Comparing Different Artificial Line Limit Methods When Simulation Cascading Failures In A Power Grid

In 2012 a cascading failure caused a loss of power to 600 million people throughout India, while the 2003 Northeastern US blackout is estimated to have cost around USD 6 billion. Using complex network analysis has become a useful tool in understanding how cascading failures propagate through the power-grid. However, obtaining detailed network data is difficult especially regarding line-limits. A popular approach to side-step this problem is to use proportional loading instead of real line-limits. Using the proportional loading method, We first calculate the flow in the power-grid and then set the line limits by adding a fixed proportion on top of the line flow. This proportion acts as a line tolerance and allows some increase in loading when the grid is attacked. However, there is very little evidence supporting the assumption that line-limits are actually proportional.
Our work uses a data set of the UK power-grid with real line limits. We find a correlation coefficient of 0.67 between real and proportional line limits, using UK base-load settings. We compare the performance of a network under attack using real line limits, proportional line limits (with seven different tolerances) and two constant line limit methods. We find that proportional line limits act as a reasonable proxy for real line limits only in the case when the damage metric is blackout size (total lost MW), and the average loading of the network is the same as the average loading of the real network. We also find that the constant line limit outperforms Proportional loading to better represent the order in which the network edges are lost due to cascading failure. We conclude that care should be used when setting the tolerance using artificial line limits and that method chosen should reflect the research goals.