Optimal Model Parameters Research Articles

Surface topography and subsurface structure evolution in laser micro polishing of monocrystalline silicon

Ultra-smooth silicon surfaces, crucial in short-wave optics, can be marred by surface defects and subsurface damage incurred during machining, thereby affecting the imaging quality and transmission efficiency of optical components. Laser micro-polishing emerges as a versatile and efficient technique for ultra-precision polishing. This study aims to scrutinize the surface topography and subsurface structural evolution of monocrystalline silicon wafers during laser micro-polishing experimentally and to formulate an accessible evaluation model that incorporates multiple process parameters for optimal polishing results. As laser fluence escalates, the polishing process can be divided into six regimes: bulge, coalescence, smoothness, groove, ripple, and ablation. Each regime's crystal structure evolution is analyzed using Raman spectroscopy, demonstrating a close relationship with the surface expansion preceding the melting of monocrystalline silicon. Surface roughness diminishes from 6 nm to 0.7 nm, and an optimal crystalline quality devoid of an amorphous layer, residual stress, or significant subsurface damage is achieved. Amorphous silicon, 20 μm thick, is converted into monocrystalline silicon through recrystallization, and subsurface distortions and dislocations are rectified due to silicon atom redistribution. Moreover, a mapping relationship between multiple process parameters and surface polishing effects is established. A dimensionless fluence considering multiple parameters is derived, and a predictive model for process parameter optimization to secure optimal polishing results is proposed.

Rock physics-based anisotropy parameters estimation of Marcellus Shale using log data acquired in a vertical well

The elastic properties of shale rock exhibit anisotropy due to the presence of highly anisotropic clay minerals and their alignment. This anisotropy has a significant impact on geophysical and geomechanical responses. However, the elastic anisotropy of shale is often overlooked due to the difficulty in measuring sufficient parameters in the field, resulting in a gap between theoretical understanding and practical implementation. To address this issue, a rock physics-based approach was tested on log data obtained from a vertical well in the Marcellus Shale to estimate anisotropy parameters. The approach utilized a rock physics model based on the Sayers-den Boer method with model parameters optimized using measured stiffness parameters C33, C55, and C66. Anisotropy parameters δ and ε were then calculated using these optimized model parameters. The optimized model parameters and estimated anisotropy parameters are consistent with existing studies, indicating the correctness of the model. Additionally, the rock physics model was used to investigate the impact of gas saturation on the vertical VP/VS ratio and ε/γ ratio. It was found that gas saturation has a significant impact on these ratios. This carries important implications for acoustic log data interpretation and seismic characterization of shales.

Coupled phase diagram study and thermodynamic modeling of the Na2O-B2O3-ZnO system

The Phase diagram of ternary Na2O-B2O3-ZnO system was investigated for the first time using the classical equilibration/quenching method and differential thermal analysis followed by the phase analysis using the Electron Probe Micro-Analysis (EPMA) and X-Ray Diffraction (XRD). The isothermal phase diagrams of this ternary system at 900, 1000 and 1100 °C were determined, and the melting temperatures of two ternary compounds were investigated using Differential Thermal Analysis (DTA). The thermodynamic modeling of ternary Na2O-B2O3-ZnO system was carried out based on the present ternary phase diagram data and the optimized model parameters of binary systems. Thermodynamic models with optimized model parameters were used to predict the phase diagram of the ternary Na2O-B2O3-ZnO system.

Fractional-Order Equivalent-Circuit Model Identification of Commercial Lithium-Ion Batteries

The precise identification of electrical model parameters of Li-Ion batteries is essential for efficient usage and better prediction of the battery performance. In this work, the model identification performance of two metaheuristic optimization algorithms is compared. The algorithms in comparison are the Marine Predator Algorithm (MPA) and the Partial Reinforcement Optimizer (PRO) to find the optimal model parameter values. Three fractional-order (FO) electrical equivalent circuit models (ECMs) of Li-Ion batteries with different levels of complexity are used to fit the electrochemical impedance spectroscopy (EIS) data operating under different states of charge (SoC) and different operating temperatures. It is found that there is a tradeoff between ECM complexity, identification accuracy, and precision.

Dynamic Environment Responsive Online Meta-Learning with Fairness Awareness

The fairness-aware online learning framework has emerged as a potent tool within the context of continuous lifelong learning. In this scenario, the learner’s objective is to progressively acquire new tasks as they arrive over time, while also guaranteeing statistical parity among various protected sub-populations, such as race and gender when it comes to the newly introduced tasks. A significant limitation of current approaches lies in their heavy reliance on the i.i.d (independent and identically distributed) assumption concerning data, leading to a static regret analysis of the framework. Nevertheless, it’s crucial to note that achieving low static regret does not necessarily translate to strong performance in dynamic environments characterized by tasks sampled from diverse distributions. In this article, to tackle the fairness-aware online learning challenge in evolving settings, we introduce a unique regret measure, FairSAR, by incorporating long-term fairness constraints into a strongly adapted loss regret framework. Moreover, to determine an optimal model parameter at each time step, we introduce an innovative adaptive fairness-aware online meta-learning algorithm, referred to as FairSAOML. This algorithm possesses the ability to adjust to dynamic environments by effectively managing bias control and model accuracy. The problem is framed as a bi-level convex-concave optimization, considering both the model’s primal and dual parameters, which pertain to its accuracy and fairness attributes, respectively. Theoretical analysis yields sub-linear upper bounds for both loss regret and the cumulative violation of fairness constraints. Our experimental evaluation of various real-world datasets in dynamic environments demonstrates that our proposed FairSAOML algorithm consistently outperforms alternative approaches rooted in the most advanced prior online learning methods.

Parameter Optimization of Frazil Ice Evolution Model Based on NSGA-II Genetic Algorithm

This study is based on the research results of frazil ice evolution in recent years and proposes an improved frazil ice evolution mathematical model. Based on the NSGA-II genetic algorithm, seven key parameters were used as optimization design variables, the minimum average difference between the number of frazil ice, the mean and the standard deviation of particle diameter of the simulation results, and the observed data were used as the optimization objective, the Pareto optimal solution set was optimized, and the importance of each objective function was analyzed and discussed. The results show that compared to previous models, the improved model has better agreement between simulation results and experimental results. The optimal parameters obtained by the optimization model reduces the difference rate of water temperature process by 5.75%, the difference rate of quantity process by 39.13%, the difference rate of mean particle size process by 47.64%, and the difference rate of standard deviation process by 56.84% during the period of intense evolution corresponding to the initial parameter group. The results prove the validity of the optimization model of frazil ice evolution parameters.

Static Equivalence Method of Power Grid Based on Genetic Algorithm

When the electromagnetic transient simulation of a large-scale power grid is carried out, because the simulation is limited by the scale of simulation software, it is necessary to divide the grid into internal and external grids, and the external grid is modeled with equivalent simplification. Aiming at resolving the difficulties of the traditional grid equivalence method, such as a cumbersome calculation process and harsh calculation conditions, a static grid equivalence method based on the genetic algorithm is proposed in this paper. The method first constructs an external grid equivalence network, which includes coupling branches between boundary nodes, and each boundary node is connected to the external grid equivalence power supply through the branches. Then, the external grid equivalence model is used to establish an optimization model for solving the equivalence network parameters based on the information of the internal network, the difference of the power input from the external grid to the boundary nodes before and after equivalence is used as the objective function, and the genetic algorithm is used to realize the iterative optimization of the objective function to obtain a set of optimal post-equivalence network model parameters so that the state variables of the internal power grid before and after the equivalence are matched. The accuracy and effectiveness of the proposed method are verified through simulation with the CERPI36V7 node system.

Analysis of the Effectiveness of Using Two-Stage Neural Network Models for Early Detection of Forest Fires

The purpose of the research – analysis of the effectiveness of two-stage neural network models for solving the problem of detecting forest fires in images obtained from unmanned aerial vehicles.Methods. А training dataset was synthesized for training neural network models for the purpose of detection and semantic segmentation of forest fires in images. Тwo-stage neural network models (“Faster R-CNN”, “Mask RCNN” and “Retina-Net”) were used for training. Тhe neural network models were trained according to the same parameters set for all models in order to ensure consistency and a common basis for experiments. Optimization of model parameters during the training process was carried out to minimize the classification loss function. Тo synthesize the test sample, we used a video sequence covering the events of forest fires in the /rkutsk region, which was filmed by an unmanned aerial vehicle. Using a specially developed script in the Рython programming language, the process of dividing this video sequence into separate frames was carried out, which were used as a test data set when assessing the quality of classification of trained neural network models.Results. Based on the analysis of the obtained values of the quality criterion, as well as visual analysis on the test data set produced as part of testing neural network models, the effectiveness of the studied models for detecting forest fires in images was assessed. Тo assess the quality of binary classification of neural network models, the quality criterion “Accuracy” (classification accuracy) was used.Conclusion. Еxperimental studies on a test data set showed that the Retina-Net model demonstrates the lowest, but acceptable, performance compared to other studied neural network models. Тhe two-stage neural network models “Faster R-CNN” and “Mask R-CNN” demonstrate similar classification accuracy values (0.9492 and 0.9521, respectively), which allows us to recommend them for use in early detection systems for forest fires.

