻓﯽ )ﻣﺴﮑﻮﻧﯽ، ﺗﺠﺎري و ﺻﻨﻌﺘﯽ( ﺑﻪ ﯾﮏ ﻣﯿﺰان ﺣﺎﺋﺰ اﻫﻤﯿﺖ ﻧﺒﺎﺷﺪ؛ ﺑﺪﯾﻦ ﻣﻌﻨﯽ ﮐﻪ اﻣﮑﺎن دارد ﺗﻨﻬﺎ ﻣﯿﺰان ﺑﺎر ﻣﺼﺮﻓﯽ ﺑﻌﻀﯽ از ﻣﻨﺎﻃﻖ و ﯾﺎ ﺗﻨﻬﺎ ﻣﯿﺰان ﻣﺼﺮف ﯾﮏ ﻧﻮع ﺑﺎر ﺑﯿﺸﺘﺮﯾﻦ ﺗﺎﺛﯿﺮ و اﻫﻤﯿﺖ را در ﺗﺼﻤﯿﻢﮔﯿﺮي داﺷﺘﻪ ﺑﺎﺷﺪ. از آﻧﺠﺎﯾﯽ ﮐﻪ ﺟﻤﻊ ﺟﺒﺮي ﭘﯿﺶﺑﯿﻨﯽﻫﺎي ﺗﻤﺎم ﻣﻨﺎﻃﻖ و ﯾﺎ اﻧﻮاع ﺑﺎر ﻟﺰوﻣﺎ ﭘﯿﺶﺑﯿﻨﯽ ﻣﻨﺎﺳﺒﯽ ﺑﺮاي ﻫﺪف ﻣﻮرد ﻧﻈﺮ اراﺋﻪ ﻧﻤﯽ دﻫﺪ، روشﻫﺎي ﻣﺨﺘﻠﻔﯽ ﺑﺮاي ادﻏﺎم ﭘﯿﺶﺑﯿﻨﯽ ﻣﻨﺎﻃﻖ ﻣﺨﺘﻠﻒ وﺟﻮد دارد. در ﺳﺎده ﺗﺮﯾﻦ ﺣﺎﻟﺖ ﻣﻤﮑﻦ، ﻣﯽﺗﻮان ﺑﻪ ﺳﺎدﮔﯽ دادهﻫﺎي ﻣﻨﺎﻃﻖ ﻣﺨﺘﻠﻒ را ﺟﻤﻊ ﺟﺒﺮي ﻧﻤﻮد و ﯾﮏ ﺳﺮي زﻣﺎﻧﯽ ﮐﻠﯽ ﺑﻪدﺳﺖ آورد و ﭘﯿﺶﺑﯿﻨﯽ را ﺑﺮ اﺳﺎس آن اﻧﺠﺎم داد. ﺑﺎ اﯾﻦ وﺟﻮد، اﯾﻦ روش ﺳﺎده ﻧﻪ ﺗﻨﻬﺎ ﻫﻤﻮاره ﭘﯿﺶﺑﯿﻨﯽ ﻣﻨﺎﺳﺒﯽ اراﺋﻪ ﻧﻤﯽدﻫﺪ ﺑﻠﮑﻪ ﻧﯿﺎزﻣﻨﺪ دﺳﺘﺮﺳﯽ ﺑﻪ ﺟﺰﺋﯿﺎت داده ﻫﺎي ﺗﺎرﯾﺨﯽ ﺑﺎر و ﻋﻮاﻣﻞ ﻣﻮﺛﺮ آن ﺑﻪ ﺗﻔﮑﯿﮏ ﻫﺮ ﻣﻨﻄﻘﻪ ﻧﯿﺰ ﻣﯽ ﺑﺎﺷﺪ. از اﯾﻦ رو، در اﯾﻦ ﻣﻘﺎﻟﻪ روشﻫﺎي ادﻏﺎم ﻣﺨﺘﻠﻔﯽ ﻧﻈﯿﺮ روش ﻫﺎي ادﻏﺎم ﭘﺎﯾﯿﻦ ﺑﻪ ﺑﺎﻻ، ﺑﺎﻻ ﺑﻪ ﭘﺎﯾﯿﻦ و ﻫﻤﭽﻨﯿﻦ روﯾﮑﺮد ﺗﺮﮐﯿﺐ ﺑﻬﯿﻨﻪ ﺑﺎ در ﻧﻈﺮ ﮔﺮﻓﺘﻦ ﻫﻤﺒﺴﺘﮕﯽ ﺑﯿﻦ ﺳﺮيﻫﺎي زﻣﺎﻧﯽ زﯾﺮﻣﺠﻤﻮﻋﻪ ﻣﻌﺮﻓﯽ ﺷﺪه اﺳﺖ. ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ اﯾﻨﮑﻪ ﺗﺎﮐﻨﻮن ﺑﻪ ﺣﻮزه ﭘﯿﺶﺑﯿﻨﯽ ﺳﻠﺴﻠﻪﻣﺮاﺗﺒﯽ در ﺻﻨﻌﺖ ﺑﺮق ﭘﺮداﺧﺘﻪ ﻧﺸﺪه اﺳﺖ؛ اﯾﻦ ﻣﻘﺎﻟﻪ ﺑﺎ ﺑﺮرﺳﯽ ﮐﺎرﺑﺮد اﯾﻦ ﻣﻮﺿﻮع ﺑﻪ ﻃﻮر ﺧﺎص ﺑﺮاي ﭘﯿﺶﺑﯿﻨﯽ ﺗﻘﺎﺿﺎي ﺑﺎر اﻟﮑﺘﺮﯾﮑﯽ در ﺷﺒﮑﻪ ﻗﺪرت از ﺗﺤﻘﯿﻘﺎت ﭘﯿﺸﯿﻦ ﻣﺘﻤﺎﯾﺰ ﺷﺪه اﺳﺖ. ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺗﻤﺮﮐﺰ اﺻﻠﯽ ﭘﮋوﻫﺶ ﺣﺎﺿﺮ ﺑﺮ ﺗﺎﺛﯿﺮ روشﻫﺎي ادﻏﺎم، در اﯾﻦ ﻣﻘﺎﻟﻪ از روشﻫﺎي ﮐﻼﺳﯿﮏ ﭘﯿﺶﺑﯿﻨﯽ ﻧﻈﯿﺮ روش ﺧﻮدﻫﻤﺒﺴﺘﻪ ـ ﻣﯿﺎﻧﮕﯿﻦ ﻣﺘﺤﺮك ﯾﮑﭙﺎرﭼﻪ )ARIMA( و ﻫﻤﻮارﺳﺎزي ﻧﻤﺎﯾﯽ اﺳﺘﻔﺎده ﺷﺪه اﺳﺖ. در اﯾﻦ ﻣﻄﺎﻟﻌﻪ، ﭘﯿﺶﺑﯿﻨﯽ ﺳﻠﺴﻠﻪ-ﻣﺮاﺗﺒﯽ ﮐﻮﺗﺎهﻣﺪت و ﺑﻠﻨﺪﻣﺪت ﺑﺮاي دادهﻫﺎي واﻗﻌﯽ ﺑﺎزار ﺑﺮق اﺳﺘﺮاﻟﯿﺎ اﻧﺠﺎم ﮔﺮﻓﺘﻪ اﺳﺖ. ﻧﺘﺎﯾﺞ ﺑﻪ روﺷﻨﯽ ﻧﺸﺎن ﻣﯽ دﻫﻨﺪ ﮐﻪ روش ﺗﺮﮐﯿﺐ ﺑﻬﯿﻨﻪ ﺑﻬﺘﺮﯾﻦ ﻧﺘﯿﺠﻪ را اراﺋﻪ ﻣﯽدﻫﺪ.

To forecast the electricity load of a city or country and facilitate the strategic decision-making, it is common to collect the historical data from different zones of the city or different cities of the country. However, normally all the zones or different sectors’ load (residential, industrial, and commercial) are not important equally. In other words, a certain zone or a sector may have the most effect on decision making. Therefore, the simple algebraic sum of the different zones’ forecasting may not be meaningful for the ultimate objective. There are different methods for aggregation of the different zones’ forecasts. The most convenient method is the simple algebraic sum of the different zones’ forecasts, which is not only inefficient but also needs more details about the effective factors on the electricity demand in each zone. In this paper, different aggregation approaches such as bottom-up, top-down, optimal combination methods are presented. It should be mentioned that any research paper in the field of the electrical power system and load forecasting have not studied the hierarchical forecasting; therefore, presenting the hierarchical method for load forecasting is a strict innovation of this paper. The Auto-Regressive Integrated Moving Average (ARIMA) and Exponential Smoothing methods are embedded in proposed aggregation approaches. The proposed methods are applied to forecast Australian electric load in short-term and long-term horizons.