This paper presents different mathematical models to integrate solar distributed generation (DG) into very-short term demand forecasting processes in Colombia. The proposed models can forecast up to 4032 5-minute periods (two weeks) of net demand. To do so, historical data, calendar effects and meteorological data have been considered to define mathematical models. Also, different strategies and models like recurrent neural networks, generalized additive models, and ordinary least squares to incorporate the effect of solar DG have been carefully evaluated and compared. Multiple scenarios of solar DG penetration levels, previously elaborated by "The Minning and Energy Planning Unit (UPME)", have been analyzed in order to explore the potential future advantages and drawbacks of the employed models and strategies. Results have shown that our models can forecast 4032 periods of net demand with MAPEs within 1.29% and 3.8%.