Power Grid Fault Diagnosis Based On Multi-source Information

The rapidly growth of the physical complexity in power systems and the running of power market bring more and more unanticipated and indeterminate factors to power systems, that render the security and economy operation of power systems face up to new serious challenges. Therefore, fault diagnosis as an important part of power dispatch, has novel request in the precision and real time performance.This dissertation tracks the main achievements in this subject, made comparison of these methods. Then this dissertation presents a novel diagnosis method combining element oriented artificial neural networks and fuzzy integral fusion. At last, a model based on the distributed characteristics is given. The main tasks of the dissertation as follows:1. A brief introduction of fault diagnosis and fault information system is given2. Particular comparisons of different models using circuit breaker and relay information are presented. To handle the Achilles' heel of ANN at getting training patterns and handling topology changes, a model combining element oriented artificial neural networks and fuzzy integral fusion. When a fault occurs, a primary diagnosis is made by element-oriented ANNs, and then the synthetic diagnosis fuses the primary diagnosis results employing fuzzy integral.3. According to the present condition of PMU installation and the features of the SCADA and PMU, this paper proposed a hybrid state estimation model. In this model, the latest nonlinear state estimation result is added to linear model as pseudo measurements, and its weight is adjusted dynamically according to relative distance, than the node state is estimated by means of interpolation method.4. At last, this paper gives a model employing the power flow fingerprint(distributed characteristics) . In this model, system create library of fault patterns according to the least sate, by matching the real-time power flow, fault could be identified.