Composite model predictive control for the boost converter and two-phase interleaved boost converter

Muhammad Adnan Samad, Yuanqing Xia, Tayyab Manzoor, Kashif Mehmood, Adeel Saleem, Ahmad H. Milyani, Abdullah Ahmed AzhariFrontiers in Energy Research 2023 · DOI

TL;DR

This paper introduces a composite control strategy called MPADRC, which combines Model Predictive Control (MPC) and Active Disturbance Rejection Control (ADRC) to stabilize DC–DC boost converters. By using MPC for optimal current prediction in the outer loop and ADRC for robust disturbance rejection in the inner loop, the researchers achieved significantly better transient response and ripple reduction compared to using either method alone.

What Problem Does This Paper Address?

The DC–DC boost converter is a staple of modern power electronics, essential for applications like electric vehicles and renewable energy systems. However, controlling them is notoriously difficult for two main reasons:

  1. Nonlinearity: Their dynamics change significantly depending on the operating point.
  2. Non-minimum Phase Behavior: In continuous conduction mode (CCM), boost converters exhibit an "inverse response"—when you try to increase the output voltage, it initially drops before rising. This characteristic often destabilizes closed-loop systems.

Traditional PID controllers often struggle with these dynamics, leading to high overshoot, slow settling times, or instability when faced with "real-world" disturbances like fluctuating input voltages or changing load resistances.

Methods

The authors propose a dual-loop "composite" architecture applied to both a standard boost converter and a two-phase interleaved boost converter.

  • The MPADRC Framework:
    • Outer Voltage Loop (MPC): The Model Predictive Controller looks ahead over a finite horizon to calculate the optimal reference current needed to reach the target voltage. It excels at handling constraints and optimizing performance based on a mathematical model.
  • Inner Current Loop (ADRC): The Active Disturbance Rejection Controller tracks the reference current provided by the MPC. Its "secret weapon" is the Extended State Observer (ESO), which treats model uncertainties and external load changes as a "total disturbance" that it estimates and cancels in real-time.
  • Interleaving: The study also applies this control to a two-phase interleaved topology, where two converter stages operate in parallel but are phase-shifted. This naturally reduces input/output current ripples and improves efficiency.
  • Modeling: The system was modeled using state-space averaging, accounting for parasitic resistances (ESR) which are often ignored in simpler models but cause significant steady-state error in practice.

Datasets

As this is a power electronics control paper, no external "big data" sets were used. Instead, the authors evaluated their design through rigorous MATLAB simulations using the following benchmark parameters:

  • Input/Output: 12V to 24V step-up.
  • Switching Frequency: 20 kHz.
  • Components: 50mH Inductance, 1,000μF Capacitance, and a 100Ω Load.

Key Insights and Ideas

  • Complementary Strengths: The core "aha moment" is that MPC and ADRC solve each other's weaknesses. MPC provides the "intelligence" to optimize future states but can be sensitive to model errors; ADRC provides the "muscle" to reject those errors and external noise without needing a perfect model.
  • Stability via Bode Plots: The paper uses frequency response analysis (Figure 10) to show that while the plant alone is unstable or poorly damped, the MPADRC controller significantly improves the phase and gain margins, effectively "taming" the non-minimum phase zero.
  • Interleaving is a Force Multiplier: While the controller improves the logic of the power delivery, the interleaved hardware topology improves the physics by canceling out ripples, making the overall system much more efficient for high-power demands.

Key Results

The simulation results demonstrated that the composite MPADRC approach is superior to standalone MPC or ADRC:

  • Transient Performance: The MPADRC showed the fastest settling time and the lowest percentage overshoot (visible in Figure 13 vs. Figures 15 and 17).
  • Steady-State Accuracy: By accounting for ESR and using the ESO to cancel disturbances, the steady-state error was virtually eliminated.
  • Ripple Reduction: The two-phase interleaved version combined with MPADRC produced the cleanest output voltage with the least amount of switching noise (ripple).
  • Robustness: The system maintained a regulated 24V output even when the input voltage was subjected to significant disturbance parameters.

Limitations and Future Work

  • CCM Focus: The study specifically focuses on Continuous Conduction Mode (CCM). The behavior of the controller in Discontinuous Conduction Mode (DCM), which occurs at light loads, was not discussed.
  • Simulation vs. Hardware: While the results are promising, they are based on MATLAB simulations. Real-world factors like sensor noise, computational latency of the MPC algorithm on a DSP/FPGA, and thermal effects remain to be validated in a physical prototype.
  • Scaling Up: The authors suggest that future work will involve scaling this to "n-phase" interleaving and testing the MPADRC approach on other complex converter topologies like SEPIC, Cuk, and Buck-Boost converters.

Summary updated

Abstract

This article compares the conventional model predictive control (MPC) and active disturbance rejection control (ADRC) with a novel MPADRC technique for controlling a non-minimum phase behavior in the DC–DC boost converter. The control of the boost converter is challenging as it is nonlinear, and it shows non-minimum phase behavior in a continuous conduction mode (CCM). Moreover, in this article, the comparison is presented for the boost converter and the two-phase interleaved boost converter using MPC and ADRC, and the effectiveness of the interleaving technique is shown. Finally, it is proved that the interleaving method has much more efficiency and less output ripple than the simple boost converter. To conclude, a novel technique has been introduced that combines both the techniques, that is, MPC and ADRC, in the outer and inner loop with a boost converter, respectively, and the response is clearly the best when compared to the said techniques individually. The overall impact of this technique includes the advantages of both the techniques, that is, the use of MPC allows us to optimize the current value by predicting the future values, and the use of ADRC ensures that the disturbance factor is well tackled and cancels the effect caused by all the disturbances including ignored quantities as well.
BibTeX
@inproceedings{samad2023composite,
  author = {Muhammad Adnan Samad and Yuanqing Xia and Tayyab Manzoor and Kashif Mehmood and Adeel Saleem and Ahmad H. Milyani and Abdullah Ahmed Azhari},
  title = {{Composite model predictive control for the boost converter and two-phase interleaved boost converter}},
  booktitle = {Frontiers in Energy Research},
  year = {2023},
  doi = {10.3389/fenrg.2022.1009812},
}
Figure