Name: Joachim Verhelst


KU Leuven

Promotor / Supervisor

Promotor: Prof. dr. ir. Lieve Helsen

Mede-promotor: Prof. dr. ir. Dirk Saelens

Copromotor: Prof. dr. ir. Geert Van Ham

Samenvatting van het onderzoek / Summary of Research

Historically, HVAC controllers in office buildings use many interlinked, but relatively simple rule-based algorithms based on feedback without predictions (e.g. thermostatic regulation valves). Nowadays, often a plethora of information about the building, building usage and even predictions of future disturbances can be made available to the BEMS. By taking this knowledge into account in monitoring systems without direct control function (fault detection and diagnosis models),  critical problems can be detected an diagnosed. When followed up by an efficient maintenance team, (manual) resolution of the identified technical issues have resulted in energy savings in the range of 10-40% in existing office buildings.

Predictive control is harder to implement than FDDE systems, but the theortical savings potential, compared to traditional HVAC control is estimated in the same range, between 1-41% (cumulative with FFDe-savings!). The savings potential is dependent on the thermal inertion of the building, type and sizing of HVAC-installations, building usage and the climate.

Before the last decade, this additional information was rarely directly used in HVAC control. Controllers based on dynamic (building-)models are in beta-phase with many BEMS-suppliers nowadays. But, model based controllers have proven to be hard to implement in practice, and often fail to fully realise the projected savings. In the worst scenarios, comfort issues arise and the energy usage even increases (up to +20%) compared to traditional control strategies. This can be attributed partially to prediction errors and parameter estimation model mismatch. Also, in existing studies about optimal HVAC control, often a perfect control of subsystems is assumed.

In real building systems, deterioration over time causes imperfect operation of control systems. In HVAC systems, deterioration occurs frequently with sensors and actuators. If these are  not critical, they often are not (or not immediately) overhauled by the maintenance staff. These types of non-critical faults (anomalies) can significantly influence the efficiency of the BEMSs HVAC control.

Further research is therefore advised, to investigate and evaluate the true costs and benefits of these model based, predictive controllers. This PhD thesis provides and implements an evaluation methodology for the determination and prediction of expected controller efficacy of model based HVAC controllers in office buildings. Instead of model mismatch and prediction errors, the focus is placed on the (weighted) impact of  frequent degradation faults on building KPI’s (such as energy and discomfort, investment cost, maintenance cost, replacement costs)

This methodology is applied on simulation results of office-building cases, with variations in thermal building inertia, control types, HVAC systems and types and severities of anomalies . Fault tolerance is a feature, describing the capability to supply a correct service (according to the goals outlined in the design intent), despite occurring disturbances.  The analysis of the results gives insight into the fault tolerance of HVAC controllers against degradation faults.

Volledige tekst van het doctoraat / full text

Examencommissie / Board of examiners

Prof. dr. ir. Lieve Helsen (promotor) 
Prof. dr. ir. Dirk Saelens (mede-promotor) 
Prof. dr. ir. Geert Van Ham (copromotor) 
Prof. dr. ir. Petrus Verbaeten (voorzitter/chairman) 
Prof. dr. ir. Karen Allacker (secretaris/secretary) 
Prof. dr. ir. Geert Deconinck 
Prof. dr. ir. Filip Logist 
Dr.Ir. Mehdi Maasoumy , C3 IoT