⚡ electrical engineering

Real-Time FPGA-in-the-Loop Validation of a Sequential Eliminative Lévy Flight MPPT Method for Global Maximum Power Point Tracking under Partial Shading

This paper presents and validates an FPGA-in-the-Loop implementation of a Sequential Eliminative Lévy Flight-based MPPT controller that utilizes adaptive mode-switching and hardware-optimized algorithms to achieve high-efficiency global maximum power point tracking with rapid convergence under partial shading conditions.

Original authors: Khadidja Bentata, Adel Ballouti, Salah AMROUNE, Eustache Hakizimana

Published 2026-07-16

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Khadidja Bentata, Adel Ballouti, Salah AMROUNE, Eustache Hakizimana

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to find the highest peak in a vast, foggy mountain range. You want to stand on the very top to get the best view, but the fog is so thick you can't see the whole map. This is the daily struggle of solar panels. They are like hungry sun-eaters, but when clouds drift by or trees cast shadows, the "sunlight" hitting them becomes uneven. This unevenness creates a tricky problem: the solar panel's energy output doesn't just have one big hill to climb; it creates a landscape with several small, fake peaks (local maximums) and one true, giant peak (global maximum). If the panel's brain gets stuck on a small hill, it thinks it's done, missing out on a huge amount of free energy.

To solve this, engineers use something called "Maximum Power Point Tracking" (MPPT). Think of MPPT as a very smart hiker with a compass. In the past, these hikers used simple rules, like "keep walking uphill." But in a foggy mountain range with fake peaks, a simple hiker gets trapped on a small hill and never finds the real summit. Newer methods use "swarm intelligence," where a group of hikers explores the terrain together, taking random, long jumps to escape the small hills. However, running these complex hiker simulations on a tiny computer chip inside a solar system is hard; it usually requires too much power or memory. This is where the paper steps in, asking: Can we build a super-fast, super-efficient hiker that fits on a small chip and never gets lost in the fog?

The authors of this paper, Khadidja Bentata and her team, say yes. They designed a new kind of hiker called the "Sequential Eliminative Lévy Flight" (SELF) MPPT. Instead of sending out a huge army of virtual hikers, this method starts with just three candidates for the best spot. It uses a clever trick called "Lévy Flight," which is like a hiker who usually takes small, careful steps but occasionally takes a giant, random leap to a completely different part of the mountain. This helps them escape the fake peaks. But the real magic is the "Sequential Eliminative" part: as the hikers get closer to the truth, the system gets stricter. It starts by testing three spots, then eliminates the worst one to focus on two, and finally narrows it down to just the single best spot. This makes the search incredibly fast and precise.

To prove this works in the real world, the team didn't just run a computer simulation; they built a "FPGA-in-the-Loop" (FIL) system. Imagine this as a video game where the solar panel and the weather are simulated on a computer, but the "hiker" (the controller) is a physical, real-time chip (an Artix-7 FPGA) that talks to the simulation instantly. This is like testing a race car driver in a simulator, but the driver is a real, physical robot sitting in the cockpit, reacting to the virtual track with zero lag.

The results of their experiments were impressive. When they tested their system under tricky conditions where the sun was blocked in different patterns (partial shading), their SELF hiker found the true global peak 98% of the time. It was also incredibly fast, reaching the top in just 0.18 seconds after the sun changed. Perhaps most importantly, once it found the top, it didn't wobble. While other methods might shake back and forth, wasting energy, this system settled down with almost no movement, changing its "duty cycle" (its control setting) by only 0.001. The team also showed that this smart hiker is very lightweight, using only about 18% of the available memory space on the chip and 5% of its math engines.

The paper explicitly rules out the idea that you need a massive, complex swarm of thousands of virtual agents to find the best spot. They argue that by using a small group and eliminating the weak candidates one by one, you get better results with less hardware. They also demonstrate that their method is superior to older "conventional" techniques like "Perturb and Observe," which they show get stuck on the fake peaks and fail to find the true maximum power.

In short, this paper presents a new, highly efficient way for solar panels to find their maximum energy output even when the sun is being blocked. By combining a smart, random-jumping search strategy with a "trim the fat" elimination process, the authors created a controller that is fast, accurate, and cheap enough to run on a small chip. Their findings, verified through both computer simulations and real-time hardware tests, suggest that this approach could make solar power systems much more reliable and efficient in the real world, especially on cloudy or partially shaded days.

Technical Summary: Real-Time FPGA-in-the-Loop Validation of a Sequential Eliminative Lévy Flight MPPT Method

Problem Statement
Photovoltaic (PV) systems face significant performance challenges under Partial Shading Conditions (PSC). When irradiance is non-uniform across a PV array, bypass diodes create multiple local maximum power points (LMPPs) on the power-voltage (P-V) characteristic curve. Conventional Maximum Power Point Tracking (MPPT) algorithms, such as Perturb and Observe (P&O), Incremental Conductance (INC), and standard Particle Swarm Optimization (PSO), often fail to distinguish between local and global maxima (GMPP), leading to suboptimal energy extraction. Furthermore, while metaheuristic algorithms offer better global search capabilities, their implementation on hardware often suffers from high computational complexity, slow convergence, and significant resource consumption, making real-time deployment difficult.

