AI-based decision support tool for storm damage prediction in Nova Scotia

Nova Scotia Power Inc. (NSPI) is the main provider of electricity in Nova Scotia. The largest disruptor to its system is the weather, which can lead to widespread equipment failure and outages across the grid. NSPI attempts to predict these damages before they occur to decide on the appropriate level of response and allocation of resources to restore service to their customers as soon as possible, during and after a weather event. Currently, NSPI uses a simple MS Excel-based tool to make predictions. However, its accuracy is relatively low. This project investigates whether modern AI/ML tools like Artificial Neural Networks, Random Forests, and Reinforcement Learning can be harnessed to improve the prediction accuracy of storm damage in Nova Scotia. The objective is to develop and test a prototype prediction tool that uses weather data like wind speed and direction, precipitation levels, ground thaw and foliage, as well as system information, e.g., the number of transformers and electricity poles and the distribution of customers across the province, to make systematic predictions about the damage to be experienced in different geographical regions as a result of forecasted extreme weather events.

Faculty Supervisor:

Ahmed Saif

Student:

Partner:

ISEN

Discipline:

Engineering

Sector:

Artificial Intelligence; Energy and Utilities; Environmental Science and Technology

University:

Dalhousie University

Program: