Integration of machine learning techniques and various empirical models for showing the impact of tilt angle optimisation

In the present scenario the solar energy acts as the best alternative for conventional energy. However, a hostile photovoltaic (PV) panel environment might result in inaccurate meta-data, which may bring challenges like processing complexity, significant biases, and data quality. Tilt angle and orientation of panels are crucial inputs to know PV efficacy. To gather metadata for distributed PV systems is challenging, resulting in high complexity in control and inspections. This research focuses on how to develop a unique technique for estimating PV system tilt angles, to predict the optimum tilt alignment of PV modules based on seasonal and time variations, the study employs three anisotropic and three isotropic empirical models. Further machine learning models, i.e., Random Forest regression (RF), multilayer perceptron (MLP), and K-nearest neighbour (KNN) are used to validate the empirical analysis, and it has been observed that RF has shown best result with R2 = 0.9644, root mean squared error (RMSE) 385.79.