Scenario selection by unsupervised learning in reliability analysis of transmission networks

In reliability assessment of deregulated power systems, an analysis of the power market should be incorporated in the assessment. This is done by using a power market model to generate power market scenarios, and use these scenarios as a basis for reliability assessment. It is shown how unsupervised learning techniques can be used to select a subset of the generated scenarios, and how the reliability assessment can be based on this subset only. Different algorithms for selecting the subset are compared, and it is also discussed how to determine the size of the subset. The results of the case studies show that the computational requirements can be reduced by about 90%, with reasonable accuracy in the reliability indices.