Studies have revealed the potential for significant energy savings in buildings by using smart predictive control systems, rather than traditional reactive-based control systems. At present, heating, ventilation and air-conditioning (HVAC) accounts for up to 50 % of building energy consumption. Moreover, analysis of HVAC systems also shows that considerable energy savings can be made by adopting automated fault detection and diagnostics (FDD). The EU-funded EINSTEIN project used FDD to develop new control strategies for use with simulation-based building control systems. The system developed not only eradicates errors between ‘as-built’ and ‘as-designed’ conditions but also uses predictive analysis to determine how the building will perform by considering impact of weather conditions and occupant use. Strategies developed for building control Project partners developed three control strategies related to FDD: building performance prediction, optimisation, and fault detection. “While FDD is generally the first step in “correcting” issues in a building, the second and third algorithms form part of what is described as a Model Predictive Control (MPC) solution,” says project coordinator Dr Ruth Kerrigan. “The MPC essentially predicts and dynamically optimises the building performance beyond typical rule-based state-of-the-art control.” Control strategies were tested and refined using a range of demonstration sites, involving residential buildings and commercial offices in Ireland and Scotland. Researchers also tested an open loop system for fault detection, signal-based fault detection for heating equipment in a typical residential building, and rule-based detection testing on real data. According to Dr Kerrigan: “All tests resulted in improved building performance in terms of energy consumption and/ or cost, while critically maintaining user comfort.” Algorithms created Researchers also created and tested FDD and MPC algorithms to determine the benefits they offered over traditional building control strategies. The algorithms developed by EINSTEIN were tested on data acquired from real buildings for FDD, as well as Advanced Calibrated Models (ACMs) based on existing buildings for MPC. “The ACMs are dynamic building simulation models calibrated to ensure the models accurately influence the real performance of the building they represent,” explains Dr Kerrigan. The MPC algorithms developed for the models resulted in energy savings in the range of 15-17 %, with energy cost savings associated with MPC algorithms ranging from 35-40 %. “Although it is difficult to associate a definite energy and cost saving with the implementation of FDD algorithms, all the tests successfully resulted in the automated identification of faults,” says Dr Kerrigan. “This leads to the avoidance of energy and cost waste and is likely to reduce the possibility of user discomfort due to mechanical faults.” Performance gap closed Advantages associated with operational models for FDD and MPC include the ability to effectively monitor and diagnose discrepancies between design intent and operational performance (often referred to as the ‘performance gap’). They are also adaptable to changes in building or system operation (compared to solely data-driven approaches). Dr Kerrigan states: “In addition, the models are capable of simulating different control scenarios, recognising actual system response and allowing optimisation of control strategies using real performance feedback and real weather data.” EINSTEIN will benefit building managers by providing them with better and more efficient building management on a continuous basis as well as greater flexibility for control through building management systems. “The project will also contribute to future research by forging partnerships between industry and academia, developing a prototype smart building control framework and applying performance prediction and control optimisation,” Dr Kerrigan points out.
EINSTEIN, energy efficiency, automated fault detection and diagnostics (FDD), Model Predictive Control (MPC), Advanced Calibrated Models (ACMs)