| The electric power security not only depends on the electrical power system movement analysis and control technology, but also is restrained by the validity and reliability of information from the system. Microcomputer relay protection, fault recording and other automatic devices in power plants and substations have provided massive useful information around the accident. Based on the communication development and the system network construction, it becomes one of the essential conditions to enhance the dispatch automation grade that these information are passed on to the dispatch end, then analyzed and processed effectively. This paper centered on the construction of fault information integration processing and diagnoses system to extend research and discussion domain.After explained its function, the paper reviews the research history and present state of fault information integration processing system, and points out the strongpoint and shortcoming in the practical systems. Thus it becomes clear about the research topic and content. In order to establish the system in a standard frame, this paper separately discusses the function requirements and realization methods of the master station and sub-stations through the basic system constitution.The power grids can bring massive fault information in an accident process, thus it will bring the paroxysm information transmission, as well as the disturbance to the network communication performance and the effective information using in the master station. In order to solve these problems, this paper develops its research on two aspects. One is that establishing a kind of mechanism for information transmission priority and the condition of information deliver to master-station, which is to guarantee the sub-station to only transmit information what the master-station wants truly to need. The other is that establishing primary fault information model to guarantee the master-station can completely collect the fault information without miss, which is brought in a power grid fault process and is associated each other. Information conjunction is realized by the time-window factor with GPS time label and other conjunction factors such as number of grid faults (NOF). On the base of two research results above, this paper has produced the fault information pretreatment process in the master and sub stations. Finally, this paper has studied the realization of power transmission network fault diagnosis as one of high-level application in the master-station. In view of the fact that neural networks have the advantage over other AI methods in the pattern recognition field, this paper take use of radial basis function network to solve the power transmission network fault diagnosis problem, which has good performance in pattern recognition and global approximation. In the current practical application of RBF network, its training algorithm is based on orthogonal least square (OSL), yet such network has the flaw in some aspects, such as fixed selection radial basis function centers with accidental rationality, non-minimum network architecture. Therefore, considering the hidden-level neuron centers'clustering characteristic in training samples, this paper introduces enhanced clustering algorithm, input-output clustering (IOC), to train RBF network. The IOC algorithm can classify the input samples by the output samples, and may speed up the training speed. Moreover, it can self-adjust the clustering center and width parameters according to the actual distribution of these samples, and only need fewer neurons to obtain a satisfied network. A RBF network simulation model has been established, which has 39 inputs and 9 outputs and 45 training-10 test samples, based on a simple test power transmission network, which has four nodes. With the simulation results, it's clear that the RBF network based on IOC algorithm has more excellent performance comparing with the network based on OLS algorithm.
|
PDF Full Text Request