Prediction Error Research Articles

Smart Grids: Integrating AI for Efficient Renewable Energy Utilization

The urgent global shift from fossil fuels to renewable energy sources necessitates innovative solutions to address energy system management challenges. Smart grids, equipped with sophisticated infrastructures, play a crucial role in this transition. This study integrates Artificial Intelligence (AI) into smart grids to enhance their efficiency and reliability, directly supporting the United Nations Sustainable Development Goals (SDGs), particularly SDG 7 (Affordable and Clean Energy), SDG 9 (Industry, Innovation, and Infrastructure), and SDG 11 (Sustainable Cities and Communities). Employing a mixed-methods approach, the research utilizes historical and real-time data, applying machine learning algorithms such as Linear Regression, Support Vector Regression (SVR), Recurrent Neural Networks (RNN), and Long ShortxTerm Memory (LSTM) for predictive accuracy in energy management. Optimization techniques like Genetic Algorithms and Particle Swarm Optimization (PSO) are also implemented for resource scheduling and grid balancing. The results demonstrate significant improvements, with an 11.76% increase in energy efficiency and grid stability, a 66.67% reduction in prediction errors, and a 20% decrease in operational costs compared to conventional systems. These enhancements highlight the transformative potential of AI in smart grids, promoting more efficient and sustainable energy utilization. The study concludes that AI-driven smart grids are pivotal in achieving the SDGs by providing scalable and efficient solutions for renewable energy integration, thereby fostering sustainable development and reducing environmental impacts.

Thermodynamic calculation-assisted design of 500 MPa high performance steel by machine learning

In order to rationally design Q500 low-alloy wind power steel with stable microstructure and properties within a large cooling rate range, this study proposed a design system that utilizes thermodynamic database, machine learning (ML), and the multi-objective genetic optimization algorithm (MOGA). Thermodynamic calculation has been used to obtain a large amount of data on phase fractions and mechanical properties for various chemical composition at different cooling rates. By comparing the predictive capabilities of different ML models, the deep neural network (DNN) and MOGA were combined to design Q500 wind power steel. Subsequently, three groups of designed chemical compositions were appropriately selected for verification. The results indicate that the yield strength of the experimental steel was over 500 MPa, and the maximum error in microstructure and mechanical property prediction was 5.0%. The prediction results demonstrated the rationality of the composition design framework. Meanwhile, the method of combining the material database and ML proposed in this study can also be applied to the design of other low-alloy steels.

High-speed rail passenger flow prediction based on crossformer and quantile regression: A deep learning approach assisted by internet search indices

Understanding potential fluctuations in passenger volume is crucial for the operation and management of high-speed rail, especially during peak times. The uncertainty caused by multiple factors is the main obstacle to accurate prediction. To quantify and mitigate the impact of these uncertainties, Internet search indices are utilized as insightful resources to grasp dynamic trends in passenger flow. Leveraging minimum redundancy maximum relevance, we identify the top search index features based on their predictive contribution to high-speed rail passenger flow. A two-level decomposition strategy is then established based on variational modal decomposition to extract significant influencing factors hidden in the Internet index and capture the dynamic uncertainty of passenger flow. By integrating Crossformer with quantile regression, we construct the upper and lower bounds of the prediction interval. Furthermore, the obtained upper and lower bounds are corrected by the error of point prediction, which allows for dynamic adjustment of the prediction intervals width based on fluctuations in uncertainty, thereby refining the precision of the prediction interval. Finally, the developed approaches effectiveness is validated through two real-world experiments, and the experimental results indicate that this method can more accurately capture variations in high-speed rail passenger flow, improving both management and service quality.

