|
چکیده
|
This study experimentally investigates a hybrid system comprising a thermal photovoltaic panel, an evacuated tube collector, and a compound parabolic concentrator (CPC). Four distinct scenarios, varying in fluid flow methodology and CPC configuration, were systematically tested over one month. Each scenario was tested for 7 days, with both experimental and theoretical data meticulously recorded. The maximum output temperature recorded during experimental trials was 54.7 ◦C, measured at 13:00 in Case 3. Additionally, sophisticated deep and wide multi-layer perceptron (MLP) and convolutional neural network (CNN) models with complex architectures were developed for modeling, forecasting, and optimizing the system. Various optimizations were performed using these models and the models’ uncertainty was assessed using forecasts derived from July 2023 data. The findings show that both deep and wide MLP and CNN models provide highly accurate predictions, with average R2-scores of 0.9982 and 0.9980, respectively. However, in terms of model uncertainty, the MLP model shows a more favorable performance. Moreover, the CPC system configured in series with the evacuated tube collectors (ETCs) achieves the most optimal performance.
|