Global sensitivity analysis for the short-term prediction of system variables

Short-term prediction of system variables with respect to load levels is highly important for market operations and demand response programs in wholesale power markets with congestion managed by locational marginal prices (LMPs). Previous studies have conducted local sensitivity analyses for LMPs at specific system operating points. This study undertakes a more global analysis of system variable sensitivities when LMPs are derived from DC optimal power flow solutions for day-ahead energy markets. The possible system states are first partitioned into subsets (“system patterns”) based on relatively slow-changing attributes. It is next established analytically that there is a fixed linear-affine mapping between bus load patterns and corresponding system variables, conditional on a particular system pattern. It is then explained how this global piecewise linear-affine mapping can be used to predict system patterns corresponding to forecasted load patterns, hence also dispatch levels, LMPs, and line flows. A 5-bus case study is used to illustrate the accuracy of the proposed prediction method.