ﭘﯿﺶﺑﯿﻨﯽ دﻣﺎي ﺧﺎك و رﻃﻮﺑﺖ ﻧﺴﺒﯽ ﻫﻮاي داﺧﻠﯽ ﮔﻠﺨﺎﻧﻪ ﻧﯿﻤﻪﺧﻮرﺷﯿﺪي ﻣﺠﻬﺰ ﺑﻪ دﯾﻮاره ﺳﯿﻤﺎﻧﯽ ﺷﻤﺎﻟﯽ ﺑﺎ اﺳﺘﻔﺎده از ﺷﺒﮑﻪﻫﺎي ﻋﺼﺒﯽ ﻣﺼﻨﻮﻋﯽ ﻣﻄﺎﻟﻌﻪ ﻣﻮردي: ﺷﻬﺮﺳﺘﺎن ﺗﺒﺮﯾﺰ

Prediction of Soil Temperature and Inside air Humidity in a Semi-Solar Greenhouse Equipped with Cement North Wall by Artificial Neural Network; Case study: Tabriz city

ﺑﻮمﺷﻨﺎﺳﯽ در ﺑﺮ ﮔﯿﺮﻧﺪه ﺑﺴﯿﺎري از ﺟﻨﺒﻪﻫﺎي ﻣﻬﻢ ﮐﺸﺎورزي اﺳﺖ ﮐﻪ ﻣﺘﺎﺳﻔﺎﻧﻪ ﺗﺎ اﻻن ﺑﻪ رﺷﺪ ﻗﺎﺑﻞ ﻣﻼﺣﻈﻪاي دﺳﺖ ﻧﯿﺎﻓﺘﻪ اﺳﺖ. در اﯾﻦ ﺷﺎﺧﻪ، ﮔﻠﺨﺎﻧﻪﻫﺎي ﮐﺸﺎورزي ﺑﻪﻋﻨﻮان ﯾﮑﯽ از ﻣﻮﺛﺮﺗﺮﯾﻦ روشﻫﺎي ﮐﺸﺖ، ﻣﻮرد ﺗﻮﺟﻪ ﺑﺴﯿﺎر اﺳﺖ. دﺳﺖﯾﺎﺑﯽ ﺑﻪ ﮔﻠﺨﺎﻧﻪﻫﺎي ﻫﻮﺷﻤﻨﺪ ﺑﺎ ﺗﻮان درﯾﺎﻓﺖ ﺑﯿﺶﺗﺮﯾﻦ ﻣﻘﺪار ﺗﺎﺑﺶ ﺧﻮرﺷﯿﺪ و ﮐﻢﺗﺮﯾﻦ ﻣﻘﺪار ﺗﻠﻔﺎت ﺣﺮارﺗﯽ ﻫﻤﯿﺸﻪ ﺑﻪﻋﻨﻮان ﯾﮏ ﻫﺪف اﯾﺪهآل ﻣﻮرد ﺗﻮﺟﻪ ﻣﺤﻘﻘﺎن اﺳﺖ. در اﯾﻦ ﺗﺤﻘﯿﻖ از ﻣﺪل ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﻣﺼﻨﻮﻋﯽ ﺑﻪﻣﻨﻈﻮر ﺗﺨﻤﯿﻦ دﻣﺎي ﺧﺎك و رﻃﻮﺑﺖ ﻧﺴﺒﯽ ﻫﻮاي داﺧﻠﯽ ﮔﻠﺨﺎﻧﻪ ﻧﯿﻤﻪﺧﻮرﺷﯿﺪي ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺑﺮﺧﯽ ﭘﺎراﻣﺘﺮﻫﺎي داﺧﻠﯽ و ﺧﺎرﺟﯽ اﺳﺘﻔﺎده ﺷﺪ. ﺑﺪﯾﻦ ﻣﻨﻈﻮر، دادهﻫﺎ از ﮔﻠﺨﺎﻧﻪاي ﻧﯿﻤﻪﺧﻮرﺷﯿﺪي واﻗﻊ در اﯾﺴﺘﮕﺎه ﺗﺤﻘﯿﻘﺎﺗﯽ داﻧﺸﮑﺪه ﮐﺸﺎورزي داﻧﺸﮕﺎه ﺗﺒﺮﯾﺰ، ﺗﻮﺳﻂ ﺣﺴﮕﺮﻫﺎي دﻣﺎ، رﻃﻮﺑﺖ، ﺗﺎﺑﺶﺳﻨﺞ و ﺑﺎدﺳﻨﺞ اﺧﺬ ﮔﺮدﯾﺪ. اﯾﻦ ﮔﻠﺨﺎﻧﻪ ﺑﺎﺗﻮﺟﻪ ﺑﻪ ﺷﮑﻞﻫﺎي راﯾﺞ در ﻣﻨﻄﻘﻪ و ﺑﺮ اﺳﺎس درﯾﺎﻓﺖ ﺑﯿﺶﺗﺮﯾﻦ ﻣﯿﺰان ﺗﺎﺑﺶ و ﮐﻢﺗﺮﯾﻦ ﻣﻘﺪار ﺗﻠﻔﺎت ﺣﺮارﺗﯽ ﺳﺎﺧﺘﻪ ﺷﺪه و در آن از ﺑﺮﺧﯽ وﯾﮋﮔﯽﻫﺎي ﯾﮏ ﮔﻠﺨﺎﻧﻪ ﺧﻮرﺷﯿﺪي اﺳﺘﻔﺎده ﺷﺪه اﺳﺖ. ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﭘﺮﺳﭙﺘﺮون ﺗﮏ ﻻﯾﻪ ﺑﺎ اﻟﮕﻮرﯾﺘﻢ آﻣﻮزش ﻟﻮﻧﺒﺮگ-ﻣﺎرﮐﻮارت (LM) و ﺗﻌﺪاد ﻧﺮونﻫﺎي ﻣﺨﺘﻠﻒ در ﻻﯾﻪ ﭘﻨﻬﺎن ﻣﻮرد اﺳﺘﻔﺎده ﻗﺮار ﮔﺮﻓﺖ. ﻧﺘﺎﯾﺞ ﻧﺸﺎن دادﻧﺪ ﮐﻪ ﺳﺎﺧﺘﺎرﻫﺎي 4-9-1 و 5-6-1 ﺑﻪ ﺗﺮﺗﯿﺐ ﺑﺎ ﺿﺮاﯾﺐ ﺗﺒﯿﯿﻦ 0/9945 و 0/9971 ﻗﺎدرﻧﺪ ﺑﻪ ﺑﻬﺘﺮﯾﻦ ﻧﺤﻮ ﺑﺮاي ﭘﯿﺶﺑﯿﻨﯽ دﻣﺎي ﺧﺎك و رﻃﻮﺑﺖ ﻧﺴﺒﯽ ﻫﻮاي داﺧﻞ ﮔﻠﺨﺎﻧﻪ ﻧﯿﻤﻪﺧﻮرﺷﯿﺪي ﻣﻮرد اﺳﺘﻔﺎده ﻗﺮار ﮔﯿﺮﻧﺪ. ﺑﺮاي ﺳﺎﺧﺘﺎرﻫﺎي ﻓﻮق، ﭘﺎراﻣﺘﺮﻫﺎي MAPE و RMSE ﺑﻪ ﺗﺮﺗﯿﺐ ﻣﻌﺎدل 1/1060 و 1/1956 درﺻﺪ، 1/0353 و 0/2502 درﺟﻪ ﺳﺎﻧﺘﯽﮔﺮاد ﻣﺤﺎﺳﺒﻪ ﺷﺪﻧﺪ. ﻧﺘﺎﯾﺞ اﺳﺘﻔﺎده از ﻣﺪل رﮔﺮﺳﯿﻮﻧﯽ ﻧﺸﺎن دادﻧﺪ ﮐﻪ ﺷﺒﮑﻪ ﻋﺼﺒﯽ اﺑﺰاري ﻗﺪرﺗﻤﻨﺪ و دﻗﯿﻖ در ﻣﻘﺎﯾﺴﻪ ﺑﺎ ﻣﺪلﻫﺎي رﯾﺎﺿﯽ اﺳﺖ. ﻧﺘﺎﯾﺞ ﮐﻠﯽ ﺗﺤﻘﯿﻖ ﻧﺸﺎن دادﻧﺪ ﮐﻪ اﺑﺰار ﺷﺒﮑﻪﻫﺎي ﻋﺼﺒﯽ ﻣﺼﻨﻮﻋﯽ ﻣﯽﺗﻮاﻧﺪ ﺑﺎ اﻃﻤﯿﻨﺎن ﺑﺎﻻ در ﺑﺤﺚ ﮐﻨﺘﺮل ﺧﻮدﮐﺎر ﮔﻠﺨﺎﻧﻪ ﺑﺎ ﻫﺰﯾﻨﻪ ﮐﻢﺗﺮ از ﻃﺮﯾﻖ ﮐﺎﻫﺶ اﺳﺘﻔﺎده از ﺣﺴﮕﺮﻫﺎ در آﯾﻨﺪه ﻣﻮرد اﺳﺘﻔﺎده ﻗﺮار ﮔﯿﺮد.

