Table 1 Challenges in existing Models.
Challenge | Limitation in Existing Approaches | How POA-GENN Addresses It |
|---|---|---|
Uncertain renewable generation due to variable irradiance and wind | ANN, LSTM, and statistical models struggle with rapid fluctuations, leading to inaccurate scheduling. | GENN captures spatial–temporal dependencies through graph-based ensemble learning, providing more accurate and robust forecasts of PV, wind, and demand. |
Metaheuristic optimizers prone to local minima and slow convergence | PSO, GA, and ABC often converge prematurely in high-dimensional scheduling problems. | POA’s scent-diffusion and adaptive trail-following balance exploration and exploitation, avoiding local optima and achieving better convergence. |
Lack of integration between forecasting and optimization | Forecasting and optimization are treated as separate tasks, reducing adaptability to fast-changing conditions. | GENN forecasts are iteratively fed into POA, creating a dynamic feedback loop that tightly couples prediction with scheduling. |
High operational cost and inefficient energy allocation | Heavy reliance on grid imports during renewable dips increases cost and reduces efficiency. | POA-GENN reduces grid dependency by aligning ESS scheduling with forecasted renewableavailability, lowering cost, and raising efficiency. |
Battery degradation from irregular charge–discharge cycles | Poor scheduling causes deep discharges and SOC volatility, accelerating ESS aging. | POA uses GENN-informed forecasts to smooth ESS charging/discharging, keeping SOC in safe bands and extending battery life. |
