Reliable and efficient detection of faults in photovoltaic systems provides pertinent information for improving their safety and productivity. However, data gathered from photovoltaic systems are generally tainted with a large amount of noise, which can swamp the most relevant features necessary to detect faults, and ultimately degrades fault detection capability of the monitoring system. Therefore, it is crucial to design a robust fault detection approach to deal with the problem of measurement noise in the data. The purpose of this study is to design a robust fault detection scheme to monitor the direct current side of a photovoltaic system and able to deal with the problem of measurement noise in the data by using multiscale representation. 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. Here, multiscale representation of data using wavelets, an efficient feature/noise separation technique, is used to enhance fault detection performance by reducing noise effect and false alarms. First, a simulation model for the monitored photovoltaic array is built. Then residuals from the simulation model are used as the input for the designed scheme for fault detection. A real data from a 9.54 kWp photovoltaic plant in Algiers, Algeria is used to evaluate the effectiveness proposed method. Also, the performance of the proposed chart to that of the conventional exponentially-weighted moving average chart has been compared and found improved sensitivity to faults and robustness to noises.