Predictive analytics for energy systems

Decentralized energy systems or distributed energy systems aim to locate energy production facilities closer to the site of consumption. Such systems are being widely adopted in multiple countries and they promise lower costs to the consumers. This can be achieved as the inefficiencies related to tr...

Full description

Saved in:
Bibliographic Details
Main Author: Tharakan, Rohan Roy
Other Authors: Zhang Jie
Format: Final Year Project
Language:English
Published: Nanyang Technological University 2022
Subjects:
Online Access:https://hdl.handle.net/10356/156690
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Decentralized energy systems or distributed energy systems aim to locate energy production facilities closer to the site of consumption. Such systems are being widely adopted in multiple countries and they promise lower costs to the consumers. This can be achieved as the inefficiencies related to transmission and distribution can be reduced compared to a single central power station. We will be focusing on a subset of the energy market, which is the electricity market. The global electric power generation, transmission, and distribution market is a 3 Trillion Dollar market and decentralized systems are on the rise. However, the challenge in such a system is to predict the direction of the flow of electricity based on the supplied from the energy production units and the demand from the neighborhoods of consumers. To solve this problem, various rule-based models have been built and most of them have not been updated with the latest technologies. Such a predictive model could save a lot of cost by helping facilities take the appropriate measures to store and transmit stored electricity. This project aims to explore a deep learning-based anomaly detection method that can be used for the prediction of the direction of the flow of electricity. The method used is called Deep one-class classification and it is ideally built to detect anomalies within the dataset.