Mitigating Load Shedding in South Africa Through Optimized Hybrid Solar–Battery Deployment: A Techno-Economic Assessment

Abstract

South Africa’s persistent electricity shortages and recurrent load shedding remain among the most pressing challenges to national economic growth and social stability. This paper presents a techno-economic framework to assess how optimized deployment of photovoltaic (PV) and battery energy storage systems (BESSs) can mitigate these disruptions under realistic grid and regulatory constraints. Despite recent operational improvements at Eskom—including a 10-month period without load shedding in 2024—energy insecurity persists due to aging coal assets, limited transmission capacity, and slow renewable integration. Using hourly demand and solar-resource data for 2023, combined with Eskom’s load-reduction records, a Particle Swarm Optimization (PSO) model identifies cost-optimal hybrid system configurations that minimize the Levelized Cost of Electricity (LCOE) while maximizing coverage of unserved energy. Three deployment scenarios are analyzed: (i) constrained regional grid capacity, (ii) flexible redistribution of capacity across six provinces, and (iii) unconstrained national deployment. Results indicate that constrained deployment covers about 86% of curtailed load at 1.88 USD kWh⁻¹, whereas flexible and unconstrained scenarios achieve over 99% coverage at ≈0.58 USD kWh⁻¹. The findings demonstrate that targeted PV–BESS expansion, coupled with selective grid reinforcement, can effectively eliminate load shedding and accelerate South Africa’s transition toward a resilient, low-carbon electricity system.

1. Introduction

South Africa’s electricity sector, dominated by the state-owned utility Eskom since its establishment in 1923, has become increasingly unstable over the past two decades. Eskom’s Generation, Transmission, and Distribution divisions supply over 86% of the country’s electricity demand and approximately 20% of the continent’s generation, primarily through high-voltage interconnections with neighboring countries [1]. Despite this scale, the system’s fragility has intensified, culminating in record levels of load shedding in recent years.

The first major warning emerged in October 2007, when widespread power shortages exposed severe supply-demand imbalances. This led to the formal introduction of “load shedding” in early 2008—a controlled rolling blackout strategy, also known as Manual Load Reduction (MLR), designed to prevent total grid collapse. By disconnecting specific areas during high-demand periods, Eskom aimed to avoid cascading failures that could trigger a nationwide blackout, which might take up to two weeks to restore. The need for load shedding stemmed from systemic challenges: delayed policy reforms, chronic underinvestment in new capacity, and an overreliance on aging coal-fired facilities [2].

The severity of load shedding escalated dramatically between 2020 and 2023. Total hours of intentional power cuts rose from 288 h in 2020 to 6837 h in 2023—an increase exceeding 2300% [1]. The Energy Availability Factor (EAF), a key indicator of fleet reliability, declined from 77% in 2011 to below 50% by 2023, with some months falling as low as 45% [3]. This decline highlights the increasing unreliability of Eskom’s coal fleet and its inability to meet rising demand, exacerbated by unplanned outages, maintenance backlogs, and labor disputes [4].

The roots of this crisis trace back to the 1998 Energy Sector White Paper, which correctly forecasted electricity shortages by 2007 and recommended reforms to introduce private investment and cost-reflective tariffs. However, these reforms were delayed, leaving the grid overwhelmingly dependent on coal—which still accounted for 82.12% of the national energy mix in 2023 [1]. Efforts to diversify through renewable energy have progressed unevenly. The Renewable Energy Independent Power Producer Procurement Programme (REIPPPP), launched in 2010, sought to attract private capital via competitive bidding processes that balanced cost with socio-economic benefits. By 2019, REIPPPP had procured 6422 MW of renewable capacity, with 5078 MW operational, yet further expansion stalled due to grid connection constraints.

Policy frameworks such as the National Development Plan (NDP) 2030 and the Integrated Resource Plan (IRP) 2018 established ambitious targets for diversification: 29,000 MW of new generation by 2030, including 17,742 MW of wind and 8288 MW of solar. However, the Generation Connection Capacity Assessment (GCCA 2025) reveals severe bottlenecks in renewable-rich provinces like the Northern and Western Cape, where transmission lines are already saturated. Conversely, provinces with unused capacity—KwaZulu-Natal, Gauteng, Limpopo—lack equivalent renewable potential, complicating balanced integration.

Against this backdrop, 2023 emerged as the most critical year since load shedding was first introduced. Eskom implemented intentional outages on 289 of 365 days, marking the deepest reliability crisis in South Africa’s history. The socio-economic impact was immense, with estimated losses of approximately USD151.35 billion [5]. This outcome was closely tied to Eskom’s chronic financial distress: while the National Energy Regulator of South Africa (NERSA) approved substantial tariff hikes between 2008 and 2013, public backlash froze further adjustments from 2013 to 2018, depriving Eskom of revenue needed for grid maintenance and upgrades.

A partial shift began in early 2024 with the emergence of private Independent Power Producers (IPPs), which started alleviating pressure on the central grid by diversifying supply sources. Nevertheless, transmission constraints and Eskom’s financial instability continue to limit the scale and speed of this transition, leaving the system vulnerable and underscoring the urgency of structural reform.

South Africa’s ongoing electricity crisis highlights the urgent need for alternative and resilient power solutions. The central research problem addressed in this paper is:

How can photovoltaic (PV) and battery energy storage systems (BESSs) be optimally deployed, within existing grid and regulatory constraints, to mitigate load shedding in South Africa?

To address this question, the study pursues the following objectives:

• Assess grid connection capacity: Identify regions with sufficient network hosting capacity to support PV installations capable of offsetting MLR events, based on Eskom’s 2023 load-shedding records and national grid-capacity reports (e.g., GCCA 2025).
• Optimize hybrid system configuration: Develop a techno-economic optimization framework to determine cost-effective sizing and spatial allocation of PV systems and BESSs. A Particle Swarm Optimization (PSO) algorithm is employed to minimize the Levelized Cost of Energy (LCOE) while satisfying reliability criteria. The framework includes an Energy Management System (EMS) to coordinate PV generation and battery operation under realistic technical constraints.
• Incorporate regulatory and infrastructure limitations: Analyze grid bottlenecks, land-use restrictions, and policy barriers that influence regional deployment potential.
• Evaluate economic feasibility: Conduct a comparative assessment of system-wide costs, land-use trade-offs, and energy not supplied (ENS) across three deployment scenarios—(i) regionally constrained, (ii) flexible redistribution, and (iii) national-scale unconstrained deployment—to determine financial viability and alignment with national energy policy targets.

The study aims to demonstrate that optimized deployment of PV and BESSs can substantially offset or eliminate load-shedding events while ensuring economic sustainability.

This study contributes to the growing body of research on sustainable energy transition in developing economies by providing a data-driven framework for mitigating load shedding through optimized renewable energy deployment. By integrating high-resolution solar resource data, grid-capacity constraints, and techno-economic optimization, the paper delivers actionable insights for policymakers, planners, and investors seeking to enhance energy security in South Africa.

The remainder of the paper is structured as follows:

• Section 2—Literature Review: Summarizes the current state of South Africa’s electricity sector, the causes and impacts of load shedding, and previous efforts to integrate renewable generation.
• Section 3—Methodology: Describes the modeling framework, datasets, and optimization algorithm used to simulate PV–BESS deployment scenarios.
• Section 4—Results and Discussion: Presents and interprets the simulation outcomes, emphasizing spatial deployment trade-offs, cost implications, and grid-constraint effects.
• Section 5—Conclusions: Highlights the main findings, policy implications, and directions for future research on renewable integration and grid resilience.

2. Literature Review

2.1. South African Energy Context

South Africa’s power sector remains dominated by Eskom, the state-owned utility responsible for about 90% of national electricity generation, transmission, and much of distribution, alongside municipalities and IPPs. Eskom alone accounts for roughly 42% of national greenhouse gas emissions, reflecting the system’s dependence on coal [1]. In 2023, coal-fired generation contributed 82% of total electricity (165.6 TWh of 201.7 TWh) [1], supplied mainly by aging plants with a median age of 39 years and an availability factor below 50% [3]. Structural failures, deferred maintenance, and high breakdown rates continue to constrain supply, prompting chronic load shedding and underscoring the urgency of a managed transition towards cleaner energy [6].

