A Real Options Analysis into the Benefits of Flexibility for the UK's Offshore Wind Industry
Aaron Carver
A ‘real option’ in design is often described as the ‘right, but not the obligation, to change a system in the face of uncertainty’; such designs can provide significant improvements (10-30%) compared to those obtained using standard deterministic optimisation methods. In response to the UK government’s commitment to increase offshore wind capacity to 40GW by 2030, this project commenced with the aim of quantifying the economic benefits of embedding flexibility into new offshore windfarms. Having concluded that flexibility provides a substantial enhancement to the expected NPV of projects, policy amendments were proposed to incentivise the development of such projects in the UK.
Supervisors:
Dr. Michel-Alexandre Cardin, Dyson School of Design Engineering, Imperial College London
A probabilistic approach for electricity load prediction using machine learning models
Benoit Putzeys
In order to support participants of the electricity market to take strategic decisions, the electricity load has to be predicted with a high degree of certainty. With machine learning and especially recurrent neural networks, the electricity load can be predicted several days in advance due to the seasonal character of the time series data. The aim of this project is to present different approaches to combine uncertainty with predictive models. In doing so, the electricity trader can be informed about periods with high variability.
Supervisors:
Dr. Samuel Cooper, Dyson School of Design Engineering, Imperial College London
Andy Hadland, Arenko Group
Steve Kench, Dyson School of Design Engineering, Imperial College London
Ben Simon, Department of Earth Science and Engineering, Imperial College London
Consumer Psychographics in Energy: The Need for a Cross-Sectoral, Principle-Based Data Regulator
Lucy Liu
Substantial shifts in domestic energy consumption behaviours will be required to meet the UK’s Net Zero targets. Individuals are leaving digital trails of their daily lives via social networks, energy meters and smart phones. These digital footprints are opportunities for psychographic profiles of consumer preferences to be developed and used to potentially influence energy consumption behaviours across a range of timescales. As a part of an ongoing psychographic research workstream, this study aims to outline the system-wide opportunities, benefits, barriers and key enablers, as well as risks to consumers of the application of psychographics in the energy industry.
Supervisors:
Dr. Mark Workman, Energy Futures Lab, Imperial College London
Freya Espir, Everoze
Energy Price Forecasting with Deep Learning for Energy Storage applications
Maria Adriana Murteira Martins
Development of Deep Learning models for Energy Price Forecasting at the Balance Mechanism for Energy Storage applications. Six models were developed and compared according with the performance on both normal price regions and spike price regions. The best hyper parameters, predictive window and external features to be included in modelling were found considering the predictive regions of interest (normal or spikes) and the model itself.
Supervisors:
Dr. Samuel Cooper, Dyson School of Design Engineering, Imperial College London
Andrea Lombardo, Department of Earth Science and Engineering, Imperial College London
Ben Simon, Department of Earth Science and Engineering, Imperial College London
Andy Hadland, Arenko Group
Designing strategies to hedge financial risks in energy projects using machine learning
Mohamed Amine Benchrifa
This project aims to use a new technique called deep hedging to hedge heat rate options, that model gas-fired power plants, using neural networks and without any mathematical models. Deep hedging has the advantage of surpassing the assumptions of mathematical models and of using different variables that impact the hedging decision.
Supervisors:
Dr. Michel-Alexandre Cardin, Dyson School of Design Engineering, Imperial College London
Benjamin Pryke, Beacon Platform
Increasing the visibility of low-voltage networks through data analytics
Ronald Marshall Monterroso Rubi
The UK’s target to become a net-zero carbon economy by 2050 will require an accelerated adoption of Low Carbon Technologies (LCT) such as rooftop solar PV and electric vehicles. The growth of connections of LCT to low-voltage (LV) networks will require Distributed Network Operators to implement cost-effective solutions to increase the visibility of LV networks. This project aimed to increase the visibility of low-voltage networks by developing an approach which applies data analytics, low-voltage modelling and satellite images. Specific objectives were to identify electrical circuits which contain solar PV systems and estimate their contribution during the PV System peak hours.
Supervisors:
Julio Perez Olvera, Department of Electrical and Electronic Engineering , Imperial College London
Dr. Adria Junyent-Ferre, Department of Electrical and Electronic Engineering , Imperial College London