Tanoto, Yusak and Ongsakul, Weerakorn and Marpaung, Charles O.P. Long-term Peak Load Forecasting Using LMFeedforward Neural Network for Java-Madura-Bali Interconnection, Indonesia. In: PEA-AIT International Conference on Energy and Sustainable Development: Issues and Strategies (ESD 2010), 2-4 June 2010, Chiang Mai, Thailand.
Abstract
This paper presents the application of artificial neural network (ANN) based on multi-layered feedforward backpropagation for long-term peak load forecasting (LTPF). A four-layered network using Levenberg-Marquardt (LM) learning algorithm is proposed to forecast annual peak load of Java-Madura-Bali interconnection, Indonesia, for the period of 2009-2018 considering 11 regional factors encompass economic, electricity statistics, and weather thought to affect the load demand. The proposed network structure is first trained over the past 11 years (1995-2005) to forecast annual peak load of 2006-2008. Afterwards, the justified network structure is trained over the past 14 years (1995-2008) to forecast annual peak load of 2009-2018. Several simulations involve changes in historical actual peak load target and variation on projected regional economic growth are carried out to observe the network adaptability. Results are then compared with that achieved by the multiple regression model and projection made by utility. In this case, forecasting result exhibited by the proposed network is the closest to actual values of 2006-2009 among others taken the average error of 0.2%. Likewise, its forecasting differences for 2010-2018 are less than 7% compared to others. In term of network adaptability, outputs generated by the network are well adjusted to the projected inputs variation.
Actions (login required)
![[img]](https://repository.petra.ac.id/style/images/fileicons/application_pdf.png)