A spatio-temporal forecasting method of fracture distribution using dynamically exposed rock images in tunnel: Methodology and application

Forecasting the geometric characteristics of rock fractures in unexcavated tunnel sections is crucial for construction arrangements. This paper presents a method for predicting fracture distribution near the tunnel face from a spatiotemporal perspective. We Innovatively analogize the tunnel excavation mileage as the time series data. Using fractal geometry and geostatistics, we extract geometric features of peripheral rock fractures and establish a spatiotemporal dataset for dynamic prediction. The SCINet model, a spatial-aware recursive neural network, is employed for deterministic-stochastic dynamic prediction. Numerical simulations are conducted to validate the predictive accuracy of the model for various distributions, including trace length, dip angle, and density. Notably, the Mean Absolute Percentage Error (MAPE) for trace length prediction is remarkably low at 5.49 %, and the Kullback-Leibler (KL) divergence is as low as 4.81 %. The method is further applied to an underground oil storage cavern, China, revealing structural surface information. By incorporating continuously exposed fracture data, prediction accuracy is progressively improved, demonstrating the method's potential as a predictive tool for enhancing tunnel construction safety and efficiency.

The Effect of Lifting Straps on the Prediction of the Maximal Neuromuscular Capabilities and 1 Repetition Maximum During the Prone Bench Pull Exercise.

Miras-Moreno, S and García-Ramos, A. The effect of lifting straps on the prediction of the maximal neuromuscular capabilities and 1 repetition maximum during the prone bench pull exercise. J Strength Cond Res 38(11): e638-e644, 2024-This study examined the effects of using lifting straps in the prone bench pull exercise on (a) the magnitude of the load-velocity (L-V) relationship variables (load-axis intercept, velocity-axis intercept, and area under the L-V relationship line) and (b) the accuracy of the L-V relationship in predicting the 1 repetition maximum (1RM). In a counterbalanced sequence, 20 resistance-trained male subjects performed 2 identical experimental sessions, 1 with and 1 without lifting straps. During each session, subjects sequentially lifted 4 loads (14 kg, 85% 1RM, and 2 intermediate loads), followed by 1RM attempts. The L-V relationship was modelled using both multiple-point (4 loads) and 2-point (14 kg-85% 1RM) methods, whereas 3 types of minimal velocity thresholds (MVT) were used for 1RM prediction: (a) general MVT (averaged across the subjects velocity of the 1RM trial), (b) individual MVT (individual velocity of the 1RM trial), and (c) average optimal MVT (averaged across the subjects velocity that eliminates the differences between the actual and predicted 1RM). Irrespective of lifting straps use, which had no impact on the L-V relationship or the accuracy of 1RM prediction, we observed: (a) greater L-V relationship variables when obtained by the 2-point method compared with the multiple-point method (p < 0.001; effect size range = 0.353-0.639) and (b) the lowest absolute errors (<4.0 kg) in 1RM prediction for the 2-point method combined with the average optimal MVT (p < 0.001). These results reinforce the 2-point method applied in field conditions and the average optimal MVT for the routine testing of the L-V relationship.

Use of a rugged mid-infrared spectrometer for in situ process analysis of liquids

The widespread application of mid-infrared (MIR) spectroscopy for process monitoring is currently limited by the poor transmission of MIR light through fibre optics. In this work, the performance of a novel and robust MIR spectrometer has been evaluated for practical deployment in a pilot plant or production environment. The spectrometer utilises a Sagnac interferometer design containing no moving parts and is directly attached to an attenuated total reflectance probe, eliminating the need for fibre optics. The quantitative performance of the spectrometer for the in situ analysis of ternary solvent mixtures was assessed. The predictions obtained by partial least squares were accurate (root mean square error of prediction of < 1% w/w) and comparable to those of a benchmark Michelson-based spectrometer with a fibre-coupled probe, which is more amenable to process development in a laboratory or pilot plant. Calibration transfer between the two spectrometers was performed using the spectral space transformation method to mimic the scenario of the scale-up of a process from the laboratory to pilot scale or from a pilot plant to production scale, where the two different MIR instruments might be deployed. The ability to perform in situ reaction monitoring with the robust Sagnac-based spectrometer was then demonstrated. Spectra acquired during an esterification reaction were resolved using multivariate curve resolution, to produce concentration profiles of each component. These results demonstrate the suitability of this rugged spectrometer for quantitative in situ monitoring of liquid processes, opening up new opportunities for process monitoring in the MIR region.

