State Estimators Applied To A White-box Geothermal Borefield Controller Model

Modelling of geothermal borefields for building energy simulations (BES) has always been a complicated task due to the challenge of implementing both their short-term and long-term responses. Besides, in model-based optimal control of geothermal systems, a simplified version of a borefield control-oriented model is desired. Typical prediction horizons used in optimal control of buildings range from hours to a few days, inviting to reduce the complexity of the controller model down to the short-term range. However, the long-term thermal behaviour of the ground is crucial with respect to the heat pump COP and availability of direct cooling. In a white-box controller model, the states keep their physical meaning. Thus, the long-term dynamics can be captured from the model used for dynamic simulation, i.e. the emulator, and updated to the controller model at each optimisation time-step. Nevertheless, since in a real implementation the availability of data is much more limited a state estimator is necessary. In this paper, three state estimators (Stationary KalmanFilter, Time-Varying Kalman Filter and Moving Horizon Estimator) for a linear borehole model are compared using real data from a building combining a geothermal heat pump and a thermally activated building system (hybrid GEOTABS). In general, all investigated linear estimators are capable of accurately estimating the ground states.