Name: CAROLINE TEDESCO SANTOS PASSOS
Publication date: 31/01/2022
Advisor:
Name |
Role |
|---|---|
| RODRIGO RANDOW DE FREITAS | Co-advisor * |
| WANDERLEY CARDOSO CELESTE | Advisor * |
Examining board:
| Name |
Role |
|---|---|
| DANIEL CRUZ CAVALIÉRI | External Examiner * |
| DANIEL JOSÉ CUSTODIO COURA | Internal Examiner * |
| HELDER ROBERTO DE OLIVEIRA ROCHA | External Examiner * |
| WANDERLEY CARDOSO CELESTE | Advisor * |
Summary: Brazil has a high rate of solar incidence in all its regions and photovoltaic solar energy is today the fourth most important source of renewable energy in the country, with a great chance of expansion and growth. This makes room for the need to search for solutions to problems related to the operation of such systems. Among these, the so-called atypical photovoltaic string (PV) operating conditions stand out, which can lead to a loss of efficiency in its generation capacity or even severe failures. Such atypical conditions can originate from dynamic effects common to the photovoltaic system, such as shading and accumulated dirt, due to natural degradation of the system over time, or even due to an abrupt failure caused by malfunction of some component of the system. Regardless of the cause of the atypical string condition, its effect is somehow registered in the electrical generation itself, with each causative factor causing an electrical signature, which makes it possible to identify it. Thus, the objective of this dissertation is to identify the operating condition of a photovoltaic string among twenty possible conditions, one being normal and the others atypical. In the case of unusual conditions from different sources, the cause of the condition must also be identified. For this, a non-intrusive monitoring method is used, based on the use of electrical voltage and current samples generated by the PV string itself, in addition to the use of different techniques based on Artificial Intelligence (AI) for the development of classifiers. A methodology based on multi-stage classification is adopted in order to divide a larger and more complex problem into fatally smaller and less complex sub-problems. Therefore, two classification stages are considered: the first whose objective is to identify one of five PV string operating conditions, that is, normal, full panel shading, partial panel shading, panel short circuit and line break ( electric arc); and the second stage, whose objective is to identify the PV panel causing an atypical condition. The classifiers used in both stages are based on kNN, SVM and MLP. The results achieved led to an average accuracy of 93.9% when using the classifier with the best performance in each subproblem treated.
