Next Article in Journal
InvMOE: MOEs Based Invariant Representation Learning for Fault Detection in Converter Stations
Previous Article in Journal
Vacuum Processability of Self-Assembled Monolayers and Their Chemical Interaction with Perovskite Interfaces
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Geospatial Planning for Least-Cost Electrification in Developing Countries

1
Department of Energy, Politecnico di Milano, 20156 Milano, Italy
2
Tractebel France, 92230 Paris, France
3
Engie Impact Belgium, 1000 Brussels, Belgium
*
Authors to whom correspondence should be addressed.
Energies 2025, 18(7), 1784; https://doi.org/10.3390/en18071784
Submission received: 1 March 2025 / Revised: 21 March 2025 / Accepted: 27 March 2025 / Published: 2 April 2025
(This article belongs to the Section F: Electrical Engineering)

Abstract

:
This paper presents two innovative procedures developed for rural electrification planning. To address the challenges of processing vast geospatial data, handling complex and computationally intensive network design, and making detailed yet accessible economic assessments, this work introduces a Buffering plugin for community identification and a Grid Routing and Cost Allocation plugin for network design and economic assessment, both integrated into the open-source QGIS platform. The first enables the identification of potential electrification zones through dual methodologies, while the second introduces three key processes: hierarchical clustering, a modified minimum spanning tree, and a novel cost allocation methodology that provides village-specific LCOE calculations. Testing in Zambia has proven that this approach is not only effective but also—compared to existing tools—offers significant advantages in terms of computational efficiency and accessibility, while providing practical solutions to large-scale challenges. This synergistic approach enables planners to move from granular geospatial data to actionable electrification decisions through a streamlined process. The analysis covered over 3 million buildings, grouped into 162,142 settlement clusters, and subsequently determined optimal electrification strategies for 3025 villages—40.4% connected to grid extensions and 59.6% to mini-grids—serving a total population of 18 million people.

