Optimal MPPT Control of a Photovoltaic System Under Non-uniform Irradiation

Meriem Halimi; Toufik Bouden; Yehia Lalili

REVIEW 4 major objections 5 minor 10 references

Optimal MPPT Control of a Photovoltaic System Under Non-uniform Irradiation

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A hybrid ANN-plus-P&O controller finds the shaded PV global peak in 0.18 seconds, faster than PSO or cuckoo search.

desk verdict Routine hybrid MPPT application whose headline speedup is an in-sample result: the ANN window is trained from the same shading pattern being tested. read the letter →

arxiv 2411.16650 v1 pith:KEY2YUMN submitted 2024-11-25 eess.SY cs.SY

Yehia Lalili , Meriem Halimi , Toufik Bouden This is my paper · ORCID

classification eess.SYcs.SY

keywords maximumpowerpointtrackingpartialshadingphotovoltaicsystemsartificialneuralnetworkperturbandobserveparticleswarmoptimizationcuckoosearchnon-uniformirradiation

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Under partial shading, a photovoltaic array's power–voltage curve develops several peaks, and classical maximum-power-point trackers such as perturb-and-observe can lock onto a local peak instead of the global one. The paper claims that this failure can be avoided by letting a small artificial neural network predict a voltage window that contains the global peak, then running a classical tracker inside that window. In simulation of a three-module array at 600, 800, and 1000 W/m², this hybrid reaches the global maximum power point in 0.18 seconds with less oscillation, while particle swarm optimization takes 1.2 seconds and cuckoo search 0.7 seconds. The practical stake is that faster, stable convergence means more captured energy and less wear on the converter during shading transients.

What carries the argument

The load-bearing mechanism is the ANN-predicted voltage window: the network maps a shading pattern (irradiance on each series module) to a narrow interval [Vmin, Vmax] known from a training table to contain the global maximum. Restricting the classical perturb-and-observe search to that interval turns a multimodal global-optimization problem into a local hill-climb inside the right valley. The two stochastic baselines are cuckoo search, which moves candidate voltage samples by Lévy flights, and particle swarm optimization, which moves a population of particles by velocity updates toward personal and swarm bests; both must explore the entire voltage range and therefore converge more slowly and oscillate more.

What would settle it

Run the hybrid controller on the same three-module array under a shading pattern absent from the training table (for example, 700, 900, and 1000 W/m²) and record whether the ANN-predicted [Vmin, Vmax] brackets the true global peak and whether the final operating power equals the global maximum. A case where the tracker settles at a local peak or the window excludes the global peak would falsify the claim that the hybrid generalizes beyond its training patterns.

Watch

Extended reading notes

Core claim

The central claim is that a hybrid controller—a three-input, ten-hidden-node ANN feeding a perturb-and-observe tracker—solves the partial-shading global-MPP problem more effectively than the two stochastic optimizers. The ANN takes the three module irradiances as inputs and outputs the window boundaries Vmin = 79.99 V and Vmax = 88 V around the global peak; perturb-and-observe then perturbs only inside that window, so it cannot be trapped by local maxima outside it. In the benchmark shading scenario, the PV curve has three local peaks and a 400 W global peak, and the hybrid settles at the global peak in t = 0.18 s, versus t = 0.7 s for cuckoo search and t = 1.2 s for particle swarm optimization, with visibly smaller oscillations. The paper also reports that plain perturb-and-observe fails to reach the global peak in the same scenario, which is the baseline failure the hybrid is designed to fix.

Load-bearing premise

The neural network's voltage window is trained and tested on the same irradiance combinations, so the 0.18 s result assumes the operating shading pattern is inside the training distribution; otherwise the window could miss the global peak.

