Accurate forecasting of renewable energy resources and load has a crucial role in the overall operation efficiency and energy system integration of microgrids. In addition to this, in comparison with conventional power systems, the behaviour of microgrids loads presents higher frequency changes, which means greater volatility and higher uncertainty. In order to improve the robustness of microgrid energy management, and define through two different prediction techniques the best model for load forecasting, this paper provides a substantial review of theoretical Short Term forecasting methodologies, specifically Artificial Neural Network and ARIMA model, for microgrids loads. Using data from a real microgrid, the ANN model demonstrated a better performance than the ARIMA model in the forecasting results evaluated through specific metrics such as RMSE or MAE.