In recent years, a large variety of short-term load forecasting methodologies have been proposed, but a common drawback of them is that require a large amount of historical data to train the model.