Adequacy Assessment of Distribution Systems using Artificial Neural Network Accounting Embedded Generations

摘要

This paper describes a methodology for obtaining failure probability of a distribution system for a specified chronological load curve for 24-h. Distribution substation capacity and load in each sub-interval has been assumed to be continuous random variable having Normal distribution. A multi-layer feed forward neural network has been trained to give reliability or probability of failure for various values of safety factor, and coefficient of variations of loads and capacity of the distribution system. The reliability of the distribution system is evaluated on-line by the trained network using predicted chronological load curve, which is available (using short term load forecasting) for each sub-interval with its mean and Standard deviation. The methodology has been implemented on a sample distribution system.