David Greenwood

Scott Angus, Jethro Browell, David Greenwood et al.

Low voltage (LV) distribution transformers face accelerating demand growth while replacement lead times and costs continue to rise, making improved utilisation of existing assets essential. Static and conservative protection devices (PDs) in distribution transformers are inflexible and limit the available headroom of the transformer. This paper presents a probabilistic framework for dynamically forecasting optimal thermal protection settings. The proposed approach directly predicts the day-ahead scale factor which maximises the dynamic thermal rating of the transformer from historical load, temperature, and metadata using clustered quantile regression models trained on 644 UK LV transformers. Probabilistic forecasting quantifies overheating risk directly through the prediction percentile, enabling risk-informed operational decisions. Results show a 10--12\% additional capacity gain compared to static settings, with hotspot temperature risk matching the selected percentile, including under realistic temperature forecast errors. These results demonstrate a practical approach for distribution network operators to take advantage of PDs with adaptive settings to maximise capacity and manage risk on operational time scales.

Hang Zhou, Sarah Taylor, David Greenwood et al.

We propose SUB-Depth, a universal multi-task training framework for self-supervised monocular depth estimation (SDE). Depth models trained with SUB-Depth outperform the same models trained in a standard single-task SDE framework. By introducing an additional self-distillation task into a standard SDE training framework, SUB-Depth trains a depth network, not only to predict the depth map for an image reconstruction task, but also to distill knowledge from a trained teacher network with unlabelled data. To take advantage of this multi-task setting, we propose homoscedastic uncertainty formulations for each task to penalize areas likely to be affected by teacher network noise, or violate SDE assumptions. We present extensive evaluations on KITTI to demonstrate the improvements achieved by training a range of existing networks using the proposed framework, and we achieve state-of-the-art performance on this task. Additionally, SUB-Depth enables models to estimate uncertainty on depth output.

Hang Zhou, David Greenwood, Sarah Taylor

Self-supervised learning for depth estimation uses geometry in image sequences for supervision and shows promising results. Like many computer vision tasks, depth network performance is determined by the capability to learn accurate spatial and semantic representations from images. Therefore, it is natural to exploit semantic segmentation networks for depth estimation. In this work, based on a well-developed semantic segmentation network HRNet, we propose a novel depth estimation network DIFFNet, which can make use of semantic information in down and upsampling procedures. By applying feature fusion and an attention mechanism, our proposed method outperforms the state-of-the-art monocular depth estimation methods on the KITTI benchmark. Our method also demonstrates greater potential on higher resolution training data. We propose an additional extended evaluation strategy by establishing a test set of challenging cases, empirically derived from the standard benchmark.