The nonlinearity property accommodation in the Monte Carlo method of generation system reliability prediction by the neural network model
One of the major concerns of the utility companies is to ensure that the generation capacity (GC) is maintained above the load growth. The demand for assessment of the GC at non-distant time intervals is thus crucial. In general, the rising load is linearly proportional to the generation deficiency. A linear regression approach had been successfully developed to predictively accommodate this demand to avert the issue of deficit supply arising from an unforeseen delay. The Monte Carlo (MC)
... te Carlo (MC) technique was used in the generation system (GS) modeling. However, due to the inherent stochasticity associated with the MC algorithm, the emerging graphical relationship between the load and the generation deficiency is generally linear but always maintains nonlinearity at various intervals along the gradient curve. This paper integrates the artificial neural network (ANN) nonlinear feature using the Levenberg-Marquardt training algorithm with the MC simulation to accommodate the MC-associated nonlinearities to improve the generation system reliability (GSR) prediction. The generalization performance of the prediction obtained on the test data was found to have been greatly improved.