A data analytics approach to evaluating a Friday effect on electrical demand daily peak in Gran Buenos Aires area during winter

Load forecasting represents a crucial topic for the power grid industry. Several activities such as grid expansion planning, generation dispatch and maintenance scheduling depend to a great extent on load behaviour prediction.To fulfill this wide range of applications, several techniques were and are currently being developed covering short-term, medium-term and long-term load forecasting.1-week-ahead and 2-week-ahead load forecasting are typically required for unit commitment and outage scheduling. For these time intervals, temperature and day type (business day / weekend / holiday) are identified as the key explanatory variables.Fridays are often considered to be a regular business day with a somewhat lower electrical demand peak. Should that statement prove to be true and relevant, a ‘Friday effect' would be identified and Friday would be considered as a day type per se. Therefore, decision making processes based on demand forecasting would have to be adapted to account for this effect.Otherwise, should the effect be non-existent or existent but non-relevant, the somewhat lower electrical demand peak consideration for Fridays should be discarded.In this paper, a data analytics approach is used to assess the existence and the degree of a Friday effect for winters comprised in the 2010-2019 period.The analysis shows the existence of a Friday effect with an expected 3.2% difference on the demand daily peak of a non-Friday when compared to its correspondent temperature-similar Friday.The statistical significance is quite high with p-values being far less than 0.001 and test assumptions were verified successfully.However, the impact of the effect is moderate and thus its relevance has to be evaluated for each application on a case- by-case basis.