Multi-Variable Thunderstorm Algorithm for Efficiency and Power Stability Optimization in Battery-Integrated Micro-Hydropower Systems

Resumo

Small-scale hydropower can support sustainable electrification, yet field performance is often limited by sub-optimal operation that yields fluctuating power and modest conversion efficiency. This study applies the Thunderstorm Algorithm (TA) as a multi-variable optimizer to jointly tune net head (H), electro-mechanical efficiency (η), and gate opening in a battery-integrated micro-hydropower system. A composite objective maximizes daily stored energy while penalizing instability via the standard deviation of hourly power and the cumulative ramp. TA follows the Cloud–Streamer–Avalanche phases (40 clouds, 250 iterations) over the bounds H∈[30,100] m, η∈[0.70,0.95], and gate ∈[0.40,1.00], using a 24-hour discharge series; battery capacity is fixed at 5 MWh, with weightsstd = wramp= 0.10. The algorithm converges to an interior optimum H = 64.95 m, η = 0.745, gate = 0.904. Because the storage ceiling is binding, stored energy equals 5,000 kWh for both the baseline  (H = 60 m, η = 0.90, gate = 0.90) and the TA solution; improvements arise from stability: the power standard deviation decreases 19.186→17.264 kW and the cumulative ramp 171.636→154.440 kW (=10% reductions). A linear surrogate fitted to TA samples attains RMSE 48.175 kWh, MAE 22.755 kWh, MAPE 0.49% (Accuracy ≈ 99.51%), R2 = 0.931, and Pearson r = 0.9701. The results confirm that TA effectively navigates the bounded design space to preserve energy yield under a storage ceiling while materially smoothing the power profile—beneficial for power conditioning, battery thermal management, and grid interfacing.