The Economic and Carbon Mitigation Potential of Manufacturing Small Modular
Lead Research Organisation:
Imperial College London
Department Name: Centre for Environmental Policy
Abstract
The Economic and Carbon Mitigation Potential of Manufacturing Small Modular
Nuclear Reactors in the UK
Nuclear Reactors in the UK
Organisations
People |
ORCID iD |
Iain Staffell (Primary Supervisor) | |
Elpiniki Avgeraki (Student) |
Studentship Projects
Project Reference | Relationship | Related To | Start | End | Student Name |
---|---|---|---|---|---|
EP/N509486/1 | 30/09/2016 | 30/03/2022 | |||
2035247 | Studentship | EP/N509486/1 | 30/09/2017 | 29/09/2020 | Elpiniki Avgeraki |
Description | To infer the location of PV installations: (1a) develop a GIS database containing variables such as population density, irradiance, affluence and proximity to grid infrastructure; (1b) develop a database of known solar PV installations, or installation density; (1c) use machine learning to develop statistical relationships between the explanatory metadata and installed locations, for countries where these are known; (1d) apply these relationships worldwide to develop a best estimate of where PV panels are installed in every country. To create a database of PV output: (2a) identify openly-available datasets of individual PV system production, such as PVOutput and the Sheffield Solar dataset; (2b) assemble data from each source either through direct relationship with the developers or automated access to their APIs; (2c) develop data cleaning and filtering tools to identify and remove spurious systems; (2d) deliver consistent and representative datasets for the hourly (or better) solar PV production from individual countries or regions, with confidence intervals. I am developing a niche database that will include information about solar PV that are not currently available (e.g. snowfall, tilt of panel, output, temperature, irradiance,etc). Next, I using machine learning algorithms I will try to minimise simulation error using these variables and reconcile the relationships developed with physical laws of solar power production to test whether correlation implies causality. |
Exploitation Route | Machine learning has been employed widely for optimising solar PV technology parameters and the design of specific installations. It has not before been used for broad global-scale assessments of either where systems are located or how they perform. A global system-level database of power output has not before been constructed, and currently researchers must either use national-aggregate estimates from system operators (such as National Grid), which are themselves estimated, or no measured data for testing their output models against. Current methods of simulating solar PV output using historic meteorological data focus on the two primary variables of irradiance and temperature, but no previous works have systematically explored secondary variables such as wind speed, rainfall, snow cover and aerosol loading. Each of the three specific objectives of this project therefore contains novel areas of research. |
Sectors | Digital/Communication/Information Technologies (including Software) Energy Environment Government Democracy and Justice |
Description | Imperial Fellowship Programme |
Amount | £500 (GBP) |
Organisation | Tsinghua University China |
Sector | Academic/University |
Country | China |
Start | |
End | 07/2018 |
Title | develop machine learning techniques to minimise simulation error |
Description | To improve the accuracy of solar PV modelling: setup an existing model of solar power production and the cluster computing infrastructure to run global-scale simulations using identify other meteorological or physical variables which could influence PV production through literature review and acquire data for them; develop machine learning techniques to minimise simulation error using these variables; reconcile the relationships developed with physical laws of solar power production to test whether correlation implies causality. |
Type Of Material | Improvements to research infrastructure |
Year Produced | 2020 |
Provided To Others? | No |
Impact | Machine learning has been employed widely for optimising solar PV technology parameters and the design of specific installations. It has not before been used for broad global-scale assessments of either where systems are located or how they perform. A global system-level database of power output has not before been constructed, and currently researchers must either use national-aggregate estimates from system operators (such as National Grid), which are themselves estimated, or no measured data for testing their output models against. Current methods of simulating solar PV output using historic meteorological data focus on the two primary variables of irradiance and temperature, but no previous works have systematically explored secondary variables such as wind speed, rainfall, snow cover and aerosol loading. Each of the three specific objectives of this project therefore contains novel areas of research. |
Title | Database for solar PV output |
Description | create a database of solar PV power production, with system-level granularity, hourly or better temporal resolution, and global scope; |
Type Of Material | Database/Collection of data |
Year Produced | 2020 |
Provided To Others? | No |
Impact | Develop a GIS database containing variables such as population density, irradiance, affluence and proximity to grid infrastructure; Develop a database of known solar PV installations, or installation density; |