Topic: ARIMA and holidays

We assume we try to make the forecast for current consumption on summer residences. There accurate recurrence - the peak fading somewhere in the environment is tracked each 7 days. Plus  on seasons - in the winter as a whole consumption above. All would Seem simply - ARIMA and rejoice. But! Is still  days of type on March, 8th, new year, on February, 23rd which break a 7-day cycle prolonging peak. How in such cases arrive? Throw out holidays from the data? Or it is possible to consider them?