The authors through this work have shown the usefulness of state-space models to adjust and forecast daily time series, and the technique of periodic cubic spline regression to model annual seasonality. A structural model is used to analyse the series of daily average demand of electricity in Argentina. This model considers the trend, the weekly and annual seasonal component, the effect of public holidays, two cycles, and the temperature as explanatory variable. The method gave satisfactory results, both at the adjustment level as well as in the forecasting and interpretability of its components. Alternative methods are recommended when the future temperature values are unknown.