The increasing deployment of photovoltaic (PV) systems in both utility and building integrated scales dominates the significance of using an appropriate protection scheme to improve the safety, reliability, and efficiency of the system. One of the main protection challenges of PV arrays is their low fault currents under low-irradiance, low-mismatch, and high-impedance faults. In addition to these conditions, the operation of maximum power point tracking (MPPT) algorithm and use of blocking diodes may lead to the faults within the PV array remain undetected, resulting in potential fire hazards and power loss. This thesis proposes three different fault detection methods that require only the output power of PV arrays. Using the sample entropy–based complexity, the first method quantifies the irregularity of the time series of the normalized fault-imposed components of PV power as the fault detection criterion. Second method consists of two stages; using the amplitude of the normalized super-imposed component of PV array power, the first stage detects a disturbance while the second stage distinguishes a fault condition from a partial shading using a wave-shape based feature. Using the Teager-Kaiser energy operator, third method proposes a novel index to detect the challenging PV fault conditions. The proposed protection schemes are capable of distinguishing the line-to-line, line-to-ground, and open circuit faults from the weather disturbances and partial shadings without the need for the prior information about the PV array and the training data set and are effective for both grid-connected and islanded PV systems. Simulation results in MATLAB/Simulink environment validate the performance of the proposed fault detection schemes.