A Study of An Image Encryption Model Based on Tent-Ushiki Chaotic Fusion

Aiming at the shortcomings of current image encryption, such as simple structure, few parameters and small key space, we propose the Tent-Ushiki chaotic mapping image encryption method based on the improved firefly optimization algorithm. First, a chaotic mapping model based on the fusion of Tent and Ushiki is proposed. Second, for the lack of parameter optimization in chaotic models in general, we introduce the firefly optimization algorithm and optimize the algorithm for the shortcomings in terms of the adaptive step size, the adjustment factor and the inertia weights. Finally, the improved firefly algorithm is used for the fusion of the chaotic parameters of Tent and Ushiki. In the simulation experiments, this paper’s algorithm performs well in the statistical analysis and adjacent element correlation of the classical image test and significantly outperforms the comparative algorithms in terms of information entropy and anti-attack, which demonstrates that the algorithm is able to optimize the image encryption effect better.

Cabbage Transplantation State Recognition Model Based on Modified YOLOv5-GFD

To enhance the transplantation effectiveness of vegetables and promptly formulate subsequent work strategies, it is imperative to study the recognition approach for transplanted seedlings. In the natural and complex environment, factors like background and sunlight often hinder accurate target recognition. To overcome these challenges, this study explores a lightweight yet efficient algorithm for recognizing cabbage transplantation states in natural settings. Initially, FasterNet is integrated as the backbone network in the YOLOv5s model, aiming to expedite convergence speed and bolster feature extraction capabilities. Secondly, the introduction of the GAM attention mechanism enhances the algorithm’s focus on cabbage seedlings. EIoU loss is incorporated to improve both network convergence speed and localization precision. Lastly, the model incorporates deformable convolution DCNV3, which further optimizes model parameters and attains a superior balance between accuracy and speed. Upon testing the refined YOLOv5s target detection algorithm, improvements were evident. When compared to the original model, the mean average precision (mAP) rose by 3.5 percentage points, recall increased by 1.7 percentage points, and detection speed witnessed an impressive boost of 52 FPS. This enhanced algorithm not only reduces model complexity but also elevates network performance. The method is expected to streamline transplantation quality measurements, minimize time and labor inputs, and elevate field transplantation quality surveys’ automation levels.

Gaussian Process Latent Variable Model-Based Multi-Output Modeling of Incomplete Data

The rapid development of sensor technologies allows the acquisition of high dimensional sensing data. Multi-output modeling techniques have been developed to leverage the data for decision making. However, the data often contain segments of missing values, which cause great information loss and thus affect the modeling performance. This study explores the missing pattern and the correlation structure of missing segments and maximally exploits useful information in the data to improve multi-output modeling accuracy. Specifically, a new multi-output modeling method is developed based on Gaussian Process Latent Variable Model (GPLVM). A decision score is developed to seek an optimal modeling strategy and then a tailored Expectation-Maximization (EM) algorithm based on GPLVM is designed to estimate the missing segments while optimizing model parameters. The proposed method demonstrates superior performance in both a simulation study and a case study, which makes it a powerful tool to enable process automation. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —In real-life applications, missing values are constantly present in multi-output sensing data, which greatly affects data-driven decision-making. Modeling of such data becomes more difficult when consecutive observations within or across different outputs are missing. Existing methods often discard the missing values and extract information only from the available observations. However, the pattern of missing values may contain important messages that can potentially boost the modeling performance. This research develops a new framework based on GPLVM to model multi-output data with segmented missing patterns. A tailored EM algorithm is developed to iteratively impute the missing values and optimize model parameters. In addition, a decision score that quantifies both the missing pattern and correlation is designed to determine an optimal modeling strategy. The proposed method can benefit many applications across different industries that require modeling of multi-output incomplete data, especially when the data have many segments of missing observations.

