Enhancing Solar Power Forecasting Accuracy Using HMPCS and Machine Learning Techniques: An Applied Study Authors: Asmaa Abdul-Hussein Aziz, Iraq T. Abbas, Ashif Mohammad, Bin Du, Bin Du, iraq abbas, Ashif Mohammad, iraq abbas Published: 2026 ABSTRACT BackgroundSolar irradiance is a nonlinear and intermittent function, which makes accurate forecasting of solar power generation a challenge. The high variability of meteorological conditions is not well represented by conventional atmospheric models, thus hampering forecasting skill and model robustness. In this work, an advanced hybridization of multi-population cuckoo search (HMPCS) algorithm with machine learning (ML) methods is developed to enhance the predicated performance of photovoltaic (PV) power forecasting with more reliability.MethodsIn this study, a hybrid modeling framework is proposed, called HMPCS–ML framework which captures the global search capacity of HMPCS and predictive power of sophisticated ML models (Long Short-Term Memory (LSTM), Light Gradient Boosting Machine (LightGBM)). Optimizing hyperparameters by balancing exploration and exploitation, the algorithm runs on multi-populations through Lévy flight randomization. Interpolation, normalization, and temporal windowing were utilized to preprocess synthetic meteorological and irradiance datasets. We evaluated the framework by comparing commonly used statistical measures (MAE, RMSE, MAPE, R 2).ResultsMoreover, experimental analyses showed that HMPCS–ML models significantly outperformed baseline approaches (Grid Search and Particle Swarm Optimization (PSO)). Results showed that the optimized LSTM+HMPCS model outperformed other models in terms of lowest RMSE (0.139) and highest R 2 (0.93), reflecting the LSTM model’s good fit with practical observations and generalization ability. The optimal LightGBM+HMPCS variant also proved to be consistently better, with reduced error (23% lower than unoptimized models).ConclusionsIn this regard, the HMPCS–ML framework is a powerful and efficient solution for the optimization of Photovoltaic (PV) power forecasting, improving the predictive performance and calculation efficiency. This research shows the potential of hybrid metaheuristic–ML integration for renewable energy prediction and smart-grid applications in general and indicates further extensions to multi-objective and Transformer-based architectures. 1. Introduction The drive for installing more renewable energy in current power systems means short-term forecasting of photovoltaic power is now urgently needed. However, the non-linear and intermittent nature of PV generation makes forecasting incredibly difficult, particularly in cases where there are variable cloud motion and atmospheric instability. Recent advanced research in deep learning has demonstrated that multi variate meteorological and irradiance time series can significantly enhance forecast performance by capturing temporal as well as non-linear dependencies. In fact, in recent surveys that record a clear transition towards deep learning (DL) architectures and hybrid pipelines for solar forecasting that exploit the potential of multivariate meteorological and irradiance time series, as they contain a wealth of spatial and temporal information. 1 – 4 In the DL family, hybrid and attention models have pushed the performance envelope of Photovoltaic (PV) power forecasting. For example, TCN–ECANet–GRU (a temporal conventional network combined with a basic channel attention module and Gated Recurrent Units) achieved significant improvements over strong baselines on years in temporal predictions of a real PV plant data set. 5 Meanwhile, Transformer-based architectures are occupying an ever-increasing space in Photovoltaic (PV) power forecasting workflows as several reviews converge on the significant, notable improvements over classical recurrent models achieved by Transformer variants specially tailored to time series forecasting (e.g. transmit at longer horizons) through modelling long-range temporal dependencies and multi-horizon outputs. 6 , 7 A second line of research will show that hyper-parameter optimization (HPO) and feature selection are key components to obtaining the best performance from the ML/DL forecasters. Moreover, while recent studies have shown that metaheuristic-based optimization of hyperparameters can lead to improved accuracy in solar forecasting models, there are still some issues. First, many existing studies focus on single-population optimizers grid search or random searches like PSO which will suffer from premature convergence or lack diversity in the search space when it is complicated. The second problem arises with models -Transformers-based, hybrid deep learning ones with high prediction performance in recent days. Forecasting machine learning with optimization strategies for multi-population metaheuristics has not yet been systemically combined and studied in practical settings, offering a new direction of research for future endeavors in this direction. 8 – 10 In addition, literature often reports improved forecasting accuracy without separating the contribution of a forecasting model from that of its hyperparameter optimization. As a result, it is currently impossible to tell whether improvements in performance are due primarily to better architecture or to better choice of hyperparameters, or both. 11 – 15 To get round these problems, this paper introduces a new Hybrid Multi-Population Cuckoo Search with Machine Learning (HMPCS–ML) framework for Photovoltaic (PV) power forecasting. Based on the tradition of diversity-preserving multi-population search methods and machine learning models for forecasting which combine Lévy flight-based exploration, the new methodology aims to improve predictive accuracy by optimizing hyperparameters. 16 Unlike conventional optimization methods, the HMPCS strategy has been specifically designed to prevent premature convergence and to foster adaptability over a wide range of different modes settings. The contributions of this paper are as follows: Proposing a hybrid HMPCS–ML framework for forecasting photovoltaic power utilizing LSTM and LightGBM models. Provide a comparative assessment of such a technique against known optimization strategies. Demonstrate the value of multi-population optimization for both improving the accuracy and integrity of a search. Terminology Clarification. In this work, the two terms Photovoltaic (PV) power forecasting and photovoltaic (PV) power forecasting are used interchangeably since both refer to predicting electrical power produced by PV systems. The term more general renewable energy prediction is reserved only when the forecast methodologies are discussed which are applicable for predicting the different renewable energy sources in general. We use Photovoltaic (PV) Power Forecasting as the main terminology for consistency throughout this manuscript. Organization of the Paper. The rest of the paper is organized as follows. Related works recently proposed in the field of photovoltaic power forecasting, machine learning techniques, and metaheuristic optimization methods are reviewed in Section 2. In Section 3, the proposed Hybrid Multi-Population Cuckoo Search (HMPCS) optimization framework combined with the machine learning models is described; In Section 4, the experimental setup with datasets and evaluation metrics is presented along with discussion of the obtained results against baselines. Finally, Section 5 concludes by summarizing the key findings and contributions of this work, as well as laying out potential avenues for future research. 2. Literature Review In the last several years, there has been a growing interest in integrating meta-heuristic optimization algorithms with machine learning (ML) for the forecasting of renewable energy systems. The hybrid approaches can improve predictive performance by searching through optimal hyper conditions for the model and working toward increasing the robustness of the model through the adaptive search traits of the optimization methods for instance. 17 Developers of the so-called hybrid Gorilla Troops Optimizer (GTO) and Beluga Whale Optimization (BWO) algorithm 15 proposed a new methodology with state-of-the-art accuracy for the modelling of photovoltaic (PV) systems. 18 Caselli et al. XL with ML-based clustering for population dynamics in bio inspired algorithms. 19 Abd El-Mageed et al. (2024) developed a PV calibration model using DE, leading to significant decreases in RMSE. 20 In Moayedi and Mosavi (2021), both Cuckoo Search (CS) and artificial neural networks (ANN) were employed to validate swarm intelligence for optimization tasks in electrical demand forecasting 21 ; however, Lotfi Nejad et al. (2023) and Mohsin et al. 22, 23 Moreover, Li et al. (2018) made a comparison between different kinds of neural architectures and suggested the same conclusion that hybrid models (such as Bat-NN and GRNN) demonstrate better performance than pure neural networks in applications such as PV and energy forecasting. 