Adaptive Fault Localization in Photovoltaic Systems

The photovoltaic (PV) technology has experienced a rapid growth in deployment over the past years. To ensure the performance and safety of the PV system, it is necessary to develop techniques that can efficiently localize faults or failures that occur among its components. In this paper, a fast fault localization algorithm is proposed for PV systems under the sequential change detection framework. In particular, multiple output signals are measured at each string, and the time correlation of the faulty signal and the signal correlation among different meters are exploited by using a vector autoregressive model to describe the fault. A fault localization algorithm is developed based on the generalized local likelihood ratio test which captures any possible patterns of abnormalities at each string. Meanwhile, a selective sampling strategy is implemented at the fusion center to tackle the constraints of communication and processing capacities. Extensive simulation results demonstrate that the proposed method achieves fast detection and high localization accuracy. Moreover, a key feature of the proposed method is that it can adaptively localize different faults without any prior knowledge about possible faulty patterns.