A simple and effective detection strategy using double exponential scheme for photovoltaic systems monitoring. (15th January 2021)

• Record Type: Journal Article
• Authors: Taghezouit, Bilal; Harrou, Fouzi; Sun, Ying; Arab, Amar Hadj; Larbes, Cherif
• Is Part Of: Solar energy. Volume 214(2021)
• Page Start: 337
• Page End: 354
• Publication Date: 2021-01-15
• Subjects: Photovoltaic systems -- Empirical models -- Anomaly detection -- Shading -- Electrical faults -- Statistical control charts; Solar energy -- Periodicals; Solar engines -- Periodicals; 621.47
• DOI: 10.1016/j.solener.2020.10.086
• Languages: English
• ISSNs: 0038-092X

Abstract

Highlights: An effective detection method for photovoltaic systems monitoring is presented. Combine empirical models and double exponential smoothing method to detect faults. This monitoring method is validated by using actual data with six types of faults. High performance are achieved via the new method with a nonparametric threshold. Abstract: Effective and efficient monitoring of a photovoltaic plant are indispensable to maintain the generated power at the desired specifications. In this work, a simple and effective monitoring method based on parametric models and the double exponentially smoothing scheme is designed to monitor photovoltaic systems. This method merges the simplicity and flexibility of empirical models and the sensitivity of the double exponentially smoothing strategy to uncover small deviations. Essentially, the empirical models are adopted to obtain residuals to detect and identify occurred faults. Here, a double exponentially smoothing scheme is used to sense faults by examining the generated residuals. Moreover, to extend the flexibility of the double exponentially smoothing approach, a nonparametric detection threshold has been computed via kernel density estimation. Several different scenarios of faults were considered to assess the developed method, including PV string fault, inverter disconnection, circuit breaker faults, partial shading, PV modules short-circuited, and soiling on the PV array. It is showed using real data from a 9.54 kWp photovoltaic system that the considered faults were successfully traced using the developed approach.