Although renewables have expanded, their contribution remains modest. As of 2023, installed renewable capacity reached 6.28 GW, representing 13.5% of South Africa’s total capacity (49.8 GW) [1]. Wind power contributed 5.8% (11.6 TWh) of total generation, largely from the Western and Eastern Cape, while solar PV and concentrated solar power (CSP) together supplied only 3.2% [1,7]. Renewable output stagnated in 2024 due to weaker resource conditions and grid limitations rather than a lack of capacity. Hydropower remains minor (≈0.5% of generation) and highly dependent on variable inflows, mainly through pumped-storage facilities such as Ingula, Drakensberg, and Palmiet [8]. Other renewables, including biomass, contribute less than 0.2% [1]. Despite this, renewables have shown strong growth since 2020, particularly in privately financed projects.

Nuclear energy plays a stabilizing but small role. The Koeberg Nuclear Power Station, Africa’s only commercial plant, supplied 4.0% (8.1 TWh) of electricity in 2023. Recent lifetime extensions for both reactor units to 2044 and beyond, and plans for an additional 2.5 GW of nuclear capacity, indicate government commitment to retaining nuclear power as a low-carbon baseload option [9].

Spatially, electricity generation is highly concentrated in Mpumalanga, where most large coal-fired stations—such as Kusile and Medupi—are located near extensive coal reserves [1]. This regional concentration has led to severe air-quality degradation, with particulate matter (PM10) levels consistently exceeding national standards [10]. Meanwhile, renewable energy facilities remain geographically scattered, and grid infrastructure in high-resource provinces (Northern, Western, and Eastern Cape) remains underdeveloped, creating a transmission bottleneck for new generation [11,12].

2.2. Load Shedding: Causes, Trends, and Implications

The most visible symptom of South Africa’s structural energy imbalance is load shedding, a controlled curtailment of demand—formally known as Manual Load Reduction (MLR)—implemented to prevent system collapse when supply falls short. It is executed in staged blocks of 2–4 h, each stage removing roughly 1 GW from the grid (up to Stage 8, or 8 GW curtailed), affecting regions in rotation to maintain stability [13].

The persistence of load shedding stems from multiple interacting factors: an aging coal fleet with frequent unplanned outages; underinvestment in maintenance and new capacity following years of below-cost tariffs; delayed integration of IPPs; and insufficient transmission capacity linking renewable-rich regions to demand centers [1,3,12,13]. Deferred maintenance during “keep-the-lights-on” policies in the 2000s further degraded reliability.

The severity of load shedding has increased dramatically. Total hours under MLR rose from 859 h in 2020 to 6837 h in 2023, with the daily profile typically peaking around 20:00, when evening demand surges and solar generation drops [1]. High-stage events have expanded from seasonal (winter) to year-round occurrences. However, operational recovery efforts are showing results: in 2024, Eskom’s Generation Operational Recovery Plan lifted the Energy Availability Factor (EAF) from ~60% to 70%, enabling a 10-month period without load shedding (March 2024–February 2025)—the longest uninterrupted supply since 2007 [3].

The economic impact has been profound. Between 2007 and 2019, blackouts cost the economy an estimated R35 billion (≈USD2.3 billion), comparable to a quarterly GDP contraction of 5% [13]. In 2023, total losses were estimated at R2.8 trillion (~60% of GDP), including production losses, diesel backup costs, and investment disruptions [14]. These figures highlight the critical importance of accelerating renewable and storage projects that can reduce curtailment during evening peaks, particularly in constrained urban regions.

2.3. Policy and Grid Integration Framework

Policy frameworks now aim to accelerate this transition. The National Development Plan (NDP) [15] targets 29 GW of new capacity by 2030—20 GW from renewables—while the Integrated Resource Plan (IRP 2019) [16] allocates 17.7 GW wind and 8.3 GW solar PV. The flagship Renewable Energy Independent Power Producer Procurement Programme (REIPPPP), launched in 2011, has procured 6.4 GW and operationalized 5.1 GW across seven bid rounds [17]. It has generated 114 TWh of clean electricity, created 83,000 job-years, and avoided 110 MtCO₂, while mobilizing over $4 billion in Black Economic Empowerment investments [17]. Despite this, limited transmission capacity has slowed new connections—only ≈50% of procured capacity was grid-connected by early 2023 [18].

To manage grid access, Eskom’s Generation Connection Capacity Assessment (GCCA) identifies transmission availability at substation, local, and supply area levels [11,12]. By 2023, grid capacity in the Northern, Western, and Eastern Cape was nearly saturated, leaving ≈20 GW of available capacity mostly in KwaZulu-Natal, Gauteng, Limpopo, Mpumalanga, North West, and Free State. Planned transmission expansion—≈14,000 km of new lines by 2033—will relieve these constraints, though most projects will only be operational after 2027 [19,20]. A USD500 million World Bank loan supports these upgrades to unlock up to 20 GW of stranded renewable projects [20].

In summary, South Africa’s power system remains heavily coal-dependent, financially strained, and geographically imbalanced. Yet, steady growth in renewables, reforms enabling private generation, and ambitious grid-expansion programs offer a pathway to reduce load shedding, diversify generation, and build a more resilient and sustainable electricity sector.

2.4. Literature Review

The growing prevalence of load shedding across developing economies—particularly in sub-Saharan Africa—has intensified research into hybrid renewable energy systems (HRES) as a pathway to enhance energy reliability and sustainability. These systems, typically integrating PV, wind, and BESSs, are increasingly being optimized through advanced energy management strategies and metaheuristic algorithms to reduce dependence on fossil fuels and unstable grids. By synthesizing recent studies on hybrid system optimization, this review identifies current methodological trends, assesses the comparative performance of optimization techniques, and highlights gaps in applying these frameworks to load-shedding mitigation through renewable deployment—the central focus of this work.

Falama et al. (2021) [21] investigated the use of hybrid PV–battery–grid systems to mitigate load shedding in northern Cameroon. Using a Firefly Algorithm–based optimization, the authors tailored system configurations to local energy needs and demonstrated that hybrid setups could substantially reduce grid dependence and, in some cases, fully meet household demand at lower long-term cost. Their results highlight the potential of hybrid renewable systems as viable solutions for power-deficient regions in sub-Saharan Africa.

In South Africa, Ghayoor et al. (2021) [22] analyzed hybrid renewable configurations for Durban, focusing on cost optimization and grid relief. Given the city’s limited wind potential, the study identified solar PV as the most suitable source, with system sizing adapted to local resource conditions. Although their consumer-scale analysis differs from the grid-scale community systems addressed in this work, it provides useful insights into component performance, scalability, and user-oriented design in distributed renewable deployment.

Energy management is defined by ISO 50001 as a “set of interrelated or interacting elements to establish an energy policy and energy objectives, and processes and procedures to achieve those objectives” [23]. The EMS is a key process to define and coordinate the flow of the HES because it balances energy supply and demand, optimizes performance, and minimizes costs.

Bakht et al. [24] developed an integrated optimization framework to determine the optimal configuration and energy-management strategy for HRES. Their approach models the EMS as the central controller coordinating PV generation, BESSs, and diesel backup (DGen) under two operating modes—grid-connected and islanded. In grid mode, surplus grid or PV power is directed to battery charging in line with time-of-use tariffs, maximizing renewable utilization. In islanded mode, the EMS prioritizes PV generation, supports it with stored energy from the BESS, and dispatches the DGen only when necessary. The study demonstrates that an optimized EMS substantially improves operational reliability, economic efficiency, and autonomy during grid outages such as load-shedding events.

The techno-economic optimization of the HRES poses considerable hurdles owing to the system’s intricate, non-linear, and non-convex characteristics. Conventional deterministic optimization techniques generally have difficulties with these complexities, frequently experiencing convergence problems when utilized in such systems. Consequently, metaheuristic algorithms have emerged as superior techniques for optimizing HRES. These algorithms can traverse the complex and high-dimensional solution spaces commonly encountered in hybrid systems, resulting in enhanced convergence and greater accuracy of solutions. Commonly utilized metaheuristic algorithms encompass Particle Swarm Optimization (PSO), Cuckoo Search (CS), Genetic Algorithm (GA), Simulated Annealing (SA), Bee Algorithm (BA), and Grasshopper Optimization Algorithm (GOA). These algorithms are adept at tackling the complexities of optimizing hybrid systems, as they can manage the non-linearity and varying interdependencies intrinsic to energy systems.

Bakht et al. [24] also applied a GOA to design an optimal hybrid PV–wind–battery–diesel system for a residential community in southwestern Pakistan, where load shedding is frequent. The optimization minimized the levelized cost of energy (LCOE), payback period (PBP), and loss of power supply probability (LPSP). The resulting configuration achieved an LCOE of 0.0664 USD/kWh and a PBP of 7.4 years, validated through PSO. Compared with standalone diesel and UPS systems, the hybrid solution yielded lower costs, shorter payback, and reduced emissions, underscoring its suitability for regions facing chronic power shortages.

A case study at Sultan Qaboos University (Oman) [25] explored the techno-economic optimization of a PV–battery system for a sustainable street lighting network using a GA. The optimization minimized life-cycle cost (LCC) and cost per kWh while ensuring reliability through loss of power supply probability (LPSP) constraints. The optimal configuration, featuring larger PV panels and battery capacity, reduced energy cost to 0.08 USD/kWh—compared to 19.9 USD/kWh for smaller systems—and enabled the transition from 400 W HPS lamps to 80 W LEDs, cutting CO₂ emissions by 133.6 tonnes annually. The results confirm that properly sized PV–battery systems can significantly enhance the economic and environmental performance of street lighting infrastructure and are transferable to similar applications worldwide.

Maleki et al. (2015) [26] evaluated several PSO variants for sizing hybrid PV–wind–battery systems in three remote regions of Iran: Rafsanjanian (Kerman), Namin (Ardabil), and Davarzan (Khorasan Razavi). The study compared PSO with constriction factor (PSO-CF), adaptive inertia weight PSO (PSO-W), Tabu Search (TS), Simulated Annealing (SA), and Harmony Search (HS) algorithms. Results showed that PSO-CF achieved the most robust optimization performance, effectively balancing cost and reliability. Given Iran’s high solar irradiance and relatively low wind speeds, the authors found PV–battery configurations most suitable for most regions, while PV–wind–battery systems were recommended where wind resources are stronger. The study highlights the superior efficiency and adaptability of PSO-based optimization for designing reliable hybrid renewable systems.

Wali et al. (2023) [27] investigated the economic optimization of hybrid PV–battery systems using the PSO algorithm to enhance cost-effectiveness and reliability. Two configurations were analyzed: an existing baseline and a PSO-optimized system. The optimized setup achieved an 80.36% reduction in the (LCOE—from 0.056 to 0.011 USD/kWh—while maintaining 100% supply reliability with a 3.3 kW PV array and a 1 kWh battery. Beyond economic gains, the optimized design also contributed to significant reductions in greenhouse gas (GHG) emissions. The study demonstrates the strong potential of PSO-based frameworks to identify cost-optimal configurations in hybrid renewable systems, promoting both financial sustainability and environmental performance.

Kefale et al. (2021) [28] applied Selective Particle Swarm Optimization (SPSO) to determine the optimal design and placement of grid-connected PV systems within Ethiopia’s radial distribution networks. Conducted at Bahir Dar University, the study addressed recurring power interruptions caused by overloading and external interferences such as wind, animals, and falling trees. The multi-objective optimization sought to minimize power losses and improve voltage profiles, demonstrating that integrating PV systems into radial feeders can significantly enhance network reliability and performance. The findings confirm SPSO’s effectiveness in optimizing PV sizing and siting for developing power grids.

Within this context, battery storage turns out to be of keen interest in the optimization problem. The role of battery storage technologies within the HRES was extensively investigated in the past. Jindal and Shrimali (2022) [29] examined the cost competitiveness of renewable energy coupled with battery storage compared to new coal-fired generation in India. Beyond assessing economic feasibility, the study explored policy frameworks and procurement mechanisms to accelerate the adoption of renewables with storage. Using forecasted LCOE for both technologies, the analysis showed that from 2022 onward, renewable energy integrated with battery storage becomes cost-competitive with new coal power, highlighting a pivotal shift toward cleaner and more sustainable generation in India’s power sector.

Hassan (2021) [30] developed a computational optimization model for a photovoltaic system integrated with battery storage and hydrogen fuel cells, achieving a significant improvement in renewable energy contribution—from 32% to 96% annually—while reducing electricity costs to approximately 0.12 USD/kWh. The results underscore the synergistic potential of hybrid PV–battery–hydrogen systems to enhance energy self-sufficiency and cost efficiency in renewable power generation.

Thango and Bokoro (2022) [31] reviewed battery energy storage applications for photovoltaic systems in South Africa, highlighting the technical and economic viability of lithium-ion (Li-ion) batteries as an effective solution to mitigate load shedding. Their analysis identified several operational challenges—such as output power smoothing, load shifting, frequency regulation, PV plant dispatchability, and energy arbitrage—that can be alleviated through advanced control and energy management algorithms. The authors further emphasized the need for comprehensive techno-economic assessments of battery storage systems under varied operating configurations to inform large-scale deployment strategies.

Understanding the techno-economic potential of PV systems in South Africa is essential for maximizing their integration with BESSs. To achieve this, geospatial assessments using Geographic Information Systems (GIS) and solar atlas data are essential. For instance, Harrucksteiner et al. (2023) [32] conducted a comprehensive geospatial analysis of solar and wind potential in Mongolia, integrating GIS tools with high-resolution global datasets such as the Global Solar Atlas and SolarGIS. The study emphasized that combining global resource databases with local topographical information significantly enhances the accuracy of renewable energy simulations and supports the identification of optimal sites for PV deployment.

Optimization frameworks for integrating renewable energy sources, particularly PV systems, into existing power grids or microgrids have been studied in different contexts. The optimization models typically aim to improve system reliability, reduce its costs, and minimize its environmental impacts. For instance, Kassab et al. (2024) [33] proposed a multi-objective mixed-integer linear programming (MILP) model to optimize microgrids integrating PV systems, BESSs, and optional grid connections. The algorithm simultaneously minimized the LCOE and Life Cycle Emissions (LCE) over a 25-year horizon, incorporating hourly energy management for both long-term planning and short-term operation. The results demonstrated that integrating renewable systems with advanced optimization frameworks substantially enhances energy reliability while reducing dependence on fossil fuels.

Although numerous studies have optimized HRES using metaheuristic algorithms, most focus on small-scale or off-grid applications. Their findings rarely account for grid integration, hosting capacity, or regulatory constraints, which are essential to large-scale deployment in contexts such as South Africa. Moreover, few works explicitly link optimization results to load-shedding mitigation, often relying on reliability metrics instead of quantifying the actual reduction in unserved energy. Spatial and policy-related factors—such as grid bottlenecks, land-use limits, and regional regulation—also remain underrepresented in techno-economic analyses.

This study addresses these gaps by integrating solar resource mapping, grid-capacity assessment, and techno-economic optimization within a unified PSO-based framework. It quantifies the potential of optimized PV–BESS deployment to offset MLR events, while evaluating economic and spatial trade-offs under real-world technical and regulatory conditions relevant to South Africa’s electricity transition.

3. Models and Methods

3.1. Overview of the Methodology

This study develops a techno-economic optimization framework for designing an HRES—integrating PV generation and BESSs—to mitigate the impacts of load shedding across South Africa in 2023. That year recorded the highest frequency and duration of load-shedding events in the country’s history, with severe socio-economic consequences [1,13]. Although 2024 was characterized by a ten-month period without load shedding, this improvement predominantly reflects short-term operational measures introduced under Eskom’s Generation Operational Recovery Plan, rather than a structural transformation of the generation and transmission portfolio. Consequently, 2023, as a conservative yet representative “stress-test” year, provides a complete record of the worst-observed reliability conditions, while demand profiles and solar resource patterns remain broadly consistent with those of the current system, ultimately making it more adequate for testing the robustness of the HRES configurations.

