Building system with probabilistic forecasting using a recurrent neural network sequence to sequence model

A method of probabilistic building data point forecasting, the method comprising:

• receiving, by a processing circuit, first building data for a building data point of a building;
• generating, by the processing circuit, training data based on the first building data, the training data comprising a probability distribution sequence comprising a first probability distribution for the building data point at each of a plurality of time-steps;
• training, by the processing circuit, a prediction model based on the training data;
• receiving, by the processing circuit, second building data for the building data point;
• and
• predicting, by the processing circuit, for one or more time-steps into the future, one or more second probability distributions with the prediction model based on the second building data, each of the one or more second probability distributions being a probability distribution for the building data point at one of the one or more time-steps.

A building management system, the building management system comprising a processing circuit configured to:

• receive building data for a building data point of a building;
• and
• predict, for one or more time-steps into the future, one or more probability distributions with the building data based on the prediction model, each of the one or more probability distributions being a probability distribution for the building data point at one of the one or more time-steps;
• wherein the prediction model is a model trained with training data, the training data comprising a probability distribution sequence comprising a particular probability distribution for the building data point at each of a plurality of time-steps.