Math, Climate and Energy meet Industry

P

• Powercor

  • Challenge number 5

    Expected customer restoration time

    A key challenge in electricity distribution network operation is estimating the time required to re-energise customers after an outage. This estimation is important because it directly affects customer communication, outage management, crew coordination, regulatory performance, and customer satisfaction. In practice, restoration time can vary significantly from one outage to another, depending on the nature of the fault, the location of the affected assets, the number of customers impacted, the availability of field crews, switching requirements, network configuration, access constraints, safety considerations, and the severity of weather conditions. During major events, such as storms, high winds, lightning, flooding, or extreme heat, restoration becomes even more complex because multiple outages may occur at the same time and operational resources must be prioritised. Therefore, the central research question for this study is: How can we estimate the time required to re-energise customers in an electricity distribution network based on operational and meteorological factors? This question aims to explore how historical outage records, operational data, asset information, crew response data, and weather-related variables can be combined to develop a reliable estimation model for customer restoration time. Operational factors may include outage type, feeder category, number of affected customers, fault location, protection operation, switching complexity, crew travel time, repair requirements, and previous restoration patterns. Meteorological factors may include wind speed, rainfall intensity, temperature, lightning activity, storm classification, and weather warnings. By analysing the relationship between these factors and actual restoration times, the research can identify which variables have the strongest influence on re-energisation performance. The outcome of this research could support electricity distribution businesses in improving outage prediction, enhancing operational decision-making, and providing more accurate estimated restoration times to customers and stakeholders. It may also assist control room operators and reliability teams in prioritising resources during normal and major event days. From an academic perspective, the research contributes to the growing field of data-driven asset and outage management by applying statistical, machine learning, or hybrid modelling techniques to a real operational problem. From an industry perspective, the findings could help reduce uncertainty in outage response, improve regulatory reporting, and support better customer experience during supply interruptions. Overall, this research question is highly relevant to both distribution network reliability and modern power system operation, particularly as electricity networks face increasing pressure from climate variability, customer expectations, and performance-based regulation.

W

• Water Sensitive Cities Australia

  • Challenge number 2

    Intelligent Monitoring: How can computer vision techniques be developed to observe and quantify real-time behavioural traits — such as motility, swimming dynamics, and community structure — of phytoplankton and zooplankton, and establish reliable correlations between these observable behaviours and shifts in the Water Quality Index (WQI) across dynamic aquatic environments, including nature-based treatment systems?

    The opportunity: Traditional water quality monitoring is expensive, slow and lab-centric. It fails to capture dynamic biological responses and makes it difficult to prove the real-time performance of NBS interventions to end-users, regulators and investors. Instead of slow, expensive lab tests, we want to use cameras + AI to observe the tiny plants and animals living in the water (phytoplankton and zooplankton). Their behaviour, movements, and which types are present can act like a real-time “health check” for the water.

    We need your help to answer:

    How might interdisciplinary teams design, train, and rigorously validate scalable computer-vision and machine-learning pipelines that convert these behavioural observations into practical, low-cost, real-time monitoring tools capable of delivering continuous water quality assessment, early-warning alerts for pollution or eutrophication events, and independent, auditable performance verification of NBS interventions?

    • Attachment
  • Challenge number 3

    Ecosystem as a service: How can we turn the multiple benefits of nature-based solutions into reliable, long-term income streams that pay for initial implementation and ongoing maintenance at scale?

    Public budgets alone cannot fund the green infrastructure needed for climate adaptation, clean water, and flood protection in fast-growing cities. Private investors and blended finance partners need clear, trustworthy evidence that these projects deliver measurable results and can generate attractive returns under real-world conditions and uncertainties.

    We need your help to explore:

    a) What practical mathematical and quantitative approaches — including ecosystem service valuation models, risk and uncertainty modelling, optimisation of incentive structures, and financial scenario analysis — can most effectively measure and assign credible financial value to the full bundle of benefits from nature-based systems (cleaner water, flood/drought resilience, carbon, biodiversity, and urban cooling)?

    b) How can we develop mathematical frameworks and optimisation models that integrate real-time performance data from AI monitoring systems into robust Monitoring, Reporting and Verification (MRV) processes, so that Payments for Ecosystem Services and planning incentives (such as extra building rights for high-performance green infrastructure) become low-risk, transparent, and bankable for private impact capital and blended finance?

    • Attachment

• Westpac Group

  • Challenge number 1

    Fast calibration methods for stochastic interest rate models for interest rate option valuation.

    To price complex interest rate options, e.g. Bermudan swaption, the term structure model is necessary to specify dynamics for evolution of all future zero coupon bonds. HJM model provides a framework to model the evolution of the instantaneous forward interest rate. To reduce the complexity, a subclass of HJM model with separable forward rate volatility, Cheyette model, are widely used in industry and also draw a lot of interest in academic research.

    When using simulation methods to shock the market data, the parameters of Cheyette model need to be calibrated for each simulation. Therefore, a fast calibration method is needed. There are already many academic research papers in this area, but a comprehensive survey on this topic with step-by-step numerical process will be very useful and valuable.

  • Challenge number 4

    Convexity adjustment for non-standard interest rate derivative, from IBOR rate to daily average rate.

    For nonstandard interest rate derivatives valuation, it is common to require a convexity adjustment for the underlying rate. Mathematically, it is because the unusual underlying rate is not a martingale for the selected measure. The typical examples are LIBOR-in-arrears, CMS caplet and quanto derivative.

    Global markets are in the phase to discontinue their benchmark IBOR rates and use overnight benchmark rates (ARR rates). The daily observed overnight rates are averaged (simple or compound) to be applied to a accrual period. The observation period might be different with the accrual period, for example, the daily rates used at each business day are from 5 business days before. The convexity correction is also needed in this case to value the derivatives with stochastic interest rates.

    It will be very important to increase the understanding of the convexity adjustments for both IBOR and ARR reference rates.