Output Prediction Research Articles

Research on Multi-scale Photovoltaic Output Prediction Method Based on GAN-Informer

Abstract The challenge of predicting photovoltaic power (PV) output is addressed in this paper, with a particular focus on enhancing prediction accuracy. A novel multi-scale prediction method for PV output is introduced, based on the integration of Generative Adversarial Networks (GANs) with the Informer model. The study’s content revolves around the leveraging of GANs for feature extraction and the development of the GAN-Informer model to forecast PV power outputs in national grid systems. Empirical investigations were conducted to validate the superior performance of the model. The outcomes reveal that the GAN-Informer model excels in ultra-short-term predictions, achieving the highest accuracy at a 15-minute scale with a mean squared error predominantly below 0.05, and a 1-hour prediction accuracy of 0.9728. These findings significantly contribute to improving PV energy utilization efficiency and promoting sustainable development.

Storm Gust Prediction with the Integration of Machine Learning Algorithms and WRF Model Variables for the Northeast United States

Abstract Wind gusts are often associated with severe hazards and can cause structural and environmental damages, making gust prediction a crucial element of weather forecasting services. In this study, we explored the utilization of machine learning (ML) algorithms integrated with numerical weather prediction outputs from the Weather Research and Forecasting (WRF) Model, to align the estimation of wind gust potential with observed gusts. We have used two ML algorithms, namely, random forest (RF) and extreme gradient boosting (XGB), along with two statistical techniques: generalized linear model with identity link function (GLM-Identity) and generalized linear model with log link function (GLM-Log), to predict storm wind gusts for the northeast (NE) United States. We used 61 simulated extratropical and tropical storms that occurred between 2005 and 2020 to develop and validate the ML and statistical models. To assess the ML model performance, we compared our results with postprocessed gust potential from WRF. Our findings showed that ML models, especially XGB, performed significantly better than statistical models and Unified Post Processor for the WRF (WRF-UPP) Model and were able to better align predicted with observed gusts across all storms. The ML models faced challenges capturing the upper tail of the gust distribution, and the learning curves suggested that XGB was more effective than RF in generating better predictions with fewer storms.

Research on Predicting the Safety Factor of Plain Shotcrete Support in Laneways Based on BO-CatBoost Model.

In general, the design of a safe and rational laneway support scheme signifies a crucial prerequisite for ensuring the security and efficiency of mining exploitation in mines. Nevertheless, the conventional empirical support system for mining laneways faces challenges in assessing the rationality of support methods, which can compromise the safety and reliability of the laneways. To address this issue, the safety factor was incorporated into research on laneway support, and a safety evaluation method for laneway support in line with the safety factor was established. In light of the data from a specific iron mine laneway in central China, the CRITIC method was employed to preprocess the sample data. Going one step further, a Bayesian algorithm was utilized to optimize the hyperparameters of the CatBoost model, followed by proposing a prediction model based on the BO-CatBoost model for evaluating laneway safety factors of plain shotcrete support. Furthermore, the performance indexes, such as the root mean square error (RMSE), the mean absolute error (MAE), the correlation coefficient (R2), the variance accounts for (VAF), and the a-20 index, were determined to examine the predictive performance of each proposed model. In contrast to the other models, the BO-CatBoost model demonstrated the optimal predictive output item for safety factors with the lowest RMSE and MAE, the largest R2 and VAF, and an appropriate a-20 index value of 0.5688, 0.4074, 0.9553, 95.25%, and 0.9167 in the test set, respectively. Therefore, the BO-CatBoost model was proven to be the most appropriate machine learning method that can more accurately predict the safety factor, which will provide a novel approach for optimizing laneway support design and laneway safety evaluation.

