Detecting Anomalous Patterns in Time Series: A Case Study of Electricity Price Dynamics in Queensland

Supervisor: Assoc Prof Chun Ouyang

This project investigates anomalous patterns in time series data through a case study of electricity price dynamics in the National Electricity Market (NEM). Following a standard data analytics pipeline, an exploratory data analysis (EDA) was conducted to identify seasonal price patterns, leading to a more focused investigation of the Queensland electricity market. The study examines daily price dynamics from 2019 to 2024, and focuses on a specific period in 2022, which experienced the longest continuous stretch of high prices and significant fluctuations recorded that year. The objective is to uncover patterns in trading prices and analyse their relationship with demand and the demand-supply difference. Two datasets from the Australian Energy Market Operator (AEMO)—one on prices and the other on supply and demand—were analysed, with an in-depth examination of the high-price period at five-minute intervals.

The results reveal a consistent daily pattern in both price and demand across multiple years, with price spikes predominantly occurring during peak demand periods. Extreme price events were typically part of price spikes rather than isolated. Furthermore, demand exhibited a stronger correlation with price fluctuations at five-minute intervals compared to the demand-supply difference. However, neither showed a consistently strong correlation with price throughout the high-price period. The findings provide valuable insights for market participants aiming to better understand price behaviour in the Queensland electricity market, particularly during periods of volatility.  Additionally, this study contributes to research on anomaly detection in time series data by analysing the intricate relationships between price, demand, and supply in a real-world setting.