The generation of electricity demand profiles is a fundamental task to determine the potential flexibility that could be introduced to an electrical system when applying demand side management. In this study we compare the results of two approaches, classical time series analysis and unsupervised clustering, to generate synthetic and anonymized electricity demand profiles. The objective is to retain the statically characteristics and descriptive value of real profiles without compromising company confidentiality. In this case the methods are applied to data obtained from a Chilean paper industry. The results reflect the complexity of the task and show the advantages and disadvantages of both methods.
Keywords demand side management, time series clustering, time series regression, electricity load profiles