An advanced kernel search optimization for dynamic economic emission dispatch with new energy sources

Integrating renewable energy sources, such as wind and photovoltaic power generation, into the power grid is crucial for sustainable power system development and mitigating pollutant emissions. However, these sources’ inherent uncertainty and randomness pose significant challenges to grid operations. Metaheuristic algorithms offer efficient solutions to optimization problems in this context and are widely employed in practice. Kernel Search Optimization (KSO) has emerged as a prominent metaheuristic algorithm due to its parameter-free nature and applicability to power dispatch problems. Nevertheless, KSO’s limited local search capabilities necessitate enhancements for improved performance. This paper introduces an enhanced variant of KSO, termed Upgraded Kernel Search Optimization (UKSO), which incorporates differential evolution techniques, including mutation, crossover, and selection mechanisms, to bolster KSO’s search capabilities and overall performance. The efficacy and feasibility of UKSO are evaluated through comprehensive testing using the CEC2017 benchmark. Comparative analysis demonstrates that UKSO outperforms other algorithms by achieving more optimal solutions. To address power dispatch challenges in the context of renewable energy sources, a Two-Time Energy Dispatch (TTED) model is proposed, leveraging a weighted summation approach to minimize total fuel costs and pollution emissions simultaneously. The application of the Prophet model, based on Bayesian fitting, facilitates accurate prediction of wind and photovoltaic power outputs. Through experimentation across small (8 and 14 units) and large (58 units) systems, the feasibility of UKSO and TTED is validated. In complex scenarios, UKSO demonstrates a 0.69% to 3.64% reduction in pollution emissions and a 0.99% to 1.23% decrease in economic costs compared to KSO under varying weight configurations. Furthermore, UKSO achieves a renewable energy generation utilization rate exceeding 90%, emphasizing its ability to iterate efficiently through complex problems and its efficacy in addressing renewable power dispatch challenges. These results highlight the practical significance of UKSO in optimizing power systems and provide valuable insights for future research and applications in this domain.

Combining the Strengths of the Explainable Boosting Machine and Metabolomics Approaches for Biomarker Discovery in Acute Myocardial Infarction.

Acute Myocardial Infarction (AMI), a common disease that can have serious consequences, occurs when myocardial blood flow stops due to occlusion of the coronary artery. Early and accurate prediction of AMI is critical for rapid prognosis and improved patient outcomes. Metabolomics, the study of small molecules within biological systems, is an effective tool used to discover biomarkers associated with many diseases. This study intended to construct a predictive model for AMI utilizing metabolomics data and an explainable machine learning approach called Explainable Boosting Machines (EBM). The EBM model was trained on a dataset of 102 prognostic metabolites gathered from 99 individuals, including 34 healthy controls and 65 AMI patients. After a comprehensive data preprocessing, 21 metabolites were determined as the candidate predictors to predict AMI. The EBM model displayed satisfactory performance in predicting AMI, with various classification performance metrics. The model's predictions were based on the combined effects of individual metabolites and their interactions. In this context, the results obtained in two different EBM modeling, including both only individual metabolite features and their interaction effects, were discussed. The most important predictors included creatinine, nicotinamide, and isocitrate. These metabolites are involved in different biological activities, such as energy metabolism, DNA repair, and cellular signaling. The results demonstrate the potential of the combination of metabolomics and the EBM model in constructing reliable and interpretable prediction outputs for AMI. The discussed metabolite biomarkers may assist in early diagnosis, risk assessment, and personalized treatment methods for AMI patients. This study successfully developed a pipeline incorporating extensive data preprocessing and the EBM model to identify potential metabolite biomarkers for predicting AMI. The EBM model, with its ability to incorporate interaction terms, demonstrated satisfactory classification performance and revealed significant metabolite interactions that could be valuable in assessing AMI risk. However, the results obtained from this study should be validated with studies to be carried out in larger and well-defined samples.

Lung Cancer Detection using Ensemble Techniques

This paper implements a system for enhancing the detection of lung cancer through an ensemble approach, which amalgamates the predictive outputs generated by three distinct convolutional neural networks (CNNs): ResNet50, EfficientNet, and InceptionNet. Leveraging the diverse architectural features and learning capabilities of these CNNs, the ensemble method aims to synergistically fuse their individual predictions to achieve heightened accuracy and robustness in identifying potential lung cancer manifestations.

