Özet:

The temperature and humidity parameters of the weather influence various fields, including living conditions, agriculture, and transportation. Hence, accurately predicting the future values of these parameters is important. In this study, temperature and humidity forecasts were made using deep learning techniques, specifically LSTM algorithms, through a model system created for the Süleymanpaşa district of Tekirdağ province. Temperature and humidity datasets were obtained from the Meteorology Provincial Directorate and integrated with data from multiple sensors to mitigate errors caused by noise in single-sensor data. Temperature and humidity data from the Tekirdağ Meteorology Provincial Directorate between 2015 and 2021 were merged with the 2020 temperature and humidity data obtained from the model system to create a fused dataset. Using this dataset, temperature and humidity data for the year 2022 were predicted using deep learning algorithms. Long Short-Term Memory (LSTM) algorithms were utilized for sequentially ordered data over time. The predicted data were then compared with actual data from the Tekirdağ Meteorology Provincial Directorate for the year 2022. The success metrics for these predictions were calculated as RMSE of 1.895, MSE of 3.547, an R-squared score of 0.952, and an MAE of 1.614. The results suggest that this algorithm can be employed for sequentially ordered data over time. The model system developed is based on PLC and SCADA technology.

Anahtar Kelimeler: