Peak Load Forecasting Using Optimal Linear Combinations of Artificial Neural Networks

摘要

A new approach for daily Peak Load forecasting using combinations of trained Artificial Neural Networks (ANNs) is presented in this study. Two different methods constrained and unconstrained are used to identify various combinations of ANNs for peak load forecasting. In this study, a set of neural networks are trained with different architecture and with different learning parameters. The neural networks are trained and tested for the actual peak load data of Chennai city (Tamilnadu-India). A set of better trained ANNs are selected to develop various combinations using these two methods instead of using a single best trained neural network. Obtained test results using the combinations of ANNs prove its validity.