ELKHA : Jurnal Teknik Elektro.
Vol.
18 No.
April 2026, pp.
33 - 39
ISSN: 1858-1463 .
, 2580-6807 .
Spatial Clustering of Electricity Consumption Patterns in Indonesian Higher Education Institutions Imam Arif Rahardjo.
Iwa Garniwa 2*).
Budi Sudiarto .
and Pidanic Jan.
1, 2,.
Department of Electrical Engineering.
University of Indonesia.
Jakarta.
Indonesia Department of Electrical Engineering and Informatics.
University of Pardubice.
Czech Republic Corresponding Email: *) iwa@eng.
Abstract Ae Higher education institutions represent a significant contributor to electricity consumption in the public sector, particularly in developing countries such as Indonesia.
This study aims to identify spatial patterns and provincial disparities in electricity consumption across Indonesian higher education institutions.
This research method uses spatial autocorrelation analysis with Moran's I and hierarchical clustering based on WardAos method.
The results show that the observed MoranAos I .
is higher than the expected MoranAos I (-0.
, and the spatial pattern of electricity consumption by higher education institutions is clustered.
This result is confirmed by the negligible p-value .
0003618787 < 0.
, indicating a strong clustered spatial pattern.
Hierarchical clustering was used to identify three groups of provinces based on electricity consumption levels.
The findings highlight significant regional disparities in electricity consumption patterns and provide a quantitative basis for energy management strategies and sustainable higher education policy planning in Indonesia.
Keywords: Electricity consumption, higher education institutions, spatial autocorrelation.
Moran's I, hierarchical clustering.
WardAos method.
Indonesia
INTRODUCTION
Higher education institutions are major contributors to public-sector electricity consumption in developing countries such as Indonesia due to the growing number of students and the expansion of campuses .
Several universities have implemented predictive models and renewable energy initiatives to manage energy consumption and reduce emissions .
, .
However, the growth of digital transformation in higher education with regard to IoT devices and data centers has led to increased energy consumption .
Ae.
Unregulated energy consumption not only increases operational requirements but also increases carbon emissions, thereby worsening the current global climate change dilemma .
Ae.
Data-driven spatial methodologies, especially Geographic Information Systems (GIS), are currently employed to detect regional differences in electricity consumption .
Ae.
MoranAos I is one of the many methods used for spatial analysis, which is widely used to test for spatial autocorrelation among data on energy Manuscript received 2026-01-29.
revised 2026-03-24.
accepted 2026-03-26 consumption .
Empirical studies have supported the view that MoranAos I is useful for revealing spatial clusters of energy consumption, which are often linked to geographic, demographic, and structural characteristics .
, .
Regional disparities in Indonesian higher education institutions influence the distribution of electricity consumption, reflecting infrastructure development and energy access .
Ae.
Investigations conducted in other nations, including China and India, show that spatial analysis provides important insights into improving sustainable energy management in the educational sector .
Ae.
Moreover, the implementation of GIS for the spatial analysis of electricity consumption has proven effective as a strategy for promoting SDGs, especially those related to energy efficiency and the reduction of carbon emissions .
Ae.
The analytical framework implemented in this study has significant potential to enhance energy management in IndonesiaAos higher education institutions, especially in view of the continuous growth in annual energy demand.
However, no study has systematically examined the spatial autocorrelation and provincial clustering of electricity consumption in Indonesian higher education Existing studies are mostly focused on aggregate sectoral consumption, forecasting models, or general regional clustering, without explicitly testing of whether electricity consumption in higher education institutions shows statistically significant spatial dependence across provinces.
This limitation leads to an important research gap, since understanding spatial dependence is crucial for designing regionally coordinated and evidence-based energy management strategies.
This study aims to .
measure the degree of spatial autocorrelation in the province-level electricity consumption of higher education institutions using MoranAos I, .
test the statistical significance of the result, and .
identify the spatial clusters of consumption using hierarchical clustering based on WardAos method.
The findings are expected to provide empirical evidence on whether electricity consumption patterns show statistically significant spatial dependence and to provide a spatially-informed analytical framework for - 33 - This work is licensed under a Creative Commons Attribution 4.
0 License For more information, see https://creativecommons.
org/licenses/by-nc-sa/4.
Spatial Clustering of Electricity Consumption (I.
A Rahardjo, et al.
higher education institutionsAo energy management at the provincial level.
ycuI = Average observation value of the electricity The resulting MoranAos I can be interpreted as representing five spatial patterns : weak, moderate, moderately strong, strong, and very strong, as shown in Table 1.
