Load forecasting based on neural network approach

Azizan, Nur Yusrina (2021) Load forecasting based on neural network approach. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

	Text (24 Pages) Load forecasting based on neural network approach.pdf - Submitted Version Download (337kB)
	Text (Full Text) Load forecasting based on neural network approach.pdf - Submitted Version Restricted to Repository staff only Download (1MB)

Abstract

This project focuses on the load forecasting using neural network approach. Load forecasting is a process of estimating the amount of power load usage in future. It is a crucial step for the electrical utility company to produce an optimum amount of power load supply. The electrical load should not be produced less than its demand as it can cause blackout due to power shortage. However, the power load supply should also not be produced more than its demand because it will be a waste as power load cannot be stored and this will cost the electrical utility company. Therefore, load forecasting should be done before the power load supply is produce. The objective of this project is to pre-process the load data taken from Peninsular Malaysia and to forecast the load based on previous load data. Neural network is the approach used in this project to forecast the load data. Long Short-Term Memory (LSTM) is a particular type of recurrent neural network that are designed to avoid the long-term dependency problem. Its strength is in remembering the information for long period of time. Therefore, it is suitable for the load data used in this project as the load data reading is for three months. The three months load data are being grouped by the same day in order to observe the load consumption pattern. Then, LSTM structure is modeled to forecast the load data more accurately and considered acceptable when the Root Mean Squared Error (RMSE) is equal or lower than 100. In this project, the energy load data for a week is successfully forecasted via Neural Network approach which is Long Short Term Memory (LSTM) technique and the best LSTM structure that give the least error and predict accurately is successfully identified

Actions (login required)

View Item