Analysis of the ESCRS calculator's prediction accuracy.

To evaluate prediction accuracy of formulas included in the ESCRS online intraocular lens (IOL) calculator using standard keratometry (K) or total keratometry (TK). Hospital-based academic practice. Retrospective case-series. 523 patients with cataract (523 eyes) were included in this study. The outcome measures were trimmed means of the spherical equivalent prediction error (SEQ-PE, trueness), precision and absolute SEQ-PE (accuracy) of all 7 formulas available on the ESCRS online IOL calculator, as well as mean (Mean-All) and median (Median-All) of the predicted SEQ refraction of all formulas. Subgroup analyses evaluated the effect of axial length on formula accuracy. Trimmed-mean SEQ-PE range of all formulas varied from -0.075 to +0.071 diopters (D) for K-based and from -0.003 to +0.147 D for TK-based calculations, with TK-based being more hyperopic in all formulas ( P < .001). Precision ranged from 0.210 to 0.244 D for both K-based and TK-based calculations. Absolute SEQ-PE ranged from 0.211 to 0.239 D for K-based and from 0.218 to 0.255 D for TK-based calculations. All formulas, including Mean-All and Median-All, showed high accuracy, with 84% to 90% of eyes having SEQ-PEs within 0.50 D.Myopic trimmed-mean SEQ-PEs significantly different from zero were observed in long eyes for Pearl DGS (-0.110 D, P = .005), Hill RBF (-0.120 D, P < .001), and Hoffer QST (-0.143 D, P = .001) and in short eyes for EVO 2.0 (-0.252 D, P = .001), Kane (-0.264 D, P = .001), Hoffer QST (-0.302 D, P < .001), Mean-All (-0.122 D, P = .038), and Median-All (-0.125 D, P = .043). Prediction accuracy of all ESCRS IOL calculator formulas was high and globally comparable. TK-based calculations did not increase prediction accuracy and tended toward hyperopia. Observations indicating formula superiority in long and short eyes merit further evaluation.

Locally linearized physics-informed neural networks for Riemann problems of hyperbolic conservation laws

In this work, we demonstrate that physics-informed neural networks (PINNs) tend to propagate predicted shock wave information bidirectionally in time, which does not align with the actual evolution direction of solutions to hyperbolic conservation laws. This mismatch results in instability and hinders the reduction of the loss of governing equations, as well as the initial condition loss by the deep neural network. In order to tackle this problem, we simplify the complexity of the problem by constructing equivalent linear transport equations in the region of shock wave generation. The speeds of these linearized waves are governed by the Rankine–Hugoniot relations of conservation laws. This approach is termed the Locally Linearized PINNs method. Specifically, an appropriate shock wave detector is initially designed to identify domains where shock waves occur. Near shock waves, the original nonlinear equations are transformed into their linearized forms, thereby modifying the residual terms of the partial differential equations. Additionally, an equilibrium factor is introduced in fluid compression regions to reduce prediction errors and stabilize the training of deep neural networks. Numerical examples illustrate that Locally Linearized PINNs effectively address the challenge of predicting global solutions in PINNs and significantly improve shock-capturing performance for hyperbolic conservation laws.

A Multitask Network Robustness Analysis System Based on the Graph Isomorphism Network.