Editorial extensions

If this is right

If the ANN window is reliable, any hill-climbing tracker (perturb-and-observe or incremental conductance) can recover the global peak under partial shading without stochastic search.
The reported settling times imply roughly a fourfold speedup over cuckoo search and a near sevenfold speedup over particle swarm optimization, cutting the energy lost during the tracking transient.
Lower steady-state oscillation means smaller power ripple delivered to the load and less stress on the DC-DC converter's switching devices.
Because the online controller is a feedforward ANN plus a simple hill-climb, it avoids tuning of swarm parameters such as inertia weight, acceleration constants, and Lévy step size.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial inference: the reported advantage is conditional on the ANN having seen the shading pattern; under a truly novel irradiance combination, a mispredicted window could exclude the global peak and the controller would degrade to plain perturb-and-observe.
Editorial inference: the comparison counts only online tracking time; the offline cost of building the training table and training the ANN is not included, so total engineering cost is higher than the 0.18 s figure suggests.
Editorial inference: a cheap robustness test would be to widen the predicted window by a safety margin and measure how much convergence time increases; the paper does not investigate this trade-off.
Editorial inference: the same window-prediction idea could be paired with particle swarm optimization or cuckoo search instead of perturb-and-observe, potentially combining the speed of the hybrid with the global-search guarantees of the stochastic methods.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Here's the one-line summary: this is a textbook application of a known ANN-windowed P&O scheme to a three-module PV array, and the central speed claim (0.18 s) is very likely an in-sample benchmark because the ANN is trained on the same optimal-coordinate table from which the tested 600/800/1000 W/m2 shading pattern is drawn.

What the paper does well: it gives a clear, if standard, account of CS and PSO, lists the parameter values (n, w, alpha, K, beta), and correctly describes the hybrid idea from Zhang et al. [6]: use an ANN to predict a voltage window around the global MPP, then run P&O inside that window. The simulations show the expected behavior: plain P&O converges to a local peak, while the hybrid method reaches the global peak quickly. For a conference paper, the writing is readable.

The soft spots are significant. The ANN training data (Table I) is not reproduced, so the reader cannot check whether the test irradiance pattern (600, 800, 1000) is part of the training set. If it is, as the stress-test note suggests, then the ANN is not predicting the window for an unseen condition; it is retrieving a stored coordinate. The comparison with PSO/CS is then biased in favor of the hybrid method. The paper also never tests a different shading pattern, so there is zero evidence of generalization. On top of that, Table II has a citation placeholder " [ ?]" and no actual KC200GT parameters, and no code or data are released, so the simulation is not reproducible. These are not minor quibbles; they directly undermine the paper's only claim.

The hybrid method itself is not new; the paper's own reference [6] is the source. The contribution here is a single simulation comparison on a standard array. That can be useful as a sanity check, but it is not a publishable research result on its own.

Who this is for: someone doing an MPPT literature survey might cite it as an example of the hybrid approach, but not as a source of reliable quantitative results. It does not deserve a serious referee in its current form. If the authors release code and data, show results on multiple unseen shading patterns, and fix the citation, it could become a modest conference contribution.

Referee Report

4 major / 5 minor

Summary. This conference paper proposes a hybrid maximum power point tracking (MPPT) method for photovoltaic systems under partial shading. The method uses an artificial neural network (ANN) to predict a voltage window (Vmin, Vmax) around the global maximum power point (GMPP), and then applies a perturbation-and-observation (P&O) algorithm within that window. The paper compares the hybrid method against particle swarm optimization (PSO) and cuckoo search (CS) in a single MATLAB/Simulink simulation with three series-connected KC200GT modules under irradiance 600, 800, and 1000 W/m². The reported results are that CS outperforms PSO (0.7 s vs. 1.2 s convergence), and the hybrid ANN+P&O method converges even faster (0.18 s) with fewer oscillations. The central claim is that the hybrid method achieves faster and more reliable GMPP tracking than the two evolutionary algorithms.

