Low Voltage (LV) distribution networks with high penetration levels of photovoltaics have to tackle various challenges such as overvoltages, voltage fluctuations, reverse power flows, and non-coincident demand and local generation. Energy Storage Systems (ESS) can help to ease these issues, if sized properly. This paper proposes a two-step methodology for sizing centralised ESS in LV networks. In the first step, a reoccurring daily pattern is detected using symbolic aggregated approximation (SAX) from the data measured at a German grid. Afterwards, high- and low-frequency components of the power signal are separated using a low-pass filter and then used for sizing different types of ESS. The effect of data resolution on the sizing outcomes is also investigated. The performance of the method was investigated using the full data set. It is concluded that ESS with the characteristics derived using this methodology can effectively be used for peak shaving, power smoothing and load balancing.