Despite various measures across different engineering and social systems, network robustness remains crucial for resisting random faults and malicious attacks. In this study, robustness refers to the ability of a network to maintain its functionality after a part of the network has failed. Existing methods assess network robustness using attack simulations, spectral measures, or deep neural networks (DNNs), which return a single metric as a result. Evaluating network robustness is technically challenging, while evaluating a single metric is practically insufficient. This article proposes a multitask analysis system based on the graph isomorphism network (GIN) model, abbreviated as GIN-MAS. First, a destruction-based robustness metric is formulated using the destruction threshold of the examined network. A multitask learning approach is taken to learn the network robustness metrics, including connectivity robustness, controllability robustness, destruction threshold, and the maximum number of connected components. Then, a five-layer GIN is constructed for evaluating the aforementioned four robustness metrics simultaneously. Finally, extensive experimental studies reveal that 1) GIN-MAS outperforms nine other methods, including three state-of-the-art convolutional neural network (CNN)-based robustness evaluators, with lower prediction errors for both known and unknown datasets from various directed and undirected, synthetic, and real-world networks; 2) the multitask learning scheme is not only capable of handling multiple tasks simultaneously but more importantly it enables the parameter and knowledge sharing across tasks, thus preventing overfitting and enhancing the performances; and 3) GIN-MAS performs multitasks significantly faster than other single-task evaluators. The excellent performance of GIN-MAS suggests that more powerful DNNs have great potentials for analyzing more complicated and comprehensive robustness evaluation tasks.

A shared multi-scale lightweight convolution generative network for few-shot multivariate time series forecasting

Time series forecasting is an important time series data mining technique. Among them, multivariate time series (MTS) forecasting has received extensive attention in many fields. However, many existing MTS forecasting models usually rely on a large amount of labeled data for model training, and data collection and labeling are difficult in real systems. The insufficient amount of data makes it difficult for the model to fully learn the intrinsic patterns and features of the data, which not only increases the prediction error, but also makes it hard to obtain satisfactory prediction results. To address this challenge, we propose a shared multi-scale lightweight convolution generative (SMLCG) network for few-shot multivariate time series forecasting by using samples generation strategy. The overall goal is to design a shared multi-scale feature generation prediction framework that generates data highly similar to the original sample and enriches the training sample to improve prediction accuracy. Specifically, the MTS is divided into different scales, and the multi-scale feature fusion module is utilized to capture and fuse the MTS information in different spatial dimensions to eliminate the heterogeneity among the data. Then, the key information in the multi-scale features is captured by a lightweight convolution generative network, and the feature weights are dynamically assigned to explore the change information. In addition, a spatio-temporal memory module is designed based on the parameter sharing strategy to capture the spatio-temporal dynamic relationship of sequences by learning the common knowledge in multi-scale features, thus improving the robustness and generalization ability. Through comprehensive experiments on four publicly available datasets and comparisons with other reported models, it is demonstrated that the SMLCG model can efficiently generate approximate samples in the few-shot case and provide excellent prediction results. The architecture of SMLCG serves as a valuable reference for practical solutions to address the few-shot problem in multivariate time series.

Visually Guided Sound Source Separation With Audio-Visual Predictive Coding.

The framework of visually guided sound source separation generally consists of three parts: visual feature extraction, multimodal feature fusion, and sound signal processing. An ongoing trend in this field has been to tailor involved visual feature extractor for informative visual guidance and separately devise module for feature fusion, while utilizing U-Net by default for sound analysis. However, such a divide-and-conquer paradigm is parameter-inefficient and, meanwhile, may obtain suboptimal performance as jointly optimizing and harmonizing various model components is challengeable. By contrast, this article presents a novel approach, dubbed audio-visual predictive coding (AVPC), to tackle this task in a parameter-efficient and more effective manner. The network of AVPC features a simple ResNet-based video analysis network for deriving semantic visual features, and a predictive coding (PC)-based sound separation network that can extract audio features, fuse multimodal information, and predict sound separation masks in the same architecture. By iteratively minimizing the prediction error between features, AVPC integrates audio and visual information recursively, leading to progressively improved performance. In addition, we develop a valid self-supervised learning strategy for AVPC via copredicting two audio-visual representations of the same sound source. Extensive evaluations demonstrate that AVPC outperforms several baselines in separating musical instrument sounds, while reducing the model size significantly. Code is available at: https://github.com/zjsong/Audio-Visual-Predictive-Coding.