24 Despite these contributions, gaps remain. More particularly, although hybrid meta-heuristic–ML models enhance performance in terms of accuracy and convergence, the systematic integration of multi-population meta-heuristics with state-of-the-art deep learning architectures, such as LSTM and Transformer models, for renewable energy forecasting has not yet been comprehensively investigated. Table 2 compares the proposed HMPCS–ML framework restatement to some representative recent studies at this age of PV power forecasting research, to give a clearer perspective of the current situation. This comparison features a few of our selection criteria the forecasting model, optimization strategy, individual dataset type, as well as major strengths and current limitations of both approaches. The proposed framework devotes to filling the gap in previous studies and is radically different by incorporating diversity-preserving multi-population optimization strategy with state-of-the-art machine learning approaches. Table 1. This table summarizes recent hybrid meta-heuristic and machine-learning models (2021–2025) used in renewable-energy forecasting.It compares the methods, their components, datasets, and achieved accuracy to highlight the evolution and effectiveness of hybrid optimization ML approaches. Author(s)YearTechniqueApplicationKey result(Zanial, 2023)2023CS + ANNphotovoltaic (PV) power forecastingOutperformed standalone ANN(Ali A., 2023)2023GTO + BWOPV ModelingImproved accuracy(Caselli, 2023)2023CSA + MLGlobal OptimizationBetter convergence via clustering(Abd El-Mageed, 2024)2024SSO + DEPV CalibrationHigh RMSE reduction(Nayak, 2025)2025Hybrid CS + Transformerphotovoltaic (PV) power forecastingSuperior RMSE and MAPE improvements This table summarizes recent hybrid meta-heuristic and machine-learning models (2021–2025) used in renewable-energy forecasting. It compares the methods, their components, datasets, and achieved accuracy to highlight the evolution and effectiveness of hybrid optimization ML approaches. Table 2. Comparison of the proposed HMPCS–ML framework with recent state-of-the-art photovoltaic forecasting studies. Study Forecasting model Optimization method Dataset type Key strength Limitation TCN–ECANet–GRU Hybrid DLNone/standard tuningReal PVStrong temporal modelingLimited optimizer robustness Transformer-based PV model TransformerStatic tuningReal PVLong-range dependency learningHigh complexity PSO-LSTM LSTMPSOPV time seriesBetter tuning than grid searchPremature convergence risk GA-based Photovoltaic (PV) power forecasting ANN/LSTMGAPV dataFlexible searchSlower convergence Proposed HMPCS–ML LSTM, LightGBMHMPCSPhotovoltaic (PV) power forecastingMulti-population diversity, robust search, improved accuracyRequires further large-scale validation Comparison of the proposed HMPCS–ML framework with recent state-of-the-art photovoltaic forecasting studies. Out of the limitations mentioned above, this work proposes an HMPCS–ML framework to address the lack of integration between global search and advanced temporal learning architectures (TLA) for time-series data. See Table 2 for a detailed Comparative Review of Hybrid Meta-heuristic–ML. However, most recent photovoltaic forecasting studies are presented based on general optimization methods (or static hyperparameter tuning strategies), as shown in Table 2. Despite achieving relatively good prediction accuracy, these approaches often suffer from premature convergence, limited exploration ability and insufficient population diversity in the optimization process. By contrast, the HMPCS–ML proposed herein implements a multi-population cuckoo search strategy with both Lévy-flight exploration and elite information exchange to enhance small-step (local search) and large-step (global search) performance trade-off using harmonically structured data. 25 As a result, the proposed algorithm enables more reliable hyperparameter tuning and consistently delivers better prediction performance than its counterparts. In order to present proposed framework’s relationship to current technology, recent studies can be divided into three main categories: (1) deep learning for weather prediction, including LSTM, GRU, CNN–RNN, and Transformer varieties; (2) metaheuristic hyperparameter optimization methods by Particle Swarm Algorithm (PSO), Genetic Algorithms (GA), Grey Wolves Optimization (GWO) and Cuckoo Search (CS); and (3) Hybrid weather prediction-forecasting relocation planning that combines deep models and solver-based strategies. Existing deep learning research predicts solar irradiance more accurately, particularly when attention mechanisms and Transformer-based transmission models are used. However, most such reports employ standard or static hyperparameter tuning methods to obtain results. On the other hand, metaheuristic optimization techniques enhance tuning flexibility; but many publications abbreviated the studies to one-population strategies without explicitly discussing how to maintain variety or the sturdiness of the optimizer in Photovoltaic (PV) power forecasting. Thus, this study is not just an attempt to bring machine learning and optimization together. The research debut is the development of a multi-population cuckoo based framework that is specially designed both to maintain population diversity, maintain a balance between exploration and exploitation, and to offer more reliable hyper parameter tuning for photovoltaic forecasting models. Study Limitations. The received prediction results were rousing, but the R 2 values can still be affected by the variation of meteorological conditions, complexity of photovoltaic generation behavior and characteristics of benchmark datasets in this study. In the future, we will evaluate more extensive datasets collected from various geographical areas and combine transformer-based deep learning architectures with the proposed HMPCS optimization framework to enhance forecasting accuracy. As summarized in Table 3, the proposed HMPCS–ML framework differs from recent photovoltaic forecasting studies by combining machine learning models with a diversity-preserving multi-population optimization strategy rather than relying on single-population or static hyperparameter tuning approaches. Table 3. The table lists the main environmental and historical PV variables used as model inputs to improve solar-power prediction accuracy.Feature SourceDescription Unit Global Horizontal Irradiance (GHI) NASA POWERSolar radiation incident on a horizontal surfaceW/m 2 Ambient Temperature NASA POWERAir temperature measured near the surface°C Wind Speed NASA POWERWind velocity at 10 m heightm/s Relative Humidity NASA POWERAtmospheric moisture content% Timestamp PV WattsHourly temporal index– The table lists the main environmental and historical PV variables used as model inputs to improve solar-power prediction accuracy. 3. Methodology We propose a methodical and systematic approach that consists of five main steps: (i) data acquisition, (ii) data processioning, (iii) model designing, (iv) HMPCS (Hybrid Multi-Population Cuckoo Search)-based parameter optimization, and (v) comprehensive model evaluation. Data Acquisition ensures that meteorological and irradiance variables, including global horizontal irradiance (GHI), temperature, relative humidity, and wind speed, are acquired from established sources such as NASA POWER and PV Watts. 26, 27 In addition, the raw dataset used in this study is publicly available through the Zenodo repository under Reference 28. This dataset provides the meteorological and photovoltaic observations used in the development and evaluation of the HMPCS–ML framework, supporting data transparency and reproducibility of the reported experiments. 28 These variables are to take into consideration the atmospheric and solar phenomenon and they constitute the most primal input possible to any prediction model. Determining the basic operations required for processioning like dealing with the missing values, outlier detection, normalizing the data, and temporal sequence building to make the data more qualitative. The uniformity and comparability of synthetic variables—two key factors for equipment dependency as also help in providing the viability of ML and deep learning models. Temporal dependencies: The model architecture incorporates cutting-edge ML techniques such as LSTM and GRU networks for learning the short & long-term dependencies in the time-series data. Carry out feature engineering to elicit the physical relations that determine impact on PV generation. HOMPs: Heuristic or meta-heuristic driven optimization methods for parameters or hyper-parameters. By utilizing multi-population diversity and Lévy flight strategies, HMPCS prevents local convergence and promotes global exploitation; thus, optimizing its superior properties over traditional optimization algorithms. Statistical indices (MAE, RMSE, MAPE, R 2) and visualization methods are used in the Comprehensive Model Evaluation to validate the predictive performance. This compares the proposed framework to baseline ML methods and challenges robustness with multiple meteorological conditions. Finally, since we believe that resilience against missing data and noise is what makes the way we expressed robustness, that clear steps to follow so that others can use them to replicate our results bring reprehensibility, and that the fact that different datasets and forecasting horizons can be implemented leads to adaptability, we propose the framework as a scientifically-grounded contribution for the fields of renewable energy forecasting. 