1. Introduction and Background

Access to electricity stands at the heart of modern civilization and is essential for residential, commercial, and infrastructural operations. Despite its vital contribution to human development, this service remains inaccessible to a substantial portion of humanity, particularly in Sub-Saharan Africa [1]. The United Nations has recognized this challenge under Sustainable Development Goal 7 (SDG7), which targets universal access to affordable, reliable, and modern energy services by 2030 [2]. The African continent possesses abundant energy resources, particularly renewable sources such as wind and solar, with an estimated potential of 350 GW of hydroelectric power, 110 GW of wind energy, 15 GW of geothermal energy, and 1000 GW of solar energy [3]. Despite this potential, the continent’s electricity access rate remains among the lowest worldwide, with a strong imbalance between urban and rural areas. Only selected countries in West Africa, such as Ghana and Cape Verde, have achieved more equitable access between rural and urban regions, while the overall situation remains challenging [3].
Universal access to modern, reliable, sustainable, and affordable energy services requires innovative and holistic strategies that surpass conventional methods. Manual planning processes cannot address the scale of this challenge, given the hundreds of millions without electricity access. Geospatial analysis and complex algorithms have therefore become essential components of automated planning tools. These tools can process large volumes of data, run different scenarios, and provide optimized electrification strategies fitted to local conditions. Geographic information systems (GISs)—particularly QGIS, an open-source platform with extensive plugin capabilities—have emerged as essential tools for integrating, analyzing, and visualizing the complex spatial data required for informed electrification planning. GIS-based energy system modeling in the literature falls into three main categories. First, energy models optimize generation and the energy mix, ranging from top-down macroeconomic to bottom-up technical approaches. Second, electric models focus on network optimization through power flow studies. Finally, comprehensive models, such as GISEle [4,5,6], integrate both generation optimization and network design, enabling comparisons between grid and off-grid solutions using LCOE calculations. These integrated approaches recognize the importance of proper spatial organization in the planning process, where clustering algorithms are essential for organizing potential customers into energy communities [7]. The methodology encompasses center-based methods using central points to represent clusters (e.g., K-means, K-medians, K-medoids) [8], as well as hierarchical clustering, which builds a tree-like structure of clusters through either agglomerative or divisive approaches, offering flexibility in cluster formation without requiring predefined cluster numbers [9]. Additionally, density-based clustering methods like DBSCAN identify clusters of different shapes and sizes based on point density [7,10,11], while model-based clustering assumes data generation from probability distributions [12,13]. Recent studies demonstrate the practical application of these clustering techniques in energy planning. Ref. [14] implemented clustering algorithms as part of a combinatorial optimization method for selecting participants and optimizing solar energy allocation in renewable energy communities (RECs), achieving improved energy sharing and reduced CO2 emissions. The study demonstrated how clustering can effectively organize consumers while considering multiple objectives, including self-consumption and economic viability. In the context of Sub-Saharan Africa, clustering algorithms have proven particularly valuable. Ref. [15] applied clustering techniques to analyze technology adoption patterns in rural communities of Northern Ghana, revealing important correlations between community empowerment and renewable energy acceptance. Their work highlights how clustering can inform both technical planning and social aspects of electrification. Additionally, Ref. [16] utilized clustering methodologies to rank hybrid renewable energy systems for Ghanaian industrial applications, demonstrating how these techniques can optimize system configurations across different scales and contexts.
Grid-routing algorithms, particularly graph theory-based approaches, have played a vital role in optimizing network design for rural electrification. The minimum spanning tree (MST) algorithm, implemented through either Prim’s or Kruskal’s algorithm, has emerged as a reliable solution for minimizing total network distance while ensuring connectivity, becoming a consolidated approach in the literature [17,18,19]. This theory has been validated through several practical implementations. In Zimbabwe’s Chegutu district, Ref. [18] demonstrated the effectiveness of Kruskal’s algorithm by optimizing an electricity grid network, achieving a reduction of approximately 74 km in the length of 11 kV power lines. Similarly, in Kenya, Ref. [19] achieved savings of over 25% in system cabling costs compared to classical shortest-path routing designs through a modified MST approach. Kruskal’s algorithm has proven to be especially effective for sparse networks typical of rural areas, while Prim’s algorithm performs better with dense networks. However, these approaches have important limitations: they do not incorporate technical constraints within the procedure itself, potentially leading to electrically infeasible solutions, and the basic MST algorithm only considers point-to-point connections between villages, not accounting for possible connections to intermediate points along existing lines. Recent work proposed by [17] has addressed these limitations through artificial intelligence search methods evolved from the A* algorithm, enabling efficient inter-village power connections even in areas with complex topography and allowing trade-offs between computation speed and optimality. Alternatively, Ref. [20] presented a constraint-based local search approach for solving the edge-disjoint rooted distance-constrained MST problem. This method effectively addresses the routing problem while incorporating path-length constraints and budget limitations.
Cost analysis methodologies are critical for assessing the economic viability of energy projects in rural electrification. The Levelized Cost of Electricity (LCOE) emerges as the primary metric, reflecting the average cost per unit of electricity generated over a system’s lifetime. This enables informed decisions between grid extension and off-grid solutions. The authors of [21] used this metric to evaluate rural electrification options in North Korea, analyzing the trade-off between grid extension and hybrid energy systems (HESs). The study revealed a clear economic breakpoint: for villages located beyond 9.69–20.57 km from the existing grid, hybrid systems combining solar PV, wind turbines, lead–acid batteries, and diesel generators proved to be more cost-effective than grid extension. This type of analysis demonstrates how LCOE calculations can provide concrete decision-making thresholds for selecting between centralized and decentralized electrification solutions.
In the literature, several comprehensive tools have been developed that integrate these clustering and routing methodologies for rural electrification planning. These tools combine energy modeling (generation optimization) with electric infrastructure modeling (network design) to provide integrated planning approaches. They typically incorporate optimal generation mix determination, distribution network design, grid versus off-grid assessments, and economic evaluation through LCOE calculations. GISEle, developed by Politecnico di Milano, represents a comprehensive approach that merges energy and electric models. It employs GIS, graph theory, and terrain analysis to develop optimal electric network topology [4,5]. The tool addresses the following three critical aspects of rural electrification planning: determining optimal electrification solutions between grid and off-grid options, sizing generation portfolios for off-grid systems, and designing electric networks. The reference electrification model (REM) offers another sophisticated approach, supporting both large-scale planning and local electrification projects. REM utilizes advanced clustering and optimization algorithms to create the lowest-cost electrification systems, grouping consumers into mini-grid, grid extension, or stand-alone system clusters that minimize total system costs. The tool processes diverse inputs such as building locations, solar irradiance, topography, and infrastructure costs to generate comprehensive solutions [22]. Similarly, the open-source spatial electrification tool [23] is a decision support tool that determines the least-cost electrification technology solutions to connect communities lacking electricity access; it also estimates the required investment required to deliver the projects that will facilitate these electrification connections. Compared to other methodologies, it uses a tree search algorithm complemented by locality-sensitive hashing to identify the nearest neighbor that may be qualified for grid extension. Other tools take different methodological approaches. The open-source network planner employs a two-step process, first by calculating minimum-cost grid expansion using MST methodology and then by identifying least-cost electrification options for each community [24]. Similarly, GEOSIM, used by the World Bank in African projects, incorporates social and economic development potential in its analysis methodology [25]. This diverse ecosystem of planning tools offers important context for understanding the relative advantages and limitations of different approaches to automated electrification planning. In this paper, a novel procedure is proposed; in order to validate it and highlight its strengths and limitations—a direct comparison with GISEle is conducted.
Manual planning activities cannot address the scale of this challenge, given the complex interplay among the technical, geographical, and economic factors revealed in the literature review. While existing tools have demonstrated the effectiveness of clustering and routing algorithms individually, there remains a critical need for an integrated solution within the widely used QGIS platform. The present work aims to fill this gap by developing automated, user-friendly planning tools that make advanced methodologies accessible to all stakeholders while leveraging QGIS’s extensive geospatial capabilities and plugin ecosystem. The approach will automate workflows, streamlining electrification planning while allowing flexibility for local constraints and preferences. It will enable rapid large-scale assessments by efficiently processing geospatial data—essential for vast unelectrified regions like Sub-Saharan Africa. Data-driven decision-making will be enhanced by systematically comparing grid extension and mini-grid solutions based on technical and economic factors, leveraging QGIS’s visualization tools for clear communication with stakeholders. Integration into QGIS ensures interoperability with existing datasets and workflows, combining specialized electrification tools with broader spatial analysis capabilities.
The structure of this paper is organized as follows: Section 2 presents the modeling framework and methodology, detailing both the Buffering plugin for data pre-processing and the Grid Routing and Cost Allocation plugin for network design and economic assessment. Section 3 demonstrates the implementation and validation of these tools through a comprehensive case study in Zambia. Section 4 provides a comparative analysis with GISEle, highlighting the complementary strengths of different planning approaches. Finally, Section 5 presents conclusions and suggests future improvements for the developed tools.

2. Material and Methods

The methodology developed in this work encompasses two distinct procedures working in synergy to optimize rural electrification planning, as shown in Figure 1. To validate their effectiveness and demonstrate their practical application, these procedures have been implemented in Zambia, a country facing significant electrification challenges representative of the Sub-Saharan African context. This section delineates both the methodological framework and the specific geographical context in which it was applied for assessment purposes.