Improved stochastic configuration network ensemble methods for time-series forecasting

Stochastic configuration network ensemble (SCNE) learning is an effective method for handling large datasets and reducing the number of input window size adjustments. However, this model faces challenges in effectively capturing the temporal features of time-series datasets and is constrained by certain hyperparameters, leading to longer training times. To address these issues, this paper proposes an improved stochastic configuration network ensemble (ISCNE) algorithm. Firstly, this algorithm retains the end data of the original dataset during the dataset partitioning process, effectively reducing prediction errors. Secondly, the direct ensemble algorithm eliminates redundant hyperparameters in SCNE, simplifying the model construction process. Lastly, the use of an improved normal distribution sampling strategy for input weights and biases, which provides values close to the mean and relatively small, reduces errors and further shortens the training time. The method was tested on 8 benchmark time-series forecasting datasets. Experimental results indicate that, it effectively improves the accuracy of time-series forecasting and model training efficiency while having a more streamlined structure.

Neural correlates of sensorimotor adaptation: Thalamic contributions to learning from sensory prediction error

Understanding the neural mechanism of sensorimotor adaptation is essential to reveal how the brain learns from errors, a process driven by sensory prediction errors. While the previous literature has focused on cortical and cerebellar changes, the involvement of the thalamus has received less attention. This functional magnetic resonance imaging study aims to explore the neural substrates of learning from sensory prediction errors with an additional focus on the thalamus. Thirty participants adapted their goal-directed reaches to visual feedback rotations introduced in a step-wise manner, while reporting their predicted visual consequences of their movements intermittently. We found that adaptation initially engaged the cerebellum and fronto-parietal cortical regions, which persisted as adaptation progressed. By the end of adaptation, additional regions within the fronto-parietal cortex and medial pulvinar of the thalamus were recruited. Another finding was the involvement of bilateral medial dorsal nuclei, which showed a positive correlation with the level of motor adaptation. Notably, the gradual shift in the predicted hand movement consequences was associated with activity in the cerebellum, motor cortex and thalamus (ventral lateral, medial dorsal, and medial pulvinar). Our study presents clear evidence for an involvement of the thalamus, both classical ‘motor’ and higher-order nuclei, in error-based motor learning.

Sample Efficient Deep Reinforcement Learning With Online State Abstraction and Causal Transformer Model Prediction.

Deep reinforcement learning (RL) typically requires a tremendous number of training samples, which are not practical in many applications. State abstraction and world models are two promising approaches for improving sample efficiency in deep RL. However, both state abstraction and world models may degrade the learning performance. In this article, we propose an abstracted model-based policy learning (AMPL) algorithm, which improves the sample efficiency of deep RL. In AMPL, a novel state abstraction method via multistep bisimulation is first developed to learn task-related latent state spaces. Hence, the original Markov decision processes (MDPs) are compressed into abstracted MDPs. Then, a causal transformer model predictor (CTMP) is designed to approximate the abstracted MDPs and generate long-horizon simulated trajectories with a smaller multistep prediction error. Policies are efficiently learned through these trajectories within the abstracted MDPs via a modified multistep soft actor-critic algorithm with a λ -target. Moreover, theoretical analysis shows that the AMPL algorithm can improve sample efficiency during the training process. On Atari games and the DeepMind Control (DMControl) suite, AMPL surpasses current state-of-the-art deep RL algorithms in terms of sample efficiency. Furthermore, DMControl tasks with moving noises are conducted, and the results demonstrate that AMPL is robust to task-irrelevant observational distractors and significantly outperforms the existing approaches.