II.
METHODOLOGY
This study employed a quantitative methodology based on spatial analysis to measure and explain electricity consumption patterns in IndonesiaAos higher education The methodological framework was combined with statistical data.
GIS technology, and electricity consumption clustering in higher education This GIS and clustering method is computationally efficient because it uses clustering algorithms to analyze geolocation data, similar to previous studies, although it focuses on assisting occupants .
The procedural steps taken in this research are outlined Data Collection This research method begins with the collection of electricity consumption data in higher education in some Indonesian provinces.
Furthermore, data on the administrative boundaries of provinces were obtained to enable spatial analysis.
Geographic coordinates were used only for mapping and spatial visualization.
Spatial Weight Matrix After obtaining electricity consumption data for each of the provinceAos higher education institutions, a spatial weight matrix was built using a Queen contiguity approach based on provincial administrative boundaries to define spatial adjacency relationships between provinces.
Under the queen criterion, two provinces are neighbors if they share either a common boundary or a common vertex.
binary adjacency matrix was first created in which neighboring provinces were assigned a value of 1 and all other entries were assigned a value of 0.
The matrix was then row-standardized to make the province comparisons The resulting standardized row matrix was used in the calculation of MoranAos I.
Moran's I Analysis MoranAos I is an index used to detect spatial patterns in a variable.
In the current analysis.
Moran's I is computed to assess the degree of electricity consumption concentration in a given area.
In the analysis of spatial patterns, two types of Moran's I are commonly used.
They are observed MoranAos I and the expected MoranAos I.
The observed MoranAos I is the Moran index usually calculated directly from the data.
It is calculated using the following ya= ycu ycu Ocycu ycn=1 Ocyc ycycnyc .
cuycn OeycuI ).
cuyc OeycuI ) Ocycu ycn=1.
cuycn OeycuI ) Table 1.
Interpretation of MoranAos I for Spatial Autocorrelation.
Moran's I 0Ae0.
Interpretation 2Ae0.
Moderate 4Ae0.
Moderately Strong 6Ae0.
Strong 8Ae1.
Very Strong Weak Expected MoranAos I is the theoretical value of MoranAos I assuming that the data were distributed randomly.
The expected MoranAos I is given by the following equation:
= Oe a .
ycuOe1 If the observed MoranAos I is greater than the expected MoranAos I, then the observed MoranAos I is autocorrelated and forms a cluster.
However, if the observed MoranAos I is less than the expected MoranAos I, the spatial pattern is If the calculated MoranAos I is equal to the expected MoranAos I, then the spatial pattern is random.
After analyzing MoranAos I, a significance test was performed using the p-value.
This value is important in proving that the detected autocorrelation pattern is Spatial Cluster Development Spatial patterns of electricity consumption are reflected in clusters indicating differences in consumption among provincesAo higher education institutions.
These clusters help to show general patterns of electricity consumption, thereby providing an understanding of how it is used across regions.
Spatial clustering was conducted using hierarchical clustering with WardAos method to identify provinces with similar electricity consumption The squared Euclidean distance was used as a similarity measure, and all variables were standardized using Z-scores before clustering to ensure comparability.
The dendrogram was analyzed to determine the optimal number of clusters by determining a significant increase in the linkage distance .
usion leve.
, which represents a natural break between groups.
Based on the dendrogramAos cut-off criteria, three clusters were selected as they provided the most interpretable and statistically distinct grouping structure without excessive fragmentation.
Overall, the integration of spatial statistical analysis and hierarchical clustering offers a strong methodological framework for understanding regional electricity consumption patterns and supporting data-driven policy development in Indonesian higher education institutions.
A A A A A .
I = MoranAos I n = Number of observation areas .
rovinces and higher education institution.
ycycnyc = Elements of the row-standardised Queen contiguity ycuycn = Observation value of electricity consumption at the location .
-th row are.
ycuyc = Observation value of electricity consumption at the location .
he j-th column are.
- 34 - Spatial Clustering of Electricity Consumption (I.
A Rahardjo, et al.
Table 2.
Longitudes and Latitudes of Indonesian Provinces.