Detailed effects of reservoir permeability distribution differences on enhanced geothermal systems performance

In enhanced geothermal systems, fractured reservoir permeability significantly affects geothermal exploitation efficiency. However, the detailed effects are not fully understood while most previous literature ignored the spatial differences of reservoir permeability because of the complexity and heterogeneity of fracture distribution. This study aims to reveal the quantitative relationship between the geothermal system’s heat extraction performance and the distributed permeability, through investigating heat extraction ratio and flow impedance by building a 3D thermal-hydraulic coupled model with discrete fracture network. The current study also compared the effects of twenty-nine kinds of fracture network distributions at various depths. It is found that higher fracture permeability around production well (800×10−12 m2) more significantly improves heat extraction rate by 30% rather than higher permeability in other areas in the first year. It is also found that fracture permeability changes on both sides of predominant flow regions cause a lower heat extraction rate. Lower fracture permeability at bottom and top layers (25×10−12 m2) increases the heat extraction rate by 1.17 MW in the 30th year. The proppant distribution affects the pressure distribution in fractured reservoirs. Permeability distribution variation has small effects on the heat extraction ratio. These results provided theoretical basis for fractured reservoir construction, optimal proppant pumping scheme and reservoir thermal output prediction.

Optimizing high-entropy alloys using deep neural networks

Exceptional mechanical properties of multi-principal element alloys have been typically achieved through manual trial-and-error approaches. The advances in artificial neural networks (ANN) have recently allowed for the development of predictive alloy design ANN models that take into account a large variety of prior experimental findings. However, when mechanical performance is considered, prior processing protocols (e.g. annealing, cold working) are almost impossible to track across different research groups and material classes. In this work, we propose to reverse the use of the yield stress point, setting it as an explicit input (instead of predictive output) parameter that is then used for the quantification of the material state, towards predicting material properties, such as the Ultimate Tensile Strength (UTS). This novel approach specifically addresses the challenge of predicting and improving ductility, a common issue in existing methods focused on yield stress, hardness, and phases. By combining available, validated databases of eleven elements and 734 total materials, we develop an ANN that receives as input the compositions out of 11 possible elements and the "observed" yield stress. Then, we utilize this model to perform predictions for 5-element concentrated alloys with exceptional UTS. We show that this ANN modeling strategy leads to exceptional confidence scores, and then, we use this ANN model for the design of HEAs, which we further discuss.

The effect of restrictive compared to liberal intravenous fluid volume on hypotension in adults undergoing major abdominal surgery

In a cardiac output (CO) sub-study of the Restrictive versus Liberal Fluid Therapy in Major Abdominal Surgery (RELIEF) trial, it was shown that restrictive fluid management was associated with lower cardiac index at the end of surgery. However, the association of the fluid protocol with intraoperative blood pressure was less clear. This paper primarily compares rates of hypotension between the two fluid regimens. The haemodynamic effects of these protocols may increase our understanding of perioperative fluid prescription. Using a data set of arterial pressure and cardiac output measurements, this observational cohort study primarily compares intraoperative hypotension rates defined by a mean arterial pressure < 65 mmHg between liberal and restrictive fluid protocols. Secondary analyses explore predictors of invasive mean arterial pressure and doppler-derived cardiac output, including fluid volume regimens and surgical duration. 105 patients had a combined total of 835 haemodynamic data capture events from the beginning to the end of the surgery. Here we report that a restrictive regimen is not associated with a greater proportion of participants who experience at least one episode of hypotension than the liberal regimen 64.1% vs. 61.5% (mean difference 2.6%, 95% CI − 15.9% to 21%, p = 0.78). Duration of surgery was associated with an increased risk of hypotension (OR 1.05, 1 to 1.1, p = 0.038). A fluid restriction protocol compared to liberal fluid administration is not associated with lower blood pressure.

SpliceAPP: an interactive web server to predict splicing errors arising from human mutations

BackgroundSplicing variants are a major class of pathogenic mutations, with their severity equivalent to nonsense mutations. However, redundant and degenerate splicing signals hinder functional assessments of sequence variations within introns, particularly at branch sites. We have established a massively parallel splicing assay to assess the impact on splicing of 11,191 disease-relevant variants. Based on the experimental results, we then applied regression-based methods to identify factors determining splicing decisions and their respective weights.ResultsOur statistical modeling is highly sensitive, accurately annotating the splicing defects of near-exon intronic variants, outperforming state-of-the-art predictive tools. We have incorporated the algorithm and branchpoint information into a web-based tool, SpliceAPP, to provide an interactive application. This user-friendly website allows users to upload any genetic variants with genome coordinates (e.g., chr15 74,687,208 A G), and the tool will output predictions for splicing error scores and evaluate the impact on nearby splice sites. Additionally, users can query branch site information within the region of interest.ConclusionsIn summary, SpliceAPP represents a pioneering approach to screening pathogenic intronic variants, contributing to the development of precision medicine. It also facilitates the annotation of splicing motifs. SpliceAPP is freely accessible using the link https://bc.imb.sinica.edu.tw/SpliceAPP. Source code can be downloaded at https://github.com/hsinnan75/SpliceAPP.