Methodology
The paper proposes a Sequential Eliminative Lévy Flight (SELF) MPPT algorithm designed specifically for hardware implementation. The methodology integrates three core components:

Algorithmic Strategy: Unlike traditional swarm intelligence methods that maintain large populations, SELF operates with a minimal set of three duty-cycle candidates. It employs a deterministic sequential elimination mechanism driven by real-time metrics of particle dispersion ( $|\Delta d|$ ) and relative power variation ( $\Delta P$ ).
- Regime-Based Adaptation: The algorithm dynamically switches between three regimes based on thresholds:
  - Transitional Regime 1 (Wide-range search): All three candidates are updated using Lévy-flight perturbations when power variation is high and dispersion is wide.
  - Transitional Regime 2 (Focused refinement): The weakest candidate is eliminated, and the remaining two are refined when power variation decreases but dispersion remains.
  - Steady-State Regime: Only the best candidate is retained and locally refined when the system converges.
- Reinitialization: A specific condition ( $|\Delta d| < 10^{-8}$ and $\Delta P > 20\%$ ) triggers a full reinitialization to detect and respond to sudden changes in shading patterns.
Hardware Implementation: The algorithm is implemented on an Artix-7 FPGA using a Finite State Machine (FSM) architecture in VHDL.
- Fixed-Point Arithmetic: To minimize resource usage, the design utilizes fixed-point arithmetic (UQ format) rather than floating-point, avoiding complex operations like Gamma functions.
- Lévy Flight Approximation: Instead of computationally expensive random number generation, the system uses a 16-bit Linear Feedback Shift Register (LFSR) with a piecewise approximation technique to generate pseudo-random steps that mimic Lévy distributions.
- Parallel Processing: The architecture exploits the FPGA's parallel capabilities to update duty cycles and measure power simultaneously, reducing latency.
Validation Framework: The system is validated using an FPGA-in-the-Loop (FIL) approach. The PV system model (four panels, boost converter, resistive load) runs in MATLAB/Simulink on a host PC, while the MPPT controller runs in real-time on the FPGA. Data exchange occurs via UART, allowing for rigorous testing of the hardware logic against a dynamic software model.

Key Contributions

Algorithmic Efficiency: The introduction of a "Sequential Eliminative" mechanism that reduces the search space from three candidates to one without sacrificing the ability to escape local maxima, significantly lowering computational overhead compared to standard swarm algorithms.
Hardware Optimization: The successful translation of a stochastic metaheuristic into a resource-efficient fixed-point VHDL design. The use of LFSRs for Lévy step generation replaces complex mathematical functions, making the algorithm viable for low-cost FPGAs.
Real-Time Validation: Providing a comprehensive FIL validation that bridges the gap between theoretical simulation and physical hardware performance, demonstrating the feasibility of deploying complex MPPT strategies in real-time embedded systems.

Results
The proposed system was tested under dynamic irradiance conditions, including rapid transitions between uniform irradiance and complex partial shading patterns (SP1, SP2, SP3).

Tracking Performance: The algorithm successfully tracked the GMPP in all tested scenarios, achieving a tracking efficiency of approximately 98%.
Convergence Speed: The system demonstrated a mean convergence time of 0.18 seconds to reach 98% of the GMPP following irradiance changes.
Stability: In steady-state operation, the system exhibited ultra-low oscillations with a duty cycle ripple ( $\Delta D$ ) of only 0.001 (32 LSBs in UQ1.15 format), significantly reducing steady-state power losses compared to software-based metaheuristics.
Resource Utilization: The implementation on the Artix-7 FPGA was highly efficient, utilizing only 18.21% of Lookup Tables (LUTs) and 5% of DSP slices, leaving ample resources for auxiliary control functions.

Significance and Claims
The paper claims that the proposed SELF-MPPT method offers a robust solution to the partial shading problem by balancing global exploration and local exploitation through a deterministic elimination process. The authors emphasize that the hardware-oriented design, utilizing fixed-point arithmetic and LFSRs, overcomes the latency and resource constraints typically associated with metaheuristic MPPT algorithms. The close agreement between MATLAB/Simulink simulations and FIL co-simulation results validates the accuracy of the hardware model. Ultimately, the work demonstrates that high-performance, global MPPT tracking can be achieved in real-time on moderate-cost FPGAs, making it suitable for practical, cost-sensitive photovoltaic applications where energy extraction efficiency is critical.

Technical Summary: Real-Time FPGA-in-the-Loop Validation of a Sequential Eliminative Lévy Flight MPPT Method

More like this