RESULTS AND DISCUSSION
Province Aceh North Sumatra West Sumatra Riau Jambi South Sumatra Bengkulu Lampung Bangka Belitung Riau Islands DKI Jakarta West Java Central Java Yogyakarta East Java Banten Bali West Nusa Tenggara East Nusa Tenggara West Kalimantan Central Kalimantan South Kalimantan East Kalimantan North Kalimantan North Sulawesi Central Sulawesi South Sulawesi Southeast Sulawesi Gorontalo West Sulawesi Maluku North Maluku West Papua Papua Results In the first stage of the research, electricity consumption data for higher education institutions were collected by province.
These data provide a clear overall picture of variation in electricity consumption in Indonesia and of disparities among higher education institutions.
Figure 1.
Distribution of higher education institutions and average electricity consumption .
in Indonesian provinces Data on electricity consumption in higher education institutions in various provinces in Indonesia were obtained from the official Statistik Captive Power publication distributed by Badan Pusat Statistik (BPS).
This publication offers a large amount of information related to the number of higher education institutions, which includes universities, institutes, colleges, academies, community academies, and polytechnics, measured in units, in association with the average electricity consumption measured in kilowatt-hours .
As shown in Figure 1, the bar graph depicts the number of higher education institutions, such as universities, academic institutions, community academic institutions, colleges, and polytechnics.
The line graph indicates the electricity usage across various higher education institutions in each province, by type, such as universities, institutes, academies, community academies, colleges, and This means that provinces with an urban center, such as Jakarta.
West Java, and East Java, have significantly higher electricity consumption than other Jakarta has the highest average electricity Provinces located in eastern Indonesia, such as North Maluku.
West Papua, and Papua, have a lower average electricity consumption than other regions.
The analysis of differences in electricity consumption across higher education institutions in each province is an advanced indicator of the need for spatial analysis.
Next, to obtain a geographical overview of the provinces, provincial administrative boundary data were used within a Geographic Information System (GIS) Table 2 presents data on the locations of the provinces for mapping, based on their latitude and longitude coordinates.
Longitude Latitude Spatial relationship evaluation was performed based on shared administrative boundaries between provinces using the Queen contiguity criterion.
A binary adjacency matrix was created according to this criterion, and the provinces with a common boundary or a common vertex were given the value of 1, while the non-neighboring provinces were given the value of 0.
The spatial matrix serves as the main instrument for understanding electricity consumption patterns in higher education institutions in a geographic and spatial context.
Each element in the matrix is a binary spatial adjacency indicator .
if the provinces share a common boundary or vertex, 0 if no.
Then, the matrix was row standardized to allow comparability across Row standardization was applied so that the weights of neighboring provinces were set to 1 for each With 34 provinces, a 34-by-34 matrix was developed.
To model the spatial adjacency between provinces, a 34by-34 row-standardized Queen contiguity matrix was Each element of the spatial weight matrix is the standardized adjacency relationship between provinces under the queenAos contiguity.
This spatial weighting approach is the foundation for the autocorrelation analysis by accurately calculating MoranAos I for electricity consumption in higher education institutions in each - 35 - Spatial Clustering of Electricity Consumption (I.
A Rahardjo, et al.
The calculation of MoranAos I was performed based on the row-standardized Queen contiguity spatial weight Table 3 shows the results of the calculation of the observed MoranAos I, the expected MoranAos I, and the pvalue.
Spatial analysis was performed to determine electricity consumption patterns in IndonesiaAos higher education institutions and to form homogeneous groups by province.
Variables that affect electricity consumption were included as dependent variables in each province, along with the effects of independent variables such as The clusters were produced by hierarchical cluster analysis using WardAos method to classify Indonesian provinces based on similar electricity consumption patterns.
Then, a dendrogram shows the measured distances between clusters of provinces with similar electricity consumption characteristics.
Table 3.
Statistical Results of MoranAos I Spatial Autocorrelation Analysis Parameter Observed MoranAos I Expected MoranAos I p-value Value According to the calculation results, the observed Moran's I value was 0.
This value indicates the fairly strong positive spatial autocorrelation between electricity consumption and the geographic locations of higher education institutions in each province in Indonesia.
This positive spatial autocorrelation also implies that provinces with similar electricity consumption levels .
ow or hig.
tend to be geographically close.
After obtaining the observed Moran's I value, it was compared with the expected Moran's I value to determine whether the spatial pattern of electricity consumption of higher education institutions in Indonesia is a random, clustered, or dispersed.
