Within the GARPUR project, the mid-term horizon ranges from a few months to a few years ahead in time with respect to real-time operation; thus, it bridges the gap between short-term and long-term operation planning and real-time operation. The most important decisions in the mid-term time horizon concern with asset management, whose objective is to extend the mean time to the next failure and reduce the frequency and duration of service interruption. This requires TSOs to adopt strategies that include actions such as planned or unplanned maintenance and replacement of assets. However, budget constraints will require a tradeoff between the amount of maintenance necessary and the cost of maintenance.

Although various heuristic maintenance and deterministic model based methods exist (e.g., N-k problem), we are interested in a probabilistic model-based approach, where uncertainties about component deteriorations or improvements are quantitatively assessed and used for choosing maintenance strategies. In this talk we will present a short introduction to the GARPUR project and to our preliminary strategies for developing such a probabilistic model from highly multidimensional and sparse data. Our method incorporates ideas from probabilistic risk assessment of rare events, importance sampling, sequential decision making, and scenario optimization.

Host: Anatoly Zlotnik