2.1. Study Setting

The framework’s application in Zambia provided an ideal testing ground for the developed tools, given the country’s significant electrification challenges. With only 47.8% of the population having access to electricity and rural electrification rates as low as 14.5%, Zambia represents the challenges faced by many Sub-Saharan African nations [26]. The country’s unique energy landscape, characterized by heavy dependence on hydropower (87.8% of electricity generation [27]) and significant untapped renewable energy potential, particularly solar power with average irradiance levels of 5.5 kWh/m2/day [28], presents both challenges and opportunities for electrification planning. While other countries such as the Central African Republic and Mauritania were also considered for this implementation based on Tractebel Engineering’s prior engagements, Zambia was ultimately selected due to the availability of high-quality geospatial data, representative established national electrification challenges strategy, and the opportunity to complement ongoing electrification efforts by the national utility Zambia Electricity Supply Corporation (ZESCO).
The implementation began with an analysis of Zambia’s existing electrical infrastructure, which comprises over 11,100 km of high-voltage transmission lines and an extensive distribution network divided into four operational divisions: Lusaka, Copperbelt, Northern, and Southern. In Figure 2, the described network is presented. The characteristics of the existing network, including 175 primary substations and approximately 18,862 MV/LV transformers, provided crucial baseline data for planning the extension of the network [29]. ZESCO’s ambitious development program, requiring investments ranging from USD 44 to 223 million annually through 2020–2039, provided a relevant context for testing the tools’ capabilities in supporting strategic planning decisions.

2.2. Proposed Geospatial Tools

The methodology developed in this work implements two specialized QGIS plugins that provide a comprehensive framework for rural electrification planning, integrating data processing, spatial analysis, and economic assessment capabilities. Each plugin addresses distinct but complementary aspects of the planning process, enabling a systematic approach to electrification decision-making.
The first plugin, the Buffering plugin, represents the initial step in the electrification planning process, designed to process raw geospatial data into meaningful settlement groups that can inform electrification strategies. The plugin supports two distinct methodologies, each tailored to different types of input data. The first methodology implements a population-based approach, which utilizes high-resolution population distribution data from Columbia University [30]. This dataset offers granular population information with a resolution of approximately 30 m, enabling accurate demographic analysis for electrification planning. The second methodology employs a building-based approach, leveraging Open Building data [31]. This comprehensive dataset includes building footprints across Africa, South Asia, and Latin America, providing detailed information about building sizes and confidence scores, which allows for infrastructure-based settlement identification. Both approaches employ a sophisticated buffering process, though with different parameters. Data quality was ensured through appropriate filtering measures for both datasets. The Columbia University population data incorporated census information and high-resolution satellite imagery to maintain demographic accuracy. For the Open Building dataset, quality control was implemented by filtering the initial 7.9 million buildings to include only those with confidence levels exceeding 80%, resulting in 3.2 million buildings for analysis while preserving comprehensive spatial coverage.
Population-based processing considers minimum inhabitant thresholds within buffer zones to determine potential electrification areas, while building-based processing evaluates minimum square meter thresholds of building areas within buffer zones for the same purpose, as illustrated in Figure 3. Despite the data source, the plugin generates geospatial layers of settlement groups representing potential service zones, which can be used for subsequent electrification planning stages. Additionally, it provides population and building density metrics for each identified group. These settlement groups serve as the foundation for planning both local infrastructure development through LV lines and larger infrastructure developments using MV lines. The processed data become critical inputs for the second plugin’s network design and cost allocation processes, as shown in Figure 1. The pseudo-code for this procedure is presented in Algorithm 1.
Algorithm 1 Pseudo-code for the Buffering plugin—building methodology
  1:
Input point building layer and set parameters (buffer distance, CRS, minimum area)
  2:
Convert buffer distance from meters to appropriate units based on CRS
  3:
Create buffer zones around building points with the specified distance
  4:
Create a spatial index for buffer zones to optimize performance
  5:
Convert multipart geometries to single-part geometries
  6:
Count number of buildings within each buffer polygon
  7:
Create a spatial index for the counted polygons
  8:
Calculate the total building area within each buffer polygon
  9:
Create a spatial index for polygons with area information
10:
Extract only buffer zones exceeding the minimum square meters threshold
11:
Output filtered building groups
Following the output from the Buffering plugin, the Grid Routing and Cost Allocation plugin advances the electrification planning process with sophisticated network design and economic assessment capabilities. This second phase begins with hierarchical clustering through agglomerative methods, enabling adaptive grouping of villages based on spatial relationships. The clustering algorithm can be refined through various linkage methods (single, complete, average, weighted) to optimize settlement grouping based on local conditions and geographical constraints. Secondly, network reticulation represents a fundamental redefinition of the traditional MST approach. While sharing some characteristics with conventional MST algorithms, this innovative approach enables connections to any point along the existing network through an iterative procedure, rather than limiting connections to predefined nodes. Formally, the modified MST approach is defined as follows: given a set of unconnected village clusters, V = { v 1 , v 2 , . . . , v n } , an existing network, N, consisting of connected segments, and a distance function, d ( a , b ) , representing the Euclidean distance between points a and b, our algorithm iteratively connects each unconnected cluster to either the existing network or to a cluster previously connected.
For each iteration, i, we select the connection ( v i , p i ) that minimizes the following:
min v i V u n c o n n e c t e d ; p i { N V c o n n e c t e d } d ( v i , p i )
where V u n c o n n e c t e d is the set of currently unconnected clusters, V c o n n e c t e d is the set of already-connected clusters, and p i can be any point along the existing network, N, or any point representing a connected cluster.
This differs from traditional MST algorithms in that the connections can be made to any point along existing line segments, not just to nodes; the algorithm follows a greedy approach based on the proximity to both the existing network and already-connected clusters; finally, the process incorporates both inter-cluster and intra-cluster connections hierarchically. Although this does not guarantee global optimality in the mathematical sense of a traditional MST, it provides a practical solution that better reflects the physical realities of grid extension planning, where intermediate connections along power lines are both technically feasible and economically advantageous.
Finally, the plugin introduces a novel allocation cost methodology that distributes grid extension costs among electrified villages based on their relative peak power demand, leading to a fair allocation mechanism, as present in Equation (2), which results from Equations (3)–(6). The allocation methodology considers several key parameters across the project’s lifetime. Each village’s energy profile is characterized by its annual energy demand ( E v , t , measured in kWh) and peak power demand ( P village , i , in kW), which directly influence the power flow in the connected line segments ( P Line , k , in kW). The economic assessment incorporates both temporal and financial factors; the analysis spans a period of years (y), applying a discount rate, r, to account for the time value of money. The cost structure includes initial investment costs ( I t ), annual maintenance expenses ( M t ), and ongoing energy costs ( C E t ), all expressed in dollars throughout the project timeline. This comprehensive approach ensures that both technical requirements and economic considerations are properly weighted in the final allocation of costs among villages.
The village-specific LCOE for grid extension is calculated as follows:
Grid LCOE v i l l a g e , i = k = 1 n % p v i , L k Line Cost NPV Energy v i l l a g e , i , NPV
The present value of the village’s annual energy demand is given by the following:
Energy village , i NPV = t = 1 y E v , t ( 1 + r ) t
The village peak power demand is calculated considering the load factor:
P village , i = Energy Demand village , i 8760 h y · Load Factor
The net present value of line costs includes investment, maintenance, and energy costs:
Line C o s t N P V = t = 1 y I t + M t + C E t ( 1 + r ) t
Finally, the cost allocation factors are determined by the ratio of a village’s peak power demand to the total peak power demand flowing through each line segment:
% p v i , L k = P village , i P Line , k
By distributing costs based on these weighting factors, % p v i , L k , each village’s contribution to the grid extension cost is directly proportional to its use of the infrastructure. This approach can be formally justified firstly by the cost recovery, that is, the total costs of each line segment are fully allocated among the villages served, ensuring financial sustainability:
i V % p v i , L k = 1 , k L
Secondly, usage proportionality, that is, villages with higher energy demands bear proportionally higher costs:
Cost allocated to village i Cost allocated to village j = P village , i P village , j
Finally, iterative optimization. This approach calculates village-specific LCOEs through an iterative process that considers various factors including energy demand, load factors, and operation and maintenance costs. The methodology incorporates a sophisticated comparison between grid extension and mini-grid options, with iterations continuing until a stable solution is reached where the cheapest option remains unchanged for all villages, as shown in Figure 4.
This comprehensive approach considers the complex interplay between individual village demands, spatial distribution, and overall network structure, providing policymakers with village-specific LCOE values rather than a single grid LCOE for entire extensions. While these values serve as investment indicators rather than actual tariffs, they enable informed decision-making about electrification strategies. The pseudo-code for this procedure is presented in Algorithm 2.
Algorithm 2 Pseudo-code for Grid Routing and Cost Allocation
  1:
Input unelectrified villages with inhabitant data and preexisting network
  2:
Perform hierarchical clustering on building centroids using distance criterion
  3:
Split clusters with fewer than three villages into individual clusters
  4:
Assign unique IDs to villages and calculate energy demand and peak power
  5:
Iteratively connect unconnected clusters to the closest connection point
  6:
for each unconnected cluster do
  7:
      Find the minimum distance to the connected network or already-connected cluster
  8:
      Add connection line to the grid-routing layer
  9:
      Connect villages within the cluster using MST
10:
      Update the connected network
11:
end for
12:
Identify lines directly connected to the preexisting network
13:
Find interconnected line groups and assign unique extension IDs
14:
Split lines at intersections and assign node IDs
15:
Calculate line lengths, energy transported, and peak power
16:
Compute NPV costs for each line segment
17:
for each extension do
18:
      Create a cost allocation table for villages and lines
19:
      Update grid LCOE for each connected village
20:
end for
21:
Compare grid LCOE with mini-grid LCOE to determine the cheapest electrification option
22:
while grid extension features change do
23:
      Recalculate with only villages where grid extension is cheapest
24:
      Update allocation tables and LCOE values
25:
      Recompare grid and mini-grid costs
26:
end while
27:
Filter the final grid to include only cost-effective connections