An improved prediction model for in-stream water wheel performance

As global efforts to decarbonise industries continue, there is increasing pressure to identify and proliferate sustainable methods of generating electricity, in-line with the United Nations’ Sustainable Development Goal 7. The in-stream water wheel is a pico-hydropower technology that generates electricity using only the kinetic energy in the fluid, rather than both potential and kinetic, like most forms of pico-hydropower. Current methods of estimating the power output of these turbines lack validation, robustness and proper justification for their assumptions. This study has developed a new method for predicting the power output of in-stream water wheels, based on fluid dynamics principles, and has been compared with a range of experimental results to confirm its robustness. The new model identifies novel characteristics of the power stroke of the turbine blade, including unsteady effects and negative torque production at certain blade angles. In addition, the model can be tuned to match experimental data, improving its accuracy in specific applications. The better prediction of power output aims to encourage the use of in-stream water wheels as part of the global decarbonisation strategy.

Global self-sustaining and local inheritance for source-free unsupervised domain adaptation

In this paper, we investigate a practical problem called source-free unsupervised domain adaptation, which adapts a source-trained model to the target domain with unlabeled target data. To address this problem, we propose a novel GlObal self-sustAining and Local inheritance (GOAL) method. GOAL contains three components. (1) A backbone follows a mean teacher scheme. The teacher model serves as a smoothing functionality, facilitating a more consistent convergence of the student model. This capability alleviates the student model’s sensitivity to minor input data variations and enhances the overall robustness of the model. Additionally, disparities in predictions between the student and teacher models can be leveraged to identify potential noise in the data. (2) A Global Consistency Self-Sustaining mechanism for learning a stable, discriminative, and diverse prediction space. On the one hand, we employ neighbor samples and mean-teacher schemes to enhance the discriminability and stability of model predictions. On the other hand, non-neighbor samples are leveraged to augment the diversity of model predictions. Furthermore, to mitigate the impact of potential negative neighbors, we derive a weighting factor by incorporating both neighbor entropy and the top-nd similarity of features. (3) A Local Topology Inheritance mechanism to improve the semantic structure of the feature space. We construct a semantic topology graph based on the output predictions of the teacher model and subsequently transmit the teacher topology to the feature space of the student utilizing a local topology inheritance loss. Combining these three components, GOAL can effectively solve the source-free unsupervised domain adaptation. To the best of our knowledge, GOAL is the first attempt to perform topology inheritance for global consistency domain adaptation. Comprehensive experiments illustrate the effectiveness and superiority of GOAL in addressing source-free unsupervised domain adaptation.

Lightweight strip steel defect detection algorithm based on improved YOLOv7

The precise identification of surface imperfections in steel strips is crucial for ensuring steel product quality. To address the challenges posed by the substantial model size and computational complexity in current algorithms for detecting surface defects in steel strips, this paper introduces SS-YOLO (YOLOv7 for Steel Strip), an enhanced lightweight YOLOv7 model. This method replaces the CBS module in the backbone network with a lightweight MobileNetv3 network, reducing the model size and accelerating the inference time. The D-SimSPPF module, which integrates depth separable convolution and a parameter-free attention mechanism, was specifically designed to replace the original SPPCSPC module within the YOLOv7 network, expanding the receptive field and reducing the number of network parameters. The parameter-free attention mechanism SimAM is incorporated into both the neck network and the prediction output section, enhancing the ability of the model to extract essential features of strip surface defects and improving detection accuracy. The experimental results on the NEU-DET dataset show that SS-YOLO achieves a 97% mAP50 accuracy, which is a 4.5% improvement over that of YOLOv7. Additionally, there was a 79.3% reduction in FLOPs(G) and a 20.7% decrease in params. Thus, SS-YOLO demonstrates an effective balance between detection accuracy and speed while maintaining a lightweight profile.

