Research On Power System Of Transmission Line Protection And Cascading Failure Prediction Based On Multi-Agent

As human life activity of electricity demand continues to increase, the size of the power system is more and more substantial. In the field of the entire power system, protection is one of the most important part, how fast and efficient removal of faults, to ensure the normal electricity users has become the primary prerequisite for the normal operation of the power system.There are three traditional three-stage relay protection, depending on the line asked to choose a different way to protect traditional current protection of the main advantages of a single electrical quantities only need to protect the principle is simple, but with the longitudinal distance protection Electrical amount of protection needed ones, partial complex structure, the high failure rate, processing large volumes of data. Therefore, this dissertation is to improve the selection of target-current protection, use a full-line speed based on RBF neural network off protection, that protection can reach85%of full wire speed off the line, after a delay of15%in40ms. On this basis, a combination of multiple neural networks as a sub-Agent, to join in communication with the logic circuit breaker combination of multi-Agent systems. Including tissue layer Agent, coordination layer Agent, and execution Agent, communication via optical fiber each substation, when each interval line failure, instantaneous reach full speed off the line, Sec improved protection of traditional current protection delay, by the logic Relations prove its feasibility. Finally, the pilot protection compared to prove that the improved current protection can be used as the main protection330KV lineThe second part of this dissertation is to study the power grid cascading failure. The complexity of the grid determines the complexity of fault propagation chain, from the current status of research perspective, the best way to predict the cascading failure that all aspects of the evaluation system as much as possible, little research at home and abroad, taking into account the network topology, before and after class correlation, overload factors, mostly to create a simplified model such as:OPA model, CASCADE model, small-world model. In this dissertation, after considering level Markov probability model correlation to the grid vulnerability index based topology analyzed considering the load of the system, and to calculate the matrix expression trend analysis of the topology of the grid, the grid linkage comprehensive analysis failure, compared to a cascading failure load transfer most likely whereabouts of the neural network convergence by learning to compare. Eventually IEEE30-bus system prove its feasibility. The innovation of this dissertation is based on neural networks, multi-Agent coordination, in line protection, based on the failure prediction module added to the chain, and clearly in the chain of neural network input and output fault prediction. And verify its feasibility. Has high research value in the field of electric power systems..