Demand forecasting in adverse scenarios, such as the COVID-19 pandemic, is essential to ensure the supply of electricity and the functioning of basic services in a metropolitan region. In this work, a deep learning model is proposed to predict demand for the next day, using the “IEEE DataPort Competition Day-Ahead Electricity Demand Forecasting: Post-COVID Paradigm” competition database. Two deep neural network architectures are proposed, one convolutional and one recurrent network. The performance of the proposed models was compared with models developed in the competition, through a benchmark analysis. The best result achieved was using the recurrent network, obtaining an absolute average error of 2530.15 kW, surpassing the first place in the competition.