Outlier Detection Using Gaussian Mixture Model Clustering to Optimize XGBoost for Credit Approval Prediction

Credit approval prediction is one of the critical challenges in the financial industry, where the accuracy and efficiency of credit decision-making can significantly affect business risk. This study proposes an outlier detection method using the Gaussian Mixture Model (GMM) combined with Extreme Gradient Boosting (XGBoost) to improve prediction accuracy. GMM is used to detect outliers with a probabilistic approach, allowing for finer-grained anomaly identification compared to distance- or density-based methods. Furthermore, the data cleaned through GMM is processed using XGBoost, a decision tree-based boosting algorithm that efficiently handles complex datasets. This study compares the performance of XGBoost with various outlier detection methods, such as LOF, CBLOF, DBSCAN, IF, and K-Means, as well as various other classification algorithms based on machine learning and deep learning. Experimental results show that the combination of GMM and XGBoost provides the best performance with an accuracy of 95.493%, a recall of 91.650%, and an AUC of 95.145%, outperforming other models in the context of credit approval prediction on an imbalanced dataset. The proposed method has been proven to reduce prediction errors and improve the model's reliability in detecting eligible credit applications.

Nonlinear Radiative Response to Patterned Global Warming due to Convection Aggregation and Nonlinear Tropical Dynamics

Abstract Climate responses to global warming exhibit large dependence on the spatial pattern of sea surface temperature (SST) changes. A Green’s function (GF) approach—predicting the global climate response to a complex SST pattern change as the linear superposition of the response to finite-area SST perturbations evaluated in isolation—has been used to systematically evaluate and understand the top of atmosphere (TOA) radiation response to patterned warming. However, in general, the linear GF approach fails for TOA radiation reconstruction under future global warming scenarios in coupled models. Here, we show that the linear superposition of global mean TOA radiation responses to large SST warming perturbations at multiple patches (the GF approach prediction) overestimates the actual response to simultaneous multipatch perturbation. This is because the linear superposition overestimates tropical large-scale convection aggregation strengthening upon localized heating that enhances longwave radiative cooling, which is further explained by the overestimation of circulation response and associated horizontal water vapor transport. The nonadditivity of TOA radiation response is caused by the nonadditivity of convection aggregation, ultimately rooted in nonlinear tropical dynamics. We also demonstrate that the prediction error of the GF approach grows with decreasing patch size (equivalently, increasing patch number). We conclude that using the GF approach to predict future climate change could overestimate longwave radiative cooling and underestimate (effective) climate sensitivity. Numerical experiments may be used to identify specific perturbation patterns where the errors are smaller than the signal. Our research also highlights that an increase in the degree of large-scale convection aggregation has a nonlinear and negative feedback for mean warming.

Dynamic Surface Intelligent Robust Control of Nonlinear Systems With Fixed-Time Sliding-Mode Observer.

The high tracking control precision and fast finite-time convergence for nonlinear systems is a significant challenge due to complex nonlinearity and unknown disturbances. To address this challenge, a dynamic surface intelligent robust control strategy with fixed-time sliding-mode observer (DSIRC-SMO) is proposed to improve the tracking control performance in a finite time. First, adaptive fuzzy neural network based on a predictor (P-AFNN) is designed to imitate the complex nonlinearity. In particular, the weight adaptive law is formulated by utilizing the prediction error information, which improves the accuracy of approximating the nonlinear system. Second, the fixed-time sliding-mode observer (SMO) is integrated into the dynamic surface control technique to deal with unknown disturbances and modeling errors in a fixed time. This integration allows for timely updates the dynamic surface using observation information, thereby enhancing the system's anti-interference capability. Then, the fixed-time convergence of SMO is proven. Third, the proposed DSIRC-SMO can be effectively implemented and the finite-time convergence of DSIRC-SMO is proven in detail based on the fixed-time convergence of SMO. Finally, numerical simulation and actual wastewater treatment processes simulation are conducted to validate the effectiveness of DSIRC-SMO.