3. Implementation and Validation

The Buffering plugin successfully processed two distinct data sources, namely, Columbia University’s population density data and the Open Building data, achieving significant data reduction while maintaining essential information for planning purposes. The population-based methodology reduced the initial 1,570,281 groups of inhabitants to 175,302 significant settlement groups. This consolidation process maintained population distribution patterns while identifying viable electrification zones, achieving an input data reduction of 88.8%, which facilitated the subsequent analysis steps. Figure 5 shows the graphical input and output of the procedure.
The building-based methodology processed a dataset of 3.2 million buildings with confidence levels above 80%, filtered from an initial dataset of 7.9 million. The methodology consolidated these into 162,142 buffers, achieving an input data reduction of 94.9%, with an average of 19 buildings per cluster. The total building area analyzed decreased marginally from 280.7 km2 to 277.9 km2, indicating the effective preservation of significant structures while eliminating noise in the data. Figure 5 shows the graphical input and output of the procedure.
Regarding the second plugin, named Grid Routing and Cost Allocation, the hierarchical clustering analysis took the ‘localities’ layer as input, which contained concentrated areas of population or settlements made of centroid points of clustered populations needed to optimize the electrification at the MV level. Using a 1 km distance threshold and the single linkage method, the clustering algorithm revealed a key characteristic of Zambian rural settlements: their extreme dispersion caused each village to form its own cluster, even when the threshold was increased to 2 km. This choice of 1 km is justified by its technical alignment with the maximum allowable length of the LV network, which is a standard index commonly used by DSOs in rural electrification projects and supported by relevant literature [4]. This outcome highlights the unique challenges of rural electrification in Zambia, where low population density significantly impacts infrastructure planning. The grid-routing algorithm adapted to this dispersed settlement pattern by focusing on inter-cluster connections without using intra-cluster networks, which connect the different villages inside a cluster, due to each village forming its own cluster. The algorithm successfully generated a network of MV line extensions based on minimum Euclidean distance principles, connecting villages either to the existing network or to previously connected settlements, as shown in Figure 6a. This initial grid routing represents the complete network design that would be required to connect all unelectrified communities to the grid, without yet considering economic viability. The network topology is optimized to minimize the total length of lines while ensuring connectivity for all settlements.
Following this initial network design, the cost allocation analysis was performed using parameters calibrated to Zambian conditions (including an energy cost of 0.08 USD/kWh and a line cost of 25,000 USD/km), based on industry expertise and relevant references [32,33]. This analysis revealed clear patterns in optimal electrification strategies by calculating village-specific LCOEs for grid extension and comparing them against mini-grid alternatives. Villages near existing infrastructure generally favored grid extension, as their LCOEs remained below the 0.20 USD/kWh mini-grid threshold. This threshold is supported by a real-life study conducted within the United Nations Development Programme [34]. In contrast, the trend indicates that more remote settlements are better suited for mini-grid solutions, as their grid extension LCOEs often exceed this threshold. Figure 6b shows the final optimized results after the economic assessment, where only economically viable grid extensions are retained (blue lines), while more distant communities are identified for mini-grid solutions (green dots).
This analysis converged after three iterations, providing a stable and economically justified electrification strategy for each village in the study area as reported in Table 1. These results demonstrate the plugins’ effectiveness in processing complex spatial data, identifying viable electrification zones, and generating cost-optimized infrastructure plans at a country scale. The tools successfully automated critical planning decisions while maintaining transparency in the decision-making process.