The proposed framework quantifies how strategically deployed PV and storage resources can offset curtailed demand while accounting for spatial differences in solar potential, demand patterns, and grid hosting capacity [12].

The national-scale HRES model distributes PV and BESS capacities across multiple supply areas, each characterized by specific solar resources, hourly load-shedding profiles, and transmission connection limits defined in the Generation Connection Capacity Assessment (GCCA 2025) [12]. The objective is to minimize the LCOE while maximizing coverage of MLR demand—representing unserved energy during load-shedding periods—under realistic technical and regulatory constraints.

Step 1—Data Collection and System Definition.

Hourly MLR data from Eskom [1] provide the national unmet demand profile, while regional PV generation potential is simulated using the pvlib Python library with meteorological inputs (GHI, ambient temperature, and wind speed). Spatial constraints include maximum grid connection capacity per region [12] and land-availability restrictions derived from the EGIS South Africa National Portal [34]. Technical and economic parameters for PV and BESS technology (efficiency, costs, degradation, and depth of discharge) are obtained from recent literature and industry reports.

Step 2—HRES Modeling and Energy Management.

The system comprises three components: PV generation, BESS storage, and the national grid. During load-shedding events, the HRES supplies curtailed demand—PV generation is dispatched first, followed by BESS discharge if required. When PV output exceeds demand, surplus energy charges the batteries. The dispatch strategy operates at an hourly time step to balance renewable generation with MLR demand while adhering to grid capacity constraints.

Step 3—Optimization Using Particle Swarm Optimization (PSO).

A PSO algorithm determines the optimal regional capacities of PV systems and BESSs by minimizing the LCOE subject to reliability, grid, and land-use constraints. PSO is well suited to this multi-dimensional, non-linear problem, as it efficiently explores the solution space without requiring differentiable objective functions.

Step 4—Scenario Evaluation.

Three deployment scenarios are evaluated to explore trade-offs between technical feasibility and cost:

• Scenario 1—Constrained Grid and Land Availability: Enforces all grid and land-use limits, reflecting current conditions.
• Scenario 2—Relaxed Grid Constraints: Maintains land restrictions but allows greater inter-regional energy transfers.
• Scenario 3—Unconstrained Deployment: Removes all spatial limitations, representing a theoretical upper bound for performance and cost minimization.

Each scenario yields optimal PV and BESS configurations per region and corresponding key performance indicators (ENS, LCOE, NPC, and land footprint), allowing for comparison of technical and economic outcomes across varying infrastructure assumptions.

Figure 1 shows a schematic of the entire model.

3.2. Data Collection

3.2.1. Load Profile

The analysis covers the full 2023 calendar year using hourly demand and supply data from Eskom [1]. South Africa’s electricity demand is largely met through Eskom’s mix of dispatchable generation and self-dispatched renewables. When Available Dispatchable Capacity falls below the Republic of South Africa Contracted Demand, MLR is implemented as a last resort to maintain grid stability.

The MLR dataset quantifies the hourly curtailed demand—the portion of electricity unmet due to supply shortages—and thus defines the operational space the proposed HRES must address. While pre-emptive demand-side management actions are not explicitly recorded, their effects are embedded in the residual demand and MLR values.

Seasonal variations in demand, generation availability, and renewable output contribute to fluctuations in the supply gap. This hourly unmet load serves as the target demand for HRES compensation, enabling the model to quantify how PV and battery systems can reduce reliance on load shedding and enhance grid reliability.

From a data-quality perspective, the 2023 MLR series used in this study is obtained directly from Eskom’s operational data portal via a data request platform (https://www.eskom.co.za/dataportal/data-request-form/, accessed on 4 December 2025). The dataset consists of 8760 hourly records (one for each hour of the calendar year) that have already undergone internal validation by the utility. Timestamps were checked, and the series was screened for negative or repeated entries; no gaps, duplicates, or anomalous values were detected, so no additional cleaning, gap-filling, or temporal aggregation was required.

Although load-shedding events are triggered at local or sub-regional level, the MLR data released by Eskom represent the national shortfall that the system is unable to serve after standard dispatch and demand-management measures have been exhausted. The 2023 MLR time series is therefore used as a single aggregate unmet-load profile in the optimization. Spatial differentiation in the model arises instead from the regional allocation of PV and BESS capacity (Section 3.2.2 and Section 3.3.2).

3.2.2. Regional Available Capacity

Regional analysis focuses on areas with available grid connection capacity and land suitability for PV deployment. According to the Generation Connection Capacity Assessment (GCCA 2025) [12], South Africa has approximately 19.94 GW of available connection capacity distributed across key provinces: KwaZulu-Natal (5500 MW), Gauteng (4680 MW), Limpopo (3360 MW), Mpumalanga (3320 MW), North West (1660 MW), and Free State (1420 MW).

Geospatial and environmental constraints were identified using the Environmental Geographic Information System (E-GIS) from the Department of Forestry, Fisheries, and the Environment (DFFE) [34]. From an initial land area of 60.26 million ha, successive exclusions for conservation, protected, and urbanized zones yielded a final usable area of 45.99 million ha for potential PV development. KwaZulu-Natal exhibits the largest available area (11.5 million ha), while Gauteng has the least (2.62 million ha) due to dense urbanization.

Spatial filtering of suitable land was implemented via a vector-based GIS workflow using polygon feature classes provided by the E-GIS platform (land cover, protected and conservation areas, urban/settlement fabric, water bodies and other restricted categories). All layers were reprojected to a common coordinate reference system and checked for topology errors (e.g., slivers, overlaps) prior to analysis. Starting from the national land polygon, a sequence of overlay operations was applied to remove non-eligible areas: (i) exclusion of polygons tagged as protected or conservation areas; (ii) exclusion of urban and peri-urban classes, including their prescribed buffer zones; and (iii) exclusion of residual non-buildable classes (water bodies, excessive slopes and other restricted land-cover types). The resulting polygon layer defines environmentally and technically available land for utility-scale PV. To incorporate grid accessibility, this layer was intersected with a 5 km Euclidean buffer generated around high-voltage substations and transmission lines, retaining only polygons within this buffer. These “PV-eligible” polygons were then spatially aggregated by supply area to derive regional land-availability constraints used in the optimization.

This spatial filtering process ensures that subsequent optimization results are technically feasible and environmentally compliant, balancing land availability, grid proximity, and solar resource potential across South Africa’s supply regions.

In this study, grid constraints are represented as fixed and empirical limits, derived directly from the Generation Connection Capacity Assessment (GCCA 2025). These limits specify the maximum allowable renewable injection capacity at each transmission supply area. The model treats these capacities as exogenous and time-invariant throughout the 2023 operational simulation, reflecting current grid-connection conditions rather than dynamic or forecasted future states.

3.3. Technical Model

3.3.1. PV Output Model

Hourly PV generation for each region was simulated using the Photovoltaic Geographical Information System (PVGIS) developed by the Joint Research Centre of the European Commission [35]. PVGIS provides satellite-derived and ground-validated solar radiation data of high spatial resolution, making it suitable for large-scale modeling in South Africa. Simulations were performed through the pvlib Python library, which interfaces with the PVGIS API and implements the PVWatts performance model. The PVGIS Typical Meteorological Year (TMY) product is adopted, which returns a continuous hourly time series; as a result, no missing values occurred and no temporal gap-filling or interpolation was required.

For each supply area, the PV-eligible land defined in Section 3.2.2 is used as the spatial domain for site selection. Within this domain, a set of representative sample locations is drawn, ensuring that each point lies on land that is technically suitable for utility-scale PV deployment. At each sample location in region r, an hourly unit power profile $P_{unit}\left(t\right)$ is simulated for a fixed-tilt monofacial crystalline-silicon system (tilt = 30°, azimuth = 0°, north-facing) using the SARAH3 irradiance database. This configuration follows the recommendations by Asowata et al. (2012) [36] for South African conditions.