Goal-driven long-term marine vessel trajectory prediction with a memory-enhanced network

Enhancing the precision of marine vessel trajectory prediction (VTP) is crucial for collision avoidance, intelligent navigation, and crisis alert in maritime safety. Most RNN-based methods typically face memory weakening issues during long-sequence propagation, leading to the discarding of some key features and significant predictive error accumulation over extended time intervals. Moreover, they struggle to forecast those complex trajectories involving abnormal maneuvers such as sudden acceleration or deceleration, sharp turns, or U-turns, resulting in poor generalization capabilities. To address these pivotal challenges, this paper proposes a novel Memory-Enhanced Network (MENet) for VTP, catering to intricate sailing intention modeling with long-term motion pattern perception. Specifically, we design an embeddable memory-enhanced block (MEB) that adaptively aggregates memory vectors across multiple temporal scales to assist in better prediction without disrupting the original backbone structure. Also, a goal-driven vessel trajectory decoder (GD-VTD) is developed to facilitate reliable model inferences by combining vessel type and destination variables as guidance information. Furthermore, we reconstruct the traditional loss function based on relative distance metrics, incorporating predicted headings into the optimization process to generate consistent trajectories that comply with realistic vessel dynamics. Ultimately, MENet could learn diverse sailing intentions by assembling the above parts to predict long-term marine vessel trajectories. Extensive experimental results on Automatic Identification System (AIS) datasets from three coastal regions in the US demonstrate that our model exhibits superior accuracy and robustness compared to other baselines. Specifically, on the Everglades Port (EP) dataset, our method reduces MAE, RMSE, and MAPE errors by 7.25%, 7.82%, and 7.62%, respectively, compared to the existing best results during this experiment. This is another piece of evidence for the effectiveness of goal-driven trajectory prediction in real-world maritime settings.

Automatic control algorithm for V2G frequency response mode of electric vehicle charging station

Abstract To solve the problem of excessive fluctuation of the V2G load of the electric vehicle charging pile, this paper proposes an automatic control algorithm for the V2G frequency response mode of the electric vehicle charging pile. The FPGA is used as the core processing unit to collect and process the V2G frequency data of the charging pile; the multiple linear regression model is used to predict the charging and discharging loads, and the automated two-layer control model is established; the sparrow search algorithm is used to optimize the weights and thresholds of the BP neural network to realize the control of the V2G frequency response mode. The experimental results show that the load prediction error of this method is small. After the automatic control, the load fluctuation of the charging pile is small, and the electric vehicle charging pile’s V2G frequency response control effect is better.

Intraocular lens tilt and decentration after primary and delayed implantation in phacovitrectomy for macula-off rhegmatogenous retinal detachment repair.

To evaluate the tilt and decentration of intraocular lenses (IOLs) in eyes that underwent phacovitrectomy for macula-off rhegmatogenous retinal detachment (RRD) repair. In this retrospective cohort study, 63 patients with macula-off RRD who underwent phacovitrectomy were enrolled and divided into a primary implantation group (group A) and a delayed implantation group (group B) according to the time of IOL implantation. Routine ocular examinations, including optometry, were performed, and the IOL tilt and decentration were measured. Postoperative refractive prediction error and IOL tilt and decentration were compared between the two groups, and the relevant factors of IOL positions were analyzed. The mean spherical refractive prediction error (MSE) was - 0.53 ± 0.39 D in group A and - 0.09 ± 0.55 in group B. The mean absolute spherical refractive prediction error (MASE) was 0.61 ± 0.23 D in group A and 0.50 ± 0.22 D in group B. The difference in MSE ( t = -3.623, P < 0.05) but not in MASE (t = 1.866, P = 0.067) between the two groups was significant. The mean IOL tilt and mean decentration were 8.54° ±3.65° and 0.55 ± 0.18 mm in group A and 10.62° ±3.29° and 0.66 ± 0.14 mm, respectively, in Group B. Differences in tilt ( t = -2.373, P < 0.05) and decentration ( t = -2.698, P < 0.05) between the two groups were both significant. In group B, the mean time interval between phacovitrectomy and IOL implantation (T V-I ) was 25.2 ± 6.9 weeks, and the absolute values of both IOL tilt and decentration were positively correlated with T V-I . In phacovitrectomy for macula-off RRD repair, primary IOL implantation leads to a postoperative myopic shift, while delayed IOL implantation may cause greater IOL tilt and decentration.