4. Comparative Analysis with Existing Optimization-Based Open-Source Tools

This section presents a detailed comparison between the developed procedures and GISEle, an existing open-source tool for rural electrification planning. Through specific case studies and examples from Zambia, we examine methodological differences, computational approaches, and their practical implications for addressing rural electrification challenges.
GISEle, a tool developed by Politecnico di Milano represents a comprehensive approach to rural electrification planning that merges elements of energy and electric models. It employs GIS, graph theory, and terrain analysis to develop optimal electric network topology. The tool covers the necessary steps for rural electrification planning: finding the optimal electrification solution by choosing between on-grid and off-grid solutions, sizing the generation portfolio for off-grid systems, and electric network design. The optimization problem is formulated as an MILP model with the objective function of minimizing the overall NPC costs for the expansion project:
i , j A d N P C A d , i j x i j + c N c N P C m g , c y c + s N c p N P C c p , s z s + c C ( 1 y c ) D e , c c o e + c C ( 1 y c ) N P C g , c
where x i j , y c , and z s , respectively, represent the construction of a link between i and j, the MG electrification option for cluster c, and the effective connection to the primary substation, s. Consequently, N P C A d , i j , N P C m g , c , N P C c p , s , and N P C g , c denote the NPC for the connection lines i , j , the cost for the MG in the community, c, and the cost of using the primary substation, s. From a methodological standpoint, the optimization framework in GISEle incorporates electrical grid constraints, including maximum loading limits for power lines and substations, as well as a linearized nodal voltage formulation. The voltage drop, Δ E , in lines i-j is computed using the linear approximation in Equation (10):
Δ E = R · P + X · Q E n
where R and X denote the line resistance and reactance, P and Q denote the active and reactive power, and E n denotes nominal voltage. A complete description of the chosen comparative tool is reported in [4].
A key distinction between the approaches lies in their connection point methodology. GISEle requires predefined substation locations as starting points for network extensions, following a traditional power system planning approach. In contrast, the proposed QGIS plugins show more flexibility by allowing connections to any point along the existing MV network. This fundamental difference means the presented approach can identify potentially more cost-effective connection points anywhere along the existing infrastructure, representing a significant advantage over conventional planning methodologies. Additionally, while GISEle’s total cost calculations incorporate penalty factors as relative additional costs to the baseline cost per electric line kilometer, the proposed approach maintains a simpler, more direct cost structure. However, the present approach has limitations, namely, the cost allocation based on line usage may not fully represent infrastructure development requirements, as many lines must be built regardless of individual village decisions due to network interconnections. This contrasts with holistic optimization approaches, such as GISEle, which consider the entire system simultaneously but require significantly more computational resources and face scalability issues (see Table 2 for a detailed comparison of the two approaches).
To enable meaningful comparison, a specific study area in Zambia was selected with standardized input data for both tools. The initial grid-routing solutions revealed distinct approaches to network expansion. The QGIS plugins favored shorter, more direct routes between unelectrified villages and the existing MV network, creating a distributed expansion pattern with multiple connection points. GISEle’s approach demonstrated a more centralized planning strategy, with proposed MV lines forming longer main feeders emanating from specific substation points. Two representative examples illustrated when the additional complexity of GISEle’s approach provides essential benefits versus when the presented approach’s simpler methodology suffices:
  • Figure 7a shows a case where both tools propose essentially the same order of village connections. The figure highlights a specific region where the comparison between GISEle’s routing (shown in red) and the plugin’s routing (shown in blue) revealed striking similarities in connection patterns. Despite GISEle’s more complex approach, which follows terrain contours and likely road networks, the plugin’s simpler direct-path approach achieved comparable results, with a total length of 48 km, versus GISEle’s 53 km. This 10% difference in total length demonstrates that in areas with relatively straightforward geography, the added computational complexity of terrain-aware routing may not significantly impact the routing solution.
  • On the other hand, Figure 7b demonstrates a scenario where terrain characteristics significantly affect routing decisions. In this region, characterized by more complex topography, GISEle’s solution (red lines) follows natural corridors and road networks to avoid geographical barriers. The plugin’s approach (blue lines) proposes not only a different order of village connections but also direct paths that would be impractical to implement in the actual terrain. As a result, the plugin’s estimated extension length of 21 km substantially underestimates the required MV line length compared to GISEle’s more realistic terrain-aware solution of 30 km. This 30% underestimation highlights cases where terrain-aware routing provides critical insights into infrastructure planning.
This terrain-aware routing explains the divergences between the two solutions. This is particularly noticeable in the western portion of Figure 8, where the complex terrain clearly influences the routing decisions, leading to paths that follow more accessible corridors despite not being the shortest possible connections.
The economic assessment phase further refined the solutions of both tools, with the presented approach generally favoring a more conservative stance on grid extension decisions. This approach prefers mini-grid solutions for communities beyond a certain distance threshold, while GISEle continues to support some longer extensions that follow existing routes, as shown in Figure 7. Numerically, this can be derived from Table 3, which presents aggregated results for the two solutions. Comparing the simple point-to-point connection with an exact routing solution reveals a potentially significant cost underestimation as reported in Figure 9.
This difference arises from the tools’ distinct approaches to cost methodology and mini-grid LCOE calculation. In the presented approach, the mini-grid LCOE is fixed at 0.200 USD/kWh, while GISEle calculates village-specific mini-grid LCOEs based on local conditions. GISEle’s dynamic LCOE calculation can result in higher mini-grid costs for low-density populations, potentially making the grid extension the more economical option. The analysis concludes that both tools serve complementary purposes in rural electrification planning. The Grid Routing and Cost Allocation plugin excels in rapid, country-scale assessments with reduced data requirements, while GISEle provides more detailed analysis, particularly where terrain complexity significantly impacts implementation. This comparison suggests an optimal two-stage approach, as follows: first, using the proposed plugins for initial large-scale assessment and prioritization, which is crucial in rural electrification, followed by detailed analysis with GISEle in areas where terrain complexity requires more sophisticated planning approaches.

