AI Solutions for Peak Demand Mitigation

Team Name

Team Waikato

Team Members

Project Description

This presentation proposes a machine learning-based system to predict electricity demand, generation capacity, and spot pricing in New Zealand. By leveraging neural networks and SOKNL (Streaming Online Kernelized Neural Learning), the system provides real-time forecasts that allow for more efficient grid management. Additionally, we conduct a cost-benefit analysis at the infrastructure level, assessing payback periods for investments, enabling policymakers to model various scenarios. On a consumer level, we propose a website that highlights simple, actionable steps for households to reduce peak demand and lower energy costs, ultimately contributing to grid stability and sustainable energy use.

Data Story

We used the data from emi.ea.govt.nz to model the demand and supply as well as the spot pricing for electricity. We also used information from the HEEP survey to understand the household energy consumption pattern in a typical NZ household, as well report from branz https://www.branz.co.nz/energy-efficiency/ (as well as census data from stat new zealand) to identify avenues for improvement in a typical new zealand home. We use the energy efficiency calculation from eeca (https://www.eeca.govt.nz/co-funding-and-support/products/?audience=4).