Performance Comparison of Bio-Inspired Algorithms for Optimizing an ANN-Based MPPT Forecast for PV Systems

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence in Engineering
• Optimization Algorithms

Background:

• Photovoltaic (PV) systems face challenges in maintaining peak energy output under partial shading conditions.
• Artificial Neural Networks (ANNs) are used for Maximum Power Point Tracking (MPPT), but their performance can be suboptimal without advanced optimization.
• Bio-inspired algorithms offer potential for enhancing ANN-based MPPT system accuracy and efficiency.

Purpose of the Study:

• To compare the effectiveness of four bio-inspired optimization algorithms: Grey Wolf Optimizer (GWO), Particle Swarm Optimization (PSO), Squirrel Search Algorithm (SSA), and Cuckoo Search (CS).
• To enhance an ANN-based MPPT system for improved performance under simulated partial shading conditions in PV systems.
• To identify the most computationally efficient and accurate algorithm for real-world application.

Main Methods:

• Implemented and evaluated GWO, PSO, SSA, and CS to optimize an ANN for MPPT in PV systems.
• Augmented a dataset with perturbations to simulate realistic partial shading scenarios.
• Assessed algorithm performance based on prediction accuracy (Mean Squared Error - MSE, Mean Absolute Error - MAE) and computational execution time.

Main Results:

• The standard ANN exhibited poor performance (MSE: 159.94, MAE: 8.08).
• GWO achieved the best prediction accuracy (MSE: 11.95, MAE: 2.46) with good computational efficiency (1199 s).
• SSA was the fastest (987 s) with competitive accuracy (MSE: 12.15, MAE: 2.70), while PSO minimized MAE (2.17) but was slower (1418 s). CS showed the least favorable results.

Conclusions:

• Grey Wolf Optimizer (GWO) provides the optimal balance between prediction accuracy and computational speed for ANN-based MPPT under partial shading.
• The optimized ANN using GWO significantly improves energy harvesting efficiency in photovoltaic systems.
• This methodology holds promise for real-world applications in solar farms and residential solar panels, with potential for integration into smart grid systems.