Opposite Normalized Trust-Region Reflective (ONTRR): A New Algorithm for Parameter Extraction of Single, Double and Triple Diode Solar Cell Models

• Published in: Applied sciences Vol. 13; no. 14; p. 8199
• Main Authors: Gao, Xiankun, Hou, Qingwen, Yao, Sen, Zhou, Kaiyuan
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.07.2023
• Subjects: Algorithms; Analysis; Diodes; Methods; min-max normalization; opposition-based learning; parameter extraction; Photovoltaic cells; solar cell model; trust-region reflective
• ISSN: 2076-3417; 2076-3417
• DOI: 10.3390/app13148199

To accurately, efficiently and reliably extract the parameters of single, double and triple diode solar cell models, this paper proposes a randomly initialized opposite normalized trust-region reflective (ONTRR) algorithm. The novelty of ONTRR lies primarily in two amendments to the standard TRR search. (1) Random opposite initialization is added to decrease the initial point sensitivity of TRR and thus reduce the possibility of being trapped in local optima. (2) Min-max normalization is embedded to eliminate the negative effects arising from different magnitudes of model parameter values and thus drive the derivative-dependent TRR search in an efficient manner. The proposed ONTRR algorithm is evaluated and compared to other state-of-the-art algorithms using four benchmarked I–V datasets with two commonly used objective functions. To be objective and reproducible, the comparative experiments are carried out with default random seeds for 1000 independent runs instead of the 30, 50, or 100 runs used in most studies in the literature. The comparison results demonstrate that for all 12 of the test cases, the proposed ONTRR algorithm consistently achieves the highest accuracy with the least computational effort, and is often superior to the best-performing algorithms reported in the literature in terms of convergence speed, average accuracy and statistical stability.