Open AccessShort term forecasting of electrical consumption using a neural network: joint approximate diagonal eigenvalue
Author(s) -
Mashitah Mohd Hussain,
Zuhaina Zakaria,
Nofri Yenita Dahlan,
Nur Iqtiyani Ilham,
Mohamad Zhafran Hussin,
Noor Hasliza Abdul Rahman,
Md Azwan Md Yasin
Publication year - 2022
Publication title -
indonesian journal of electrical engineering and computer science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.241
H-Index - 17
eISSN - 2502-4760
pISSN - 2502-4752
DOI - 10.11591/ijeecs.v26.i1.pp56-66
Subject(s) - computer science , load profile , profiling (computer programming) , artificial neural network , electricity , term (time) , electrical load , joint probability distribution , jade (particle detector) , data mining , engineering , mathematics , statistics , artificial intelligence , electrical engineering , physics , particle physics , quantum mechanics , voltage , operating system
This article aims to estimate the load profiling of electricity that provides information on the electrical load demand. In achieving this research implemented the neural network algorithm of joint approximate diagonalisation of eigen-matrices (JADE) to describe the load profile pattern for each point. Nowadays, utility providers claim that natural sources are used to generate power by rising consumer demands for energy. However, occasionally utility workers need to know the demand at certain location, corresponding to maintenance issues or for any shutdown area involved. A distribution pattern based on the data can be predicted based on the incoming data profile without having detailed information of certain load bus, the concept of derivatives was relevant to forecast the types of distribution data. The model was constructed with load profile information based on three different locations, and the concept of derivative was recognized, including the type of incoming data. Historical data were captured from a selected location in Malaysia that was proposed to train the JADE algorithm from three different empirical distributions of consumers, recording every 15 minutes per day. The results were analyzed based on the error measurement and compared with the real specific load distribution feeder information of needed profiles.