A predictive load shifting controller has been developed and deployed in a low-carbon house near Glasgow, UK. The house features an under floor heating system, fed by an air-source heat pump. Based on forecast air temperatures and solar radiation levels, the controller firstly predicts the following day's heating requirements to achieve thermal comfort; secondly, it runs the heat pump during off peak periods to deliver the required heat by pre-charging the under floor heating. Prior to its installation in the building, the controller's operating characteristics were identified using a calibrated building simulation model. The performance of the controller in the house was monitored over four weeks in 2015. The monitored data indicated that the actual thermal performance of the predictive controller was better than that projected using simulation, with better levels of thermal comfort achieved. Indoor air temperatures were between 18 degrees C and 23 degrees C for around 87% of the time between 07:00 and 22:00. However, the performance of the heat pump under load shift control was extremely poor, with the heat being delivered primarily by the unit's auxiliary immersion coil. The paper concludes with a refined version of the controller, that should improve the day-ahead energy predictions and offer greater flexibility in heat pump operation for future field trials. (C) 2017 The Authors. Published by Elsevier Ltd.