Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa

arXiv:2604.06371 · eess.SY, cs.SY · Submitted 2026-04-07 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: We're looking at the paper: "Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa".

Dev: Here's the gist of the paper. This paper presents a multiobjective, multiperiod global optimization framework designed for the comprehensive design, sizing, and operational dispatch of islanded hybrid microgrids situated in remote, off-grid communities within sub-Saharan Africa.

Rosa: The optimization spans a one-year horizon for system sizing and a representative day for operational dispatch, utilizing hourly resolution.

Dev: The core objective is to simultaneously minimize several critical metrics—lifecycle levelized cost of energy (LCOE), emissions, lost load, and dumped energy—while maximizing the penetration of renewable energy sources.

Rosa: The research employs Particle Swarm Optimization (PSO) as the primary metaheuristic solver, which demonstrated superior performance in terms of runtime (63 s) and solution quality (minimum normalized objective value of 0.146), leading to its adoption for all subsequent sizing results.

Title and authors: Dev: Moving on to what they actually did in this study, this paper presents a multiobjective, multiperiod global optimization framework for the design, sizing, and dispatch of islanded hybrid microgrids. It’s a comprehensive approach across one year for sizing and a representative day for operations at an hourly resolution.

Rosa: They are simultaneously minimizing lifecycle levelized cost of energy, emissions, lost load, and dumped energy while trying to maximize the penetration of renewable energy sources within that system. That’s a lot of objectives to juggle at once.

Taro: The study systematically looked at four different generation and storage combinations before settling on the optimal one. They tested different ways to mix the solar, wind, battery, and diesel components.

Dev: The key finding they highlighted is that this optimal setup achieves a normalized levelized cost of energy of zero point four six per kilowatt hour while having a renewable energy penetration exceeding ninety-four percent, which they say outperforms all other configurations they tested.

Rosa: So, the main practical implication here is that decentralized renewable energy is economically favorable in low-demand areas characterized by high connection costs, favoring systems with high renewable penetration and significantly lower emissions than relying only on fossil fuels.

Dev: They also looked at sensitivity analyses concerning fuel prices and financial discount rates specifically calibrated for sub-Saharan Africa conditions, showing how those external factors influence the design choices. That part is really important for real-world planning.

Taro: And they found that the optimal wind turbine count is robustly determined to be four across all solvers they tested, which suggests that number is a solid starting point for anyone building a similar system.

The paper's summary: Rosa: So the authors suggest a few major ways to improve this framework for real-world use. They want to move away from just one single optimization stage and integrate sizing and dispatch into one combined optimization process.

Dev: That makes sense because it lets them jointly treat technology selection, capacity sizing, the dispatch strategy, and even robustness analysis all at once in a single study tailored for sub-Saharan Africa conditions. That’s what we need for a truly integrated system model.

Taro: They also developed a two-stage sequential framework where the first stage handles the sizing using a rule-based dispatch inner loop and then the second stage solves the full multi-objective economic and environmental dispatch problem for that specific design.

Rosa: That decoupling helps manage computational complexity significantly; solving a full multi-objective dispatch optimization every single time you test a different candidate configuration in the outer loop would be too much work.

Dev: They also included parametric sensitivity analysis across fuel prices, financial factors, and storage costs calibrated to sub-Saharan Africa conditions to show designers how optimal design is affected by variations in external parameters like diesel fuel price and nominal interest rate.

Taro: That analysis led them to find that a diesel generator rating of at least approximately eleven kilowatt is needed to ensure one hundred percent reliability for this system setup, which gives engineers a concrete requirement for sizing the backup component.

Rosa: And they used those Pareto-optimal solution sets derived from these multiobjective optimization solvers to let designers visualize how the optimal minimum attainable values of those objectives vary with respect to each other.

Dev: That visualization is helpful because it shows trade-offs clearly, illustrating exactly how making a more aggressive emissions reduction or improving reliability pushes up the system's cost.

The paper's improvements: Rosa: So to wrap up this paper on "Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa," it gives us a practical blueprint for designing these systems intelligently by covering the whole lifecycle.

Dev: It shows that for these remote areas in sub-Saharan Africa we need a really thorough approach that covers everything from the initial sizing of the solar panels to how they actually run on a day-to-day basis under uncertainty.

Taro: I mean it’s interesting because it tackles uncertainty, testing what happens when things go wrong, like when there's unexpected weather or load spikes.

Rosa: Yeah, and they found that the particle swarm optimization method works really well here for finding that best balance between those competing goals in a complex way.

Dev: And they gave us some very specific numbers too, like needing a diesel generator rating of at least eleven kilowatt to make sure you have one hundred percent reliability in their baseline setup.

Taro: That's good because it gives an engineer something concrete to size the backup part based on the simulation results.

Rosa: And they show how important it is to look at things like fuel prices and financial discount rates when you’re designing for sub-Saharan Africa, because those are the biggest drivers of change in their system choices.

Dev: And that sensitivity analysis is important because it tells designers exactly how much wiggle room they have when external economic factors shift.

Taro: It confirms that decentralized renewable energy is a solid choice over just extending the main grid when you’re dealing with low demand and high connection costs around there.

Rosa: So to wrap up this paper on "Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa" gives us a practical blueprint for designing these systems intelligently.

Dev: It’s a solid framework that handles the complexity of sizing and dispatch simultaneously under uncertainty.

Taro: I just think it sets the stage well for looking at how this kind of optimization thinking applies to other complex problems in robotics, like real-time navigation under unexpected challenges.

Rosa: Exactly, because when you’re designing a system to survive in a harsh environment, that same type of optimization logic is really useful for figuring out how robots should react when the world misbehaves.

Conclusion: Rosa: So we've gone through how they used multiobjective optimization to design and size these islanded hybrid microgrids for remote, off-grid communities in sub-Saharan Africa.

Dev: Right, it’s a whole year of sizing plus a representative day of dispatch at an hourly resolution, and they were juggling so many things at once.

Taro: I still think the fact that they tested seven different optimization algorithms just to pick particle swarm optimization was pretty telling for how tricky this problem is.

Rosa: Yeah, PSO won out because it gave them that normalized objective value of zero point one hundred forty-six and it ran fast at sixty-three seconds.

Dev: Fast runtime is crucial when you need something that can actually support real-time decision making in a system like this, not just run overnight in a super slow loop.

Taro: And they found that the optimal setup uses solar panels, wind turbines, battery storage, and a diesel engine backup.

Rosa: And for the baseline setup, they ended up with fifteen solar modules and four wind turbines and one hundred sixty point five three kilowatt hours of initial battery energy capacity with ten diesel engine units.

Dev: That’s a lot of specific numbers to remember when you're trying to build out a real system, but it grounds the whole study.

Taro: What really stuck with me is that they also did this sensitivity analysis on fuel prices and financial discount rates specifically for sub-Saharan Africa conditions.

Rosa: That analysis showed how much those external economic factors affect the design choices, which is super important because those are what drive actual investment decisions in these remote places.

Dev: They found that a diesel generator rating of at least eleven kilowatt is needed to make sure you have one hundred percent reliability for this setup, which gives engineers something concrete to size the backup part.

Taro: That’s a practical requirement, not just some abstract model number from the paper.

Rosa: And they showed how important it is to look at things like fuel prices and financial discount rates when you’re designing for sub-Saharan Africa, because those are what drive actual investment decisions in these remote places.

Dev: That sensitivity analysis is important because it tells designers exactly how much wiggle room they have when external economic factors shift.

Taro: It confirms that decentralized renewable energy is a solid choice over just extending the main grid when you’re dealing with low demand and high connection costs around there.

Rosa: So to wrap up this paper on "Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa," it gives us a practical blueprint for designing these systems intelligently by covering the whole lifecycle.

Dev: It shows that for these remote areas in sub-Saharan Africa we need a really thorough approach that covers everything from the initial sizing of the solar panels to how they actually run on a day-to-day basis under uncertainty.

Taro: I mean it’s interesting because it tackles uncertainty, testing what happens when things go wrong, like when there's unexpected weather or load spikes.

Rosa: Yeah, and they found that the particle swarm optimization method works really well here for finding that best balance between those competing goals in a complex way.

Dev: And they gave us some very specific numbers too, like needing a diesel generator rating of at least eleven kilowatt to make sure you have one hundred percent reliability in their baseline setup.

Taro: That's good because it gives an engineer something concrete to size the backup part based on the simulation results.

Rosa: And they show how important it is to look at things like fuel prices and financial discount rates when you’re designing for sub-Saharan Africa, because those are the biggest drivers of change in their system choices.

Dev: That sensitivity analysis is important because it tells designers exactly how much wiggle room they have when external economic factors shift.

Taro: It confirms that decentralized renewable energy is a solid choice over just extending the main grid when you’re dealing with low demand and high connection costs around there.

Rosa: So to wrap up this paper on "Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa" gives us a practical blueprint for designing these systems intelligently.

Dev: It’s a solid framework that handles the complexity of sizing and dispatch simultaneously under uncertainty.

Taro: I just think it sets the stage well for looking at how this kind of optimization thinking applies to other complex problems in robotics, like real-time navigation under unexpected challenges.

Rosa: Exactly, because when you’re designing a system to survive in a harsh environment, that same type of optimization logic is really useful for figuring out how robots should react when the world misbehaves.

Massachusetts Institute of Technology

eess.SY, cs.SY

Submitted: 2026-04-07

Updated: 2026-11-11

Comments: Under revision

Code: https://github.com/vineetjnair9/microgrid_optimization

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 98/100

The gist: This paper presents a multiobjective, multiperiod global optimization framework designed for the comprehensive design, sizing…

Key concepts

Multiobjective Optimization
This method is used to find solutions that balance conflicting goals at the same time. In this study, it means finding a system configuration that is simultaneously cheap (low LCOE), clean (low emissions), and reliable (low lost load) without compromising other important factors like renewable energy penetration.
Particle Swarm Optimization (PSO)
PSO is the main computational tool used to search for the best possible design settings. It works like a flock of birds searching for food; it iteratively refines potential solutions by learning from good previous results, allowing the system to efficiently find a high-quality solution among many complex possibilities.
Islanded Microgrid
An islanded microgrid is an independent power system that operates entirely off the main utility grid. It must generate all its own electricity using local sources like solar panels, wind turbines, and batteries to serve the community reliably without external connection.
LCOE (Lifecycle Levelized Cost of Energy)
LCOE calculates the total cost of building and operating a power system over its entire lifespan. It is a crucial economic metric used here to compare different microgrid designs, helping designers choose the most affordable and sustainable energy solution for the off-grid community.

Terminology

Summary

This paper presents a multiobjective, multiperiod global optimization framework designed for the comprehensive design, sizing, and operational dispatch of islanded hybrid microgrids situated in remote, off-grid communities within sub-Saharan Africa. p. 1

Where the paper says it

Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa Vineet Jagadeesan Naira aDepartment of Mechanical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, 02141, MA, USA Abstract Reliable, affordable electricity remains inaccessible to over 600 million people in sub-Saharan Africa (SSA), where islanded microgrids combining renewable sources with battery storage and diesel backup represent the most viable electrification pathway. This paper presents a multiobjective, multiperiod global optimization framework for the design, sizing, and dispatch of such islanded hybrid microgrids, with a detailed case study for a remote community in Kenya.

The optimization spans a one-year horizon for system sizing and a representative day for operational dispatch, utilizing hourly resolution. p. 1
Where the paper says it

This paper presents a multiobjective, multiperiod global optimization framework for the design, sizing, and dispatch of such islanded hybrid microgrids, with a detailed case study for a remote community in Kenya. System sizing is optimized over a one-year horizon and operational dispatch over a representative day, both using hourly resolution.

The core objective is to simultaneously minimize several critical metrics—lifecycle levelized cost of energy (LCOE), emissions, lost load, and dumped energy—while maximizing the penetration of renewable energy sources. p. 1
Where the paper says it

The formulation simultaneously minimizes lifecycle levelized cost of energy (LCOE), emissions, lost load, and dumped energy, while maximizing renewable energy penetration. Among seven optimization algorithms benchmarked, particle swarm optimization (PSO) achieves the best combined runtime (63 s) and solution quality (minimum normalized objective of 0.

The research employs Particle Swarm Optimization (PSO) as the primary metaheuristic solver, which demonstrated superior performance in terms of runtime (63 s) and solution quality (minimum normalized objective value of 0. p. 1

Where the paper says it

The formulation simultaneously minimizes lifecycle levelized cost of energy (LCOE), emissions, lost load, and dumped energy, while maximizing renewable energy penetration. Among seven optimization algorithms benchmarked, particle swarm optimization (PSO) achieves the best combined runtime (63 s) and solution quality (minimum normalized objective of 0.

146), leading to its adoption for all subsequent sizing results. p. 1
Where the paper says it

Among seven optimization algorithms benchmarked, particle swarm optimization (PSO) achieves the best combined runtime (63 s) and solution quality (minimum normalized objective of 0. 146), and is used for all subsequent sizing results.

The optimal technology configuration identified is a combination of solar photovoltaic (PV), wind turbines, lithium-ion battery storage (LI), and a diesel engine (DE) backup system. p. 1
Where the paper says it

146), and is used for all subsequent sizing results. The optimal technology combination – solar PV with wind turbines, lithium-ion battery storage (LI), and diesel engine (DE) backup – achieves a normalized LCOE of 0.

This optimal setup achieves a normalized LCOE of 0. p. 1
Where the paper says it

146), and is used for all subsequent sizing results. The optimal technology combination – solar PV with wind turbines, lithium-ion battery storage (LI), and diesel engine (DE) backup – achieves a normalized LCOE of 0.

46/kWh alongside renewable energy penetration exceeding 94%, outperforming all alternative configurations explored. p. 1
Where the paper says it

The optimal technology combination – solar PV with wind turbines, lithium-ion battery storage (LI), and diesel engine (DE) backup – achieves a normalized LCOE of 0. 46 $/kWh with greater than 94% renewable penetration, outperforming all alternative configurations.

The study systematically addresses several key aspects: technology selection across four generation and storage combinations; optimal capacity sizing via a three-dimensional PSO-based global optimization over an annual 8760-hour simulation; multi-objective economic-environmental dispatch optimization for a representative day; and robustness analysis under uncertainty concerning renewable generation and load. p. 7

Where the paper says it

We present, to the best of our knowledge, the first comprehensive multi-objective microgrid sizing and dispatch optimization study specifically developed for sub-Saharan Africa conditions, with a detailed case study for a remote community in Kenya. The framework jointly addresses: (i) technology selection among four generation and storage combinations; (ii) optimal capacity sizing via a three-dimensional PSO-based global optimization over an annual 8760-hour simulation; (iii) a multi-objective economic-environmental dispatch optimization for a representative day; (iv) a systematic solver benchmark comparing seven global optimization algorithms; (v) robustness analysis under uncertainty in renewable generation and load; and (vi) a parametric sensitivity analysis across fuel prices, financial factors, and storage costs with ranges calibrated to SSA conditions.

Sensitivity analyses reveal that fuel prices and financial discount rates are the most influential parameters under Sub-Saharan Africa (SSA) conditions. p. 1
Where the paper says it

Pareto-optimal solution sets reveal key trade-offs among economic, environmental, and reliability objectives, demonstrating that cost-only optimization leads to suboptimal outcomes in emissions and reliability. Sensitivity analyses show that fuel prices and financial discount rates are the most influential parameters for SSA conditions.

The framework provides a detailed set of contributions, including the first comprehensive study tailored to SSA conditions, joint optimization for technology selection, capacity sizing using a three-dimensional PSO approach over an annual simulation, a multi-objective economic-environmental dispatch analysis for daily operations, systematic benchmarking of seven global optimization algorithms, and robustness analysis under uncertainty. p. 7

Where the paper says it

We present, to the best of our knowledge, the first comprehensive multi-objective microgrid sizing and dispatch optimization study specifically developed for sub-Saharan Africa conditions, with a detailed case study for a remote community in Kenya. The framework jointly addresses: (i) technology selection among four generation and storage combinations; (ii) optimal capacity sizing via a three-dimensional PSO-based global optimization over an annual 8760-hour simulation; (iii) a multi-objective economic-environmental dispatch optimization for a representative day; (iv) a systematic solver benchmark comparing seven global optimization algorithms; (v) robustness analysis under uncertainty in renewable generation and load; and (vi) a parametric sensitivity analysis across fuel prices, financial factors, and storage costs with ranges calibrated to SSA conditions.

The resulting framework is directly applicable to the expanding pipeline of microgrid projects across Africa. p. 7
Where the paper says it

This integrated treatment in a single study, tailored to the SSA context, provides insights and design guidelines not available from existing work. The resulting framework and findings are directly applicable to the rapidly expanding pipeline of microgrid projects across Africa.

Specific optimal system parameters for the baseline LI+DE system were found to be [ns, nw, Eb, init] = [15, 4, 106. p. 2
Where the paper says it

Nomenclature Acronyms BS Battery storage system CRF Capital recovery factor DE Diesel engine DG Distributed generator (backup) DPSP Deficiency of power supply probability DR Demand response DSM Demand-side management EED Economic and environmental dispatch GA Genetic algorithm LA Lead-acid battery LCOE Levelized cost of electricity ($/kWh) LI Lithium-ion battery MG Microgrid MILP Mixed-integer linear program MPC Model predictive control MT Microturbine PSO Particle swarm optimization PV Solar photovoltaic array REF Renewable energy fraction (penetration) REPG Relative excess power generated (dump fraction) RES Renewable energy sources SA Simulated annealing SOC State of charge SSA Sub-Saharan Africa TAC Total annualized cost ($/yr) WT Wind turbine Key variables Eb,init Initial rated battery energy capacity (kWh) ns Number of solar PV modules nw Number of wind turbines PBS (t) Battery power output at time t (kW) PDG(t) Distributed generator power output at time t (kW) PRES (t) Total renewable power output at time t (kW) wi Weight on objective i 1. Introduction and problem background Over 1 billion people or 13% of the world’s population still lacks access to electricity, with most of these people living in Africa, India and other parts of southeast Asia.

53], with a notable finding that the optimal wind turbine count (n w) is robustly determined to be 4 across all solvers. p. 2
Where the paper says it

Nomenclature Acronyms BS Battery storage system CRF Capital recovery factor DE Diesel engine DG Distributed generator (backup) DPSP Deficiency of power supply probability DR Demand response DSM Demand-side management EED Economic and environmental dispatch GA Genetic algorithm LA Lead-acid battery LCOE Levelized cost of electricity ($/kWh) LI Lithium-ion battery MG Microgrid MILP Mixed-integer linear program MPC Model predictive control MT Microturbine PSO Particle swarm optimization PV Solar photovoltaic array REF Renewable energy fraction (penetration) REPG Relative excess power generated (dump fraction) RES Renewable energy sources SA Simulated annealing SOC State of charge SSA Sub-Saharan Africa TAC Total annualized cost ($/yr) WT Wind turbine Key variables Eb,init Initial rated battery energy capacity (kWh) ns Number of solar PV modules nw Number of wind turbines PBS (t) Battery power output at time t (kW) PDG(t) Distributed generator power output at time t (kW) PRES (t) Total renewable power output at time t (kW) wi Weight on objective i 1. Introduction and problem background Over 1 billion people or 13% of the world’s population still lacks access to electricity, with most of these people living in Africa, India and other parts of southeast Asia.

The analysis confirms that renewable energy sources (RES) can reliably power an islanded microgrid without external grid support, provided they are paired with either dispatchable fossil fuel backups or sufficiently large battery storage, ideally both. p. 9
Where the paper says it

• Standby, fossil fuel distributed generators (DG) for backup which are controllable, deterministic units: Diesel engines (DE) and Microturbines (MT) • Battery energy storage (BS): Lithium-ion (Li+ ) (LI) and Lead-acid (LA) batteries • Charge controllers to regulate the state of charge (SOC) and charging and discharging of batteries • Power conditioners to regulate and improve the quality of power delivered from RES to the DC bus • Bidirectional power converters that can function either as an inverter (DC-to-AC conversion) or rectifier (ACto-DC conversion) depending on the system requirement This is a hybrid microgrid since it consists of both conventional and renewable generation sources. It is an isolated system in islanded mode serving a remote, rural community implying that there are no connections or power transfers with any larger, external macrogrid.

The break-even distance analysis estimates this break-even distance to be 0. p. 1
Where the paper says it

Sensitivity analyses show that fuel prices and financial discount rates are the most influential parameters for SSA conditions. A break-even distance analysis confirms the microgrid is economically competitive with grid extension at the chosen location.

855 km for the chosen location in Kenya. p. 5
Where the paper says it

much lower number of wind turbines relative to solar PV modules have the lowest energy costs and net equivalent CO2 emissions. This resulted from the strong solar potential in the region considered in Algeria [12] which is also likely to be the case for the location of Timbila, Kenya chosen here.

Overall, the study concludes that decentralized RES is economically advantageous compared to grid extension in areas characterized by low demand and high connection costs, favoring systems with high renewable penetration and significantly lower emissions and costs than those reliant solely on fossil fuels. p. 3
Where the paper says it

This implies that connection costs are extremely high across the continent, averaging around $136 (for single-phase connections), and can be significantly higher than monthly household incomes. This makes such projects unprofitable, especially for sparsely populated rural areas suffering from high marginal costs and less concentrated electricity demand.

Furthermore, the structure of Pareto trade-offs between cost, emissions, and reliability—specifically that cost-only optimization tends to underestimate emissions and reliability costs—is a robust phenomenon independent of location. p. 1
Where the paper says it

46 $/kWh with greater than 94% renewable penetration, outperforming all alternative configurations. Pareto-optimal solution sets reveal key trade-offs among economic, environmental, and reliability objectives, demonstrating that cost-only optimization leads to suboptimal outcomes in emissions and reliability.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Integration of multi-objective optimization into a single stage for sizing and dispatch allows for simultaneous treatment of technology selection, capacity sizing, dispatch strategy, and robustness analysis in one study. This enables the creation of a framework that jointly treats technology selection, capacity sizing, dispatch strategy, and robustness analysis tailored to SSA conditions.

  2. Implementation of particle swarm optimization (PSO) as the primary global optimization algorithm for microgrid design allows for achieving a minimum normalized objective of 0.146, providing superior runtime and solution quality compared to other benchmarked algorithms like Genetic algorithm (GA). This ensures that Particle swarm optimization is best suited to multiobjective microgrid design for practical application.

  3. Development of a two-stage sequential framework, where Stage 1 (sizing) uses a rule-based dispatch inner loop and Stage 2 (dispatch) solves a multi-objective economic-environmental dispatch problem, allows for decoupling the computational complexity while maintaining high operational realism. This addresses the limitation that solving a full multi-objective dispatch optimization at each of the thousands of outer loop function evaluations (e.g. PSO candidates) would be prohibitive.

  4. Inclusion of parametric sensitivity analysis across fuel prices, financial factors, and storage costs calibrated to SSA conditions allows designers to make informed decisions by showing how the optimal design and energy management is affected by variations in external parameters such as diesel fuel price and nominal interest rate. This informs the selection of backup DG size based on the finding that a DG rating of at least approximately 11 kW is needed to ensure 100% reliability.

  5. Utilization of Pareto-optimal solution sets derived from multiobjective optimization solvers allows designers to visualize how the optimal, minimum attainable values of the objectives vary with respect to each other, specifically identifying trade-offs such as how more aggressive emissions reductions or reliability improvements raise system costs. This helps in understanding that cost-only optimization leads to suboptimal outcomes in emissions and reliability.

  6. Extension of modeling approaches to include Combined Heat & Power (CHP) and Combined Cooling, Heating & Power (CCHP) microgrids, including thermal loads. This allows the model to be extended beyond electric power output to include thermal requirements, providing a more comprehensive design tool for various applications in remote communities.

Abstract

Sizing an islanded hybrid microgrid requires evaluating candidate designs through a year-long hourly simulation. The resulting objective is non-smooth, which limits gradient-based methods and requires approximation for mixed-integer linear programming. We size a solar photovoltaic (PV), wind, lithium-ion battery and diesel microgrid for an off-grid community in Timbila, Kenya, using five objectives: levelized cost of electricity (LCOE), emissions, unmet load, dumped energy and fossil share. We benchmark eight solvers, including particle swarm optimization (PSO), over 30 runs each with a budget of 2,000 evaluations. PSO, the PSO--pattern-search hybrid, pattern search and surrogate optimization finish within 0.5% of the exhaustively computed integer optimum in all 30 runs; PSO is fastest (expected running time 2.9 s), and gradient-based multistart methods perform worst. We then approximate the Pareto front, select designs with knee-point and compromise rules and the technique for order of preference by similarity to ideal solution (TOPSIS), trace a cost--reliability front over the generator rating, and re-rank candidates over 500 Monte Carlo resource and load scenarios. The equal-weight design, 73 kW of PV with a 239 kWh battery and no wind turbines, supplies 97% of energy from renewables at 0.184 USD/kWh, 63% below diesel-only supply. The selected design that serves all load uses a 12.2 kW generator, and accepting 1% unmet load lowers LCOE to 0.163 USD/kWh. Under uncertainty, a design with a 31% larger PV array lowers the 95% conditional value-at-risk of LCOE by 10% for a 2% higher expected cost. LCOE is most sensitive to the interest rate over the ranges tested.

Related papers