Mathematical Baseline and error Detection Techniques for the Analysis of Unaccounted For Gas (UAG)
Lead Research Organisation:
University of Manchester
Department Name: Mathematics
Abstract
Management of natural gas pipelines is a difficult and complex task. The network operator must monitor the flow through the network of pipelines to correctly allocate costs to each user, guarantee safety and reduce emissions. The balancing of input/output should be a trivial task since leakages are unlikely but the reliability of measuring equipment at both ends of the system means that there is often a difference - the Unaccounted for Gas (UAG).
The aim of this project is to use advanced stochastic and statistical theory to build up a dynamic model for expected measurement error, which will depend on the flow rates through the system. Using this model, we will be able to build up an idea of what is normal so that unexplained errors can be quickly identified. Once measurement errors are understood, we will be able to use Mathematical Finance techniques to determine optimal investment opportunities in the operation of the network.
Working closely with National Grid UK, the student will have access to a full set of data and visits to the measurement stations. The students will need a strong background in both applied mathematics and statistics.
The aim of this project is to use advanced stochastic and statistical theory to build up a dynamic model for expected measurement error, which will depend on the flow rates through the system. Using this model, we will be able to build up an idea of what is normal so that unexplained errors can be quickly identified. Once measurement errors are understood, we will be able to use Mathematical Finance techniques to determine optimal investment opportunities in the operation of the network.
Working closely with National Grid UK, the student will have access to a full set of data and visits to the measurement stations. The students will need a strong background in both applied mathematics and statistics.
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| EP/N509280/1 | 01/10/2015 | 31/03/2021 | |||
| 1868767 | Studentship | EP/N509280/1 | 01/10/2016 | 31/03/2020 | Ludomir Botev |
| Description | The underlying problem presented by UAG has proved to be a multi-faceted issue, and therefore the research carried out so far has been broad in scope. Key aspects are detailed below; Statistical process control techniques, such as change-point analysis have successfully been employed to identify systematic errors occurring in the national transmission network and appropriate monitoring tools have been implemented into the National Grid workflow procedure for high UAG analysis. The tool, based on the Shiny/R technology stack also provides advanced data visualisation techniques. A number of baseline models, based on a broad spectrum of statistical methods has also been included in the tool. These range from simplistic features such as rolling standard deviation to 1-step forecasts to ground-up system modelling. This gives practitioners a wide variety of measures, and empowers them to make more confident decisions regarding the merit of further data validation. A review of competing algorithms, in both online and offline cases has been carried out, and the scope for the use of statistical inference on UAG to determine system state has also been analysed. The results are due for submission in an upcoming paper. Research into optimising assurance schedules with a view to minimise systematic error detection time has been undertaken with some initial simulations suggesting such optimisation might reduce the total expected cost in the case of partial equipment failure. However, there has not proved to be a significant business case to warrant further research down this avenue. Multivariate anomaly detection, relating to single day errors techniques have been incorporated into the aforementioned tool. This is an area of current research, with an emphasis on supervised-learning based approaches. Systematic error attribution, also known as Gross error detection has proved to be the most challenging aspect, due to the highly varied nature of the time series making up the NTS. It is unlikely an omnibus approach can be adopted. Due to the lack of a strong business case, further research in this area will not be prioritised in the project. |
| Exploitation Route | UAG is encountered in virtually all gas transmission grids, and system operators are all incentivised to take measures to minimise it. The techniques, models and operational best practices developed by this project will naturally prove to be of interest to such parties. Indeed, systems transporting compressible gasses of any nature may benefit from the work. The industrial partner, National Grid will benefit directly from the work by virtue of the tool (app) that has been developed. |
| Sectors | Energy |
| Description | The use of our finidings is twofold: -National Grid use our fitted models and methodologies through the deployed Shiny App; -Several of our recommendations towards further investigating UAG have been acted upon, namely revisiting the linepack energy calculation process and the volumetric calculation of UAG. |
| First Year Of Impact | 2020 |
| Sector | Energy |
| Impact Types | Economic |
| Title | UAGMS |
| Description | The Unaccounted For Gas Management Suite (UAGMS) provides functionality to aid in all parts of the statistical UAG management process used by National Grid. We have implemented both existing and novel methods from R into a Shiny app, and embedded additional functionality in reporting tools, quick reference dashboards and the ability to use both online and custom datasets. |
| Type Of Technology | Webtool/Application |
| Year Produced | 2020 |
| Impact | This product allows UAG to base their UAG-related decisions on a robust statistical basis. Evaluation of the efficacy of the models is ongoing. |
| URL | https://lbotev.shinyapps.io/ChangepointDet/ |