Daily Sea Water Temperature Forecasting Using Machine Learning Approaches

OpenAlex English

The efficiency of turbines in seaside nuclear or coal-fired power plants is directly proportional to sea water temperature (SWT). The cooling medium temperature is critical in the design of any power plant when considering long-term average climatic conditions. As a result, the deviation in the SWT affects the efficiency of electricity generation. Accurate SWT estimation is critical for electrical output from power plant applications in this regard. Three different data-driven models such as long short-term memory (LSTM) neural network, adaptive neuro-fuzzy inference system (ANFIS) with fuzzy c-means (FCM) and grid partition (GP) were used to perform one-day ahead short-term SWT prediction, in this paper. The analyses were performed using 5-year daily mean SWTs measured by the Turkish State Meteorological Service in Canakkale Province between 2014 and 2018. The measured data was also used to validate the data produced by the proposed techniques. Performance criteria for the techniques suggested are mean absolute error (MAE), root mean square error (RMSE) and correlation coefficient (R). With the ANFIS-FCM technique, the best outcomes for MAE, RMSE and R values were obtained as 0.113oC, 0.191oC, and 0.9994, respectively, according to daily SWT forecasting.