3.1 Data Preprocessing Equations Normalization (Min–Max Scaling): The pre-processing equations used in this study are widely used standard normalization formulations adopted from machine learning and photovoltaic forecasting applications. The equations are not being derived for the first time but rather presented to describe what happens prior to training the models. The formulation is according to popular Min–Max normalization method which has been already widely discussed in machine learning literature. 29 , 30 Xscaled(t)=(X(t)−Xmin)/(Xmax−Xmin) Z-score Standardization: Xnorm(t)=(X(t)−μX)σX Sliding Window Representation: St={X(t−L+1),…,X(t)} 3.2 LSTM Model Equations The equations of Long Short-Term Memory (LSTM) network are the same as given in the common architecture. 30 Authors present these equations to fully describe the forecasting model employed in this study, and they were not derived newly by authors. 30 Forget Gate: ft=σ(Wf·[ht−1,xt]+bf) Input Gate: it=σ(Wi·[ht−1,xt]+bi) Candidate Cell State: C^t=tanh(WC·[ht−1,xt]+bC) Updated Cell State: Ct=ft⊙Ct−1+it⊙C^t Output Gate: ot=σ(Wo·[ht−1,xt]+bo) Hidden State: ht=ot⊙tanh(Ct) 3.3 HMPCS Optimization Equations The proposed Hybrid Multi-Population Cuckoo Search (HMPCS) framework is built upon the original Cuckoo Search algorithm presented by Yang and Deb. 31 This paper inherits the base Lévy-flight search equations from original formulation, while the chosen multi-population strategy, adaptive information exchange mechanism and hybrid optimization solution highlight methodological contributions. Lévy Flight Update Rule: X(k+1)=X(k)+α·Lévy(λ) Fitness Function (RMSE Minimization): RMSE=sqrt((1/N)Σ(yt−y^t)2) Population Diversity Update: P(g+1)=P(g)+β·(Pbest−Prand) Exploration–Exploitation Switching: Ifr 3) 36 and substitute outliers with the mean numbers of their respective neighborhood based on data type and vendor. Standardization of Attributes: Given that the model learns inside a local context constrained between 0 and 1, all features were normalized to the relevant range using Min-Max scaling, which is effective for adjusting attribute ranges that differ from the input. 37 Partitioning of Training and Testing Data: In accordance with the acceptable literature, the data is divided into learning (70%) and validation (30%) sets. 38 Temporal Structuring: Input data were transformed into temporal windows of 24–48 hours to accommodate possible Dependencies across time scales, facilitating the integration of the LSTM/GRU models. 29, 39 The proposed HMPCS–ML framework is depicted in Figure 1 and consists of six key stages: data collection, imputation of missing values, outlier detection, normalization, a data splitting stage for train and test, and the temporal sequencing to match the time index to the model input. This schematic summarizes the preprocessing steps applied before model training: raw data collection, missing value imputation, outlier removal, normalization, train–test splitting, and temporal sequencing for time-series forecasting. 3.7 Machine Learning Models To this end, the performance of the proposed hybrid framework was evaluated in two different machine learning (ML) models: Long Short-Term Memory (LSTM): LSTM networks are one of the recurrent neural networks (RNN) variations capable of learning long-term temporal dependencies in sequential data. Furthermore, they retain the temporal information, making them most suitable for time forecasting issues, such as photovoltaic (PV) power forecasting, in which some meteorological variables exhibit diurnal and seasonal periodontitis. 31, 32 To address the vanishing gradient problem, a disadvantage of classical recurrent neural networks (RNNs), LSTMs employ gated mechanisms (input, forget, and output gates), which enable them to learn both long-term and short-term dependencies effectively. 33 Light Gradient Boosting Machine (Light GBM): It is a tree-based ensemble learning algorithm that utilizes gradient boosting decision trees (GBDT), which employs an ensemble of weak learners. Light GBM can achieve fast training and efficient memory usage. Light GBM supports low-latency inference and is known to perform better in the high-dimensional tabular data space than deep neural architectures. 30 Because of this feature, it can consider both categorical and continuous features simultaneously, making it a powerful associative model compared to LSTM in solar forecasting tasks. Light GBM is also well-known for its low computational cost in many renewable energy applications, as well as its inherent resilience to over-fitting, provided suitable regularization is added. 34 We built a system that combines long short-term memory (LSTM) as a neural sequence model to exploit globally optimal attribute combinations, with Light GBM as an ensemble method to efficiently process heterogeneous meteorological inputs, by leveraging the complementary advantages of high-performing modelling methods. 3.8 Optimization Using HMPCS The HMPCS approach was combined with both ML models as a hyper-parameter optimization engine to enhance their forecasting performance. While the standard Cuckoo Search (CS) focuses solely on replacing the worst-performing cuckoos identified as such, HMPCS differs in that it maintains multiple evolving sub-populations throughout a single optimization run. This allows for an effective trade-off between exploration (i.e. global search over parameter space) and exploitation (i.e. local search around promising solutions). So, the way the optimization process works is: Each sub-population evolves in isolation but periodically exchanges elite solutions to avoid premature convergence. → Lévy flight randomization to diverge the searching course of the path that can be used by candidate solutions to get out of the local minima. HMPCS encapsulates large population structures (with associated diversity) at the cost of speed, due to the need for many reproducible runs, while retaining the most optimal hyper-parameters. In this work, we directly optimize predictive accuracy and generalization by using the negative cross-validation score of the target ML model as the fitness function. HMPCS Parameter Settings: Number of subpopulations: 3 Per subpopulation ( N = 20): α: Lévy flight parameter: 1.5 Maximum generations: 50 It was focused on important hyper-parameter optimization. These were: LSTM: learning rate, hidden layers, neurons on layer, dropout and sequence length Regularization limits and n estimators , max depth , learning rates for Light GBM We propose a framework that utilizes HMPCS to automatically fine-tune LSTM models for both seen and unseen data, 8 achieving high accuracy over manually tuned or grid-searched baselines. 3.9 Synthetic Data Generation Various synthetic datasets were created as well in addition to the real meteorological data, to assess robustness of proposed HMPCS–ML framework. Synthetic data generation is a common method, applied to the task of simulating uncertainty and variability in various systems (especially renewable energy forecasting). The synthetic data were formed by adding controlled Gaussian noise and perturbations to actual meteorological variables|that is irradiance, temperature, wind speed. This maintains the sequential nature of the data plus variability necessary (to simulate Uncertain real world). The synthetic dataset generation can be expressed as: Xsynthetic=Xreal+ε where ε is Gaussian noise with zero mean and controlled variance. This is a way to assess model stability under Varying degrees of uncertainty while still being consistent with realistic photovoltaic patterns. 3.10 Experimental Setup and Comparison Four experimental analyses were conducted to assess the stability and the generalization performance of the proposed hybrid HMPCS–ML framework. Use of publicly available meteorological datasets that guarantee reprehensibility and transparency of the results. This experiment measures the performance of the proposed framework against the best existing hyper-parameter tuning. Besides Classic Methods we compared also HMPCS with Grid Search (high computational cost, due to large search area) and Particle Swarm Optimizations (PSO, fast and popular swarm intelligence technique, more suitable for low-non-adaptability in high-dimension search space). The analysis was performed using typical error factors: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination (R 2). Finally, as an additional exploration analysis, convergence curves (i.e., the optimization dynamics of HMPCS in contrast to PSO & Grid Search were also; ~ 10 lines for each comparison + maps + figs). Using this comparative framework, HMPCS also has the following advantages: Parallel populations to speed up convergence System Architecture-periodic sharing of elite solutions to avoid local minima-adaptive hyper-parameter tuning for a few critical hyper-parameters enabling Forecasting accuracy improvement for LSTM and Light GBM models. Thus, the evidence from these experiments, in addition to direct demonstration of how HMPCS outperforms the previous strategies in optimizing, indicates that the ability of HMPCS to maintain the balance between exploration and exploitation yields the most accurate predictions with the lowest amount of cost in computation. Table 4 clearly shows that HMPCS outperformed both Grid Search and PSO in all model configurations. The LSTM + HMPCS model achieved the lowest RMSE (0.139) and highest R 2 (0.93), indicating superior predictive performance. This validates the use of HMPCS as a robust optimizer for