Fault detection for large scale indoor distributed antenna system based on time series similarity

With the advancement of the fifth-generation (5G) mobile communication networks, the number of subscribers in the interior environment continues to grow. The large-scale indoor distributed antenna system (DAS) is one of the critical approaches for bringing macro base station signals indoors. As the DAS becomes larger and the composition becomes more and more complex, the probability of system failure gradually increases. Therefore, it is very important to detect the failure of the DAS. Through actual research, limited by the user’s usage pattern, distribution, and regional functions, the daily power slave data of the room distribution system has a certain periodicity and similarity, but when a fault occurs, it will break this rule, and then be detected. However, the similarity and periodicity of the data are also affected by the randomness of users, which brings difficulties to fault detection. This paper will use the fault detection method based on N-dimensional Euclidean distance to mine the anomalies in the DAS detection data, and then carry out fault detection. To solve the influence of user randomness on the detection results, this paper will introduce a sliding window and a selection window. Although the filtering reduces the timeliness, it greatly reduces the false alarm rate. Finally, the simulation data and real data at DAS will be used to verify the method proposed in this paper.