Intelligent control architectures for fault diagnosis in electrical power distribution networks

We present three architectures, drawn from intelligent control approaches, that represent our first investigations into the design of a fast-acting fault diagnosis system for electrical power distribution networks. We present experiments on artificial neural network and fuzzy logic based approaches, then make proposals for a hybrid combination of the two that could have promising potential. All three techniques appear to be capable of providing good performance, in terms of the delay between fault inception and its detection by these systems. However, none of the systems is yet in a sufficiently mature state to be used. More validation tests and design optimization is required to make them capable of robustly distinguishing all fault scenarios. However, the work to date shows promise, with a large number of the fault categories being successfully detected and disambiguated from each other. Further work will focus on more complex fuzzy systems and on the hybrid approach. Early experimental results indicate that the latter could provide a good compromise between linguistic interpretability and performance.