fine-tuning ML-based solar forecasting models. Table 4. This table compares how LSTM and LightGBM perform under different hyper-parameter optimization methods.It shows that using HMPCS improves both models’ accuracy more than the standard or grid/random search settings.ModelRMSE (mean ± std)MAE (mean ± std)MAPE (mean ± std) R 2 (mean ± std) LSTM 0.182 ± 0.0050.141 ± 0.0049.8 ± 0.30.87 ± 0.01 LSTM + PSO 0.161 ± 0.0040.125 ± 0.0038.9 ± 0.20.90 ± 0.008 LSTM + HMPCS 0.139 ± 0.0030.110 ± 0.0027.6 ± 0.20.93 ± 0.006 LightGBM 0.175 ± 0.0060.136 ± 0.0059.5 ± 0.40.88 ± 0.01 LightGBM + HMPCS 0.148 ± 0.0040.118 ± 0.0038.1 ± 0.30.91 ± 0.007All results are reported as mean ± standard deviation over 10 independent runs. This table compares how LSTM and LightGBM perform under different hyper-parameter optimization methods. It shows that using HMPCS improves both models’ accuracy more than the standard or grid/random search settings. All results are reported as mean ± standard deviation over 10 independent runs. We adopted a cross-validation strategy over time-series to achieve sound evaluation and overcome oversaturation. Specifically, we followed a rolling-origin (walk-forward) validation approach in which the training window is gradually extended and the model evaluated on subsequent unseen segments of data. A 5-fold time-series cross-validation scheme was adopted (with gaps between each period) so that the order of observation is kept for temporal stability. Quite unlike normal k-folds cross-validation. Each experiment was repeated 10 independent times with fixed random seeds to ensure reproducibility and stability of the results. The average performance metrics (MAE, RMSE, MAPE, R 2) were reported with standard deviations to give a relatively solid verdict on model performance. 3.10.1 Statistical Validation To guarantee that the claimed improvements made by the proposed HMPCS-ML architecture are statistically real, a series of non-parametric statistical tests were carried out. Specifically, using the Wilcoxon signed-rank check, the efficiency of HMPCS–ML algorithms was compared to those of other optimization methods in wide use, such as Grid Search and Particle Swarm Optimization (PSO), over multiple independent runs. Each test was done ten times and the resulting performance indicators (RMSE and R 2) were recorded. Where differences in performance occurred, a Wilcoxon test was applied to find out whether they were statistically important. Most of the obtained p-values were less than 0.05, proving that this HMPCS-ML system has made significant improvements over the benchmark systems compared with it. These findings confirm that the beneficial performance gains observed are robust and not simply due to random fluctuation. As Table 5 documents, stated p-values confirm the new HMPCS–ML approach has performed much more significantly than previous types of approaches. Table 5. Wilcoxon signed-rank test results for HMPCS–ML versus baseline methods.Comparisonp-value Result HMPCS vs Grid Search 0.011Significant HMPCS vs PSO 0.018Significant Wilcoxon signed-rank test results for HMPCS–ML versus baseline methods. 3.11 Ablation Study Moreover, the study design was employed to examine the respective roles of different components in the proposed framework. Specifically, three different model configurations were contrasted: a.baseline LSTM model without optimizationb.LSTM optimized through standard CSc.LSTM optimized using the proposed Hybrid Multi-Population Cuckoo Search (HMPCS) a.baseline LSTM model without optimization baseline LSTM model without optimization b.LSTM optimized through standard CS LSTM optimized through standard CS c.LSTM optimized using the proposed Hybrid Multi-Population Cuckoo Search (HMPCS) LSTM optimized using the proposed Hybrid Multi-Population Cuckoo Search (HMPCS) The results show the baseline LSTM model offers reasonable forecasting performance; however, optimization via CS will increase accuracy by tuning parameters better. And conversely, proposed HMPCS approach reaches best record, indicating that multi-population strategy has significant impact on balance between exploitation and exploration, themselves Combining to ensure system doesn’t mistakenly jump into a local minimum too fast before data points suggest an altogether different picture. These results demonstrate that it is not only due to the forecasting model that the proposed HMPCS approach, outperforms (Optimized)’s CS–ML above) shows the above performance. As shown in Table 6, the proposed HMPCS optimization provides the best performance compared to both the baseline model and the standard CS-based optimization. Table 6. Statistical performance of models (mean ± std over 10 runs).ModelRMSE (mean ± std) R 2 (mean ± std) LSTM (baseline) 0.182 ± 0.0050.87 ± 0.01 LSTM + CS 0.156 ± 0.0040.90 ± 0.008 LSTM + HMPCS 0.139 ± 0.0030.93 ± 0.006The results are reported as mean ± standard deviation over 10 independent runs. Statistical performance of models (mean ± std over 10 runs). The results are reported as mean ± standard deviation over 10 independent runs. According to Table 7, the HMPCS method slightly increases computational time when compared to traditional methods. Nonetheless, this is offset by significantly greater accuracy and robustness. Table 7. Computational time comparison of optimization methods.Method Average Runtime (s) Grid Search 45.2 PSO 38.6 HMPCS 52.4 Computational time comparison of optimization methods. We compared the predictive performance of our proposed HMPCS–ML framework against several representative photovoltaic power forecasting studies published in recent years as shown in Table 8. While some studies report R 2 values at least slightly higher than ours, they also need to be interpreted with caution in direct comparison of numerical performance, since forecasting performance depends heavily on multiple factors such as dataset characteristics, forecasting horizon, climate zone products used for validation and training data preprocessing strategies such as feature selection and evaluation protocols. Most existing high-performance studies have been created and assessed based on unique datasets collected under relatively stable environmental situations, while the suggested framework was intended to provide a strong generalized optimization approach to use in various PV forecasting scenarios. Table 8. Comparison State-of-the-Art PV Forecasting Studies.StudyMethodDatasetR 2 Main LimitationZhang et al. (2022)LSTMSingle PV Plant0.95Site-specificLi et al. (2021)Deep EnsemblePV Farm0.96High computational costZhang et al. (2023)Hybrid AIPublic Dataset0.94Static optimization Proposed HMPCS–ML HMPCS + LSTM + LightGBMPublic Dataset 0.93 General-purpose framework Comparison State-of-the-Art PV Forecasting Studies. Also, in contrast to simply obtaining the highest prediction accuracy, this research mainly contributes by providing an artificial intelligence optimization framework designed to optimize hyperparameters more efficiently and improve convergence performance as well as solution robustness using a better balance of exploration–exploitation behaviors throughout model training. Therefore, the HMPCS framework is consistently effective in enhancing forecasting performance and serves as a practical computationally efficient and scalable optimization strategy for renewable-energy based applications. 3.12 Results and Analysis Experimental results show that the proposed HMPCS–ML hybrid framework outperforms both baseline models and traditional optimization strategies. In all experiments, models optimized with HMPCS consistently outperformed both their standalone models and fine-tuned models with Grid Search or Particle Swarm Optimization (PSO). With improved by approximately 23% on average across multiple independent runs, with statistically significant differences confirmed using the Wilcoxon test, exemplifying the gains through hyper-parameter tuning. Likewise, LSTM+HMPCS also showed an improved RMSE and MAE, indicating that, as only the hyper-parameters were optimized, effective hyper-parameter optimization can decrease error rates and improve the model’s capability to generalize in changing weather conditions. 25 We further visualized the improvements using bar charts and comparative plots. Results from these showed that both Grid Search and PSO improved upon baseline performance to a degree; however, HMPCS provided the best trade-off between accuracy and speed. The better convergence of HMPCS was due to the use of a multi-population strategy and elite solution exchange in HMPCS, as it helps to escape from local minimum and to search the parameter space more thoroughly. Overall, the results validate HMPCS as a fast, powerful and robust optimizer with measurable benefits in RMSE and robustness over the current state of the art optimization methods in the renewable energy forecasting space. 3.13 Experimental Setup To ensure the reproducibility and transparency of the HMPCS–ML system presented in this study, we give a detailed description of our used experiments in this section. Datasets: The data used for our study were real-world photovoltaic datasets obtained from NASA POWER and PVWatts platforms. The dataset consists of meteorological variables such as Global Horizontal Irradiance (GHI), temperature, wind speed, and humidity. The total dataset consists of approximately 8,760 samples collected over a one-year period. Data Handling: All input features were normalized using Min-Max Scaler. Missing values were filled in using linear interpolation. Model setting: LSTM: Number of layers: 2, Hidden units: 64 Batch size: 32, Epochs: 100, Optimizer: Adam, Learning rate: 0.001 LightGBM: Number of leaves: 31, Learning rate: 0.05, Number of estimators: 100 HMPCS Parameters: Population size: 25, Subpopulations count: 3, Discover rate (pa): 0.25 Initial step size (a): 0.01, Maximum number of generations to run: 100 Hardware and Procedure: All the experiments in this paper were implemented using Python, TensorFlow and LightGBM libraries on a system with an Intel Core i7 processor, 16GB of memory, and the option to use a GPU for acceleration. Reproducibility: To help assure the stations and reliability of our experimental results, we used random seeds which were then fixed and repeated each of our different experiments 10 times in a row. Description of Data Sets: The data set used in this study was an hourly one consisting of 8,760 samples of meteorological and photovoltaic observations that were collected over the course of a year. Input features data Global Horizontal Irradiance (GHI), temperature, wind speed and humidity. Before incorporating data in the final study, establish a threshold value. As such, further designs and parameters of lysis methods (compart-mental, concurrent response or parametric) would be left entirely to user preference at this stage. The data split into training (70%), validation (15%) and testing (15%) sets while maintaining the original order of observations. 3.14 Evaluation Metrics Due to the various facets of model performance assessed, we applied the following evaluation metrics: • RMSE (Root Mean Squared Error): It computes the root of the mean squared error between predicted and observed alike values and is more severe at penalizing large discrepancies. We also preferred a low RMSE for the sake of good forecasting performance.• Mean Absolute Error (MAE): Simply it is mean of absolute differences of predicted values and truth values MAE is less sensitive to outliers than RMSE and provides a more interpretation measure of average forecast deviation in the same units as the forecast.• Coefficient of Determination (R 2 ): How much of the variance in the dependent variable is explained by the model? Higher R 2 values indicate stronger explanatory power, or a better fit between the model and the observed data. • RMSE (Root Mean Squared Error): It computes the root of the mean squared error between predicted and observed alike values and is more severe at penalizing large discrepancies. We also preferred a low RMSE for the sake of good forecasting performance. RMSE (Root Mean Squared Error): It computes the root of the mean squared error between predicted and observed alike values and is more severe at penalizing large discrepancies. We also preferred a low RMSE for the sake of good forecasting performance. • Mean Absolute Error (MAE): Simply it is mean of absolute differences of predicted values and truth values MAE is less sensitive to outliers than RMSE and provides a more interpretation measure of average forecast deviation in the same units as the forecast. Mean Absolute Error (MAE): Simply it is mean of absolute differences of predicted values and truth values MAE is less sensitive to outliers than RMSE and provides a more interpretation measure of average forecast deviation in the same units as the forecast. • Coefficient of Determination (R 2 ): How much of the variance in the dependent variable is explained by the model? Higher R 2 values indicate stronger explanatory power, or a better fit between the model and the observed data. Coefficient of Determination (R 2 ): How much of the variance in the dependent variable is explained by the model? Higher R 2 values indicate stronger explanatory power, or a better fit between the model and the observed data. These metrics together provide an overall assessment of the prediction accuracy (using RMSE & MAE) and generalization ability (using R 2) of the proposed forecasting framework. The multiple complementary measures used in the study ensure that the benefits of HMPCS are not restricted to one performance measure, but rather comprise strong, generalization, and practically essential improvements in performance across several dimensions of evaluation. Furthermore, sensitivity analysis provides additional insight into the robustness of optimization-based models by evaluating the influence of parameter variations on solution quality and forecasting stability. Such analyses have been successfully applied in mathematical optimization and engineering applications to support reliable decision-making. 34 Figure 2 the forecasting (RMSE and MAE) performance of baseline models (LSTM, Light GBM) and combined with HMPCS. HMPCS-enabled optimization resulted in the most pronounced increases in accuracy for both LSTM and Light GBM, as shown in the results section. Especially for LSTM+HMPCS, there is a significant reduction in RMSE and MAE compared to the version without optimization, while Light GBM+HMPCS achieves a 23% reduction in RMSE. The results highlight the effectiveness of HMPCS in hyper-parameter tuning, convergence acceleration, and generalization of machine learning models for renewable energy prediction. The figure shows that adding HMPCS to both LSTM and LightGBM reduces RMSE and MAE, meaning the optimization improves forecasting accuracy in all cases. 3.14 Limits of the Studying Though indicating some good results this paper has several limitations. At First, the experiments were performed on just a few sets of data, little different in any actual photovoltaic routine could possibly not be simulated accurately enough. Second, when compared to other solutions with simpler computations, the computational cost of our HMPCS framework is higher. Thirdly, this performance was tested under controlled experimental conditions and additional testing in the field is necessary to deal with very large installations in industrial sites. 4. Conclusion In this research work, we established a Hybrid Multi-Population Cuckoo Search By Means of Machine Learning Framework (HMPCS-ML), running from a fresh start to sustainably improve the forwardness performance of photovoltaic power generation. The results show that when the multi-population optimization mechanism is mated with LSTM and LightGBM models, both check RMSE (average variance-mean square deviation) increase accompanied by a drop in R 2 in comparison to traditional methods such as Grid Search and PSO. More importantly, these changes are consistent across various aspects of experimentation and can be statistically verified using techniques within statistical analyses. repeated tests, Wilcoxon methods, the improvements we saw are not to be regarded as mere happenstance. In addition, when the multi-population mechanism was removed from the optimization process and we carried out an exclusion test study, it also indicated that having populations had an enhancing effect upon balance between exploration and exploitation as well as upon deterring premature convergence. However, it must be pointed out that this study so far has taken place solely under controlled experimental conditions and with a combination of real and artificially reconstructed data streams; while the results suggest a big difference in performance, the move to practical deployment across large real grid systems will require more rigorous scrutiny. Furthermore, comparing its computational complexity to that of traditional optimization methods, the HMPCS framework with its multiple subsystems takes greater energy demand to calculate. But the advantage is only made up for by improved convergence behavior. The next stage will extend the modified approach out to bigger datasets; integrate it with real-time forecasting pipeline tests; and explore how well systems could perform under varying environmental conditions in parallel to this one-off study. Ethics and Consent Ethical approval and consent were not required for this study because it did not involve human participants, animals, or sensitive personal data. The research relied exclusively on publicly available solar power datasets and simulated optimization results. I have reviewed the revised version of the manuscript and the authors’ responses to the previous reviewer comments. The authors have addressed the major concerns by improving the clarity of the methodology, adding explanations regarding the HMPCS framework, expanding the discussion of comparative studies, and providing additional details on experimental evaluation and limitations. The revised manuscript is substantially improved in terms of organization and presentation. The authors have also clarified the distinction between standard equations and their proposed methodological contributions, which improves the transparency of the work. I have no further major comments at this stage. Minor editorial improvements, such as careful proofreading for grammatical errors, typographical issues, and consistency of terminology throughout the manuscript, may still be considered before final indexing. Is the work clearly and accurately presented and does it cite the current literature? Partly If applicable, is the statistical analysis and its interpretation appropriate? Partly Are all the source data