This paper presents the development of a deep neu-ral network architecture based on stacked LSTM and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$T$</tex> ime2Vec layers for predicting electricity prices several steps ahead (8 hours) to feed future decision-making tools. The proposed model was tested with hourly wholesale electricity price data from Colombia, and the results were compared with some state-of-art time series based statistical forecasting models as SARIMA and Holt-Winters. The results showed that the proposed model outperformed these techniques by modeling nonlinearity and explicitly characterizing the data behavior. Specifically, the pro-posed model was able to capture complex patterns and depen-dencies in the data, resulting in more accurate price predictions. The Time2Vec layer was particularly useful in characterizing the temporal relationships between the input and output variables. The proposed architecture has the potential to significantly improve the accuracy of electricity price predictions, which can have important implications for decision-making in the energy sector.