TY - JOUR
T1 - Data optimization on the accuracy of forecasting electricity energy sales using principal component analysis based on spatial
AU - Iswan, Iswan
AU - Garniwa, Iwa
AU - Surjandari, Isti
N1 - Publisher Copyright:
© 2021, Econjournals. All rights reserved.
PY - 2021
Y1 - 2021
N2 - It is very important to make forecasts to support future planning. In electricity field, for estimating the demand for electrical energy, there are several influential factors to be considered, e.g. economic growth, increased demand for electricity, and the capacity of power and electrical energy providers. The limited availability of data and variables causes the predictions made to be inaccurate. This paper focuses on the accuracy of forecasting with various numbers of variables to optimize the data held. The initial stage of this research is the division of clusters using the hierarchical clustering method to divide 24 administrative regions into 6 clusters, and to increase the accuracy of forecasting using principal component regression. Based on the results obtained, it can be seen that the MAPE values vary in each cluster. The use of 7 variables in forecasting, in general, shows better accuracy than the use of 6 or 5 variables. However, the difference between the number of these variables is narrow. In cluster 6, the MAPE value in 7 variables is 0.88% while in 5 variables the MAPE value is 0.91%. In cluster 1 and cluster 4, the use of 5 variables has a better value than the use of other variables. Thus, this model can be used and developed to do forecasting even though it uses limited data and variables.
AB - It is very important to make forecasts to support future planning. In electricity field, for estimating the demand for electrical energy, there are several influential factors to be considered, e.g. economic growth, increased demand for electricity, and the capacity of power and electrical energy providers. The limited availability of data and variables causes the predictions made to be inaccurate. This paper focuses on the accuracy of forecasting with various numbers of variables to optimize the data held. The initial stage of this research is the division of clusters using the hierarchical clustering method to divide 24 administrative regions into 6 clusters, and to increase the accuracy of forecasting using principal component regression. Based on the results obtained, it can be seen that the MAPE values vary in each cluster. The use of 7 variables in forecasting, in general, shows better accuracy than the use of 6 or 5 variables. However, the difference between the number of these variables is narrow. In cluster 6, the MAPE value in 7 variables is 0.88% while in 5 variables the MAPE value is 0.91%. In cluster 1 and cluster 4, the use of 5 variables has a better value than the use of other variables. Thus, this model can be used and developed to do forecasting even though it uses limited data and variables.
KW - Clustering
KW - Principal Component Analysis
KW - Spatial Forecasting
UR - http://www.scopus.com/inward/record.url?scp=85105140485&partnerID=8YFLogxK
U2 - 10.32479/ijeep.11010
DO - 10.32479/ijeep.11010
M3 - Article
AN - SCOPUS:85105140485
SN - 2146-4553
VL - 11
SP - 215
EP - 220
JO - International Journal of Energy Economics and Policy
JF - International Journal of Energy Economics and Policy
IS - 3
ER -