underlying the results available to ensure full reproducibility? Partly Is the study design appropriate and is the work technically sound? Partly Are the conclusions drawn adequately supported by the results? Partly Are sufficient details of methods and analysis provided to allow replication by others? Partly Reviewer Expertise: AI, Macine Learning, Solar I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard. The authors have made substantial revisions to the manuscript in response to the reviewers’ comments and have added considerable new content. Each reviewer comment has been addressed point by point, and the responses provided are generally clear and effective in resolving the concerns raised during the previous round of review. The revised manuscript further strengthens and improves the research content on the basis of the original version and provides a more comprehensive and coherent presentation of the work. In particular, the revised manuscript presents the research in a complete, well-organized, and logically structured manner, allowing readers to clearly understand how the study was conducted, what research gap the work aims to address, and what advances have been achieved compared with previous studies. The manuscript is well written and follows appropriate academic conventions. The literature review covers a broad range of relevant studies, while the figures and tables are clear and well integrated with the corresponding textual descriptions, effectively conveying the main ideas and finding of the study. The research objectives and novel aspects of the work are described in sufficient detail, and the research methodology and procedures are presented in a clear and step-by-step manner. The results obtained appear to be reliable and credible. Overall, the study provides useful contribute to both the engineering application and theoretical development of the research field. Based on the substantial revisions and improvements made by the authors, I think that the manuscript is now in a condition suitable for indexing. Is the work clearly and accurately presented and does it cite the current literature? Partly If applicable, is the statistical analysis and its interpretation appropriate? Partly Are all the source data underlying the results available to ensure full reproducibility? Yes Is the study design appropriate and is the work technically sound? Yes Are the conclusions drawn adequately supported by the results? Yes Are sufficient details of methods and analysis provided to allow replication by others? Yes Reviewer Expertise: Renewable Energy, Artificial Intelligence, Solar energy, energy systems modeling and optimization, energy storage I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard. The manuscript developed an advanced hybrid model by integrating the hybridization of multi-population cuckoo search (HMPCS) algorithm with machine learning (ML) methods to predict solar power. The superiority of the proposed model was demonstrated through comparisons of various statistical performance indicators. However, I believe that the following suggestions and questions should be addressed before the manuscript can be reconsidered for indexing. the manuscript repeatedly uses the terms “solar power forecasting”, “renewable energy prediction”, “PV forecasting” and “photovoltaic power generation forecasting”. Could the authors clarify the differences among these concepts in the context of the present study. Specifically, what does each term refer to, and what role does each play in the proposed research framework. For the sake of consistency and clarify, the authors are also encouraged to adopt unified terminology through out the manuscript where appropriate;It appears that table 2 is not adequately introduced or discussed in the main text. The authors are encouraged to provide a clear description of the purpose of this table and discuss its key contents and significance to improve the readability and logical flow of the manuscript;The manuscript presents a number of equations and computational procedures in the sections entitled “Data preprocessing equations”, “LSTM model equations”, “HMPCS optimization equations” and “evaluation metrics”. It is not clear whether these equations were newly derived by the authors or adopted from previous studies. If they are original derivations, the authors should provide a brief derivation or theoretical justification. If they are based on existing methods, the corresponding references should be properly cited to acknowledge the original sources;In the introduction, the authors are encouraged to include a brief outline of the manuscript by summarizing the organization of the remaining sections and the main work presented in each section. Such an overview would improve the readability of the manuscript and help readers better understand the overall structure of the study;The reported values of the performance metrics ( R 2 ) in several tables range from approximately 0.80 to 0.93. although these results indicate reasonably good predictive performance, they do not appear to be particularly high compared with those reported in some recent studies on photovoltaic (PV) power forecasting, where higher prediction accuracies have been achieved. The authors are encouraged to provide a more detailed comparison with representative studies from the literature and discuss the possible reasons for the differences in predictive performance. Such a discussion would help readers better understand the advantages and limitations of the proposed method. Please send the revised manuscript to reviewers to review again. the manuscript repeatedly uses the terms “solar power forecasting”, “renewable energy prediction”, “PV forecasting” and “photovoltaic power generation forecasting”. Could the authors clarify the differences among these concepts in the context of the present study. Specifically, what does each term refer to, and what role does each play in the proposed research framework. For the sake of consistency and clarify, the authors are also encouraged to adopt unified terminology through out the manuscript where appropriate; the manuscript repeatedly uses the terms “solar power forecasting”, “renewable energy prediction”, “PV forecasting” and “photovoltaic power generation forecasting”. Could the authors clarify the differences among these concepts in the context of the present study. Specifically, what does each term refer to, and what role does each play in the proposed research framework. For the sake of consistency and clarify, the authors are also encouraged to adopt unified terminology through out the manuscript where appropriate; It appears that table 2 is not adequately introduced or discussed in the main text. The authors are encouraged to provide a clear description of the purpose of this table and discuss its key contents and significance to improve the readability and logical flow of the manuscript; It appears that table 2 is not adequately introduced or discussed in the main text. The authors are encouraged to provide a clear description of the purpose of this table and discuss its key contents and significance to improve the readability and logical flow of the manuscript; The manuscript presents a number of equations and computational procedures in the sections entitled “Data preprocessing equations”, “LSTM model equations”, “HMPCS optimization equations” and “evaluation metrics”. It is not clear whether these equations were newly derived by the authors or adopted from previous studies. If they are original derivations, the authors should provide a brief derivation or theoretical justification. If they are based on existing methods, the corresponding references should be properly cited to acknowledge the original sources; The manuscript presents a number of equations and computational procedures in the sections entitled “Data preprocessing equations”, “LSTM model equations”, “HMPCS optimization equations” and “evaluation metrics”. It is not clear whether these equations were newly derived by the authors or adopted from previous studies. If they are original derivations, the authors should provide a brief derivation or theoretical justification. If they are based on existing methods, the corresponding references should be properly cited to acknowledge the original sources; In the introduction, the authors are encouraged to include a brief outline of the manuscript by summarizing the organization of the remaining sections and the main work presented in each section. Such an overview would improve the readability of the manuscript and help readers better understand the overall structure of the study; In the introduction, the authors are encouraged to include a brief outline of the manuscript by summarizing the organization of the remaining sections and the main work presented in each section. Such an overview would improve the readability of the manuscript and help readers better understand the overall structure of the study; The reported values of the performance metrics ( R 2 ) in several tables range from approximately 0.80 to 0.93. although these results indicate reasonably good predictive performance, they do not appear to be particularly high compared with those reported in some recent studies on photovoltaic (PV) power forecasting, where higher prediction accuracies have been achieved. The authors are encouraged to provide a more