AI-Based Prediction of Latent Electricity Asset Failures After Extreme Weather Events
Supervisor: Dr. Shanghai Wei
Host University: University of Auckland
Industrial Partner: Vector Ltd., New Zealand
Funding Method: Industry Funded
Project Description
Electricity distribution assets are the infrastructure used to deliver electricity from the transmission system to end users, including power lines, substations, and transformers. These assets are crucial for the final stage of power delivery, stepping down voltage to safe levels for homes and businesses. Severe weather events, including storms, strong winds, and heavy rain, can damage power lines and cause outages. According to the NZ Infrastructure Vulnerability Assessment report, a localised infrastructure failure can cause economic losses exceeding $10M, while a national-level failure can result in losses greater than $200M.
The industry partner, Vector conducted a comprehensive investigation on the aged conductors and concluded that while immediate damage is often addressed promptly, latent failures—those that manifest days or weeks later— remain a significant challenge. These failures are not captured by current asset management frameworks, which is a critical gap as electrification expands, and major storms become more frequent. Our literature review shows that aluminium conductors are well studied. However, there is no research on latent failures, particularly predictive models for their occurrence after extreme weather. The development of these failures is complex, potentially involving mechanisms such as creep, fatigue, corrosion, and aging deformation across varying temperatures and load cycles.
This project brings together a multidisciplinary team from the University of Auckland (materials engineering and digital twin expertise) and the industry partner Vector. The research aims to develop predictive models for latent asset failures by integrating materials engineering with deep learning. The ultimate deliverable is a multi-scale intelligent expert system for analysing power distribution material failures.
The primary research objectives of this project are as follows:
1) To analyse historical failure data correlated with weather events with advanced data analytics to identify patterns and risk indicators.
2) To design laboratory experiments involving long-duration creep testing under various cycles of humidity and temperature. These experiments will simulate the asset’s behaviour during extreme weather to understand latent defect formation, propagation and failure mechanisms. Additional standard testing, including resistance measurements, heat rise, and mechanical testing, will also be performed.
3) To develop physics-based materials engineering models of latent failures, establishing the feasibility and industrial applicability of artificial intelligence (AI) deep learning techniques for predicting failure behaviour.
4) To derive a predictive model for latent failure likelihood based on asset type, environmental exposure, and event characteristics.
5) To create a digital decision-support tool for post-event asset inspection and prioritisation.
6) To validate all models through laboratory testing and real-world data from industry partner Vector and subcontractors, including asset management systems and field reports.
The successful completion of this project will significantly enhance the resilience of electricity networks by enabling proactive maintenance and reducing unplanned outages. It directly supports critical climate adaptation strategies and aligns with evolving regulatory expectations for risk-based asset management. Furthermore, the research outcomes will support the broader energy sector’s transition by facilitating the integration of more renewable energy sources and the rapid uptake of electric vehicles and distributed energy storage systems.