TY - GEN
T1 - Building Deep Neural Network Model for Short Term Electricity Consumption Forecasting
AU - Chandramitasari, Widyaning
AU - Kurniawan, Bobby
AU - Fujimura, Shigeru
PY - 2019/3/22
Y1 - 2019/3/22
N2 - Electricity consumption forecasting has a main role in the energy supply management system of a power supply company. A power supply company needs to keep the balancing of the electricity demand and supply for their customers. The target is to forecast the electricity consumption in manufacturing company for each 30-minutes in the next day to prevent the lack of electricity supply from a power supply company. Due to this problem, it is the challenge for short term electricity time series consumption forecasting. In this work, we proposed the model of deep learning neural network with approach the combination of Long Short-Term Memory (LSTM) and Feed Forward Neural Network (FFNN) to perform the electricity forecasting. This proposed method (LSTM-FFNN) was implemented in the time-series data of electricity consumption on a manufacturing company. In our experiment, we used LSTM to perform the time-series forecasting by using historical data of electricity consumption, and we performed FFNN along with additional information which represented by one-hot encoding shape to increase the forecasting performance. Experimental results showed that LSTM-FFNN gave the better result as we compared with our baseline which is the original LSTM and Moving Average (MA) based on the Root Mean Squared Error (RMSE) score.
AB - Electricity consumption forecasting has a main role in the energy supply management system of a power supply company. A power supply company needs to keep the balancing of the electricity demand and supply for their customers. The target is to forecast the electricity consumption in manufacturing company for each 30-minutes in the next day to prevent the lack of electricity supply from a power supply company. Due to this problem, it is the challenge for short term electricity time series consumption forecasting. In this work, we proposed the model of deep learning neural network with approach the combination of Long Short-Term Memory (LSTM) and Feed Forward Neural Network (FFNN) to perform the electricity forecasting. This proposed method (LSTM-FFNN) was implemented in the time-series data of electricity consumption on a manufacturing company. In our experiment, we used LSTM to perform the time-series forecasting by using historical data of electricity consumption, and we performed FFNN along with additional information which represented by one-hot encoding shape to increase the forecasting performance. Experimental results showed that LSTM-FFNN gave the better result as we compared with our baseline which is the original LSTM and Moving Average (MA) based on the Root Mean Squared Error (RMSE) score.
KW - deep neural network
KW - electricity forecasting
KW - feed-forward neural network
KW - long short-term memory (LSTM)
KW - short term electricity forecasting
UR - http://www.scopus.com/inward/record.url?scp=85064158399&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85064158399&partnerID=8YFLogxK
U2 - 10.1109/SAIN.2018.8673340
DO - 10.1109/SAIN.2018.8673340
M3 - Conference contribution
AN - SCOPUS:85064158399
T3 - Proceeding - 2018 International Symposium on Advanced Intelligent Informatics: Revolutionize Intelligent Informatics Spectrum for Humanity, SAIN 2018
SP - 43
EP - 48
BT - Proceeding - 2018 International Symposium on Advanced Intelligent Informatics
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2018 International Symposium on Advanced Intelligent Informatics, SAIN 2018
Y2 - 29 August 2018 through 30 August 2018
ER -