The regional PV profile is then obtained as the spatial average of the unit profiles associated with all sample locations in that region. In practice, the hourly capacity-factor time series for region r is computed as the mean of the simulated unit profiles over all PV-eligible sites, so that each region is represented by a single, representative profile capturing the average temporal behavior of PV output under local resource and siting constraints.

In the optimization, the installed PV capacity in region r is treated as a decision variable $C_{PV}$. For any given solution of the PSO algorithm, the corresponding hourly PV power output in region r is obtained by scaling the regional unit profile by this capacity,

$$P_{PV,r}\left(t\right)=C_{PV,r}{P}_{unit,r}\left(t\right)$$

(1)

so that changes in optimized capacity directly affect the magnitude of regional generation while preserving the underlying temporal shape determined by the PV resource and siting constraints.

Representative sites were selected within 5 km of existing high-voltage substations or transmission lines to ensure grid accessibility under the constrained-grid scenarios. Each site’s output was simulated for a fixed-tilt monofacial crystalline-silicon system (tilt = 30°, azimuth = 0°, facing north) using the SARAH3 irradiance database. The configuration follows recommendations by Asowata et al. (2012) [36].

3.3.2. Battery Energy Storage Model

Battery dynamics follow the formulation of Bakht et al. (2022) [24]. The state of charge (SOC) evolves as:

$$SOC\left(t\right)=SOC\left(t-1\right)\left(1-\sigma \right)+{\eta }_{Bat}{P}_{Bat}\left(t\right)$$

(2)

where $\sigma$ is the self-discharge rate (assumed 0), ${\eta }_{Bat}$ the charge/discharge efficiency (0.92 per process, 85% round-trip), and $P_{Bat}\left(t\right)$ the charging (+) or discharging (−) power at hour $t$.

Hourly charging and discharging limits are:

$$+P_{Bat}\left(t\right)={\displaystyle \frac{\left(SO{C}_U-SOC\left(t\right)\right)C_{Bat}{N}_{Bat}}{\mathrm{\Delta }t}}$$

(3)

$$-P_{Bat}\left(t\right)={\displaystyle \frac{\left(SOC\left(t\right)-SO{C}_L\right)C_{Bat}{N}_{Bat}}{\mathrm{\Delta }t}}$$

(4)

subject to:

$$SO{C}_L\le SOC\left(t\right)\le SO{C}_U$$

(5)

where $SO{C}_U$: Upper limit of the state of charge (90%); $SO{C}_L$: Lower limit of the state of charge (10%); $SOC\left(t\right)$: Current state of charge; $C_{Bat}$: Capacity of a single battery (in kWh); $N_{Bat}$: Total number of batteries in the BESS; $∆t$: Time step (one hour).

3.3.3. National Energy Balance

At each time $t$, total national electricity demand $P_{Load}\left(t\right)$ is met by the grid, PV generation, and battery storage, with residual shortages captured as load shedding $P_{Shed}\left(t\right)$:

$$P_{Load}\left(t\right)=P_{Grid}\left(t\right)+\sum _r{P}_{PV,r}\left(t\right)+P_{Bat}\left(t\right)+P_{Shed}\left(t\right)$$

(6)

where $P_{Grid}\left(t\right)$ is the central-grid supply, $P_{PV,r}\left(t\right)$ is the power generated by PV systems in region $r$, and $P_{Bat}\left(t\right)$ the net battery flow (positive = charge, negative = discharge). This balance quantifies the fraction of demand met by the grid, renewable and storage resources and defines load shedding as the residual unmet load, providing the key performance metric for evaluating HRES effectiveness.

3.4. Economic Model

3.4.1. Levelized Cost of Electricity (LCOE)

The economic model aims to minimize the LCOE by optimizing the sizing of the PV and BESS components, balancing investment, operational costs, and energy production to ensure a reliable and cost-effective solution compared to conventional alternatives. LCOE is an economic assessment of the average total cost to build and operate a power-generating asset over its lifetime divided by the total energy output of the asset over that lifetime. It is expressed in [$/MWh] and is calculated as:

$$LCOE={\displaystyle \frac{C_0+{\sum }_{t=1}^n\frac{O\&M_t}{{\left(1+r\right)}^t}+{\sum }_{t=1}^n\frac{D_t}{{\left(1+r\right)}^t}+{\sum }_{t=1}^n\frac{P_t}{{\left(1+r\right)}^t}}{{\sum }_{t=1}^n\frac{E_t-L_t}{{\left(1+r\right)}^t}}}$$

(7)

where $C_0$: Initial capital cost, which includes the cost of the PV system and the battery system; $O\&M_t$: Annual operating and maintenance costs, including PV and BESS technology; $D_t$: Replacement cost of the BESS, which occurs at specified intervals during the project lifetime; $P_t$: Penalty cost associated with unserved energy (load shedding) in year $t$, calculated as the product of unmet demand and a fixed penalty rate (LSC); $E_t$: Total electrical energy produced by the system in year $t$, including PV production and energy discharged from the battery; $r$: Discount rate; $n$: Project lifetime (in years).

3.4.2. Load Shedding Cost

In this study, the HRES is modeled as a replacement for the grid during outages, and its economic contribution is quantified through the avoided Cost of Unserved Energy (CoUE) rather than conventional operational savings. The CoUE, as reported by the National Energy Regulator of South Africa (NERSA) [37], represents the financial loss per unit of unsupplied energy (R/kWh) (R stands for Rand, the South Africa monetary unit) and is converted to USD/MWh for international consistency.

The analysis distinguishes between direct and indirect components of CoUE [37]. The direct CoUE reflects immediate production losses, expressed as forgone Gross Value Added (GVA) per kWh across 62 ISIC (International Standard Industrial Classification)-classified economic sectors and municipalities. The indirect CoUE accounts for secondary impacts such as supply chain disruptions and reduced economic activity.

Between 2013 and 2022, NERSA reported a significant increase in CoUE—from approximately 100 R/kWh to 140 R/kWh [37]. The 2021–2022 value, 140.37 R/kWh, corresponding to 7.76 USD/kWh, is adopted as the baseline parameter. This metric is embedded into the optimization framework through the instantaneous load shedding cost:

$$LSC\left(t\right)=CoUE E_{shed}\left(t\right)$$

(8)

where $LSC\left(t\right)$ is the cost of unserved energy at time t (USD) and $E_{shed}\left(t\right)$ is the unsupplied energy (kWh).

This term is incorporated into the multi-objective optimization to ensure that the economic implications of unmet demand are explicitly captured, guiding the design of the HRES toward configurations that minimize the overall financial impact of load shedding.

3.4.3. HRES Component Characteristics

The techno-economic analysis is based on the following parameters. The PV system is modeled with an overall installed cost of $758/kW and an O&M cost of $17/kW-year over a 20-year lifetime [38,39]. For the estimation of the photovoltaic area in South Africa, it is assumed that a 1 kWp installation covers an area of 6.5 m² [40]. The battery energy storage system (BESS) is modeled as a 4 h device with a nominal capacity of 1800 Wh per unit. Its capital cost is assumed to be $476.74/kWh (approximately $1906.95/kW for a 4 h system), with a fixed O&M cost of $47.674/kW-year, a round-trip efficiency of 85%, and an expected lifetime of 15 years [41]. In addition, a discount rate of 11% is used over a 30-year project life [42].

3.4.4. Sensitivity to Cost Assumption

A simple cost sensitivity test is carried out to assess the influence of techno-economic assumptions on the optimization results. The analysis is restricted to the scenario considered most relevant for policy and planning purposes, namely the configuration that minimizes LCOE while achieving very high load-shedding coverage (reference scenario).

Cost uncertainty is represented through a dimensionless scaling factor $s$ that is applied uniformly to all PV and BESS capital and fixed O&M cost components. This approach preserves the internal cost structure between technologies and cost categories, while isolating the effect of an overall upward or downward shift in investment costs on the optimal LCOE and capacity mix. The range $s\in \{0.8,0.9,1.0,1.1,1.2\}$ corresponds to a ±20% variation around the baseline values.

