Average value buck converter PI tuning problems (convergence and instability)

I am developing a simple Simulink model of a 100 V to 50 V buck converter operating under a constant power load (CPL) profile. The converter uses a voltage‑mode PI controller to regulate the 50 V output.

The simulation runs using the ode23t solver with a maximum step size of 10 µs (1×10⁻⁵ s), while all other solver settings remain at their default values.

During the tuning process of the PI controller, the model behaves normally at the beginning of the simulation, but after approximately 0.2 seconds, the system begins to lose convergence. The output voltage, inductor current, or controller signals start to diverge, indicating numerical instability or control-loop issues. The attached figures illustrate this behavior.

The problem started to happend when I'm trying to increase the intergral coefficient.

I am trying to understand the cause of this convergence problem and determine how to modify the controller, the model, or the solver settings to maintain stable operation throughout the entire simulation.

Answer 1

[Sam Chak]

Hi @Anh Tuan

Generally, it is difficult to determine what went wrong by inspecting the block diagram alone. It is comparable to "looking" at the motherboard in an attempt to determine why the PC does not boot up, despite the power supply unit functioning correctly.

Most of the time, by analyzing the mathematical model of the buck converter control system, a system designer should be able to understand why increasing the integral gain destabilizes the system.