Abstract/Summary

Approximately 40% of global energy use can be attributed to buildings; in commercial buildings around 20% of the total energy used comes from small power loads. In the UK,this percentage is expected to reach up to 50% in highly efficient offices in the next 20 years. This trend makes small power loads in commercial buildings one of the fastest growing load categories. Quantitative energy audits for the analysis of the energy performance of buildings are conventionally divided into two approaches, calculation, based on algorithms and equations, and measurement,which performs some level of direct monitoring. The sequantitative energy audit approaches are common tools for evaluating the potential for reducing energy demand in buildings. Small power load estimations in office buildings present challenges for both approaches due to the large number of such loads and their heterogeneous nature, and results in significant uncertainty in these estimations. This thesis investigates the sources of uncertainty of the small power energy estimations for the different audit approaches, and proposes and tests a number of methods and techniques to overcome these weaknesses in the auditing process. For the calculation approach, insufficient input parameter specifications have been identified as the main source of uncertainty, which is associated with variability in the model output. A sensitivity analysis method has been developed to identify the inputs that most contribute to such output variability and that require additional effort to strengthen their accuracy in order to minimize the likely error in calculated small power energy consumption. These influential parameters have been found to depend not only on the information sources available, but also on the calculation method used and the type of load estimated. Regarding the measurement approach, its uncertainty is related to the number of meters used, which increases the quality of the information, but also the complexity of the hardware installation. An extrapolation method for providing the relationship between the number of appliances monitored and the accuracy obtained in the final energy estimations has been proposed. Results showed a logarithmic relationship between the number of desks monitored in a case study office and the relative standard uncertainty percentage obtained in the energy estimations for the aggregated load of the PCs. The method informs about the level of metering infrastructure required in accordance with the level of uncertainty that can be accepted for the small power energy estimations. Non-Intrusive Appliance Load Monitoring (NIALM) methods, as a solution for small power individual load estimation in office buildings, have also been explored through a practical study. The disaggregation capabilities for the different electrical signatures, and their dependence on appliance type and number have been investigated. Although the overall accuracy in the disaggregation process was found to be significantly smaller for offices than for domestic scenarios, some signature combinations, such as the Root Meter Square Increments and the Steady Harmonic Increment, were found to achieve up to 90% of accuracy in the disaggregation process. The outcomes from this study contribute to the extension of the use of existing NIALM methods from domestic to office buildings in the field of small power disaggregation.