A deep-learning-based approach combines deep learning models with probability density estimation methods for day-ahead net load forecasts.