Abstract:
|
The presence of measurement errors (noise) in the data and model uncertainties degrade the quality of fault detection (FD) techniques. In addition, most engineering and environmental process data generally have multiscale properties, signifying that they include features and noise occurring at different contributions over both time and frequency. Nevertheless, the majority of FD approaches are based on time-domain data (operating on a single time scale), and thus they do not take into consideration the multiscale characteristics of the data. Multiscale representation of data using wavelets, which is a powerful feature extraction tool, has demonstrated a good capacity to efficiently separate deterministic and stochastic features. Thus, combining the advantages of multiscale representation with those of univariate monitoring statistics, such as exponentially-weighted moving average, should provide even further improvements in FD. Towards this end, a framework merging the benefits of multiscale representation of data and those of the exponentially-weighted moving average scheme to suitably detect faults is proposed and used in the context of fault detection in photovoltaic systems.
|