Abstarct: Power quality problems such as flicker (voltage fluctuation) are major concerns of electric companies and consumers. Identification of flicker source locations especially in non-radial power systems is important stage in flicker reduction process. In this thesis, some methods are proposed to determine coupling point of flicker sources, separate share of each source in final flicker, and predict non-stationary fluctuation of amplitude. For condition that frequencies of flicker sources are different, a method baxsed on designing especial correlation filter is presented. For condition that flicker sources fluctuate with similar frequency, two approaches are developed. In the first, which needs to train data, k-mean clustering and statistical correlation coefficient are utilized. In the second, which is an analytical one, by assuming flicker sources as independent variable and constructing Jacobian matrix, the criterion is extracted to single point detection. Also, a directed graph is defined to describe propagation of flicker through network and identify dominant source(s), whose dominancy is analytically interpreted. Since the amplitude fluctuation may be non-stationary, a fluctuation index is proposed to determine flicker source location in such condition. Also, by considering discretized envelope signal as a time series, it is modeled and predicted by time series classic models. In addition to these models, grey theory baxsed modes are modified to our purpose. The proposed methods during several scenarios are simulated in a typical non-radial power system. The simulation results show ability and high performance of them in the expected purposes.