Fiber Bragg grating sensor-based temperature monitoring of solar photovoltaic panels using machine learning algorithms

Abstract

Fiber Bragg Grating (FBG) sensors are an emerging and prominent optical sensing technology of accurately measuring strain, depth, temperature, density, and several physical parameters. Due to high solar radiation, the increased solar panel temperature affects photovoltaic cell efficiency. Hence, monitoring the temperature of solar panels and providing proper cooling is essential to attain optimal electrical performance. FBG sensor is used to monitor the solar panel temperature in this research. The accuracy and stability of the peak search algorithms in the acquired experimental data are analyzed. Reisz fractional-order derivative and Savitsky-Golay filter are improved using a decision tree regressor to determine the peak and noisy spectrum with different signal-to-noise ratios. The algorithm surpassed the demerits of conventional peak algorithms used for the FBG spectrum. This work implements the fractional derivatives and a machine learning algorithm initially denoised. The number of epochs for linear regression is 15. The depth allowed for decision trees is 1000, and the number of estimators for the random forest is 150. A peak is detected with a decision tree regressor trained on clean and noiseless FBG data. Machine learning algorithms such as linear regression, random forest, and decision tree are compared. The decision tree regressor with hyperparameter tuning yields the best results with high accuracy of 99.83% compared to other conventional peak detection methods. The statistical tests and metaheuristic algorithms are also performed to assess the experimental outcome.