The current work is devoted to the problem of time series analysis. One of the relevant tasks connected with time series is splitting the set of objects into individual groups – clusters for further forecasting the behaviour of time series. A various number of clustering methods exist; this work is focused on the time series clustering method which uses self-organising maps. Classical self-organising maps presume that the input vectors are of the same length, but in the most cases the real data does not satisfy this assumption. The work analyses the problems connected with self-organising maps modification aimed to enable clustering time series of various length. The algorithm that allows accomplishing the described tasks, is not only developed and presented in this work, but is also practically realized. A specialised software solution combining data transformation and clustering algorithms and also practical data analysis procedures are developed. The main result of this work is an algorithm, which allows using self-organising maps for clustering time series of various length as well as the practical use of the algorithm and analysis of the obtained results.