Banca de QUALIFICAÇÃO: ALLAN BRAZ DA SILVA SANTOS

Development of a computational system for anomaly detection and preliminary fault diagnosis in micro-generation photovoltaic systems.

Photovoltaic systems; microgeneration; anomaly detection; preliminary fault diagnosis; inverter data; external meteorological data; EWMA; CUSUM.

Photovoltaic microgeneration systems often operate with limited local instrumentation, relying mainly on inverter telemetry and external meteorological databases. This condition makes it difficult to identify persistent losses, interpret operational anomalies, and prioritize inspections. In this context, this dissertation presents the development of a computational system for anomaly detection and preliminary fault diagnosis in photovoltaic microgeneration systems, based on the integration of inverter operational data, georeferenced external meteorological data, reference modeling, and operational deviation analysis. The proposed approach compares the measured behavior of the plant with a reference estimated from plant registration data, meteorological variables, and physical models of the photovoltaic generator. For this purpose, the time series are consolidated at 15 minute intervals, qualified according to coverage, consistency, stability, and eligibility criteria, and used to calculate relative residuals associated with the electrical variables of the system. Statistical monitoring techniques, based on EWMA and CUSUM, are then applied to these residuals in order to identify persistent deviations from the expected behavior. The resulting signals are grouped into events, for which duration, severity, estimated non-generated energy, and confidence level of the preliminary diagnostic evidence are estimated. The method was implemented as a web application developed in Python and Django, with a multiplant structure, equipment registration, data source integration, diagnostic persistence, visualization dashboards, event review, and reference records. The system was applied to a case study involving a photovoltaic microgeneration plant located in the municipality of Extremoz, Rio Grande do Norte, Brazil, using operational and meteorological data from 2025.