A random spatial pattern is specified when the observed and expected MoranAos I values are the same as the expected Moran's I value.
cluster spatial pattern is defined when the observed MoranAos I value is higher than the expected MoranAos I A dispersed spatial pattern occurs when the observed MoranAos I value is below the expected MoranAos I The calculation results in the expected Moran's I value Referring to the observed MoranAos I .
as greater than the expected MoranAos I .
, the electricity consumption pattern by higher education institutions is a cluster pattern.
This result is supported by a negligible p-value .
below This provides strong statistical evidence that a significant spatial pattern exists in electricity consumption at higher education institutions .
ejecting the null hypothesis that there is no spatial autocorrelatio.
The statistical results demonstrate the importance of spatial patterns in determining electricity consumption in higher education institutions.
Figure 2 shows the visualization of electricity consumption by higher education institutions across provinces in Indonesia.
This image attempts to simplify the results and show the regional patterns.
Figure 3.
Hierarchical clustering dendrogram of provinces using WardAos method based on electricity consumption in higher education institutions This finding can serve as a basis for designing appropriate management strategies and formulating policies for electricity consumption efficiency in higher education institutions.
Therefore, the formation of clusters of provinces with similar electricity consumption profiles among higher education institutions is a crucial step in supporting more efficient energy management and effective data-driven policies.
The dendrogram in Figure 3 illustrates the Indonesian provincesAo interrelationship with electricity consumption in higher education institutions.
A hierarchical cluster analysis was performed using WardAos method and the squared Euclidean distance to measure of how different the variables were.
Cluster 1 comprises provinces with high electricity consumption.
Some provinces of this cluster are DKI Jakarta.
Banten.
West Java.
I Yogyakarta.
Central Java, and East Java.
These provinces are national centers of academic growth, marked by many continuities in higher education and research.
From a policy perspective.
Figure 2.
Spatial pattern of electricity consumption in Indonesian higher education institutions - 36 - Spatial Clustering of Electricity Consumption (I.
A Rahardjo, et al.
these provinces are well-positioned to be hotbeds for research-based projects that will encourage advanced technology and support the Triple Helix framework .
hich includes government, industry, and academi.
Cluster 2 covers provinces with moderate electricity Provinces included in this cluster are Aceh.
North Sumatra.
West Sumatra.
Riau.
Riau Island.
Bangka Belitung.
Bengkulu.
Jambi.
South Sumatra.
Lampung, and North Kalimantan.
These regions will have a high demand for middle-skilled workers, requiring the full potential of their workforce.
However, interventions are required to assist certified training programs and strengthen partnerships between vocational schools and businesses.
Cluster 3 includes provinces with a relatively few institutions and low electricity consumption.
Several provinces in this cluster are located in eastern Indonesia, including Bali.
West Nusa Tenggara.
East Nusa Tenggara.
West Kalimantan.
Central Kalimantan.
South Kalimantan.
East Kalimantan.
Sulawesi.
Gorontalo.
North Maluku.
Maluku, and Papua.
Given the low electricity consumption, access to higher education and building infrastructure are difficult.
Policy measures for this cluster should focus on affirmative action, such as the establishment of educational facilities, the provision of scholarship programs, and the capacity building of The formation of clusters 1Ae3 demonstrates the relevance of spatial analysis in capturing the heterogeneity of regions and electricity consumption in higher education Identifying clusters with homogeneous consumption patterns within each province can provide an empirical basis for developing policy strategies to optimize electricity distribution and improve the efficiency of electricity consumption programs in higher education This spatial clustering approach can also be used to make data-driven decisions in planning and implementing improvements to IndonesiaAos electricity Discussion The results show a positive relationship between the presence of higher education institutions and electricity consumption levels, which is consistent to some extent with previous results, which found an association between the level of education and energy consumption in Indonesia .
However, fundamental differences exist in the approaches and units of analysis used.
Previous studies analyzed the impact of education level on per capita energy consumption using the STIRPAT framework and a panel econometric approach, differentiating between Java and non-Java regions.
Meanwhile, the current analysis focuses on the spatial dimension and clustering patterns of electricity consumption in higher education institutions, considering the geographical proximity between The spatial clustering pattern of higher education institutions in Indonesia demonstrates that geographical location is related to similar electricity consumption levels.
This pattern indicates that electricity consumption dynamics are influenced by regional context rather than occurring in isolation.
Thus, the spatial autocorrelation discovered using MoranAos I not only indicates a statistical relationship but also reflects the structure of higher education development, which tends to concentrate in certain areas.
