Automatic classification of LCA data

Shanlin Chen

Key information

Authors:

Shanlin Chen (Shanlin Chen)

Supervisors:

Susana Margarida da Silva Vieira (Susana Margarida da Silva Vieira); Christopher Lucien Mutel

Published in

June 26, 2020

Abstract

Life Cycle Assessment (LCA) is a useful tool for environment impact assessment and decision making in energy systems, meanwhile it’s also complicated as it requires detailed information, time and computing resources. Many studies have been conducted to make LCA simpler in the categories of impact assessment like Cumulative Fossil Energy Demand, Carbon Footprints and Representativeness Index. Life Cycle Inventory (LCI), as an important part of a LCA study compiles all necessary information. This study aims to use machine learning algorithms to explore the data structures and hidden patterns in ECOINVENT and EXIOBASE LCI database, estimate the LCI of wind turbines with limited information for LCA. Clustering results from different machine learning algorithms show that all the datasets of ECOINVENT or EXIOBASE seems to be one cluster with some random outliers by using inventories as features. Location is a good feature to change the data structure but more information needed for better clustering. It’s possible to estimate the LCI of a wind turbine from partial information through machine learning algorithms and mathematical methods. It’s also practical to predict the total electricity production of wind turbines for better environmental impact assessment. Although the estimations of input materials and predicted amount of electricity may differ from the actual values, they are still good references for impact assessment as well as decision making.