Significance. If substantiated, the hybrid ANN-plus-P&O controller would be an interesting practical contribution to MPPT under partial shading, because it combines a modest-size neural network with a classical hill-climbing method to avoid the local-MPP trap while retaining fast convergence. However, the current evidence is not sufficient to support the claim. The paper provides no code or data, reports a single simulation run for each method, omits the ANN training set and training details, and does not test on any shading pattern outside the training distribution. The paper also contains unresolved production issues, such as a missing citation placeholder in Table II and a referenced but absent Table I. The contribution is therefore only potentially significant; its validity rests on several unverified assumptions about the ANN's generalization and the statistical representativeness of the reported convergence times.

major comments (4)

[Section III-A, Table I] Table I, which is stated to contain the ANN 'input data and optimal coordinates,' is not reproduced in the manuscript. The single test scenario (600, 800, 1000 W/m², Section IV-B) may or may not be one of the training samples; without Table I this cannot be checked. If the tested shading pattern is in the training set, the reported t=0.18 s reflects retrieval of stored coordinates rather than prediction for an unseen condition, invalidating the comparison against PSO and CS. The authors must report the training data, the train/test split, and at least one out-of-sample shading scenario, along with the ANN's prediction error for Vmin and Vmax.
[Section IV-B] The central comparison rests on one deterministic-looking simulation per method. PSO and CS are stochastic algorithms whose convergence time and oscillations vary between runs, and the hybrid method's performance depends on the ANN output for one specific shading pattern. No multiple runs, standard deviations, or statistical tests are reported. To support the claim of faster convergence and lower oscillations, the authors need to show results over several random seeds or over an ensemble of shading patterns, including patterns outside the ANN training set.
[Table II] Table II, 'Specification of the PV module KC200GT,' contains an unresolved citation placeholder '[ ?]' and its values are not listed in the text. The simulation parameters (open-circuit voltage, short-circuit current, etc.) are necessary to reproduce the P-V curve in Fig. 6 and to assess whether the reported GMPP power of 400 W is consistent with the module datasheet. This is a load-bearing omission for an experimental (simulation) paper.
[Sections III-A and IV-B] The ANN design is described only as '3 input layers, 10 hidden layers and 2 output layers' (Fig. 4 caption and surrounding text), with no information on the training algorithm, number of training samples, epochs, activation functions, or performance measures such as mean squared error. Since the hybrid method's entire advantage depends on the ANN returning a window (Vmin=79.99 V, Vmax=88 V) that brackets the GMPP, the absence of any accuracy or generalization metric for the ANN is a critical gap. A quantification of the Vmin/Vmax prediction error is needed.

minor comments (5)

[Introduction, paragraph 3] The sentence 'A comprehensive review of optimization techniques for renewable energy systems is presented in this article' overstates the paper's content; the paper is a short simulation study, not a review. Please rephrase.
[Abstract] The abstract uses the future tense ('MATLAB simulations will be used'), but the results are already reported; change to past or present tense.
[Section IV, first paragraph] There is a typographical inconsistency: 'Incon' appears in the abstract and introduction, while 'IncCon' appears later; please use one consistent abbreviation.
[Section IV-B, bullet list] 'Vmax = 88 V' has a missing space in 'andVmax'; also 'PP M' appears with an odd space in multiple places, likely a formatting artifact.
[Section II-A, Eq. (3)] The symbols 'u' and 'v' are introduced but not explicitly defined in the sentence following Eq. (3); they are later identified as normal random variables in Eq. (4), but the wording 'while and are determined' is incomplete. Please define all variables at first use.

Circularity Check

1 steps flagged · score 6.0 of 10

The ANN-predicted GMPP voltage window is trained on the same optimal-coordinate table used for the test scenario, making the reported hybrid speedup an in-sample fit rather than an independent prediction.

fitted input called prediction [Section III-A (Table I, Figs. 4-5) and Section IV-B]
"The illumination intensity G on each module is introduced into the nodes of the input layer (the input data and optimal coordinates are represented in Table. I), then transferred to the hidden layer, which passes their output to the nodes of the output layer, getting the optimal voltage areas of GMPP. ... The ANN allowed the P&O algorithm to limit its search to the GMPP’s vicinity by providing it with the values of Vmin = 79.99 V andVmax = 88 V."

