Building management system for forecasting time series values of building variables

A forecasting system configured to forecast time series values in a building management system (BMS), the forecasting system comprising:

• a load/rate predictor configured to use one or more deterministic models trained with historical BMS data to predict a deterministic value for the time series;
• wherein the load/rate predictor is configured to use a stochastic model trained with differences between actual values for the time series and deterministic values predicted by the one or more deterministic models to predict a stochastic value for the time series;
• wherein the load/rate predictor comprises a forecast adjuster configured to adjust the deterministic values predicted by the one or more deterministic models using the stochastic value predicted by the stochastic model to generate an adjusted forecast for the time series.

A method for forecasting time series values in a building management system (BMS), the method comprising:

• using one or more regression models trained with historical BMS data for time series to forecast a deterministic value for the time series;
• using a history of residuals for the time series to forecast a residual for the time series, wherein the residuals are differences between forecasted deterministic values for the time series and actual values for the time series;
• using the forecasted residual for the time series to adjust the forecasted deterministic value.