Search

There is a growing trend for engaging edge computing to help smart homes to improve the comfort of living for residents. The rapidly rising interest in such deployments, however, does not usually correspond with energy management schemes. In this thesis, we will propose a unified system framework for enabling a sustainable edge computing, while meeting the needs of data processing for smart home applications. Besides, we also designed an edge-based energy management framework for powering smart appliances at home, which aims to enable the use of renewable energy to greatly reduce energy costs for households. To make better use of renewable energy, a modified forecasting approach will be applied in both frameworks, to constantly forecast renewable energy generation. Furthermore, two optimal energy scheduling algorithms for edge devices and one job scheduling algorithm for appliance jobs are also proposed in this thesis. By employing advanced energy disaggregation technique in the edge-based energy management framework, usage pattern and users preference on each appliance can be derived, and thus, each appliance job can be rescheduled based on this information. To evaluate the performance of two edge computing paradigms, two prototype systems have been implemented by using low-cost and easy-to-get hardware. The experiment results demonstrate that renewable energy is completely capable of supporting the reliable running of edge computing devices and smart appliances, and the utilisation of renewable energy increased significantly when our proposed frameworks were employed, compared with when they were not.

The change in width of sea ice cracks seriously affects the transportation during human's activities in the Antarctic sea, but there are few automatic technologies for accurate monitoring of such cracks. To solve this problem, an automatic detecting apparatus for sea ice cracks was designed, and its feasibility and monitoring principles were studied. Furthermore, a practical methodology was developed and such apparatus was applied to the testing of the Antarctic sea ice cracks. The testing results showed that the data were reliable and stable for the analysis of icing behaviour.

When there're clear changes in environmental temperature, the electrode itself of the capacitive ice layer thickness sensor also varies and its capacitance changes, causing mistakes in determining the ice/snow-air interface with the electrode. Based on analysis on relation between electrode capacitance and temperature changes, a differential temperature compensation circuit composed by a thermistor is designed in this paper. Experiments indicate that the electrode voltage is less influenced by temperature changes after adding the compensating circuit, improving the accuracy of the sensor.

Due to wild values or singular values existing in each set of data sampled with the capacitive ice layer thickness detection system, bigger errors always occur in determining ice layer thickness. To improve the accuracy, denoising methods for data based on Segmented Average Value (SAV) and Singular Spectrum Analysis (SSA) are designed. By comparison, the latter is better in term of effect, filtering wild values and singular values and improving the accuracy.