For each value of $s$, all PV and BESS cost components (capital expenditure and fixed O&M) are multiplied by the corresponding factor, while all technical parameters, demand and PV profiles, and the penalty cost for unserved energy are kept unchanged. The PSO-based optimization is then repeated using the same hyperparameters and stopping criteria as in the main analysis, yielding an updated set of regional PV and BESS capacities together with the corresponding LCOE and annual load-shedding coverage. The case $s=1.0$ coincides with the reference scenario reported in the results, whereas $s=0.8$ and $s=1.2$ represent a ±20% perturbation around the baseline cost assumptions.

3.5. Integrated National Energy Management Flow

The integrated system optimizes power use from the grid, PV, and battery, ensuring the load is met while maintaining safe battery operation and prioritizing grid power.

3.5.1. Hierarchy-Based Flow

At any time $t$ the respective hierarchy is followed as shown also in the Flow Chart in Figure 2.

This integrated energy management flow effectively balances the usage of grid power, PV generation, and battery storage to meet the load demand. It prioritizes grid power, utilizes PV generation when available, and resorts to battery discharge only when necessary. The system ensures that the battery operates within safe SOC limits, preventing over-discharging and overcharging. It also adjusts charging and discharging actions based on the availability of surplus power and the load requirements.

3.5.2. Optimization Formulation

The optimization framework seeks to minimize the total economic cost of the HRES by jointly considering the LCOE and the CoUE. This dual-objective approach ensures that energy generation costs are minimized while the economic impact of load shedding is reduced, promoting both affordability and reliability.

The optimization simultaneously adjusts the installed capacities of the PV system and BESS, along with their operating parameters, to balance supply and demand throughout the year. The model allows for slightly higher generation costs when this results in significant reductions in unmet demand.

Key assumptions include constant cost structures for PV and BESS technologies, a single battery replacement at year 15 within a 30-year project lifetime (discounted to present value), and fixed regional grid connection capacities over the study period.

Objective Function

$$\min LCOE\left(x\right),x\in X$$

(9)

where vector $x\in X$ represents the collection of all decision variables that can be adjusted to minimize the LCOE.

Constraints

The optimization is subject to the constraints presented in Equation (2) (Battery State-of-Charge Dynamics), Equation (5) (Battery Capacity Limits, Equation (6) (National Energy Balance), and the Non-negativity Constraints (Equation (10))

$$P_{PV}\left(t\right)\ge 0, P_{Bat}\left(t\right)\ge 0, P_{Shed}\left(t\right)\ge 0$$

(10)

Decision Variables

The decision variables include:

• Total installed PV capacity in each region (in kW). This capacity is adjusted based on the solar potential of each region.
• Total installed battery storage capacity in each region (in kWh). This determines the energy storage available to balance the PV output and load demand.
• Total power output from the PV systems (in kW), which depends on both the installed capacity and the solar irradiance.
• Power charged to or discharged from the battery (in kW). This controls how the battery storage is used to balance supply and demand.
• State of charge of the battery (in kWh), which evolves dynamically during charging and discharging and must remain within specified limits.

3.5.3. Implementation of the PSO Algorithm

The HRES optimization is performed using the Particle Swarm Optimization (PSO) algorithm, a metaheuristic method effective for determining the optimal sizing of PV and BESS capacities. The objective is to minimize the LCOE while satisfying all technical and regional constraints.

Each particle in the swarm represents a candidate solution defined by PV and BESS capacities for each region. The algorithm iteratively updates particle positions and velocities based on individual and global best solutions, guiding the swarm toward the optimal configuration. The fitness of each particle is evaluated through energy dispatch simulations, with the LCOE serving as the objective function.

In this study, the PSO hyperparameters are configured as follows. The swarm consists of 30 particles ($N_{PARTICLES}=30$), and the maximum number of iterations is set to 30 ($MA{X}_{ITER}=30$), in line with previous applications of PSO to hybrid renewable-energy systems. The inertia weight and acceleration coefficients are fixed to $w=0.5$ and $c_1=c_2=1.5$. Each particle encodes the regional PV and BESS capacities, with decision variables bounded between $P{V}_{min}=0$ and ${PV}_{max}=9\times {10}^9$ kW for PV, and between $BES{S}_{min}=0$ and ${BESS}_{max}=9\times {10}^9$ kWh for storage. Particle positions are initialized uniformly at random within these bounds, while initial velocities are drawn from a uniform distribution scaled to the corresponding decision-variable range. At each iteration, particle positions and velocities are updated according to the standard global-best PSO velocity and position update equations, and any infeasible solution is projected back onto the feasible region defined by the technical and grid constraints.

Convergence is reached when successive iterations yield negligible improvements in LCOE or when the maximum number of iterations is attained. The final global best solution identifies the PV and BESS capacities that minimize the system’s overall cost. The implementation follows the configuration and parameterization approach proposed by Lo et al. [43].

3.6. Validation/Model Robustness

Although no real-world PV–BESS-based technology comparable in scale and purpose to the national scenarios modeled in this study currently exists in South Africa or internationally, each component of the modeling framework builds upon methodologies that have been validated in previous peer-reviewed studies. PV generation is simulated using PVGIS, which has been extensively validated across multiple climatic regions. Battery degradation and dispatch models follow widely adopted formulations tested in both academic and industrial applications. The optimization layer relies on PSO, an algorithm that has been benchmarked in numerous techno-economic studies of hybrid renewable systems. For these reasons, while direct empirical comparison is not feasible at national scale, the proposed framework remains grounded in established and validated modeling foundations.

4. Results and Discussion

This section evaluates the performance of the HRES under three deployment scenarios designed to capture the influence of grid and land-availability constraints on system outcomes. The analysis compares total HRES generation (PV + battery) with the load to be supplied—represented by Eskom’s MLR—while grid supply is omitted from the figures for clarity.

• Scenario 1—Constrained grid: Represents current national conditions, where generation connection capacity follows the GCCA 2025 limits. The scenario tests how effectively PV and BESS technology can alleviate load shedding within existing grid constraints.
• Scenario 2—Partially relaxed grid: Assumes unlimited grid connection capacity for the six identified constrained supply areas but maintains realistic land-use limits for PV deployment.
• Scenario 3—Fully unconstrained grid: Extends the unlimited-capacity assumption nationwide, allowing unrestricted PV installations across South Africa to reveal the theoretical upper bound of renewable integration for mitigating load shedding.

4.1. Scenario 1: Constrained Grid Conditions

Under the grid limitations established by the Generation Connection Capacity Assessment (GCCA 2025) [12], the HRES operates near the upper connection limits in regions such as Mpumalanga, Limpopo, and KwaZulu-Natal. The optimization yields (see

Table 1. Regional Breakdown of Hybrid System Components and Costs for Scenario 1.

Region	PV Capacity (MW)	Battery Capacity (MWh)	PV Cost (M$)	Battery Cost (M$)	Total Cost (M$)
Free State	14	525	11	251	262
Gauteng	4680	29,566	3547	14,095	17,643
KwaZulu Natal	55	158	42	75	117
Limpopo	3360	19,975	2547	9523	12,070
Mpumalanga	3320	26,376	2517	12,575	15,091
North West	1659	11,293	1259	5384	6642
Total	13,089	87,893	9921	41,802	51,723

) an installed capacity of 13.09 GW of PV and 87.89 GWh of battery storage, requiring approximately 13,089 ha of land (assuming 6.5 m² per kW). Despite these efforts, grid constraints restrict full utilization of South Africa’s solar potential.

The system achieves an annual load-shedding coverage of 85.8%, leaving 2.35 TWh of unserved energy (ENS) and resulting in a LCOE of 1.88 USD/kWh and a net present cost (NPC) of 227.8 billion USD. The total initial capital investment (CAPEX) for PV and BESS technology amounts to 51.7 billion USD, while the NPC includes discounted O&M costs, component replacements (batteries at year 15, PV modules at year 20), and the penalty cost of unserved energy over the 30-year lifetime.

The optimization prioritizes Gauteng, Limpopo, and Mpumalanga, each receiving between 3 and 4.7 GW of PV and over 19 GWh of storage capacity. The Free State, despite strong solar potential, receives minimal allocation (14 MW PV) due to grid bottlenecks and limited cost-effectiveness.