Estimation of Sediment Transport Parameters From Measured Suspended Concentration Time Series Under Waves and Currents With a New Conceptual Model

AbstractIn‐situ observations of hydrodynamics and suspended sediment concentrations (SSCs) were conducted on an abandoned lobe in the northern part of the modern Yellow River Delta, China. The SSC record at the site is found to be the superposition of a general trend (fast increase and slow decrease cycle) caused by storm waves (SubSSC1) and relatively smaller fluctuations caused by tidal currents (SubSSC2). Physically, this indicates that storm waves eroded the bottom sediments while tidal currents then re‐suspended and advected the suspended sediments in the study area. To further obtain the suspended sediment transport parameters, first, SubSSC1 is modeled with significant wave height which incorporates a “memory curve” to consider the remaining impacts of historical waves. It is detected that waves in the past 75 hr still influence the present SSC which is reasonable because 75 hr is roughly the typical duration of a normal storm. Second, SubSSC2 is modeled with tidal excursion and trigonometric functions with measured periodicities. Finally, some sediment transport parameters, for example, the background SSC, the horizontal SSC gradient, the tidal constituents that advect it, and their relative time lags are optimized from the best fits of the measured and modeled SSC time series. The proposed framework for model construction and parameter optimization can be extended to other sea areas for inferring sediment transport parameters from field SSC time series at a specific station.

Forecasting carbon peaking in China using data-driven rule-base model: An in-depth analysis across regional and economic scenarios

At the 2020 United Nations Climate Summit, China officially announced the goal to achieve carbon peaking by 2030. Exploring whether it is possible to reach the peak of carbon emissions earlier necessitates an urgent and imperative need for precise long-term forecasting of China's carbon emissions dynamics. However, the current carbon peaking predictions mostly depend on mechanical or mathematical models, which failed to consider the interdependence between carbon emissions and the time series-based patterns existed in carbon emission data. Therefore, this study presents a novel carbon peaking prediction method based on the data-driven rule-base model, which is implemented by the adaption of the extended belief rule base (EBRB) model for time series forecasting (TSF), and thus the proposed method is referred to as TSF-EBRB model. The TSF-EBRB model not only captures and measures the temporal correlations within the data throughout the processes of modeling and inference, but also consists of a novel parameter optimization model based on the temporal correlations. The study collected carbon emission data from 30 provinces in China for empirical analysis. It computed and predicted the carbon peaking trajectories of each province under three different scenarios from 2022 to 2030, validating the effectiveness and superiority of the TSF-EBRB model better than other existing carbon peaking prediction methods. The results indicated that, with policy interventions, the majority of provinces are projected to reach carbon peaking before 2030.

A Versatile Deposition Model for Natural and Processed Surfaces

This paper introduces a robust deposition model designed for exploring the growth dynamics of deposits on surfaces under practical conditions. The study addresses the challenge of characterizing the intricate morphology of deposits, exhibiting significant visual variations. A generative approach is deployed to create diverse natural and engineered surface textures, governed by probabilistic principles. The model’s formulation addresses key questions related to deposition initiation, nucleation point behaviour, spatial scaling, deposit growth rates, spread dynamics, and surface mobility. A versatile algorithm, relying on six parameters and employing nested loops and Gaussian sampling, is developed. The algorithm’s efficacy is examined through extensive simulations, involving variations in nucleation scaling densities, aggregate scaling scenarios, spread factors, and diffusion rates. Surface statistics are computed for simulated deposits and analyzed using Fast Fourier Transform (FFT). The resulting database enables quantitative comparisons of surfaces generated with different parameters, where the database-derived parallel coordinates offer guidance for selecting optimal model parameters to achieve desired surface morphologies. The proposed approach is validated against urea-derived deposits, exhibiting statistical consistency and agreement with experimental observations. Overall, the model’s adaptable framework holds promise for understanding and predicting deposit growth on surfaces in diverse practical scenarios.

