Transient stability assessment in longitudinal power systems using artificial neural networks

Abstract

Results of the application of Artificial Neural Networks to the problem of Transient Stability Assessment are presented. This technique is applied to a real Longitudinal Power System that includes discrete supplementary controls. Different representations of the training space patterns and neural networks architectures are investigated. Input variables include topological changes, load and generation levels and contingencies. A special organization of training patterns with a separation by type of contingency is proposed to reduce classification errors. A graphical presentation of results is suggested as an aid to help system operators to select preventive control actions.

Transient stability assessment using artificial neural networks

“…The concept of individual energy function characterizes the transient stability as a local phenomenon in contrast to the total energy function.In [6], 1916 composite indices were generated and performing correlations, 18 composite indices were selected as inputs. In [7], 12 steady state variables were used as inputs and it was reported that dynamic variables did not. improve the results.…”

“…Faults at mar?y locations have to be considered to evaluate the capability of ANN. In [7], training patterns were divided into many sets and separate ANN was used to predict stability for each set. Classification errors of less than 1% for training set and less than 5% for test set were obtained.…”

“…In [7], Aboytes et al used ANN to predict stability of a 53 generator system which is part of Mexican power system. Training patterns were organised by dividing them into different sets and using separate ANN for each set.…”

Research on Power System Transient Stability Assessment Based on Statistical Learning Theory

“…(2) the model parameter selection is difficult and training results are not stable. To solve the above problems, this paper uses support vector machine developed from statistical learning theory to build a new transient stability assessment model [1]. The proposed model contains bagging and approximate reasoning strategy, which can make accurate reasoning for the unknown sample and improves the stability and accuracy of the prediction.…”

“…But these methods practical applications still exist some problems, mainly as follows: (1) The generalization error of transient stability assessment can not be guaranteed. (2) the model parameter selection is difficult and training results are not stable.…”