5. Conclusions and Future Improvements

The development and validation of two QGIS plugins represent significant contributions to advancing universal electricity access by making sophisticated planning methodologies more accessible. The Buffering plugin has proven effective in processing both population and building-based data, significantly reducing data volume while preserving essential planning information. The Grid Routing and Cost Allocation plugin has introduced key advancements, including hierarchical clustering for dispersed rural settlements, flexible network connections leveraging existing infrastructure, and village-specific LCOE calculations to optimize grid extension strategies. These tools successfully automated complex planning decisions while ensuring transparency and scalability.
The analysis demonstrated the plugins’ effectiveness in processing spatial data, identifying viable electrification zones, and generating cost-optimized infrastructure plans at a national scale. The Zambia case study highlighted their capability to handle millions of data points, producing actionable electrification strategies that balance technical feasibility with economic viability. Compared to GISEle, the developed plugins prioritize computational efficiency and accessibility, making them particularly suitable for large-scale planning. While GISEle offers more terrain-aware routing sophistication, the plugins’ user-friendly interface and ability to analyze extensive geographical areas efficiently provide significant advantages in nationwide electrification planning.
Despite these strengths, limitations remain. The Grid Routing and Cost Allocation plugin does not account for voltage drop analysis or power losses, which are crucial for electrical network planning. It also simplifies terrain considerations and applies a uniform mini-grid LCOE across all regions, potentially overlooking location-specific variations. Additionally, the absence of road network integration in routing decisions may limit the practicality of implementation. While these limitations exist, they are often difficult to integrate into a scalable framework capable of handling country-level analyses. A key conclusion of this work is the need for complementarity between the compared tools, advocating for a two-stage approach that enhances the effectiveness of rural electrification planning. Future work should focus on enhancing these tools by incorporating dynamic mini-grid LCOE calculations, integrating sophisticated terrain analysis while maintaining computational efficiency, and refining the grid-routing algorithm to optimize settlement connections. Further improvements should include low-voltage network design capabilities and a line-type determination system to enhance overall planning accuracy and practicality.

Author Contributions

Conceptualization, N.C., C.M.C., A.D., M.P., M.C. and M.M.; methodology, N.C., C.M.C., A.D., M.P., M.C. and M.M.; software, N.C., C.M.C. and A.D.; validation, N.C., C.M.C. and A.D.; data curation, N.C.; writing—original draft preparation, N.C. and C.M.C.; writing—review and editing, N.C., C.M.C., A.D., M.P., M.C. and M.M.; visualization, N.C.; supervision, M.P., M.C. and M.M.; project administration, M.P., M.C. and M.M. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors on request.

Conflicts of Interest

Author Marina Petrelli was employed by the company Tractebel France. Author Midas Caubergs was employed by the company Engie Impact Belgium. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

