Data-driven methods and their applications to renewable energy

Abstract

Nowadays, an increasing number of household rooftop PV arrays and EVs bring a great challenge to the power grid. To ensure the stability of the power grid, it is necessary to estimate the potential PV power generation and the potential EV charging power consumption based on PV array distribution and vehicle distribution in different areas. However, collecting distribution information on PV arrays and vehicles manually is time-consuming and expensive. To overcome this problem, we proposed novel deep learning-based detectors that can detect the roof-top PV arrays and vehicles in satellite images automatically. The proposed detectors can capture the locations, quantities and sizes of PV arrays and vehicles in various urban scenes. Moreover, a novel framework is proposed to estimate the potential PV power generation and EV charging power consumption according to the detection results in different scenarios. The detectors proposed for PV detection and vehicle detection are built on top of regional-based CNN architectures such as Faster R-CNN and Mask R-CNN. Machine learning techniques and image feature extraction techniques, e.g. support vector machine Fourier descriptors, are also adopted to improve the detection performance. The satellite images used in the experiments are captured from Google Earth covering the complex urban scenes in Manchester. The PV power generation estimation and EV power consumption estimation are carried out by Monte Carlo simulation based on the PV array and vehicle detection results. In summary, this thesis significantly propels low-cost technology development in large-scale renewable energy demand estimation, offering concrete insights into PV array detection, vehicle detection as well as PV power generation and EV charging power consumption estimation. This research contributes significantly to future power grid planning and policymaking, facilitating the transmission from conventional energy to renewable energy.

Date of Award	11 Dec 2023
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Long Zhang (Main Supervisor) & Joaquin Carrasco Gomez (Co Supervisor)