(highlight:: The Use of Influence Graphs for Modeling and Predicting Network Failures
Summary:
Influence graphs are developed to model how failures propagate on power grids.
Influence graphs represent the network based on the effects one part of the network has on other parts, not through physical connections. Connections in the influence graph are determined by the tendency for the failure of one piece of equipment to be followed by the failure of another.
This approach allows failures to propagate through the influence graph more consistently with our expectations, indicating that existing topological models may not effectively predict or explain failures on the power grid.
Transcript:
Speaker 1
And this line of research has basically developed something called an influence graph, okay? To try to model how failures propagate on power grids, right? And the idea behind an influence graph is that you have some system, you have some physical network, but you're not going to represent it as a physical network with nodes and edges and Things like that. The way that you're going to think about, if I may say drawing the network and kind of visualizing its structure is not through physical connections, but the effect that one part of the Network has on other parts of the network. This is the sort of thing that you either have to build through lots and lots of data, which for the power grid can be difficult because a lot of that data sort of kept under lock and key, Or you sort of have to basically build it by running computer simulations. Basically the idea is that in an influence graph, what you do is if the failure of one piece of equipment in a number of kind of different real world examples or computer simulations tends To be followed by the failure of another piece of equipment, then you have a very strong connection in your influence graph between those two things. And so you're basically building the network, not based on what its physical on the ground topology is, but by how different parts of the network influence each other. So if you build the network that way and look at it as an influence graph, then you find that failures actually propagate through the influence graph much more like we would expect them To propagate in a topological model. And so that's really interesting because you know what it suggests is that maybe because a lot of existing topological models haven't done a very good job of predicting or explaining Failures on the power grid, we tend to blame the models.)
- Time 0:12:04
- failure_propogation, influence_graphs, network_failure, 1socialdont-post,

(highlight:: The Use of Influence Graphs for Modeling and Predicting Network Failures
Summary:
Influence graphs are developed to model how failures propagate on power grids.
Influence graphs represent the network based on the effects one part of the network has on other parts, not through physical connections. Connections in the influence graph are determined by the tendency for the failure of one piece of equipment to be followed by the failure of another.
This approach allows failures to propagate through the influence graph more consistently with our expectations, indicating that existing topological models may not effectively predict or explain failures on the power grid.
Transcript:
Speaker 1
And this line of research has basically developed something called an influence graph, okay? To try to model how failures propagate on power grids, right? And the idea behind an influence graph is that you have some system, you have some physical network, but you're not going to represent it as a physical network with nodes and edges and Things like that. The way that you're going to think about, if I may say drawing the network and kind of visualizing its structure is not through physical connections, but the effect that one part of the Network has on other parts of the network. This is the sort of thing that you either have to build through lots and lots of data, which for the power grid can be difficult because a lot of that data sort of kept under lock and key, Or you sort of have to basically build it by running computer simulations. Basically the idea is that in an influence graph, what you do is if the failure of one piece of equipment in a number of kind of different real world examples or computer simulations tends To be followed by the failure of another piece of equipment, then you have a very strong connection in your influence graph between those two things. And so you're basically building the network, not based on what its physical on the ground topology is, but by how different parts of the network influence each other. So if you build the network that way and look at it as an influence graph, then you find that failures actually propagate through the influence graph much more like we would expect them To propagate in a topological model. And so that's really interesting because you know what it suggests is that maybe because a lot of existing topological models haven't done a very good job of predicting or explaining Failures on the power grid, we tend to blame the models.)
- Time 0:12:04
- failure_propogation, influence_graphs, network_failure, 1socialdont-post,