Seasonal results show significant variability in performance (Figure 3). The system performs weakest in February and May, when PV provides only ≈55% of the demand and batteries supply ≈18%, leaving ≈28% unmet. By contrast, during October and December, PV covers more than 85% of the curtailed load, benefiting from peak irradiance.

The hourly dispatch for February (Figure 4) reveals the persistent gap between evening demand and available supply, with batteries unable to fully compensate after sunset.

Overall, Scenario 1 confirms that current grid connection limitations prevent full deployment of renewable capacity, constraining the HRES’s ability to eradicate load shedding. These findings underscore the need to explore scenarios with expanded or unconstrained grid capacity to leverage South Africa’s solar resource potential more effectively.

4.2. Scenario 2: Relaxed Grid Constraints

In Scenario 2, grid connection limits defined in the Generation Connection Capacity Assessment (GCCA 2025) [12] are removed for the six supply regions analyzed in Scenario 1, while land-use restrictions remain. This relaxation enables the hybrid renewable energy system (HRES) to exploit South Africa’s solar potential more effectively and meet nearly all load-shedding demand.

The optimization converged after 30 PSO iterations to a minimum LCOE of 0.586 USD/kWh, substantially lower than the 1.88 USD/kWh of Scenario 1. The system now achieves 99.31% load-shedding coverage, serving 16.45 TWh out of 16.56 TWh total curtailed demand, leaving only 0.69% (113.8 GWh) of unserved energy (ENS). The improvement reflects both higher renewable utilization and a sharp reduction in penalty costs associated with unserved energy.

The optimal configuration deploys 30.30 GW of PV capacity and 66.09 GWh of battery storage, occupying roughly 30,300 ha—just 0.07% of the land identified as suitable for PV installations. The regional breakdown in

Table 2. Regional Breakdown of Hybrid System Components and Costs for Scenario 2.

Region	PV Capacity (MW)	Battery Capacity (MWh)	PV Cost (M$)	Battery Cost (M$)	Total Cost (M$)
Free State	100	0	76	0	76
Gauteng	10,000	23,097	7580	11,011	18,591
KwaZulu Natal	100	0	76	0	76
Limpopo	10,000	20,013	7580	9541	17,121
Mpumalanga	10,000	22,975	7580	10,953	18,533
North West	100	2	76	1	77
Total	30,300	66,087	22,967.11	31,506.39	54,474

highlights how the absence of transmission constraints allows the algorithm to prioritize regions with the most favorable generation profiles and cost–benefit ratios, notably Gauteng, Limpopo, and Mpumalanga.

Seasonal performance, shown in Figure 5, reveals that PV generation provides 75–95% of total energy supply, with batteries covering the remainder. Unserved energy remains below 1% in the most challenging months (February and May) and drops to zero from October through December, when PV output exceeds 90% of demand.

To illustrate daily operational dynamics, Figure 6 presents the hourly dispatch for February, one of the most demanding months. Despite lower irradiance, the HRES nearly eliminates load curtailment—PV supplies most daytime demand while batteries sustain evening consumption, leaving only minimal residual shortages.

Overall, Scenario 2 demonstrates that removing regional grid connection limits enables the HRES to cover nearly all load-shedding demand, reducing the LCOE by almost 70% relative to Scenario 1. These results emphasize the crucial role of transmission reinforcement and grid flexibility in unlocking South Africa’s solar potential and achieving large-scale reliability improvements.

4.3. Scenario 3: Unconstrained National Deployment

Scenario 3 represents a fully unconstrained framework in which all South African regions are assumed to have unlimited grid connection capacity and ideal transmission infrastructure. This theoretical scenario establishes the upper boundary of PV and BESS integration, offering insight into the long-term potential of large-scale renewable deployment to eliminate load shedding nationwide.

The optimization allocates approximately 35.28 GW of PV capacity and 64.35 GWh of battery storage (

Table 3. Regional Breakdown of Hybrid System Components and Costs for Scenario 3.

Region	Optimized PV Capacity (MW)	Optimized Battery Capacity (MWh)	PV Cost (M$)	Battery Cost (M$)	Total Cost (M$)
Eastern Cape	100	0	76	0	76
Free State	10,000	0	7580	0	7580
Gauteng	10,000	20,550	7580	9797	17,377
Hydra Central	100	0	76	0	76
KwaZulu Natal	100	0	76	0	76
Limpopo	10,000	24,938	7580	11,889	19,469
Mpumalanga	100	0	76	0	76
North West	100	0	76	0	76
Northern Cape	4679	18,864	3547	8993	12,540
Western Cape	100	0	76	0	76
Total	35,279	64,352	26,742	30,677	57,419

), resulting in 99.67% load-shedding coverage—a slight improvement over the 99.31% achieved in Scenario 2. The corresponding LCOE decreases marginally (to 0.583 USD/kWh), reflecting the near-complete elimination of unserved energy and efficient spatial balancing across the system.

The optimized configuration broadens PV deployment beyond previously constrained areas, particularly into the Northern Cape, which now contributes substantially to total generation despite its distance from major demand centers. This spatial diversification improves system balance and highlights the potential of high-irradiance provinces for future grid expansion.

Seasonal performance, shown in Figure 7, indicates near-complete coverage throughout the year. Only February registers a minor 0.59% unserved energy, with PV supplying 79.5% of demand and batteries providing 19.9%. From March to December, the system achieves full load coverage with PV contributing above 85% in most months.

Hourly dispatch results for February (Figure 8) further confirm system robustness: battery discharge effectively mitigates evening demand peaks, leaving negligible unserved energy (<1%). Compared with Scenario 2, the improved temporal balancing demonstrates the benefits of expanded spatial distribution and higher aggregate storage capacity.

Overall, Scenario 3 establishes the theoretical upper limit of renewable integration in South Africa, where complete load-shedding mitigation is technically feasible under unconstrained grid conditions. While idealized, these results emphasize that achieving such performance in practice would require major transmission reinforcement, strategic regional diversification, and nationwide energy storage deployment, paving the way for a resilient, fully renewable power system.

4.4. Sensitivity to Cost Assumptions

Within the terms of the study, a sensitivity test is applied to Scenario 2, considered as the best compromise among the 3 scenarios considered. Within the term of the study, it delivers the best compromise between low LCOE and the load-shedding coverage and feasibility. The outcomes of the sensitivity experiment are illustrated in the accompanying sensitivity plot (Figure 9). For the baseline case ($s=1.0$), the reference scenario yields an LCOE of 0.586 USD/kWh with 99.31% load-shedding coverage, deploying 30.30 GW of PV and 66.09 GWh of storage.

When PV–BESS costs are reduced by 20% ($s=0.8$), the optimal LCOE falls to 0.513 USD/kWh, and coverage increases to 99.77%, while total PV capacity rises to approximately 40.24 GW and BESS capacity to 69.91 GWh. For a 10% cost reduction ($s=0.9$), the LCOE reaches 0.556 USD/kWh with 99.73% coverage, and the optimal configuration includes about 32.39 GW of PV and 76.50 GWh of storage.

When costs are increased ($s=1.1$ and $s=1.2$), the optimal LCOE rises to 0.667 and 0.849 USD/kWh, respectively. In both cases, the system still compensates more than 99.3% of curtailed demand, with total PV capacity ranging between roughly 30.30 and 36.34 GW and BESS capacity between about 69.69 and 94.16 GWh. Overall, the sensitivity results indicate that the reference scenario is robust to plausible ±20% variations in PV and BESS costs: absolute LCOE values change as expected with cost scaling, but load-shedding coverage remains above 99% and the main conclusions regarding the effectiveness of PV–BESS deployment and the relative performance of the scenarios are unchanged.

4.5. Comparative Analysis of Scenarios

The results demonstrate a clear progression in performance as grid and land constraints are relaxed, highlighting the decisive role of transmission capacity in enabling large-scale renewable deployment.

Scenario 1—which reflects current grid limitations—shows restricted PV and battery expansion, with 13.09 GW of PV and 87.89 GWh of storage achieving only 85.81% annual load coverage. The constrained configuration leads to an elevated LCOE of 1.88 USD/kWh and an NPC of 227.8 billion USD, driven by high unserved energy penalties and limited use of high-irradiance regions.