Data-driven atmospheric corrosion prediction model for alloys based on a two-stage machine learning approach

Alloy material corrosion is a comprehensive problem affected by multi-dimensional factors, including the time, environment, and material element. Its corrosion prediction has been facing the challenges of complex mechanism and small sample size. However, parametric methods are limited in terms of modeling accuracy and generalization performance for high-dimensional feature problems. Non-parametric machine learning algorithms, while having excellent modeling capabilities, need to be optimized in conjunction with the structure of the corrosion problem. Therefore, we develop a two-stage hybrid intelligent machine learning model for improving alloy corrosion prediction accuracy. The model takes the time, environment, and material information of the alloy corrosion process as inputs and the corrosion rate as output. The first stage of the model utilizes tree ensemble learning (TEL) models for initial prediction modeling and feature importance knowledge mining. In the second stage, the neural network (NN) is used to enhance the fusion of the multi-model prediction outputs, and it is optimized using feature importance modification (FIM). The proposed TEL-FIMNN model shows considerable performance improvement over the TEL and ANN models on real low-alloy steel corrosion dataset. Compared to other advanced multi-model fusion methods such as linear regression (LR), support vector regression (SVR), random forest (RF), and Bayesian model averaging (BMA), the proposed model also shows superior results. The TEL-FIMNN model provide a new structure from the perspectives of knowledge mining and fusion enhancement, which has good potential for multi-dimensional corrosion prediction.

An efficient multi-disease prediction model using advanced optimization aided weighted convolutional neural network with dilated gated recurrent unit

In the medical field, prediction accuracy over multi-diseases is significant and it is helpful for improving the patient’s health. Most of the conventional machine learning techniques concentrates only on detecting single diseases. Only a few systems are developed for predicting more than one disease. The classification of multi-label data is a challenging issue. Patients have symptoms of various diseases while analyzing the medical data and hence it is necessary to implement tools for the earlier identification of problems. The patterns in the health data have been effectively identified through deep learning-based health risk prediction models. Thus, an efficient prediction model for predicting various types of diseases is implemented in this work. Initially, the required data regarding various types of diseases will be gathered from Kaggle database. The garnered healthcare data are pre-processed for quality enhancement. The pre-processing procedures include data cleaning, data transformation, and outlier detection are performed at first. The outlier detection is done using the “Density-Based Spatial Clustering of Applications with Noise (DBSCAN)” approach. The pre-processed data is then given to the Weighted Convolutional Neural Network Feature with Dilated Gated Recurrent Unit (WCNNF-DGRU) model. Here, the pre-processed data is provided to the CNN structure for feature extraction, in which the weights are optimized by means of the Enhanced Kookaburra Optimization Algorithm (EKOA). Then the features from the weighted CNN layer are provided to the Dilated GRU structure to determine the final prediction output. Experimental verification is carried out on the implemented WCDG in predicting multiple diseases by comparing it with other conventional prediction models and optimization algorithms.

Exploring Trade-offs in Explainable AI

Machine Learning (ML) models are increasingly used in systems that involve physical human interaction or decision-making systems that impact human health and safety. Ensuring that these systems are safe and reliable is an important topic of current AI research. For many ML models it is unclear how a prediction (output) is arrived at from the provided features (input). Critical systems cannot blindly trust the predictions of such "black box" models, but instead need additional reassurance via insight into the model's reasoning. A range of methods exist within the field of Explainable AI (XAI) to make the reasoning of black box ML models more understandable and transparent. The explanations provided by XAI methods may be evaluated in a number of (competing) ways. In this paper, we investigate the trade-off between selected metrics for an XAI method called UnRAvEL, which is similar to the popular LIME approach. Our results show that by weighting the terms within the acquisition function used in UnRAvEL, different trade-offs can be achieved.