detailed comparison with representative studies from the literature and discuss the possible reasons for the differences in predictive performance. Such a discussion would help readers better understand the advantages and limitations of the proposed method. The reported values of the performance metrics ( R 2 ) in several tables range from approximately 0.80 to 0.93. although these results indicate reasonably good predictive performance, they do not appear to be particularly high compared with those reported in some recent studies on photovoltaic (PV) power forecasting, where higher prediction accuracies have been achieved. The authors are encouraged to provide a more detailed comparison with representative studies from the literature and discuss the possible reasons for the differences in predictive performance. Such a discussion would help readers better understand the advantages and limitations of the proposed method. Is the work clearly and accurately presented and does it cite the current literature? Partly If applicable, is the statistical analysis and its interpretation appropriate? Partly Are all the source data underlying the results available to ensure full reproducibility? Yes Is the study design appropriate and is the work technically sound? Yes Are the conclusions drawn adequately supported by the results? Yes Are sufficient details of methods and analysis provided to allow replication by others? Yes Reviewer Expertise: Renewable Energy, Artificial Intelligence, Solar energy, energy systems modeling and optimization, energy storage I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above. We sincerely thank the reviewer for the careful evaluation of our manuscript and for the constructive comments. We have carefully revised the manuscript according to all suggestions. Specifically, we have: Unified the terminology throughout the manuscript and added a terminology clarification section.Expanded the discussion of Table 2.Clarified the originality of the equations and added the appropriate references.Added an "Organization of the Paper" paragraph at the end of the Introduction.Included a comparison with recent photovoltaic forecasting studies and expanded the discussion of the obtained R² values. All comments have been fully addressed in Version 3 of the manuscript. We sincerely appreciate the reviewer's valuable suggestions, which have significantly improved the quality and clarity of our work. Unified the terminology throughout the manuscript and added a terminology clarification section. Unified the terminology throughout the manuscript and added a terminology clarification section. Expanded the discussion of Table 2. Expanded the discussion of Table 2. Clarified the originality of the equations and added the appropriate references. Clarified the originality of the equations and added the appropriate references. Added an "Organization of the Paper" paragraph at the end of the Introduction. Added an "Organization of the Paper" paragraph at the end of the Introduction. Included a comparison with recent photovoltaic forecasting studies and expanded the discussion of the obtained R² values. Included a comparison with recent photovoltaic forecasting studies and expanded the discussion of the obtained R² values. This study proposes a Hybrid Multi-Population Cuckoo Search (HMPCS) framework integrated with machine learning models (LSTM and LightGBM) for improving photovoltaic (PV) power forecasting accuracy. The authors compare HMPCS-optimized models against Grid Search and Particle Swarm Optimization (PSO). Results show that LSTM+HMPCS achieves the lowest RMSE (0.139) and highest R² (0.93), while LightGBM+HMPCS reduces RMSE by 23% relative to unoptimized models (Table 3, p. 9) 1. Clarity and Use of Literature — Partly Strengths The introduction references recent DL/Transformer-based forecasting studies (2023–2024), indicating awareness of current literature.Metaheuristic background is supported with multiple citations.A structured literature review is presented (Section 2). Concerns Several grammatical issues and unclear technical phrasing reduce clarity.Some citations appear misaligned or weakly integrated into the argument.The literature gap is stated but not rigorously justified through systematic comparison.Minor typographical inconsistencies (e.g., duplicated or unclear wording in Sections 1–3). Required Revisions (Mandatory) Substantial language editing.Stronger synthesis of literature gap (explicit comparison table vs state-of-the-art Transformer-based works).Clearer explanation of novelty beyond “multi-population + ML.” The introduction references recent DL/Transformer-based forecasting studies (2023–2024), indicating awareness of current literature. The introduction references recent DL/Transformer-based forecasting studies (2023–2024), indicating awareness of current literature. Metaheuristic background is supported with multiple citations. Metaheuristic background is supported with multiple citations. A structured literature review is presented (Section 2). A structured literature review is presented (Section 2). Several grammatical issues and unclear technical phrasing reduce clarity. Several grammatical issues and unclear technical phrasing reduce clarity. Some citations appear misaligned or weakly integrated into the argument. Some citations appear misaligned or weakly integrated into the argument. The literature gap is stated but not rigorously justified through systematic comparison. The literature gap is stated but not rigorously justified through systematic comparison. Minor typographical inconsistencies (e.g., duplicated or unclear wording in Sections 1–3). Minor typographical inconsistencies (e.g., duplicated or unclear wording in Sections 1–3). Substantial language editing. Substantial language editing. Stronger synthesis of literature gap (explicit comparison table vs state-of-the-art Transformer-based works). Stronger synthesis of literature gap (explicit comparison table vs state-of-the-art Transformer-based works). Clearer explanation of novelty beyond “multi-population + ML.” Clearer explanation of novelty beyond “multi-population + ML.” 2. Study Design and Technical Soundness — Partly Strengths Logical workflow: preprocessing → model building → optimization → evaluation.Explicit HMPCS parameter settings provided (Section 3.8).Comparative evaluation against Grid Search and PSO. Major Concerns Data Inconsistency Section 3.5 states meteorological data were sourced from NASA POWER and PV Watts.However, the Data Availability section states the dataset is synthetic and simulated (p. 11) ae0a0900-6882-443d-8cb0-c0bde5d… .This contradiction must be clarified. Lack of Cross-Validation Description The manuscript mentions cross-validation in the fitness function but does not specify: Number of foldsTemporal cross-validation strategy (rolling origin?)Random seed control No Statistical Significance Testing Performance improvements are reported descriptively.No hypothesis testing, confidence intervals, or repeated runs analysis. Required Revisions (Mandatory) Clarify whether data are real, synthetic, or hybrid.Specify validation protocol clearly.Provide statistical robustness analysis (e.g., multiple runs, standard deviations). Logical workflow: preprocessing → model building → optimization → evaluation. Logical workflow: preprocessing → model building → optimization → evaluation. Explicit HMPCS parameter settings provided (Section 3.8). Explicit HMPCS parameter settings provided (Section 3.8). Comparative evaluation against Grid Search and PSO. Comparative evaluation against Grid Search and PSO. Data Inconsistency Section 3.5 states meteorological data were sourced from NASA POWER and PV Watts.However, the Data Availability section states the dataset is synthetic and simulated (p. 11) ae0a0900-6882-443d-8cb0-c0bde5d… .This contradiction must be clarified. Data Inconsistency Section 3.5 states meteorological data were sourced from NASA POWER and PV Watts.However, the Data Availability section states the dataset is synthetic and simulated (p. 11) ae0a0900-6882-443d-8cb0-c0bde5d… .This contradiction must be clarified. Section 3.5 states meteorological data were sourced from NASA POWER and PV Watts. Section 3.5 states meteorological data were sourced from NASA POWER and PV Watts. However, the Data Availability section states the dataset is synthetic and simulated (p. 11) ae0a0900-6882-443d-8cb0-c0bde5d… . However, the Data Availability section states the dataset is synthetic and simulated (p. 11) ae0a0900-6882-443d-8cb0-c0bde5d… . This contradiction must be clarified. This contradiction must be clarified. Lack of Cross-Validation Description The manuscript mentions cross-validation in the fitness function but does not specify: Number of foldsTemporal cross-validation strategy (rolling origin?)Random seed control Lack of Cross-Validation Description The manuscript mentions cross-validation in the fitness function but does not specify: Number of foldsTemporal cross-validation strategy (rolling origin?)Random seed control The manuscript mentions cross-validation in the fitness function but does not specify: Number of foldsTemporal cross-validation strategy (rolling origin?)Random seed control The manuscript mentions cross-validation in the fitness function but does not specify: Number of foldsTemporal cross-validation strategy (rolling origin?)Random seed control Number of folds Number of folds Temporal cross-validation strategy (rolling origin?) Temporal cross-validation strategy (rolling origin?) Random seed control Random seed control No Statistical Significance Testing Performance improvements are reported descriptively.No hypothesis testing, confidence intervals, or repeated runs analysis. No Statistical Significance Testing Performance improvements are reported descriptively.No hypothesis testing, confidence intervals, or repeated runs analysis. Performance improvements are reported descriptively. Performance improvements are reported descriptively. No hypothesis testing, confidence intervals, or repeated runs analysis. No hypothesis testing, confidence intervals, or repeated runs analysis. Clarify whether data are real, synthetic, or hybrid. Clarify whether data are real, synthetic, or hybrid. Specify validation protocol clearly. Specify validation protocol clearly. Provide statistical robustness analysis (e.g., multiple runs, standard deviations). Provide statistical robustness analysis (e.g., multiple runs, standard deviations). 