Analisis Multivariat untuk Penentuan Pola Kebutuhan Beban Listrik secara Spasial Menggunakan Metode Geographically Weighted Regression

Abstract

Distribution system planning must be in accordance with the pattern of needs and load characteristics, so that the location and capacity of distribution substations can be determined. In general, the existing electrical load demand pattern approach is in macro and does not involve interactions between variables (multivariate), making it difficult to map and determine the existence of load centers. Therefore, a spatial approach to the pattern of Electric Load needs using the Geographically Weighted Regression (GWR) is a suitable method to overcome this. GWR is a linear regression method where the basic principle difference between GWR and ordinary linear regression lies in the weight of the location, so that each model will have a spatial effect. The linear regression model can be used to see the relationship between load density in each subdistrict and the most influential factor. This study uses the value of Variance Inflation Factors (VIF) as a criterion to determine the presence of multicollinearity between independent variables. The optimum weighting carried out in the GWR modelling is Adaptive Exponential with the Exponential kernel function because it has the smallest AIC value and the largest R Square that is 0.9853. The GWR analysis shows that the variables Housing, GDP, and Housing have a significant effect on Load Density in each village. While social factors do not have a significant effect on Load Density.