Multi-Objective Evolutionary Hybrid Deep Learning for energy theft detection

Electricity theft has emerged as a notable concern for smart grids, as fraudulent users illicitly access electricity from utilities without a contractual agreement or manipulate meter readings to evade bill payments. Consequently, the importance of electricity theft detection (ETD) in preserving the cost-effectiveness of smart grids has increased. In recent years, the utilization of deep learning models in electricity theft detection has surged in popularity, owing to their capacity to capture the periodic patterns within time-series data on electricity consumption. Nonetheless, deep learning techniques for electricity theft detection encounter several challenges, such as data imbalance, higher false positive rates, and the determination of periodic cycle timings. In this paper, we present an innovative approach to tackle these challenges: the Multi-Objective Evolutionary Hybrid Deep Learning (MOE-HDL) architecture, specifically designed to identify electricity theft. In MOE-HDL techniques, first, a novel multi-objective evolutionary algorithm based on ranking scheme (RMOE) is proposed to optimize architectural model parameters and the detection window size, with precision and recall as conflicting objectives. Next, an optimized dynamic detection window is generated to effectively capture the periodic patterns in consumption data and convert the one-dimensional data into a grid structure. Finally, a hybrid deep learning model is designed with the help of optimized model parameters to identify abnormal patterns in electricity consumption data. We conducted a comprehensive set of experiments, comparing our approach with eight baseline methods for electricity theft detection. The experimental results demonstrate that the proposed MOE-HDL approach outperforms these alternatives in terms of performance.

Simultaneous Rapid Detection of Multiple Physicochemical Properties of Jet Fuel Using Near-Infrared Spectroscopy.

Jet fuel is the primary fuel used in the aviation industry, and its quality has a direct impact on the safety and operational efficiency of aircraft. The accurate quantitative detection and analysis of various physicochemical property indicators are important for improving and ensuring the quality of jet fuel in the domestic market. This study used near-infrared (NIR) spectroscopy to establish a suitable model for the simultaneous and rapid detection of multiple physicochemical properties in jet fuel. Using more than 40 different sources of jet fuel, a rapid detection model was established by optimizing the spectral processing methods. The measurement models were separately built using the partial least-squares (PLS) and orthogonal PLS algorithms, and the model parameters were optimized. The results show that after the Savitzky-Golay second derivative preprocessing, the PLS model built using the feature spectra selected by the uninformative variable elimination wavelength algorithm achieved the best measurement performance. Compared with the PLS model without preprocessing, the range of the resulting accuracy improvement was at least 15.01%. Under the optimal model parameters, the calibration set regression coefficient (Rc2) of the 11 jet fuel property index models ranged from 0.9102 to 0.9763, with the root-mean-square error of calibration values up to 0.8468 °C (for flash points). The regression coefficient (Rp2) of the validation set ranged from 0.8239 to 0.9557, with the root-mean-square error of prediction values up to 1.1354 °C (for flash points). The ratios of prediction to deviation (RPD) values were all in the range of 1.9-3.0, indicating high accuracy and reliability of the model. The rapid NIR analysis method established in this study enables the simultaneous and rapid detection of multiple physicochemical properties of jet fuel, thereby providing effective technical support for ensuring the quality of jet fuel in the market.

Sampling Points-Independent Identification of the Fractional Maxwell Model of Viscoelastic Materials Based on Stress Relaxation Experiment Data.

Considerable development has been observed in the area of applying fractional-order rheological models to describe the viscoelastic properties of miscellaneous materials in the last few decades together with the increasingly stronger adoption of fractional calculus. The fractional Maxwell model is the best-known non-integer-order rheological model. A weighted least-square approximation problem of the relaxation modulus by the fractional Maxwell model is considered when only the time measurements of the relaxation modulus corrupted by additive noises are accessible for identification. This study was dedicated to the determination of the model, optimal in the sense of the integral square weighted model quality index, which does not depend on the particular sampling points applied in the stress relaxation experiment. It is proved that even when the real description of the material relaxation modulus is entirely unknown, the optimal fractional Maxwell model parameters can be recovered from the relaxation modulus measurements recorded for sampling time points selected randomly according to respective randomization. The identified model is a strongly consistent estimate of the desired optimal model. The exponential convergence rate is demonstrated both by the stochastic convergence analysis and by the numerical studies. A simple scheme for the optimal model identification is given. Numerical studies are presented for the materials described by the short relaxation times of the unimodal Gauss-like relaxation spectrum and the long relaxation times of the Baumgaertel, Schausberger and Winter spectrum. These studies have shown that the appropriate randomization introduced in the selection of sampling points guarantees that the sequence of the optimal fractional Maxwell model parameters asymptotically converge to parameters independent of these sampling points. The robustness of the identified model to the measurement disturbances was demonstrated by analytical analysis and numerical studies.

Prognostic value of different discretization parameters in 18fluorodeoxyglucose positron emission tomography radiomics of oropharyngeal squamous cell carcinoma.

