A novel artificial hummingbird algorithm-optimized extreme learning machine for high-resolution DC power forecasting in rooftop retrofitted photovoltaic systems

Accurate short-term forecasting of photovoltaic (PV) power output is critical for grid stability, energy management, and optimal integration of renewable energy systems. However, retrofitted rooftop PV installations present unique forecasting challenges due to site-specific constraints, including variable tilt angles, partial shading, aging infrastructure, and non-uniform configurations. This study proposes a novel hybrid model that integrates the Artificial Hummingbird Algorithm (AHA) with Extreme Learning Machine (ELM) for high-resolution DC power prediction in retrofitted PV systems. The AHA-ELM model leverages bio-inspired foraging behaviors to optimize the input weights and biases of ELM, addressing its inherent limitations of random parameter initialization and prediction instability. Real-world experimental validation was conducted using 15 min interval data from a 465 kWp rooftop retrofitted PV system at Universiti Malaysia Pahang Al-Sultan Abdullah throughout 2024. The proposed model was benchmarked against four metaheuristic-optimized ELM variants: Barnacle Mating Optimizer (BMO-ELM), Moth Flame Optimizer (MFO-ELM), African Vultures Optimization Algorithm (AVOA-ELM), and Particle Swarm Optimization (PSO-ELM). Performance evaluation using standardized metrics demonstrated that AHA-ELM achieved superior accuracy with the lowest Mean Squared Error (84.1144), Root Mean Squared Error (9.1713), and Mean Absolute Error (4.0967), while attaining the highest coefficient of determination (R² = 0.8979). The model exhibited faster convergence, enhanced stability, and better generalization capability across diverse operational conditions. These findings establish AHA-ELM as a computationally efficient and reliable forecasting framework for retrofitted PV systems, with significant implications for real-time energy dispatch, demand-response management, and smart grid applications.