Nowadays, methods and tools for reliable and fast estimation of grid voltage distortions are in great demand for different applications such as relay protection, grid power inverters; variable frequency drives during the voltage dips etc. The paper considers a novel digital algorithm for detecting and identifying the type of asymmetrical faults in a three-phase network on the basis of iterative determination of parameters of the voltage vector trajectory in a stationary reference frame with a consequent calculation of positive and negative components through estimated parameters of ellipse, i.e. major and minor axes, inclination angle. Identification of the fault is carried out by comparison of obtained results with available set of data concerning characterization and classification of symmetrical and asymmetrical faults of different types. The idea of proposed algorithm is explained in details for the case of asymmetrical three-phase voltage system with a sinusoidal waveform of the voltages. It is shown that a set of only three voltage samples is sufficient to detect the fault type, therefore the algorithm allows for revealing the mains voltage distortions nearly instantly. At the presence of high order harmonics and noisy components in a voltage waveform, the algorithm has been modified through averaging the trajectories of a generalized voltage vector using several sets of samplings with consequent minimization of a cumulative deviation error. The algorithm was implemented in MATLAB/Simulink environment and its performance was verified for sinusoidal and distorted signals. Results of simulation have demonstrated a number of advantages of proposed algorithm as compared to algorithms based on the Fast Fourier Transform and the instantaneous symmetrical components theory. The proposed algorithm contains simple calculations, and therefore may be considered as a suitable tool for real-time digital systems.