PPO-Based PV Grid Simulation
Budget: ₹1,500 – ₹12,500 INR
I need a full MATLAB/Simulink model of a single-phase, grid-connected photovoltaic system that lets me plug in my own DC-DC boost converter topology and then closes the loop with a Proximal Policy Optimization (PPO) controller. The priority is clear: the reinforcement-learning algorithm must deliver high accuracy in tracking the reference—voltage, current, or power—throughout irradiance and load changes.
Here is what I expect:
• A modular Simulink file where the PV array, my custom boost converter, the single-phase inverter, the LCL or LC filter, and the grid interface are clearly separated so I can swap devices in or out.
• A PPO training script (MATLAB RL Toolbox or compatible Python interface) that learns directly on the Simulink environment and saves the trained agent for deployment.
• A deployment variant of the model that runs the trained agent in real time without the training overhead.
• Graphs and logged data demonstrating precise reference tracking under at least two irradiation profiles and two step-load tests. Please label them so I can compare the RL controller to any baseline you choose (PI, fuzzy, etc.).
• Short documentation: block diagram, signal names, training parameters, and a quick note on how to swap in my converter if I change component values.
Acceptance criteria
1. Steady-state tracking error ≤ 1 % for both DC-link voltage and injected grid current.
2. Total harmonic distortion of grid current ≤ 5 % under IEEE-1547 test scenario.
3. All files open and run in MATLAB R2021b or later without missing dependencies.
If you have tackled RL control in power converters before, this should feel familiar. Let’s keep the focus on reproducible, well-commented code rather than fancy visuals. I’m around to answer any topology-specific questions as you build.
Here is what I expect:
• A modular Simulink file where the PV array, my custom boost converter, the single-phase inverter, the LCL or LC filter, and the grid interface are clearly separated so I can swap devices in or out.
• A PPO training script (MATLAB RL Toolbox or compatible Python interface) that learns directly on the Simulink environment and saves the trained agent for deployment.
• A deployment variant of the model that runs the trained agent in real time without the training overhead.
• Graphs and logged data demonstrating precise reference tracking under at least two irradiation profiles and two step-load tests. Please label them so I can compare the RL controller to any baseline you choose (PI, fuzzy, etc.).
• Short documentation: block diagram, signal names, training parameters, and a quick note on how to swap in my converter if I change component values.
Acceptance criteria
1. Steady-state tracking error ≤ 1 % for both DC-link voltage and injected grid current.
2. Total harmonic distortion of grid current ≤ 5 % under IEEE-1547 test scenario.
3. All files open and run in MATLAB R2021b or later without missing dependencies.
If you have tackled RL control in power converters before, this should feel familiar. Let’s keep the focus on reproducible, well-commented code rather than fancy visuals. I’m around to answer any topology-specific questions as you build.