The ANN is fitted to Table I, which contains 'input data and optimal coordinates': irradiance values as inputs and the optimal voltage bounds as targets. The paper reports no train/test split, no held-out shading pattern, and no generalization metric for the ANN. The only reported hybrid test uses modules receiving 600 W/m2, 800 W/m2, and 1000 W/m2, which is an irradiance pattern of the same type used to build Table I. Therefore the Vmin = 79.99 V and Vmax = 88 V window is a fitted retrieval from the training coordinates, not an independent prediction for an unseen condition. The claimed t = 0.18 s convergence is an in-sample benchmark against PSO and CS, which are not given access to the same precomputed optimal-coordinate table.

full rationale

The paper's central comparison in Section IV-B is that the hybrid ANN+P&O method reaches the GMPP in t = 0.18 s with fewer oscillations than PSO (t = 1.2 s) and CS (t = 0.7 s). The speedup is enabled by the ANN supplying a narrow voltage window (Vmin = 79.99 V, Vmax = 88 V) around the global peak. Section III-A states that the ANN is trained using NNTOOL on 'input data and optimal coordinates' represented in Table I, with irradiance on each module as input and the optimal voltage zones as output. Since the test irradiance pattern (600, 800, 1000 W/m2) is drawn from the same type of input data used to create that table, and since no held-out evaluation or ANN accuracy statistic is reported, the reported window is a fitted output rather than a verified prediction. However, this is not a self-citation or uniqueness-importation problem: the cited baselines and ANN references are external works. The circularity is specifically that the central 'prediction' of the GMPP vicinity reduces to the training table. The comparison against PSO/CS is therefore in-sample, and the generalizability of the hybrid method to unseen shading patterns is not demonstrated. This warrants a score of 6: the main speed comparison is partially circular because the ANN's output is fitted to the same optimal-coordinate data that defines the evaluation scenario.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities; the ANN is a function approximator, not a new mechanism. The free parameters are all manually chosen simulation settings, and the axioms are the modeling and generalization assumptions behind the claimed results.

free parameters (7)

PSO inertial weight w = 0.3
Chosen by the authors for the simulations in Section IV-A; affects convergence of the PSO baseline.
PSO acceleration constants α1, α2 = 1.2 each
Manual choice satisfying α1+α2 ≤ 4; affects the PSO trajectory.
CS Lévy multiplier K = 0.8
Set by the authors in Section IV-A; controls step size in the cuckoo search.
CS exponent β = 1.5
Standard value used in Lévy flight; still a design choice.
ANN hidden-layer node count = 10
Architecture choice stated in Section III-A; no tuning or sensitivity analysis is given.
ANN weights and biases = not provided
Trained with MATLAB NNTOOL on an unreleased dataset; the network's predictions are the crux of the hybrid method.
Boost converter elements L, C1, C2 = L=0.0005 H, C1=800 µF, C2=850 µF
Simulation component values; no justification given.

assumptions (4)

domain assumption The KC200GT PV module and boost converter models used in MATLAB/Simulink accurately represent a real photovoltaic system.
Invoked in Section IV; the entire simulation comparison depends on model fidelity.
domain assumption The GMPP of the array lies within one of the precomputed voltage zones, so restricting P&O to [Vmin, Vmax] can find it.
Section III states that the GMPP distribution under various shading patterns lands in three zones; this is taken from ref [6] and assumed.
domain assumption The ANN trained on irradiance-to-optimal-coordinate pairs generalizes to the shading patterns used in the evaluation.
Section III-A; if the test cases are outside the training distribution, the hybrid method can fail to bracket the global peak.
domain assumption The chosen PSO and CS parameter values are adequate for the GMPP search.
No sensitivity study is provided; parameter values are fixed by hand in Section IV-A.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Optimal MPPT Control of a Photovoltaic System Under Non-uniform Irradiation." pith.science (2026). https://pith.science/paper/KEY2YUMN

@misc{pith2026241116650,
  author       = {Pith},
  title        = {Pith review of: Optimal MPPT Control of a Photovoltaic System Under Non-uniform Irradiation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KEY2YUMN}},
  note         = {Machine review of arXiv:2411.16650}
}

read the original abstract

Under Partial shading conditions (PSC), traditional MPPT methods such as, P&O, IncCon and SMC, cannot track down the Global MPP. Thus, the energy conversion of the PV modules will decrease. To overcome this drawback, two evolutionary algorithms; PSO and CS, were presented. Also, the study introduces a hybrid optimization method to enhance the overall performance of the PV system under PSC. MATLAB simulations will be used in order to illustrate the efficiency of the proposed method.

