The invention provides a method and a system for detecting a fault of a photovoltaic module. The method comprises a step of comparing whether a difference between a residual of a sample to be detectedof a photovoltaic component and a mean value of samples of a historical data sample set exceeds a threshold for allowing an error or not by a pre-built fault detection function, a step of determiningthat the photovoltaic module fails if so, otherwise, the photovoltaic module runs normally, wherein the samples in the historical data sample set include historical data under normal operation, the historical data includes short circuit current, open circuit voltage, maximum power point current and maximum power point voltage, and the fault detection function is constructed by fitting the relationship among the maximum power point voltage, historical short-circuit current, historical open circuit voltage, and a historical maximum power point. According to the method and system, the fault detection of the photovoltaic component can be achieved through current and voltage changes, and the invention has higher diagnostic reliability than an artificial neural network method under a small sample condition.