No abstract

This paper presents a new optimal adaptive dynamic load-shedding scheme for a large steelmaking industry with cogeneration units. The proposed method is based on the initial rate of a frequency change (df0/dt) and is coordinated with tie-lines frequency protection relays. An adaptive network-based fuzzy inference system (ANFIS) with a new training algorithm is developed in order to increase the speed of the load-shedding scheme and to have an optimum response at different loading conditions. To overcome the ANFIS training difficulties, a new hybrid approach composed of particle swarm optimization and gradient decent algorithms is used. The training data set for the ANFIS is prepared by a transient stability analysis to determine the minimum load shedding for various operation scenarios without causing the tripping problem of cogeneration units. By using an accurate dynamic modeling of the Mobarakeh steelmaking company in Esfahan Regional Electrical Company network, the performance of the proposed method is compared with the traditional ANFIS learning algorithms, adaptive artificial neural network load-shedding scheme and transient stability analysis. Simulation results show the effectiveness of the proposed method.

No abstract

In this paper, a new wide area neural network-based method is presented for the accurate detection of necessity and the time of controlled islanding execution in large interconnected power systems. By performing coherency analysis at different conditions, the initial coherent groups of the network are determined. To account different stability margins between areas at different conditions and network topologies, we introduced the new parallel neural network (P-NN) structure. The proposed P-NN consists of different individual recurrent neural networks between each of adjacent initial groups. The P-NN is trained with respect to selected wide area signals and generated database through comprehensive stability studies. After the online detection of possible asynchronous oscillations and the corresponding final coherency determination, the alarm signal is sent to the designed P-NN to investigate the network stability between initial groups in real time. The proposed method is applied to New England 39-and 118-bus power systems at different cases and is compared with another intelligent method. It is shown that the proposed method is able to detect islanding necessity and related islands accurately for different disturbances. The speed of the islanding detection, as an important aspect in intelligent controlled islanding, is increased by the proposed method. This will in turn help the system keep stability. In addition, the proposed method could distinguish large stable swings from unstable ones in different contingencies.

No abstract