Case Studies

Outage planning is currently based on a worst-case scenario for each outage. There is limited accounting for the potential impact of increasingly changing system conditions (generation, weather, etc.) or of changes to one outage as a result of other outages. This has historically been done using “rules of thumb”. With the rapid pace of change, the current planning methods are starting to show their limitations. In particular, a lot of work is devoted to reacting and re-planning. The Edinburgh team developed a tool that facilitates efficient economic decision-making and further reduces the amount of work for the National Access Planning process to react and re-plan.

An important task of National Grid ESO is balancing supply and demand of electricity in real-time. To keep the balance, the control room engineers can take actions, for instance buying and selling energy from so-called balancing units. Making the decision of how much energy to buy or sell from which units at the least expensive way is a real time optimization problem which involves a large amount of data and needs to be solved with low latency. Operational research consultancy unit contributed to the optimization tool which solves the described unit balancing problem.

The generation of a digital twin of a construction project from beginning to end is essential for the efficient management of a construction project. In the construction site, allocation of human resources and equipment to activities and scheduling the order of activities are the main challenges. An optimal decision with the minimum cost and duration must be made considering the availability of resources, precedence relation of activities, time loss due to material and equipment transfer between different construction zones, continuous allocation of workers to the similar tasks in order to preventing the productivity loss. Operational research consultancy unit developed an optimization tool which works as an integrated module of a bigger digital twin platform for construction projects.

It is important both for efficiency and safety to be able to detect and predict states of fluid flow, such as stratified flow, wavy flow, and slugging flow. Various features of flow were measured by sensors in several modalities, and a decision tree model was developed to explain transitions in flow state in terms of feature thresholds.

Financial institutions are using deep neural networks as decision support systems for fraud detection. However some predictions may be untrustworthy because predictive uncertainty is not easily quantifiable. Recent advances in measurement of epistemic and aleatoric uncertainty were applied to the general task of anomaly detection in image data.

Mitigating threats to plant species is important for maintaining diversity for biomedical use and food security, especially in the presence of climate change. The project aimed to develop an R package for point process models of area interactions that would be free and easy to use, to inform the reforestation/replanting processes and facilitate the prediction of the response of forests/plantations to future climatic change.

Diseases that spread from animals to people (zoonotic diseases) are an increasing concern to public health. Some (e.g. Ebola virus disease, E.coli O157) are responsible for serious epidemics, but outbreaks are unpredictable. The Vietnam Initiative on Zoonotic Infections (VIZIONS) collected clinical and behavioural data from hospital patients in Vietnam. We analysed these data using multivariate statistical methods and mixed models to 1) identify patients most likely to have an unknown zoonotic virus; 2) identify risk factors associated with zoonotic disease. We identified one patient with a novel potentially zoonotic virus from pigs, and found that contact with pigs was a particular risk factor for zoonotic disease in Vietnam.

Cognitive ability changes over time and declines in later years. In several studies we fit models of the trajectory to summarise the overall shape of change over time in a useful way. These models enable change to be compared between groups in order to address questions around the influence of social, demographic, lifestyle, and genetic factors on age-related change in cognitive ability. The models were used to evaluate risk factors and predictive factors for change in cognitive ability. More complex models were constructed to explore the reciprocal influence of changes in factors such as wellbeing and physical fitness on change in cognitive ability.

A new smartphone-based instrument was developed at Edinburgh Royal Infirmary to assess patients in intensive care. It was designed to measure levels of arousal in order to screen for post-operative delirium and evaluate risk of dementia and death. The instrument was validated by comparison against existing gold standard assessment methods.

An important research question concerns whether change in aging brain structure due to tissue wearing away is due to narrowing arteries that supply blood or to an as yet unknown cause. This question has direct consequences for cognitive decline, stroke, and dementia. Models were developed to show the association between progressive changes in brain structure shown by image analysis of MRI scans over a period of six years, and progressive arterial narrowing over the same period, and also change in key cognitive abilities over the same period. Results suggested that changes in brain structure, associated with cognitive decline, were not caused by narrowing arteries.

A novel model of personality was developed based on the idea of a state-space of personality in which each individual is represented as a multidimensional point. The model was based on the proposition that individuals move within this space under the influence of various attractors, such as other individuals for example, with the overarching purpose of striving for an equilibrium state. This model enabled simulations of various complicated social phenomena, and was able to demonstrate interesting complex interplay between individuals and their situational experiences.

Movements of mobile animals such as seabirds are often changeable and difficult to predict. We tested whether oceanographic and environmental variables can be used to predict offshore areas used by seabirds and whether these predictions are consistent over time. We used spatial models to relate foraging locations, identified using GPS data collected from seabird foraging trips, with variables such as sea surface temperature. We found that seabird foraging areas can be predicted from environmental data, but the accuracy of these predictions do not remain consistent between years or at different stages of the breeding season.

Red grouse are an important bird species in the UK, with grouse shooting supporting rural economies. Recently, there are concerns that grouse numbers are declining in some areas with potential repercussions for local economies. Biologists at the Game and Wildlife Conservation Trust used long-term harvesting data to estimate regional changes in grouse populations and factors influencing these changes. Generalised additive models were fitted to account for temporal trends in the data and showed that grouse populations had declined in Scotland and Wales in recent decades due to changes in land management.