Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Μοντελοποίηση φωτοβολταϊκών στοιχείων με τεχνικές μηχανικής μάθησης

Perraki Maria

Full record

---

Summary

Predicting the output power that a photovoltaic system can produce isof major importance for evaluating its efficiency, but also for predicting itslong-term operation.Plenty of equivalent-circuit models can be used for the evaluation of aphotovoltaic system’s electrical behavior. In the present work, the singlediode model (SDM) and the double diode model (DDM) are studied underdifferent levels of solar irradiation and temperature. The use of exclusivelyone of the two models to model the photovoltaic system, limits the accuracyof its performance prediction. For this reason, targeting to improve thedesign methods of photovoltaic systems, the combination of the single andthe double diode model, with the classification algorithms in the process ofthe machine learning framework, is evaluated. The purpose is to predict theoutput power of a system, under different climatic conditions with higheraccuracy.The machine learning classification algorithms that are implemented toidentify which model among the single and the double diode model, providesa more accurate estimation of the output power, for given values of solarirradiation and temperature are: classification trees, discriminant analysis, knearest neighbors algorithm, Naive Bayes algorithm, support vector machinesand classification ensembles algorithms.In the present work, two geographical areas are examined. The first ischaracterized by the Mediterranean climate and is located in southern Italy,while the second one by the temperate oceanic climate and is located innorthern Germany.The implementation of both single- and double-diode models and classification algorithms was done using the MATLAB environment.

Available Files

Services