Figures

Figures reproduced from arXiv: 2411.16650 by the authors.

**Figure 2.** GMPP distribution of the PV array under different PSC [6]. TABLE I. INPUT DATA AND OPTIMAL COORDINATES. analysis, the basic idea of the method is to predict the optimal voltage zones, using artificial neural networks, namely Vmin and Vmax, from the light intensity on different modules. Then, classical algorithm is applied to the zones to achieve the global maximum power point. A. Artificial neural network (ANN) Arti… view at source ↗

**Figure 5.** ANN trained structure of the PV system simulated in MATLAB/Simulink [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗

Figures from the paper (6 more)

**Figure 6.** Figure 6: P-V Curve under partial shading [PITH_FULL_IMAGE:figures/full_fig_p004_6.png]

**Figure 7.** Figure 7: Input and output voltages of the Boost converter with [PITH_FULL_IMAGE:figures/full_fig_p005_7.png]

**Figure 11.** Figure 11: Input and output voltages of the Boost converter with [PITH_FULL_IMAGE:figures/full_fig_p005_11.png]

**Figure 12.** Figure 12: Input and output currents of the Boost converter with [PITH_FULL_IMAGE:figures/full_fig_p005_12.png]

**Figure 13.** Figure 13: GPV power and the transmitted power to the load with ANN [PITH_FULL_IMAGE:figures/full_fig_p006_13.png]

**Figure 14.** Figure 14: Evolution of the Duty Cycle D with ANN. C. CONCLUSION In this paper, we have focused on the problem of nonuniform irradiation in photovoltaic systems, known as partial shading. An MPPT control based on the use of intelligent algorithms CS and PSO was first studied. W…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 10 canonical work pages

[6]

A modified hybrid maximum power point tracking method for photovoltaic arrays under partially shading condition,

W. Zhang, G. Zhou, H. Ni, and Y . Sun, “A modified hybrid maximum power point tracking method for photovoltaic arrays under partially shading condition,” IEEE Access, vol. 7, pp. 160091-160100, 2019

work page 2019
[1]

A comprehensive review and classified comparison of MPPT algorithms in PV systems,

M. Sarvi and A. Azadian, “A comprehensive review and classified comparison of MPPT algorithms in PV systems,” Energy Systems, vol. 13, no. 2, pp. 281-320, 2022

work page 2022
[2]

A new array reconfiguration scheme for solar PV systems under partial shading conditions,

Nihanth, M. S. S., Rajasekar, N., Pillai, D. S., & Prasanth Ram, J. (2020). “A new array reconfiguration scheme for solar PV systems under partial shading conditions,”. in Intelligent Computing Techniques for Smart Energy Systems: Proceedings of ICTSES 2018 (pp. 387-396). Springer

work page 2020
[3]

Comparative study of P&O and InC MPPT algorithms,

W. Christopher and R. Ramesh, “Comparative study of P&O and InC MPPT algorithms,” American Journal of Engineering Research (AJER), vol. 2, no. 12, pp. 402–408, 2013

work page 2013
[4]

Comparison of Swarm Optimization Methods for MPPT in Partially Shaded Photovoltaic Systems,

H. S. Moreira, J. L. S. Silva, G. C. S. Prym, E. Y . Sak ˆo, M. V . G. dos Reis, and M. G. Villalva, “Comparison of Swarm Optimization Methods for MPPT in Partially Shaded Photovoltaic Systems,” in 2019 International Conference on Smart Energy Systems and Technologies (SEST), IEEE, 2019, pp. 1–6

work page 2019
[5]

A Maximum Power Point Tracking (MPPT) for PV system using Cuckoo Search with partial shading capability,