3. Replicability — Partly Strengths Zenodo repository links provided.Hyperparameter settings listed.Evaluation metrics clearly defined. Missing Critical Details Dataset size (number of samples) not clearly specified.Training epochs and batch sizes not reported.Hardware specifications missing.No ablation study to isolate HMPCS contribution.Convergence curves are mentioned but insufficiently documented. Required Revisions (Mandatory) Provide full experimental configuration table.Add reproducibility checklist.Include ablation analysis (e.g., single-population vs multi-population). Zenodo repository links provided. Zenodo repository links provided. Hyperparameter settings listed. Hyperparameter settings listed. Evaluation metrics clearly defined. Evaluation metrics clearly defined. Dataset size (number of samples) not clearly specified. Dataset size (number of samples) not clearly specified. Training epochs and batch sizes not reported. Training epochs and batch sizes not reported. Hardware specifications missing. Hardware specifications missing. No ablation study to isolate HMPCS contribution. No ablation study to isolate HMPCS contribution. Convergence curves are mentioned but insufficiently documented. Convergence curves are mentioned but insufficiently documented. Provide full experimental configuration table. Provide full experimental configuration table. Add reproducibility checklist. Add reproducibility checklist. Include ablation analysis (e.g., single-population vs multi-population). Include ablation analysis (e.g., single-population vs multi-population). 4. Statistical Analysis — Partly Strengths Standard regression metrics (RMSE, MAE, MAPE, R²) used appropriately.Comparative table (Table 3) is clear. Weaknesses No uncertainty quantification.No repeated trials reported.No statistical test comparing optimizers.Percentage improvement claims lack statistical backing. Required Revisions (Mandatory) Perform ≥10 independent runs per optimizer.Report mean ± standard deviation.Conduct statistical significance test (e.g., Wilcoxon signed-rank test). Standard regression metrics (RMSE, MAE, MAPE, R²) used appropriately. Standard regression metrics (RMSE, MAE, MAPE, R²) used appropriately. Comparative table (Table 3) is clear. Comparative table (Table 3) is clear. No uncertainty quantification. No uncertainty quantification. No repeated trials reported. No repeated trials reported. No statistical test comparing optimizers. No statistical test comparing optimizers. Percentage improvement claims lack statistical backing. Percentage improvement claims lack statistical backing. Perform ≥10 independent runs per optimizer. Perform ≥10 independent runs per optimizer. Report mean ± standard deviation. Report mean ± standard deviation. Conduct statistical significance test (e.g., Wilcoxon signed-rank test). Conduct statistical significance test (e.g., Wilcoxon signed-rank test). 5. Data Availability — Partly Zenodo links are provided (p. 11) ae0a0900-6882-443d-8cb0-c0bde5d… , which is positive. However: Data described as “synthetic and simulated.”Yet earlier sections imply NASA/PV Watts sourcing.Reproducibility depends on clarity regarding data generation process. Required Revision Provide full data generation protocol.Clarify real vs synthetic usage.Include dataset dimensions and time span. Data described as “synthetic and simulated.” Data described as “synthetic and simulated.” Yet earlier sections imply NASA/PV Watts sourcing. Yet earlier sections imply NASA/PV Watts sourcing. Reproducibility depends on clarity regarding data generation process. Reproducibility depends on clarity regarding data generation process. Provide full data generation protocol. Provide full data generation protocol. Clarify real vs synthetic usage. Clarify real vs synthetic usage. Include dataset dimensions and time span. Include dataset dimensions and time span. 6. Conclusions — Partly The conclusions align directionally with reported results. However: Claims regarding robustness and real-world grid deployment are speculative.No computational time comparison table provided.Policy implications are overstated relative to experimental scope. Required Revisions Temper policy and deployment claims.Add computational efficiency comparison.Explicitly state study limitations. Claims regarding robustness and real-world grid deployment are speculative. Claims regarding robustness and real-world grid deployment are speculative. No computational time comparison table provided. No computational time comparison table provided. Policy implications are overstated relative to experimental scope. Policy implications are overstated relative to experimental scope. Temper policy and deployment claims. Temper policy and deployment claims. Add computational efficiency comparison. Add computational efficiency comparison. Explicitly state study limitations. Explicitly state study limitations. Minor Issues Typographical inconsistencies.Occasional unclear phrasing in equations formatting.Some redundant explanation of evaluation metrics.Figures lack detailed captions and axis labeling explanation. Typographical inconsistencies. Typographical inconsistencies. Occasional unclear phrasing in equations formatting. Occasional unclear phrasing in equations formatting. Some redundant explanation of evaluation metrics. Some redundant explanation of evaluation metrics. Figures lack detailed captions and axis labeling explanation. Figures lack detailed captions and axis labeling explanation. Overall Recommendation Major Revision Required The study has promising technical contribution, but before being scientifically sound, it must: Resolve data inconsistency (real vs synthetic).Strengthen reproducibility and statistical robustness.Provide proper validation protocol.Include statistical testing.Clarify novelty relative to recent Transformer-based PV forecasting literature.Improve language and technical precision. Resolve data inconsistency (real vs synthetic). Resolve data inconsistency (real vs synthetic). Strengthen reproducibility and statistical robustness. Strengthen reproducibility and statistical robustness. Provide proper validation protocol. Provide proper validation protocol. Include statistical testing. Include statistical testing. Clarify novelty relative to recent Transformer-based PV forecasting literature. Clarify novelty relative to recent Transformer-based PV forecasting literature. Improve language and technical precision. Improve language and technical precision. Is the work clearly and accurately presented and does it cite the current literature? Partly If applicable, is the statistical analysis and its interpretation appropriate? Partly Are all the source data underlying the results available to ensure full reproducibility? Partly Is the study design appropriate and is the work technically sound? Partly Are the conclusions drawn adequately supported by the results? Partly Are sufficient details of methods and analysis provided to allow replication by others? Partly Reviewer Expertise: EnergySolar EnergyArtificial IntelligenceMPPT I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above. We sincerely thank the reviewer for the valuable and constructive feedback. All comments have been carefully addressed in the revised version of the manuscript. In particular, we have clarified the data sources by distinguishing between real-world datasets (NASA POWER and PVWatts) and synthetic data used for robustness analysis. We have also strengthened the methodological section by clearly defining the validation protocol, including time-series cross-validation and repeated experimental runs. Furthermore, statistical robustness has been improved by reporting mean ± standard deviation over multiple runs and by applying the Wilcoxon signed-rank test to confirm the significance of the observed improvements. Additional enhancements include a detailed experimental setup, an ablation study to isolate the contribution of the HMPCS framework, and improvements in language, clarity, and presentation. The conclusions have also been revised to better align with the experimental scope, including a clear discussion of limitations and computational efficiency. We believe that these revisions have significantly improved the quality, clarity, and reproducibility of the manuscript.