Scenario 2, removing connection restrictions in the six primary supply areas, markedly improves both cost and reliability. The optimized system installs 30.30 GW of PV and 66.09 GWh of storage, attaining 99.31% load coverage and reducing the LCOE to 0.586 USD/kWh. The NPC falls to 82.2 billion USD, indicating a 64% cost reduction relative to Scenario 1. Optimization favors Gauteng, Limpopo, and Mpumalanga, where solar potential and load profiles are best aligned, confirming the efficiency of targeted regional deployment strategies once infrastructure bottlenecks are alleviated.

Scenario 3 removes all national constraints, extending deployment to all provinces and revealing the theoretical upper bound of system performance. With 35.28 GW of PV and 64.35 GWh of storage, the system reaches 99.67% load coverage and achieves a further decline in LCOE to 0.583 USD/kWh. Although only marginally more cost-effective than Scenario 2, this configuration demonstrates the long-term benefits of a fully interconnected and spatially diversified grid, with significant contributions from the Free State and Northern Cape.

Overall, the comparison confirms that grid expansion and spatial flexibility are far more influential than total installed capacity in determining system performance. While Scenario 3 offers an idealized vision of nationwide optimization, Scenario 2 represents a realistic and economically balanced pathway toward eliminating load shedding through regional reinforcement and optimized PV–battery integration.

4.6. Discussion

4.6.1. Cost–Reliability Trade-Offs

Scenario 2 provides the most realistic balance between reliability and affordability. Although its LCOE (0.586 USD/kWh) exceeds current utility-scale PV–battery benchmarks (≈0.23–0.24 USD/kWh), this premium reflects the near-complete reliability (99.31% coverage) and inclusion of penalty costs for unserved energy (7.76 USD/kWh). Even minimal shortages—such as the 0.69% unmet load—exert a strong influence on total cost. Nevertheless, the economic impact of load shedding, estimated at 151 billion USD in 2023, indicates that the social and economic benefits of avoiding blackouts outweigh the incremental generation cost.

Comparing the LCOE of Scenario 2 with Eskom’s retail tariff of 0.12 USD/kWh highlights that these figures represent fundamentally different cost structures: tariffs are end-user prices including distribution and taxes, whereas LCOE reflects producer-side investment and operational costs. The comparison, however, remains valuable for policy design—it shows that paying a higher LCOE may be economically justified when it purchases reliability and prevents large-scale economic losses.

4.6.2. Policy, System, and Non-Technical Barriers

The modeled results align closely with South Africa’s Integrated Resource Plan (IRP 2019) and the Just Energy Transition Investment Plan (2023–2027), both of which target large-scale renewable deployment and significant storage investment. The optimized configuration in Scenario 2 mirrors these national objectives, indicating that strategically located PV-BESS platforms, combined with selective transmission reinforcement, can deliver near-complete mitigation of load shedding under realistic grid conditions.

However, the feasibility of scaling such deployments is not determined solely by techno-economic performance. Several non-technical barriers may also affect the pace and success of implementation. First, regulatory and institutional processes remain complex, particularly regarding project permitting, environmental approval, and grid-connection licensing, often leading to multi-year delays. Second, political and governance constraints, including uncertainty in energy policy direction, Eskom’s financial instability, and fragmented responsibilities across national and municipal entities, can hinder coordinated expansion of renewable capacity. Third, financing constraints pose challenges: although Scenario 2 is economically justified at the national scale, upfront investment requirements are substantial, and investor confidence may be affected by evolving tariff structures, sovereign-risk perceptions, and currency volatility. Finally, social and distributional considerations, such as land-use conflicts, local opposition, and equitable access to grid upgrades, can shape the spatial allocation and acceptance of large PV deployments.

Taken together, these factors indicate that eliminating load shedding requires not only optimized system design but also strengthened regulatory clarity, streamlined permitting, predictable long-term procurement mechanisms, and targeted public–private financing frameworks. The results of this study therefore support a policy emphasis on:

• Strategic grid reinforcement in high-irradiance and high-demand provinces (Limpopo, Gauteng, Mpumalanga, Northern and Western Cape).
• Accelerated deployment of large-scale battery storage, which is critical for mitigating evening shortfalls.
• Institutional reforms and stable regulatory conditions that reduce approval times and enhance investment certainty.
• Integration of reliability valuation into tariff and procurement frameworks, recognizing the substantial socio-economic cost of load shedding.

4.6.3. Concluding Remarks

In summary, the comparative analysis shows that Scenario 2 represents the most pragmatic pathway for South Africa’s near-term energy transition—balancing cost, reliability, and infrastructure feasibility. While Scenario 3 defines the long-term vision of a fully interconnected renewable grid, the results collectively emphasize that eliminating load shedding does not necessarily require a theoretical ideal but rather a strategically optimized HRES supported by focused grid investments and policy alignment.

This study adopts simplified and static grid-connection capacity constraints based on the GCCA 2025 report. While these represent the best available planning estimates, they do not capture dynamic operational factors such as short-term congestion, maintenance outages, or future transmission reinforcements not yet reflected in official forecasts. As a result, the optimization outcomes may under- or overestimate feasible PV and BESS deployment in specific regions. Future work should incorporate more granular and time-varying grid constraints to refine these estimates and better represent operational uncertainty.

5. Conclusions

South Africa’s 2023 electricity crisis, marked by nearly 16,500 GWh of unserved energy across 6940 h of load shedding, underscored the urgent need for resilient and decentralized power solutions. This study developed and applied a techno-economic optimization framework based on PSO to evaluate the potential of hybrid PV and BESS technology to mitigate load shedding under varying infrastructure and land constraints. Three deployment scenarios were analyzed to quantify the trade-offs between cost, reliability, and feasibility, providing an integrated assessment of how renewable-based hybrid systems can strengthen South Africa’s energy resilience.

Under Scenario 1, which reflects existing grid limitations, the system installed 13.09 GW of PV and 87.89 GWh of storage, covering 85.81% of annual load-shedding demand. The high LCOE (1.88 USD/kWh) and NPC (227.8 billion USD) highlight the economic inefficiencies of constrained infrastructure and overreliance on battery storage. In contrast, Scenario 2, which relaxes grid capacity limits while maintaining land restrictions, achieved 99.31% coverage, an LCOE of 0.586 USD/kWh, and an NPC of 82.2 billion USD, demonstrating the benefits of spatially optimized PV–BESS deployment in key provinces such as Gauteng, Limpopo, and Mpumalanga. The fully unconstrained Scenario 3 marginally improved coverage (99.67%) and cost (0.583 USD/kWh) but at the expense of significantly greater land-use and infrastructure complexity, suggesting diminishing returns beyond regional optimization.

The results confirm that optimized hybrid renewable deployment can offset nearly all MLR under realistic assumptions. Scenario 2’s configuration—requiring roughly 30 GW of PV and 66 GWh of storage—represents the most balanced and economically viable pathway to eliminating load shedding. Expanding grid capacity in high–solar-resource regions such as the Free State and Northern Cape further enhances long-term potential, though incremental, strategically phased transmission investments remain the most cost-effective approach. Although Scenario 2’s LCOE exceeds current retail tariffs (0.12 USD/kWh), it delivers near-complete reliability and prevents the estimated 151 billion USD annual economic loss associated with blackouts, making such investments socio-economically justified.

Ultimately, this study demonstrates that a regionally optimized HRES, supported by targeted grid reinforcement and policy alignment with South Africa’s Just Energy Transition Investment Plan (JET-IP), provides a technically sound and economically justified route toward a more resilient and low-carbon power sector.

Future work will extend the proposed framework to incorporate additional renewable sources (such as wind and concentrated solar power), and to model the progressive reinforcement of transmission infrastructure through a phased expansion approach. Furthermore, we plan to adopt a stochastic formulation that captures uncertainty in solar resources, demand fluctuations, and equipment availability. These extensions will enable a more comprehensive evaluation of long-term planning options for South Africa’s renewable-driven power system.