ISSN (Online): 3108-1584
Volume 1, Issue 6 - May 2026
Maximum power point tracking (MPPT) is a fundamental requirement for achieving optimal energy extraction from photovoltaic (PV) systems, whose output characteristics vary nonlinearly with solar irradiance and cell temperature. Despite an extensive literature on individual MPPT algorithms, comprehensive comparative performance evaluations spanning classical, bio-inspired, and deep learning-based approaches across multiple irradiance and shading conditions remain scarce. This paper presents a rigorous comparative evaluation of nine MPPT algorithms Perturb and Observe (P&O), Incremental Conductance (INC), Fractional Open-Circuit Voltage (FVoc), Fractional Short-Circuit Current (FIsc), Fuzzy Logic Control (FLC), Artificial Neural Network (ANN-MPPT), Particle Swarm Optimization (PSO-MPPT), Grey Wolf Optimizer (GWO-MPPT), and a proposed Deep Learning MPPT (DL-MPPT) using a CNN-LSTM architecture implemented in Python using Tensor Flow 2.x and simulated under MATLAB/Simulink for hardware validation. Evaluation is performed under uniform irradiance, step-change irradiance, and partial shading conditions using a 5 kW PV array parameterised for Northern Nigerian climatic conditions (Bauchi State, lat. 10.3°N). Key findings establish that the proposed DL-MPPT achieves 99.7% steady-state tracking efficiency and 98.4% dynamic efficiency under step-change irradiance outperforming all classical and metaheuristic baselines. Crucially, DL-MPPT achieves a convergence time of 0.31 s, 2.6 times faster than PSO-MPPT (0.81 s), with a ripple power of only 0.4 W under steady-state operation. Under partial shading with multiple local maxima, DL-MPPT successfully identifies the global MPP in all 50 test scenarios, whereas P&O and INC fail to escape local optima in 38% and 31% of cases respectively. Annual energy yield simulations for Bauchi demonstrate a cumulative advantage of 6.4% (283 kWh/yr) for DL-MPPT over P&O for a 5 kW residential system.
MPPT, Photovoltaic, Solar Energy, Deep Learning, CNN-LSTM, Partial Shading, P&O, PSO, GWO, Python, Tensor Flow, Nigeria
Mohammed Adamu Sule, Muhammad Bello, Shehu Abdullahi, Aliyu Muhammad Bello, Usman Musa, Salisu Muhammad, "Performance Evaluation of Different MPPT Algorithms for Solar Energy Systems: A Comprehensive Python-Based Experimental and Simulation Study", Cosmo Research & Science International Journal, vol. Jul-25, no. 1, pp. 59-71, 2026.
Mohammed Adamu Sule, Muhammad Bello, Shehu Abdullahi, Aliyu Muhammad Bello, Usman Musa, Salisu Muhammad (2026). Performance Evaluation of Different MPPT Algorithms for Solar Energy Systems: A Comprehensive Python-Based Experimental and Simulation Study. Cosmo Research & Science International Journal, Jul-25(1), 59-71.
Mohammed Adamu Sule, Muhammad Bello, Shehu Abdullahi, Aliyu Muhammad Bello, Usman Musa, Salisu Muhammad. "Performance Evaluation of Different MPPT Algorithms for Solar Energy Systems: A Comprehensive Python-Based Experimental and Simulation Study." Cosmo Research & Science International Journal, vol. Jul-25, no. 1, 2026, pp. 59-71.
@article{CRSIJ26000129,
author = {Mohammed Adamu Sule, Muhammad Bello, Shehu Abdullahi, Aliyu Muhammad Bello, Usman Musa, Salisu Muhammad},
title = {Performance Evaluation of Different MPPT Algorithms for Solar Energy Systems: A Comprehensive Python-Based Experimental and Simulation Study},
journal = {Cosmo Research and Science International Journal},
year = {2025},
volume = {1},
number = {6},
pages = {59-71},
issn = {3108-1584},
url = {https://cosmorsij.com/published/CRSIJ26000129.pdf},
abstract = {Maximum power point tracking (MPPT) is a fundamental requirement for achieving optimal energy extraction from photovoltaic (PV) systems, whose output characteristics vary nonlinearly with solar irradiance and cell temperature. Despite an extensive literature on individual MPPT algorithms, comprehensive comparative performance evaluations spanning classical, bio-inspired, and deep learning-based approaches across multiple irradiance and shading conditions remain scarce. This paper presents a rigorous comparative evaluation of nine MPPT algorithms Perturb and Observe (P&O), Incremental Conductance (INC), Fractional Open-Circuit Voltage (FVoc), Fractional Short-Circuit Current (FIsc), Fuzzy Logic Control (FLC), Artificial Neural Network (ANN-MPPT), Particle Swarm Optimization (PSO-MPPT), Grey Wolf Optimizer (GWO-MPPT), and a proposed Deep Learning MPPT (DL-MPPT) using a CNN-LSTM architecture implemented in Python using Tensor Flow 2.x and simulated under MATLAB/Simulink for hardware validation. Evaluation is performed under uniform irradiance, step-change irradiance, and partial shading conditions using a 5 kW PV array parameterised for Northern Nigerian climatic conditions (Bauchi State, lat. 10.3°N). Key findings establish that the proposed DL-MPPT achieves 99.7% steady-state tracking efficiency and 98.4% dynamic efficiency under step-change irradiance outperforming all classical and metaheuristic baselines. Crucially, DL-MPPT achieves a convergence time of 0.31 s, 2.6 times faster than PSO-MPPT (0.81 s), with a ripple power of only 0.4 W under steady-state operation. Under partial shading with multiple local maxima, DL-MPPT successfully identifies the global MPP in all 50 test scenarios, whereas P&O and INC fail to escape local optima in 38% and 31% of cases respectively. Annual energy yield simulations for Bauchi demonstrate a cumulative advantage of 6.4% (283 kWh/yr) for DL-MPPT over P&O for a 5 kW residential system.},
keywords = {MPPT, Photovoltaic, Solar Energy, Deep Learning, CNN-LSTM, Partial Shading, P&O, PSO, GWO, Python, Tensor Flow, Nigeria},
month = {May}
}