Combining theoretical knowledge and artificial neural networks for power system topology verification

Topology errors, i.e. errors resulting from incorrect modelling of power network connections may be a source of serious errors in various power system applications. Errors in power-system topology are difficult to handle with classical methods based on state estimation. The problem of power system topology verification has been also solved using artificial neural networks (ANNs). It should be pointed out that implementation of pure neural models, where theoretical knowledge on power system is neglected, are not too effective. The paper presents new approach to topology verification based on combining ANN´s technique and larger theoretical knowledge on the power system (and its model) than it is in the case of other approaches. After utilized knowledge on the power system (and its model) is characterized the idea of the approach is outlined. Then, a computational example of the verification using the new approach is given. At the end, the features of the considered approach are discussed.