Publicação
Anomaly detection in photovoltaic systems
| Resumo: | Photovoltaic (PV) solar energy is the fastest-growing renewable source of energy, and poised to become the world’s largest source of electricity by 2050. To maximize efficiency and remain a viable alternative energy source, PV systems should ideally operate seamlessly without anomalies. In reality, however, several kinds of anomalies may occur that prevent PV systems from operating at their full capacity. Here, we address this problem by developing five algorithms for the detection of several PV-system anomalies, and establishing metrics to determine the severity of daytime shading and suboptimal orientation. Specifically, our algorithms are used to detect brief and sustained daytime zero-production, daytime and sunrise/sunset shading, low maximum production and suboptimal orientation. We apply these detection algorithms to several time-series of electricity production, which were obtained for two periods with contrasting weather conditions. When weather conditions were favorable, our algorithms successfully detected the majority of time-series labeled with either sustained or brief daytime zero-production, and either daytime or sunrise/sunset shading. Furthermore, these algorithms also produced a relatively low percentage of false positives, which indicates that most anomaly detections are correct. When weather conditions were adverse, the detection rate of our algorithms was similarly high, if not higher, than when weather conditions were favorable. However, the percentage of false positive anomaly detections is also substantially higher under adverse weather conditions, which indicates that the algorithms are generally more robust under favorable weather conditions. Our results suggest that, on the one hand, daytime shading is a relatively rare anomaly, although it may have a severe impact on PV-system efficiency that warrants its detection. On the other hand, suboptimal orientation appears to be relatively common, and our orientation index can therefore be useful to determine the severity of this prevalent type of anomaly. |
|---|---|
| Autores principais: | Branco, Pedro Miguel Mayer |
| Assunto: | Anomaly detection Photovoltaic systems Renewable energy Time-series |
| Ano: | 2019 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | Photovoltaic (PV) solar energy is the fastest-growing renewable source of energy, and poised to become the world’s largest source of electricity by 2050. To maximize efficiency and remain a viable alternative energy source, PV systems should ideally operate seamlessly without anomalies. In reality, however, several kinds of anomalies may occur that prevent PV systems from operating at their full capacity. Here, we address this problem by developing five algorithms for the detection of several PV-system anomalies, and establishing metrics to determine the severity of daytime shading and suboptimal orientation. Specifically, our algorithms are used to detect brief and sustained daytime zero-production, daytime and sunrise/sunset shading, low maximum production and suboptimal orientation. We apply these detection algorithms to several time-series of electricity production, which were obtained for two periods with contrasting weather conditions. When weather conditions were favorable, our algorithms successfully detected the majority of time-series labeled with either sustained or brief daytime zero-production, and either daytime or sunrise/sunset shading. Furthermore, these algorithms also produced a relatively low percentage of false positives, which indicates that most anomaly detections are correct. When weather conditions were adverse, the detection rate of our algorithms was similarly high, if not higher, than when weather conditions were favorable. However, the percentage of false positive anomaly detections is also substantially higher under adverse weather conditions, which indicates that the algorithms are generally more robust under favorable weather conditions. Our results suggest that, on the one hand, daytime shading is a relatively rare anomaly, although it may have a severe impact on PV-system efficiency that warrants its detection. On the other hand, suboptimal orientation appears to be relatively common, and our orientation index can therefore be useful to determine the severity of this prevalent type of anomaly. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.