[Paper Review] Machine Learning of Public Sentiments toward Wind Energy in Norway

Oskar Vågerö, Anders Bråte+4 more|arXiv (Cornell University)|Apr 5, 2023

Social Acceptance of Renewable Energy33 references4 citations

TL;DR

This study applies a Norwegian BERT-based NLP model (NorBERT) to analyze public sentiment toward wind energy in Norway using 68,828 geotagged Twitter tweets from 2006 to 2022. It reveals a significant rise in negative sentiment peaking at 32.5% in 2020, with nationwide discussion patterns showing weak regional clustering, offering a scalable method to track public opinion dynamics and complement traditional survey research.

ABSTRACT

Across Europe negative public opinion has and may continue to limit the deployment of renewable energy infrastructure required for the transition to net-zero energy systems. Understanding public sentiment and its spatio-temporal variations is as such important for decision-making and socially accepted energy systems. In this study, we apply a sentiment classification model based on a machine learning framework for natural language processing, NorBERT, on data collected from Twitter between 2006 and 2022 to analyse the case of wind power opposition in Norway. From the 68828 tweets with geospatial information, we show how discussions about wind power intensified in 2018/2019 together with a trend of more negative tweets up until 2020, both on a regional level and for Norway as a whole. Furthermore, we find weak geographical clustering in our data, indicating that discussions are country wide and not dominated by specific regional events or developments. Twitter data allows for detailed insight into the temporal nature of public sentiments and extending this research to additional case studies of technologies, countries and sources of data (e.g. newspapers, other social media) may prove important to complement traditional survey research and the understanding of public sentiment.

Motivation & Objective

• To understand the spatio-temporal evolution of public sentiment toward onshore wind energy in Norway.
• To evaluate the utility of Twitter data and NLP for tracking public opinion on renewable energy, beyond traditional survey methods.
• To assess the feasibility and limitations of using a fine-tuned Norwegian BERT model (NorBERT) for sentiment classification in the context of energy policy.
• To compare NLP-derived sentiment trends with existing survey data and identify key events influencing public discourse.
• To explore the potential of social media data for informing energy transition policies and public engagement strategies.

Proposed method

• Fine-tuned the Norwegian BERT model (NorBERT) for binary sentiment classification on Norwegian-language tweets.
• Collected 68,828 geotagged tweets from Twitter spanning 2006 to 2022 using keyword-based queries related to wind energy.
• Applied data preprocessing steps including text cleaning, normalization, and filtering for geospatial relevance.
• Used a binary sentiment classification framework, labeling tweets as 'negative' or 'non-negative' based on model predictions.
• Evaluated model performance using an F1 score of 0.88 on a held-out test set.
• Conducted spatial and temporal analysis to identify regional trends and temporal shifts in sentiment.

Experimental results

Research questions

• RQ1 How has public sentiment toward wind energy in Norway evolved over time between 2006 and 2022?
• RQ2 What is the geographical distribution of sentiment, and to what extent are discussions regionally clustered or nationwide in scope?
• RQ3 How do sentiment trends from Twitter data compare with findings from traditional public opinion surveys?
• RQ4 What role do specific events (e.g., protests, policy changes) play in shaping public sentiment as reflected in social media?
• RQ5 To what extent can NLP-based analysis of Twitter data complement or contrast with survey-based research on energy policy acceptance?

Key findings

• Public discussion on wind energy on Twitter intensified significantly after 2018/2019, with activity levels increasing fourfold compared to earlier years.
• Negative sentiment toward wind energy peaked in 2020 at 32.5% of all tweets, marking the highest share of negative sentiment in the study period.
• Despite the rise in negative sentiment, the overall trend shows a decline in negativity after 2020, indicating a potential stabilization or shift in public discourse.
• Spatial analysis revealed weak geographical clustering, suggesting that discussions about wind energy are widespread across Norway rather than driven by localized events.
• The sentiment trends on Twitter align with key national events, such as the 2020 protests at Haramsøya and reports on bird fatalities, indicating event-driven sentiment shifts.
• The model achieved an F1 score of 0.88, demonstrating strong performance for binary sentiment classification in Norwegian text using NorBERT.