Constraint Graph: Power Supply Restoration

This is a picture of the logical structure of a problem arising in the course of reconfiguring an electrical power grid to resupply all non-faulty lines after a short circuit has caused a circuit breaker to trip off. The specific problem, here in a form suitable for a finite domain constraint solver, is to find a plan - a sequence of opening and closing of switches - at the end of which the faulty line is isolated and the rest are fed. This is only a small part of the real problem, which involves interleaving diagnosis and planning, and must also take account of loads on the lines, priority customers and so forth.

Here is an account of the PSR problem, which serves as a benchmark in both the AI planning and the diagnosis communities.

The constraint graph shows areas of very tight clustering, associated in the problem representation with calculation of which lines or switches are upstream or downstream from which others, and areas of much looser connection representing the fluents saying which switches are open and which are closed in the possible successive states.

For the drawing, the lines are made partially transparent so as not to obscure the nodes. This results in darker colour where there are many lines on top of each other or where they represent relatively strong constraints. The colours of the dots are assigned in accordance with the names of the variables, thus corresponding to quantities that play different roles in the problem and the search for a solution.


Dr J K Slaney                      Phone (Aus.):  (026) 125 8607
Automated Reasoning Project        Phone (Int.): +61 26 125 8607
Australian National University     Fax (Aus.):    (026) 125 8651
Canberra, ACT, 0200, AUSTRALIA     Fax (Int.):   +61 26 125 8651
John.Slaney@anu.edu.au