Accuracy of solar cell model parameters is now a requirement for the manufacturing of appropriate photovoltaic models. Consequently, the methodologies of the parameters had already attracted immense interest among researchers over the years. This paper deals about the approximation of the characteristics curves of a solar cell using the Bézier curve method. This technique is selected to overcome the problems of the non-linearity, complexity, multivariate and multimodal of the characteristic's curves. The Bézier curves method is a parametric curve used throughout computer graphics and related disciplines. It is a robust methodology for constructing free-form curves, surfaces, shape design, and geometric representation of various products. Six different cases are carried out to test the validity of the proposed method, they are: (i) organic flexible dual junction amorphous silicon solar cell, (ii) silicon solar cell, (iii) Schutten Solar STP6-120/36 polycrystalline photovoltaic module, (iv) Schutten Solar STM6-40/36 monocrystalline photovoltaic module, (v) Photowatt-PWP 201 polycrystalline photovoltaic module and (vi) 230-W multi-silicon photovoltaic module under partial shading conditions. The technique is implemented and adapted by the mean of the De Casteljo algorithm to optimize the control points and minimize the distance between the experimental and calculated points. It is based on subdividing the curve into various segments and controlled by control points using the De Casteljo algorithm to minimize the distance between the experimental and theoretical points. The performance of the approximated characteristics curves is compared with experimental data as well as recent algorithms and techniques from the literature. For more validity, statistical error metrics from literature are computed and compared to ensure the accuracy of the approximated results and the performance of the proposed method. For instance, the comparison is carried out by resorting to the Individual Absolute Error (IAE) and the Relative Error (RE). Further, the Bias metrics and accuracy are calculated using the Autocorrelation Function (ACF), Tracking Signal (TS), and Normalized Forecast Metric (NFM). The proposed method provides a low error for the six case studies and better than that obtained by the other algorithms.
Read full abstract