The use of Geographic Information Systems (GIS) technology to analyse spatial patterns of electricity consumption in higher education institutions is very important because this integration supports data-driven decision-making, especially in priority areas.
Combining spatial methods with energy policy in Indonesia can improve electricity consumption efficiency not only in the higher education sector but also in other sectors.
Integration, a key element of this research methodology, is also relevant in strengthening national energy policies, including greenhouse gas reduction strategies.
However, a direct relationship between the management of electricity consumption in higher education and national carbon emissions cannot be directly concluded from the present analysis and requires further analysis.
The results of the provincial cluster analysis in Indonesia, based on electricity consumption in higher education institutions, were divided into three clusters.
The number of clusters aligns with the studies of Y.
Arvio, who clustered electrical load consumption using k-Means clustering .
, and Y.
Berliana et al.
, who clustered electricity consumption levels by province in Indonesia using k-Means clustering .
Thus, the objects of analysis and the variables used are substantively different from the current investigation, which focuses specifically on electricity consumption in higher education institutions.
In addition, it is also strengthened by R.
Utari et al.
who clustered electricity distribution using the mean shift algorithm .
Mulyadi et al.
, who used the mini-batch k-Means clustering algorithm .
, and M.
Farid et al.
using Density Based Spatial Clustering of Application With Noise (DBSCAN) algorithm .
, also strengthen this approach.
However, previous studies focused on clustering based on numerical similarity.
However, the approach in this study incorporates geographic linkages into the analysis structure.
Therefore, the contribution of this analysis is the integration of cluster analysis with spatial dimensions through geographical weight and autocorrelation analysis, so that grouping is determined not only by numerical similarities in electricity consumption but also by the geographical connectedness between regions.
Thus, the current research opens up the perspective of clustering in the study of electricity consumption in Indonesia by situating the higher education institutions in a more contextual spatial analysis Overall, the spatial clustering methodology of this study provides a more detailed picture of how location and educational outcomes on electricity consumption.
These findings must be placed in the context of an initial analytical basis to support the formulation of more focused policies, rather than as final or comprehensive policy These insights are helpful in developing strategies for sustainable electricity consumption, equitable educational growth, and long-term improvement planning.
While this analysis provides an overview of the spatial patterns of electricity consumption in higher education - 37 - Spatial Clustering of Electricity Consumption (I.
A Rahardjo, et al.
institutions, some limitations should be considered.
The data used are secondary data at the provincial level, so it does not fully capture the variations in electricity consumption across the institutions of one province.
Therefore, further research with more detailed provinciallevel data is required to gain a better understanding of the determinants of electricity consumption in higher education institutions.
for his help in validating the statistics used in this research.
Finally, thanks for my family for their support and motivation in finishing this research.
REFERENCES
IV.
CONCLUSION
This analysis examines the spatial distribution of electricity consumption in IndonesiaAos higher education The observed MoranAos I value is 0.
512, which is higher than the expected MoranAos I value of -0.
This implies a huge spatial clustering pattern in electricity Furthermore, the p-value 0.
-value < .
indicates a statistically significant spatial pattern.
GIS-based thematic maps reveal clusters of high electricity consumption, mainly in cities with large higher educational institutions.
Conversely, low electricity consumption clusters occur in areas with limited access to energy or operational activity.
Further analysis shows that these consumption patterns result from a combination of geographic, demographic, and operational factors.
The results show the potential utility of the geoinformation system (GIS)-based methods for understanding electricity consumption patterns on a larger scale .
, at the regional The results also categorized higher education institutions into three groups based on electricity The formation of these clusters can offer initial insights for the creation of more specific energyefficiency policies, especially in regions with a relatively high electricity consumption, while also promoting the consideration of electrification strategies in regions with a low energy consumption.
This methodological framework has the potential to make analytical contributions to datadriven energy policy discussions.
However, its use at the national level still requires further study by considering other heterogeneous variables across provinces and testing in other research contexts.
However, the present analysis has a number of limitations, in particular the use of provincial-level aggregate data, which does not adequately reflect internal variation across institutions, and the limitations of the explanatory variables.
Future research could combine other factors such as university accreditation level, lecturer functional position composition, and regional socio-economic variables to better understand the determinants of electricity Furthermore, future research could develop an analysis of energy use intensity to obtain a more comprehensive understanding of electricity consumption patterns across the types and characteristics of higher education institutions based on each provinceAos type of education.
ACKNOWLEDGMENT