We aim to interrogate the role of positron emission tomography (PET) image discretization parameters on the prognostic value of radiomic features in patients with oropharyngeal cancer. A prospective clinical trial (NCT01908504) enrolled patients with oropharyngeal squamous cell carcinoma (; mixed HPV status) undergoing definitive radiotherapy and evaluated intra-treatment 18fluorodeoxyglucose PET as a potential imaging biomarker of early metabolic response. The primary tumor volume was manually segmented by a radiation oncologist on PET/CT images acquired two weeks into treatment (20Gy). From this, 54 radiomic texture features were extracted. Two image discretization techniques-fixed bin number (FBN) and fixed bin size (FBS)-were considered to evaluate systematic changes in the bin number ({32, 64, 128, 256} gray levels) and bin size ({0.10, 0.15, 0.22, 0.25} bin-widths). For each discretization-specific radiomic feature space, an LASSO-regularized logistic regression model was independently trained to predict residual and/or recurrent disease. The model training was based on Monte Carlo cross-validation with a 20% testing hold-out, 50 permutations, and minor-class up-sampling to account for imbalanced outcomes data. Performance differences among the discretization-specific models were quantified via receiver operating characteristic curve analysis. A final parameter-optimized logistic regression model was developed by incorporating different settings parameterizations into the same model. FBN outperformed FBS in predicting residual and/or recurrent disease. The four FBN models achieved AUC values of 0.63, 0.61, 0.65, and 0.62 for 32, 64, 128, and 256 gray levels, respectively. By contrast, the average AUC of the four FBS models was 0.53. The parameter-optimized model, comprising features joint entropy (FBN = 64) and information measure correlation 1 (FBN = 128), achieved an AUC of 0.70. Kaplan-Meier analyses identified these features to be associated with disease-free survival ( and , respectively; log-rank test). Our findings suggest that the prognostic value of individual radiomic features may depend on feature-specific discretization parameter settings.

RL-SeqISP: Reinforcement Learning-Based Sequential Optimization for Image Signal Processing

Hardware image signal processing (ISP), aiming at converting RAW inputs to RGB images, consists of a series of processing blocks, each with multiple parameters. Traditionally, ISP parameters are manually tuned in isolation by imaging experts according to application-specific quality and performance metrics, which is time-consuming and biased towards human perception due to complex interaction with the output image. Since the relationship between any single parameter’s variation and the output performance metric is a complex, non-linear function, optimizing such a large number of ISP parameters is challenging. To address this challenge, we propose a novel Sequential ISP parameter optimization model, called the RL-SeqISP model, which utilizes deep reinforcement learning to jointly optimize all ISP parameters for a variety of imaging applications. Concretely, inspired by the sequential tuning process of human experts, the proposed model can progressively enhance image quality by seamlessly integrating information from both the image feature space and the parameter space. Furthermore, a dynamic parameter optimization module is introduced to avoid ISP parameters getting stuck into local optima, which is able to more effectively guarantee the optimal parameters resulting from the sequential learning strategy. These merits of the RL-SeqISP model as well as its high efficiency are substantiated by comprehensive experiments on a wide range of downstream tasks, including two visual analysis tasks (instance segmentation and object detection), and image quality assessment (IQA), as compared with representative methods both quantitatively and qualitatively. In particular, even using only 10% of the training data, our model outperforms other SOTA methods by an average of 7% mAP on two visual analysis tasks.

Ultra-short term wind power prediction based on quadratic variational mode decomposition and multi-model fusion of deep learning

To address the challenges associated with the intricate selection of decomposition technical parameters and the adverse impact of high-frequency non-stationary components on prediction accuracy, a differential evolution (DE) and sparrow search algorithm (SSA) parameter optimization model is formulated. Through the optimization of variational mode decomposition (VMD) and bidirectional short and long-term memory (BiLSTM) neural networks, enhancements are made to the overall performance of the prediction method. Subsequently, DESSA–VMD is employed to quadratically decompose the high-frequency unstable component, mitigating the detrimental effects of randomness and volatility on the prediction accuracy of the primary component. Following this, the DESSA–BiLSTM neural network is applied to predict the high-frequency strong non-stationary component, whereas the autoregressive integrated moving average is utilized for predicting the low-frequency periodic component. Finally, a case analysis is conducted using actual data comprising 2000 sets of wind power in a province. When the test set consisted of 600 groups, the root mean square error index value of the proposed prediction model exhibited reductions of 1.95 MW, 2.86 MW, and 1.16 MW, respectively, compared with the BP, SVM, and LSTM methods. Additionally, the mean absolute percentage error index showed decreases of 2.26 %, 2.96 %, and 1.73 %, respectively.