Modeling & implementation of DRLA based partially shaded solar system integration with 3-ϕ conventional grid using...

Area of Science:

• Electrical Engineering
• Renewable Energy Systems
• Artificial Intelligence in Energy

Background:

• Global environmental protection drives the adoption of Renewable Energy Resources (RERs).
• Solar photovoltaic systems are crucial for electrical energy generation, reducing reliance on nonrenewable fuels.
• Extracting Maximum Energy Point (MEP) from solar modules under varying atmospheric conditions requires advanced Maximum Power Point Tracking (MPPT) techniques.

Purpose of the Study:

• To propose a novel Deep Reinforcement Learning Algorithm based MPPT (DRLAMPPT) for solar photovoltaic systems, particularly under partial shading conditions (PSC).
• To address the limitations of traditional reinforcement learning (RL) by enabling operation with continuous state spaces using deep deterministic policy gradient (DDPG).
• To integrate DRLAMPPT with a Constant Current Controller (CCC) for enhanced performance in a 2 kW solar photovoltaic power plant.

Main Methods:

• Development of DRLAMPPT, combining reinforcement learning algorithm (RLA) and deep learning algorithm (DLA), utilizing an artificial neural network (ANN) for control signal generation.
• Implementation of deep deterministic policy gradient (DDPG) to handle continuous state spaces for reaching the Global Maximum Energy Point (GMEP).
• Experimental setup of a 2 kW solar photovoltaic power plant, including a photovoltaic array, DC/DC converter, and a 3-Φ PWM-VSI integrated with the grid via CCC.

Main Results:

• The proposed DRLAMPPT demonstrated superior efficiency and faster adaptation to environmental changes compared to existing MPPT techniques.
• DRLAMPPT successfully reached the GMEP within 0.8 seconds under partial shading conditions (PSC).
• The CCC with an LC filter ensured the inverter output voltage and grid voltage were in phase with low total harmonic distortion (THD) of 1.1% and 0.98% respectively.

Conclusions:

• The DRLAMPPT, integrated with CCC, offers a highly effective solution for maximizing energy extraction from solar photovoltaic systems, especially under challenging partial shading conditions.
• The proposed method significantly enhances tracking speed and efficiency, outperforming traditional MPPT techniques.
• Experimental validation confirms the robustness and effectiveness of the DRLAMPPT and CCC for grid-connected solar photovoltaic power systems.