The ecological domain includes some aspects of agriculture that has not a good development until now. Commercial greenhouse is one of the most effective cultivation methods need to more attentions. Researches are on intelligent greenhouses with some important aspects such as receiving the maximum solar radiation and having the minimum heat loss. In recent study, the application of Artificial Neural Network models to predict inside air humidity and soil temperature beside on some inside and outside parameters were investigated. For this purpose, data was recorded from a semi-solar greenhouse located at agricultural faculty of Tabriz University using several temperature and humidity sensors, solar meter and hot wire. This greenhouse has the best structure between all typical kinds of greenhouse and furthermore has the best situation about heat lost and gain the most solar radiation between all types of greenhouses in this region. Single layer multilayer perceptron (MLP) models with LM learning rule and different neurons in hidden layer were used. The results showed that 4-9-1 and 5-6-1 topology with R2=0.9945 and 0.9971have the highest precision to predict the soil temperature and inside air humidity in semi-solar greenhouse. For theses topologies, MAPE and RMSE were 1.1060, 1.1956% and 1.0353, 0.2502°C, respectively. Comparison between the results of ANN and multiple linear regression (MLR) models showed that ANN is more powerful than mathematical models in this subject. Also results showed that Artificial Neural Network can be used to control automatically the greenhouse environmental parameters with minimum cost in the future.