References

  1. World Bank Group. Tracking SDG 7: The Energy Progress Report—Access to Electricity. Data on Population Without Access to Electricity. 2023. Available online: https://trackingsdg7.esmap.org/results?p=Access_to_Electricity&i=Population_without_access_to_electricity,_million_people_(Total) (accessed on 26 March 2025).
  2. UNCTAD. Commodities at a Glance: Special Issue on Access to Energy in Sub-saharan Africa; Technical Report; United Nations Conference on Trade and Development; UNCTAD: Geneva, Switzerland, 2023. [Google Scholar]
  3. Agoundedemba, A.; Kim, C.K.; Kim, H. Renewable energy development in Africa: Progress, challenges and the way forward. Energy Strategy Rev. 2023, 45, 101029. [Google Scholar]
  4. Dimovski, A.; Corigliano, S.; Edeme, D.; Merlo, M. Holistic MILP-based approach for rural electrification planning. Energy Strategy Rev. 2023, 49, 101171. [Google Scholar] [CrossRef]
  5. Vinicius, G.T.F.; Silvia, C.; Aleksandar, D.; Massimo, B.; Marco, M. Rural electrification planning based on graph theory and geospatial data: A realistic topology oriented approach. Sustain. Energy Grids Netw. 2021, 28, 100525. [Google Scholar] [CrossRef]
  6. Caminiti, C.M.; Dimovski, A.; Edeme, D.; Carnovali, T.; Corigliano, S.; Gadelha, V.; Ragaini, E.; Merlo, M. The GISEle Framework: Innovations in Rural Electrification Planning. In Proceedings of the 2024 IEEE International Humanitarian Technologies Conference (IHTC), Bari, Italy, 27–30 November 2024; pp. 1–7. [Google Scholar] [CrossRef]
  7. Teixeira, V. GIS-Based Strategy for Distribution Grid Planning in Rural Areas. Master’s Thesis, Politecnico di Milano, Milano, Italy, 2023. [Google Scholar]
  8. Kazakovtsev, L.A.; Rozhnov, I. Application of Algorithms with Variable Greedy Heuristics for k-Medoids Problems. Informatica 2020, 44, 2–6. [Google Scholar] [CrossRef]
  9. Lattanzi, S.; Lavastida, T.; Lu, K.; Moseley, B. A Framework for Parallelizing Hierarchical Clustering Methods. In Lecture Notes in Computer Science; Springer: Berlin/Heidelberg, Germany, 2020; pp. 73–89. [Google Scholar] [CrossRef]
  10. Deng, D. DBSCAN Clustering Algorithm Based on Density. In Proceedings of the 2020 5th International Conference on Intelligent Informatics and Biomedical Sciences (ICIIBMS), Naha, Japan, 18–20 September 2020. [Google Scholar] [CrossRef]
  11. Prakash, N.; Rangasamy, S. The Concepts Directly Density Reachability, Density Reachability and Density Connectivity. Available online: https://www.researchgate.net/figure/The-concepts-directly-density-reachability-density-reachability-and-density_fig1_259461372 (accessed on 16 September 2023). [CrossRef]
  12. Renjith, S.; Sreekumar, A.; Jathavedan, M. Performance evaluation of clustering algorithms for varying cardinality and dimensionality of data sets. Mater. Today Proc. 2020, 27, 627–633. [Google Scholar] [CrossRef]
  13. Oyelade, J.; Isewon, I.; Oladipupo, O.; Emebo, O.; Omogbadegun, Z.; Aromolaran, O.; Uwoghiren, E.; Olaniyan, D.; Olawole, O. Data Clustering: Algorithms and Its Applications. In Proceedings of the 2019 19th International Conference on Computational Science and Its Applications (ICCSA), St. Petersburg, Russia, 1–4 July 2019. [Google Scholar] [CrossRef]
  14. Lazzari, F.; Mor, G.; Cipriano, J.; Solsona, F.; Chemisana, D.; Guericke, D. Optimizing planning and operation of renewable energy communities with genetic algorithms. Appl. Energy 2023, 338, 120906. [Google Scholar] [CrossRef]
  15. Agozie, D.Q.; Afful-Dadzie, A.; Gyamfi, B.A.; Bekun, F.V. Does psychological empowerment improve renewable energy technology acceptance and recommendation? Evidence from 17 rural communities. Renew. Energy 2023, 219, 119449. [Google Scholar] [CrossRef]
  16. Odoi-Yorke, F.; Owusu, J.J.; Atepor, L. Composite decision-making algorithms for optimisation of hybrid renewable energy systems: Port of Takoradi as a case study. Energy Rep. 2022, 8, 2131–2150. [Google Scholar] [CrossRef]
  17. Leonard, A.; Wheeler, S.; McCulloch, M. Geospatial Clustering and Network Design for Rural Electrification in Africa. In Proceedings of the 2020 IEEE PES/IAS PowerAfrica, Nairobi, Kenya, 25–28 August 2020. [Google Scholar] [CrossRef]
  18. Nyakudya, R.; Mhlanga, S. An Efficient Method for Rural Electrification Grid Design; Technical Report; CERN European Organization for Nuclear Research; Zenodo: Kocani, Republic of Macedonia, 2014. [Google Scholar] [CrossRef]
  19. Kebir, N.; Ahsan, A.; McCulloch, M.; Rogers, D.J. Modified Minimum Spanning Tree for Optimised DC Microgrid Cabling Design. IEEE Trans. Smart Grid 2022, 13, 2523–2532. [Google Scholar] [CrossRef]
  20. Arbelaez, A.; Mehta, D.; O’Sullivan, B.; Quesada, L. A constraint-based parallel local search for the edge-disjoint rooted distance-constrained minimum spanning tree problem. J. Heuristics 2018, 24, 359–394. [Google Scholar] [CrossRef]
  21. Kang, D.; Jung, T.Y. Renewable Energy Options for a Rural Village in North Korea. Sustainability 2020, 12, 2452. [Google Scholar] [CrossRef]
  22. Ciller, P.; Ellman, D.; Vergara, C.; Gonz’alez-Garc’ia, A.; Lee, S.J.; Drouin, C.; Brusnahan, M.; Borofsky, Y.; Mateo, C.; Amatya, R.; et al. Optimal Electrification Planning Incorporating On- and Off-Grid Technologies: The Reference Electrification Model (REM). Proc. IEEE 2019, 107, 1872–1905. [Google Scholar] [CrossRef]
  23. Isihak, S.; Akpan, U.; Bhattacharyya, S.C. Evolution of GIS-based rural electrification planning models and an application of OnSSET in Nigeria. Renew. Sustain. Energy Transit. 2022, 2, 100019. [Google Scholar] [CrossRef]
  24. Kemausuor, F.; Adkins, E.; Adu-Poku, I.; Brew-Hammond, A.; Modi, V. Electrification planning using Network Planner tool: The case of Ghana. Energy Sustain. Dev. 2014, 19, 92–101. [Google Scholar] [CrossRef]
  25. GEOSIM—Planification Géospatiale de L’électrification Rurale. GEOSIM. Available online: https://www.ied-sa.com/solutions/planification/geosim#distributed-energy (accessed on 17 March 2024).
  26. World Bank Group. Zambia—Development News. 2025. Available online: https://www.worldbank.org/en/country/zambia (accessed on 26 March 2025).
  27. International Energy Agency. Zambia—Countries & Regions. 2024. Available online: https://www.iea.org/countries/zambia (accessed on 17 March 2024).
  28. World Bank Group. Zambia—Solar Radiation Measurement Data. 2020. Available online: https://wbwaterdata.org/dataset/zambia-solar-radiation-measurement-data-0 (accessed on 26 March 2025).
  29. Market, E.; Consultants, R. Zambia Electricity Cost of Service Study; Final Report; Energy Regulation Board: Lusaka, Zambia, 2021. [Google Scholar]
