Data-driven building retrofit using a wireless sensor network

Building retrofits are essential for the mitigation of greenhouse gas emissions in Switzerland. Unfortunately, numerous uncertainties often surround the retrofitting process, which then lead to performance gaps between expected and realized energy and emissions savings. However, there is the potential to increase both energy and emission savings from building retrofitting without increasing the cost. This thesis introduces a retrofit process based on a novel wireless sensor kit to collect in-situ measurements of individual buildings. A low-cost, modular wireless sensor network is introduced, which automatically collects information on thermal state of a building and reduces the need for in-person building audits. The sensor network is able to provide measurement data, from which key parameters of building performance can be derived, such as U-values and heat loss coefficients. In addition, thermal building model parameters can be estimated, which allows the assessment of retrofit measures.The presented wireless sensor network is cheaper and more versatile than other commercial solutions for three main reasons: the software and hardware design files for the sensor network presented in this work are completely open-source, the costs are transparently discussed in detail, and it the capability of the system for continuous measurements without errors is demonstrated. This is particularly important as cheaper and increasingly versatile sensor networks are required to enable large scale deployments and to answer research questions about the required quantity and the properties of sensors for improved retrofit processes.Subsequently, the benefits and challenges of conducting in-situ measurements in occupied buildings are demonstrated by means of case studies. The wireless sensor network has been deployed in occupied single-family houses, and measurements were taken across a period of several months with high temporal resolution. With these measurements, an estimate of the overall heat loss coefficient of residential buildings has been obtained.Following data collection, the impact of measurement uncertainties on the performance of retrofit measures was then examined. The versatile and cheap sensor hardware allowed for sensor-rich deployments, which resulted in rich data sets that allowed quantifying of measurement uncertainties. The results thus show that measurement uncertainties influence the predicted cost and emissions of retrofit measures. However, this does not affect the ranking of retrofitting interventions based on the cost or emissions to a large extent, and only in cases where the sets of retrofit measures perform similarly do rank changes occur due to measurement uncertainties. It was also observed that measurement uncertainties impact the effectiveness of retrofitting interventions in terms of cost and performance due to their impact on the sizing of said retrofit interventions. Several practical challenges have been discovered and documented. The acquisition of heating energy demand turned out to be challenging due to the diverse heating system setups encountered in the case study buildings. In addition, there is a lack of accessible energy metering infrastructure in Switzerland. That being mentioned, the versatile hardware architecture of the WSN made it possible to both make adjustments and integrate new water flowmeter sensors with little additional effort. It was, therefore, possible to measure space heating demand independent from both the heat generation system and domestic hot water use. This also emphasized the importance of uninterrupted measurement data. In summary, the successful determination of thermal building characteristics requires the monitoring of space heating input with few time gaps, among other parameters. This is challenging to implement in any building. The occurrence of time gaps in the measurements is a phenomenon that has been addressed with more reliable hardware components. The amount of lost data was reduced significantly, allowing the introduction of new thermal modeling metho s. The results described in this work are intended for practitioners and researchers to reduce uncertainties in the building retrofit process, increase the planning reliability of retrofits, and thus increase the building retrofit rate.For practitioners, this work contains instructions on how to build and operate a wireless sensor network for building performance assessment. Building parameters can be derived, such as U-value, heat loss coefficient, and heating set-point. In addition, this work outlines a sensor deployment process and highlights practical challenges likely to be encountered in occupied residential buildings. Researchers could use the low-cost sensing hardware to explore research questions in the built environment, which require the deployment of many sensors in one or more buildings. Due to the low cost and the flexible layout, the sensor hardware can be deployed at a large scale and adapted to the specific research. Further, the measurement uncertainty assessment method can be used to quantify the quality of the measured data for the existing task.