J. Ahmed and Z. Salam, “A Maximum Power Point Tracking (MPPT) for PV system using Cuckoo Search with partial shading capability,” Applied energy, vol. 119, pp. 118–130, 2014

work page 2014
[7]

Cuckoo search via L ´evy flights,

X. S. Yang and S. Deb, “Cuckoo search via L ´evy flights,” in 2009 World congress on nature & biologically inspired computing (NaBIC), IEEE, 2009, pp. 210–214

work page 2009
[8]

A new optimizer using particle swarm theory,

R. Eberhart and J. Kennedy, “A new optimizer using particle swarm theory,” in Proceedings of the sixth international symposium on micro machine and human science (MHS’95), IEEE, 1995, pp. 39-43

work page 1995

Show all 10 references

[9]

ANN based MPPT al- gorithm design using real operating climatic condition,

Bouakkaz, Mohammed Salah, Ahcene Boukadoum, Omar Boudebbouz, Ahmed Bouraiou, Issam Attoui, and others. “ANN based MPPT al- gorithm design using real operating climatic condition,” 2020 2nd international conference on mathematics and information technology (ICMIT). IEEE, 2020

2020
[10]

KC200GT datasheet,

Kyocera, “KC200GT datasheet,” Energy Matters, https://www.energymatters.com.au/images/kyocera/KC200GT.pdf (accessed May 4, 2023)

2023

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [6]

A modified hybrid maximum power point tracking method for photovoltaic arrays under partially shading condition,

W. Zhang, G. Zhou, H. Ni, and Y . Sun, “A modified hybrid maximum power point tracking method for photovoltaic arrays under partially shading condition,” IEEE Access, vol. 7, pp. 160091-160100, 2019

work page 2019

[2] [1]

A comprehensive review and classified comparison of MPPT algorithms in PV systems,

M. Sarvi and A. Azadian, “A comprehensive review and classified comparison of MPPT algorithms in PV systems,” Energy Systems, vol. 13, no. 2, pp. 281-320, 2022

work page 2022

[3] [2]

A new array reconfiguration scheme for solar PV systems under partial shading conditions,

Nihanth, M. S. S., Rajasekar, N., Pillai, D. S., & Prasanth Ram, J. (2020). “A new array reconfiguration scheme for solar PV systems under partial shading conditions,”. in Intelligent Computing Techniques for Smart Energy Systems: Proceedings of ICTSES 2018 (pp. 387-396). Springer

work page 2020

[4] [3]

Comparative study of P&O and InC MPPT algorithms,

W. Christopher and R. Ramesh, “Comparative study of P&O and InC MPPT algorithms,” American Journal of Engineering Research (AJER), vol. 2, no. 12, pp. 402–408, 2013

work page 2013

[5] [4]

Comparison of Swarm Optimization Methods for MPPT in Partially Shaded Photovoltaic Systems,

H. S. Moreira, J. L. S. Silva, G. C. S. Prym, E. Y . Sak ˆo, M. V . G. dos Reis, and M. G. Villalva, “Comparison of Swarm Optimization Methods for MPPT in Partially Shaded Photovoltaic Systems,” in 2019 International Conference on Smart Energy Systems and Technologies (SEST), IEEE, 2019, pp. 1–6

work page 2019

[6] [5]

A Maximum Power Point Tracking (MPPT) for PV system using Cuckoo Search with partial shading capability,

J. Ahmed and Z. Salam, “A Maximum Power Point Tracking (MPPT) for PV system using Cuckoo Search with partial shading capability,” Applied energy, vol. 119, pp. 118–130, 2014

work page 2014

[7] [7]

Cuckoo search via L ´evy flights,

X. S. Yang and S. Deb, “Cuckoo search via L ´evy flights,” in 2009 World congress on nature & biologically inspired computing (NaBIC), IEEE, 2009, pp. 210–214

work page 2009

[8] [8]

A new optimizer using particle swarm theory,

R. Eberhart and J. Kennedy, “A new optimizer using particle swarm theory,” in Proceedings of the sixth international symposium on micro machine and human science (MHS’95), IEEE, 1995, pp. 39-43

work page 1995