Statistical analyses for support of SPRS on Sellafield site, UK

The Sellafield Products and Residues Store (SPRS) is a passively-cooled storage building. National Nuclear Laboratory (NNL) has carried out statistical analyses to further develop understanding of how external weather conditions affect SPRS cooling system. The data supplied consisted of around 9.6 million measurements collected from Sellafield weather station and SPRS itself over a period of just over 2 years. This data included various meteorological measurements and data collected from within the building. The technical challenge was to manipulate and analyse the dataset to learn more about the correlations between external and internal conditions, and the periodicity and lags involved. Two main statistical methods were employed; wavelets and cross-correlations. Wavelet analyses were used to a two-fold advantage. First, they were used to compress the data to a size that could be more easily analysed. This was done by cutting out frequencies in the data that were deemed to be noise, to leave a signal that could be represented by summing Daubechies wavelets. Second, they identified frequencies present in the data along with their location, over and above what is possible with a Fourier analysis. Cross-correlations were then used on the de-noised series to identify lags between variables. A change in an external weather condition may result in a change in an internal condition some time later, and cross-correlations helped to quantify this. These methods allowed for an analysis that not only identified correlations between weather conditions and SPRS's internal conditions, but also gave quantitative estimates of the time-frames involved. These technical advances improved NNL's capabilities and gave a better-informed understanding of processes in the industry. This article is followed by the slides of the presentation. (author)