Accurate sea water temperature (SWT) estimation is an important parameter for the efficiency of electricity generation in coastal nuclear power plants and thermal power plants and is also a crucial parameter in determining the optimum conditions for the desalination process. In addition, it is critical for understanding the thermal regimes of oceans and rivers in the context of climate change. Machine-learning methods provide an experimentally based way of forecasting water temperatures with great accuracy. In this paper, 3 different data-driven methods, such as adaptive neuro-fuzzy inference system (ANFIS) accompanied by fuzzy c-means (FCM), ANFIS with grid partition (GP) and ANFIS with subtractive clustering (SC) methods and 3 different deep learning methods such as multilayer perceptron (MLP), long short-term memory (LSTM) and bidirectional long short-term memory (BiLSTM) neural networks were applied to make sea water temperature forecasting using real observed data. In this regard, analyses were conducted using 5-year daily mean sea water temperature values measured by the Turkish State Meteorological Service for Antalya province between 2014 and 2018 years. All of the models were evaluated using mean absolute error (MAE), root means square error (RMSE), and correlation coefficient (R). According to applied models, it was observed that the most sensitive results were obtained with the ANFIS SC model.
Anahtar Kelimeler: Sea water temperature, Data-driven method, Deep learning method
|
