Predicting the Energy Performance of Buildings : A Method using Probabilistic Risk Analysis for Data-driven Decision-support

As most countries worldwide continue to mitigate climate change, one area of focus is a building's energy efficiency. Buildings currently account for approximately 40 per cent of the total energy use in the European Union. One solution for reducing the energy demand from buildings is by continually implementing more stringent building regulations. Consequently, the importance of accurate and efficient building performance simulations to predict the energy performance of a building design increases with more stringent regulation reducing the margin of error. However, a performance gap exists between the predicted energy performance and a building's actual energy performance.This research uses the concept of risk, which consists of probability and consequence, to bridge the knowledge gap and explain and predict the energy performance gap. This research aimed to develop and test a method for using a predictive model to quantify a building design's risk level and evaluate design options. The developed method uses probabilistic risk analysis for quantifying and predicting the energy performance of buildings, resulting in a data-driven decision-support when deciding the building design. The method was developed in steps, presented in several studies. The studies used the research strategy of case studies to test the method and compared the outcome against field measurements and was designed to develop, test, verify and validate the method.One case study used the traditional deterministic approach for quantifying energy performance focused on identifying financially viable renovation packages to passive house level and comparing and optimising different energy performance levels for the building stock. The results showed the potential and limitations of the current method for building performance simulations in evaluating the design space and quantifying the energy performance and costs of different design options.An explorative phase followed, identifying alternative methods and comparing these to the traditional deterministic approach based on the knowledge gained. Finally, qualitative methods were used to evaluate and discern when and for what purpose the alternative probabilistic methods might be advantageous to apply.Based on the outcome of this work, the development of the probabilistic risk analysis method began, resulting in several studies focusing on different aspects of the method. The first study developed and tested an experimental version of the method in a case study, verifying the resulting predictive model against field measurements for the energy performance. The outcome showcased the potential with the developed method and identified several aspects needing further investigation and development. One aspect was regarding the accuracy of the predictive model, requiring further investigation into the data quality used as a basis for the model. The other was how to quantify the consequences of not attaining the design criteria and develop the method to compare design options and support the decision-making process using a data-driven approach. However, the data resolution from the field measurements was coarse, reducing the analytical potential by only comparing the aggregated total.Thus, one purpose with the following case study was to increase the resolution to enable additional analytical comparisons and focused on the data quality. The study evaluated how different levels of data quality impact the predictive model's accuracy. The outcome showcased the importance of data quality and the impact on the predictive model.The subsequent study evaluated how to develop the method to include optimising the building design. The implemented optimisation step compares design options and how the stakeholder viewpoints impact the analysis and models' scope when quantifying the risk. Also evaluated was how the selected design criteria impact a project's risk level. The evaluation illustrates how quantifying the risk level could improve the decision-making process, either when deciding on a building design or the design criterion to use.With the outcomes of this project, the probabilistic methods for quantifying energy performance and consequences of decisions and how to apply them are more accessible to the building sector. The main findings from these studies were the benefits of implementing the probabilistic approach to quantify building designs' energy performance and how this could support the decision-makers during the design process while also clarifying the current limitations to overcome. However, this method adds several new dimensions for data gathering, performing the simulations and analysis, and visualising and conveying the results using different plots. Furthermore, instead of providing a single value based on deterministic approaches for predicting the energy performance of a building design, the probabilistic approach provides a distribution of possible outcomes based on empiric data of uncertainties from which to quantify the probability of failure. Finally, the results also showed the importance and impact of data quality and a structured process for gathering data