Predicting water prices using machine learning techniques
Comprehensive analysis of Victorian water markets using advanced predictive modeling and optimization techniques. Includes time series forecasting of water allocation prices, portfolio optimization for water assets, and decision-making algorithms for water purchases. Utilizes machine learning methods like ARIMA, LSTM, and Bayesian neural networks to analyze market trends and optimize trading strategies in the Murray-Darling Basin.
Optimising Water Resource Management: Advanced Predictive Modeling for the Victorian Water Market
As part of the "Optimisation and Machine Learning" subject at the University of Melbourne, my teammates Meredith, Sultan, and I embarked on an extensive exploration of the Victorian water market, with a focus on the Murray-Darling Basin. Our goal was to apply and integrate advanced machine learning and optimisation techniques to this complex real-world system.
We developed a range of predictive models and optimization strategies aimed at enhancing decision-making in water resource management. Our approach included time series forecasting, portfolio optimization, and decision-making algorithms for water purchases.
While this paper primarily serves to demonstrate our application of machine learning and optimisation techniques and has not undergone peer review, the potential impact of such work is significant. Efficient management of water resources is critical in regions like the Murray-Darling Basin, which face recurring droughts and environmental pressures. Our models, if further developed and validated, could assist government entities and irrigators in making more informed decisions about water purchases and investments, potentially leading to substantial cost savings and more sustainable water use practices.
This project provided us with valuable insights into the complexities of modeling water markets. We grappled with the challenges of incorporating multiple factors into price predictions and learned to navigate highly variable and sometimes unreliable data. Perhaps most importantly, we gained practical experience in applying various machine learning techniques to a complex economic and environmental problem, bridging the gap between theoretical knowledge and real-world application.