  30. Center for International Earth Science Information Network (CIESIN). High Resolution Settlement Layer (HRSL). 2023. Available online: https://www.ciesin.columbia.edu/data/hrsl/ (accessed on 17 September 2023).
  31. Google Research. Open Buildings. 2023. Available online: https://sites.research.google/gr/open-buildings/ (accessed on 17 September 2023).
  32. Energy Regulation Board. Zambia Electricity Supply Corporation (ZESCO). 2024. Available online: https://www.erb.org.zm/electricity (accessed on 26 March 2025).
  33. Energy Regulation Board with the Support of the African Development Bank. Zambia Electricity Cost of Service Study Least-Cost Distribution Expansion Plan. 2021. Available online: https://www.erb.org.zm/wp-content/uploads/files/CoSS/Cost-of-Service-Study-Task-4-Least-cost-Distribution-Expansion-Plan.pdf (accessed on 26 March 2025).
  34. United Nations Development Programme. To Improve Availability of, and Access to, Financing for Renewable Energy and Energy Efficiency Initiatives in the Energy Generation and End-Use Sectors in Papua New Guinea. 2021. Available online: https://erc.undp.org/api/download?filePath=/documents/9909/mgmtresponse/keyaction/doc_5763556989437054914Report-comparison-of-costs-of-grid-extension-Vs-RE-mini-grid.pdf (accessed on 26 March 2025).
Figure 1. Flowchart of the methodology.
Figure 1. Flowchart of the methodology.
Energies 18 01784 g001
Figure 2. Existing network in Zambia representing HV and MV lines [29].
Figure 2. Existing network in Zambia representing HV and MV lines [29].
Energies 18 01784 g002
Figure 3. Comparison of settlement identification workflows: (a) uses population data, while (b) is based on building data.
Figure 3. Comparison of settlement identification workflows: (a) uses population data, while (b) is based on building data.
Energies 18 01784 g003
Figure 4. Selection procedure between the grid extension and mini-grid option.
Figure 4. Selection procedure between the grid extension and mini-grid option.
Energies 18 01784 g004
Figure 5. Population density map (left) and the buffer results in a locality in Zambia (right).
Figure 5. Population density map (left) and the buffer results in a locality in Zambia (right).
Energies 18 01784 g005
Figure 6. Comprehensive electrification planning results: (a) grid extension routing, and (b) optimal electrification strategy.
Figure 6. Comprehensive electrification planning results: (a) grid extension routing, and (b) optimal electrification strategy.
Energies 18 01784 g006
Figure 7. Comparison of electrification routing strategies: QGIS plugin (blue) versus GISEle (red).
Figure 7. Comparison of electrification routing strategies: QGIS plugin (blue) versus GISEle (red).
Energies 18 01784 g007
Figure 8. Weighted grid visualization with GISEle.
Figure 8. Weighted grid visualization with GISEle.
Energies 18 01784 g008
Figure 9. Length distribution by category after the selection of the least-cost option between the grid extension and mini-grid option.
Figure 9. Length distribution by category after the selection of the least-cost option between the grid extension and mini-grid option.
Energies 18 01784 g009
Table 1. Summary of electrification planning results for the Zambia case study.
Table 1. Summary of electrification planning results for the Zambia case study.
Case Study Statistics
Total area analyzed752,614 km2
Total population18,028,125 people
Population with grid extension9,961,217 (55.3%)
Population with the mini-grid8,066,908 (44.7%)
Settlement Statistics
Number of villages3025
Connected settlements by grid1223 (40.4%)
Off-grid Systems1802 (59.6%)
Grid Infrastructure
Total line length5235.95 km
Annual energy demand2988 GWh/year
Total NPV cost (40 years)5.7 billion $
Average grid LCOE0.107 USD/kWh
Table 2. Comparison between the Grid Routing and Cost Allocation plugin and GISEle.
Table 2. Comparison between the Grid Routing and Cost Allocation plugin and GISEle.
AspectGrid Routing and Cost Allocation PluginGISEle
Scale of AnalysisCountry-wide rapid assessment (>3000 villages)Local-scale detailed planning (100 villages)
Computational ComplexitySimplified approach for faster processingHigher complexity with detailed terrain analysis
Routing StrategyAdvanced MST with straight-line distancesLeast-cost path considering terrain and roads
Technical AnalysisBasic power demand calculationsComplete electrical analysis (voltage, loading)
Cost AnalysisVillage-specific LCOE and iterative optimizationMILP optimization with terrain-based costs
Input RequirementsBasic population and network dataExtensive geographical and technical data
Processing TimeFast computation for preliminary assessmentLonger processing for detailed solutions
User InterfaceIntegrated QGIS plugin, user-friendlyOpen-source without a user-friendly interface
Terrain ConsiderationNot considered (simplified approach)Detailed terrain analysis and weighting
Planning OutputsGrid extension vs. mini-grid decisionGrid extension vs. mini-grid decision; design of electric network; sizing of off-grid system
Table 3. Comparison between QGIS and GISEle network solutions after the economic assessment. The total cost for GISEle incorporates penalty factors, which are relative additional costs added to the baseline cost per kilometer of electric line.
Table 3. Comparison between QGIS and GISEle network solutions after the economic assessment. The total cost for GISEle incorporates penalty factors, which are relative additional costs added to the baseline cost per kilometer of electric line.
ParameterQGIS PluginsGISEleRelative Difference
Total Length [km]128.92306.89+138.05%
Total Cost [M$]3.228.98+178.75%
Number of Segments2740+48.15%
Avg Segment Length [km]4.777.67+60.80%
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Ceccato, N.; Caminiti, C.M.; Dimovski, A.; Petrelli, M.; Caubergs, M.; Merlo, M. Geospatial Planning for Least-Cost Electrification in Developing Countries. Energies 2025, 18, 1784. https://doi.org/10.3390/en18071784

AMA Style

Ceccato N, Caminiti CM, Dimovski A, Petrelli M, Caubergs M, Merlo M. Geospatial Planning for Least-Cost Electrification in Developing Countries. Energies. 2025; 18(7):1784. https://doi.org/10.3390/en18071784

Chicago/Turabian Style

Ceccato, Nicolò, Corrado Maria Caminiti, Aleksandar Dimovski, Marina Petrelli, Midas Caubergs, and Marco Merlo. 2025. "Geospatial Planning for Least-Cost Electrification in Developing Countries" Energies 18, no. 7: 1784. https://doi.org/10.3390/en18071784

APA Style

Ceccato, N., Caminiti, C. M., Dimovski, A., Petrelli, M., Caubergs, M., & Merlo, M. (2025). Geospatial Planning for Least-Cost Electrification in Developing Countries. Energies, 18(7), 1784. https://doi.org/10.3390/en18071784

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop