South Africa's electricity sector is transitioning toward competitive balancing and ancillary-service markets, requiring sophisticated tools to schedule generation and consumption across geographically dispersed assets. This research evaluates whether nature-inspired computational methods can match traditional optimization for managing virtual power grids comprising renewables, battery storage, cogeneration, and flexible loads.
The study compared three approaches: mixed-integer linear programming (MILP), genetic algorithms, and gorilla troop optimization—a newer biologically-inspired technique. Researchers tested all methods on a simplified South African network derived from actual 400 kV transmission corridors, renewable independent power producer interconnections, and distribution-level load connections. The analysis used real South African generation and consumption patterns across 24 hourly scheduling periods.
The deterministic MILP benchmark achieved the lowest total cost, establishing a reference for comparison. However, the genetic algorithm executed faster while maintaining comparable voltage stability metrics. Gorilla troop optimization produced marginally lower costs than the genetic algorithm despite slower computation. Both metaheuristic approaches successfully converged across all 24 hourly scenarios, demonstrating reliability for practical grid applications.
A particularly promising finding emerged when flexible demand combined with battery storage: this integration significantly reduced the overall energy deficit that the system had to address. This suggests that virtual-grid operators can better manage supply constraints by strategically timing consumption and storage discharge.
The simplified 35-bus grid structure proved effective for comparing algorithm performance without sacrificing relevance to South African conditions. This reduced-order model transfers existing academic benchmarks to local contexts while remaining computationally tractable for industrial applications. The authors conclude that metaheuristic algorithms warrant serious consideration for operational scheduling in emerging South African markets, particularly when computational speed and adaptability matter more than achieving mathematically optimal solutions. Future work should validate these approaches on larger networks and under real market conditions.