Unveiling EMG semantics: a prototype-learning approach to generalizable gesture classification

Objective. Upper limb loss can profoundly impact an individual’s quality of life, posing challenges to both physical capabilities and emotional well-being. To restore limb function by decoding electromyography (EMG) signals, in this paper, we present a novel deep prototype learning method for accurate and generalizable EMG-based gesture classification. Existing methods suffer from limitations in generalization across subjects due to the diverse nature of individual muscle responses, impeding seamless applicability in broader populations. Approach. By leveraging deep prototype learning, we introduce a method that goes beyond direct output prediction. Instead, it matches new EMG inputs to a set of learned prototypes and predicts the corresponding labels. Main results. This novel methodology significantly enhances the model’s classification performance and generalizability by discriminating subtle differences between gestures, making it more reliable and precise in real-world applications. Our experiments on four Ninapro datasets suggest that our deep prototype learning classifier outperforms state-of-the-art methods in terms of intra-subject and inter-subject classification accuracy in gesture prediction. Significance. The results from our experiments validate the effectiveness of the proposed method and pave the way for future advancements in the field of EMG gesture classification for upper limb prosthetics.

Insight generation for improving e-governance: towards crop yield prediction model

ABSTRACT Agriculture is one crucial industry that expects to profit from these improvements in maintaining financial stability and availability of food. Accurate agricultural output prediction is critical for efficient policy formation, allocation of resources, and preparedness for catastrophes. This work intends to propose a novel insight generation for improving e-Governance towards crop yield prediction model. In the first stage, the data is preprocessed by conducting improved data normalisation and data augmentation. The outlier handling, custom weighted Min-Max scaling and using a custom scaling range are the steps to be followed in the improved data normalisation. Then the augmented data is subjected to a second stage called feature extraction. From the augmented data, the raw feature, Unbiased Estimator based Correlation (UEC) feature, correlation feature, and statistical features are extracted. In the UEC-based feature, the unbiased estimator is deployed to estimate outliers and sparse data. These extracted features are fed to a hybrid DL-based prediction model, wherein it contains Bidirectional Long Short Term Memory (Bi-LSTM) and Shifted Beamish Activation-Deep Convolutional Neural Network (SBA-DCNN) models. The SBA activation function is employed in SBA-DCNN at each convolutional layer. By taking the mean of both models, this hybrid prediction model accurately predicts the yield.

Wholesale price forecasts of green grams using the neural network

PurposeAgriculture commodity price forecasts have long been important for a variety of market players. The study we conducted aims to address this difficulty by examining the weekly wholesale price index of green grams in the Chinese market. The index covers a ten-year period, from January 1, 2010, to January 3, 2020, and has significant economic implications.Design/methodology/approachIn order to address the nonlinear patterns present in the price time series, we investigate the nonlinear auto-regressive neural network as the forecast model. This modeling technique is able to combine a variety of basic nonlinear functions to approximate more complex nonlinear characteristics. Specifically, we examine prediction performance that corresponds to several configurations across data splitting ratios, hidden neuron and delay counts, and model estimation approaches.FindingsOur model turns out to be rather simple and yields forecasts with good stability and accuracy. Relative root mean square errors throughout training, validation and testing are specifically 4.34, 4.71 and 3.98%, respectively. The results of benchmark research show that the neural network produces statistically considerably better performance when compared to other machine learning models and classic time-series econometric methods.Originality/valueUtilizing our findings as independent technical price forecasts would be one use. Alternatively, policy research and fresh insights into price patterns might be achieved by combining them with other (basic) prediction outputs.

Traffic Flow Prediction with Random Walks on Graph and Spatiotemporal Bidirectional Attention Transformer

Traffic flow prediction is crucial in intelligent transportation systems. Considering the severe disruptions caused by traffic accidents or congestion, a time series model is developed for traffic flow prediction based on multiple random walks on graphs (MRWG) and the bidirectional spatiotemporal attention mechanism (BSAM), which can adapt to both normal and exceptional situations. The MRWG mechanism is applied to capture spatial features of urban areas during traffic accidents and congestion, especially the spatial dependencies among neighboring regions. Further, a local position attention module is applied to acquire the spatial correlations between different regions to investigate their impact on the global area, while a local temporal attention module is adopted to extract short-term periodic time correlations from traffic flow data. Finally, a spatiotemporal bidirectional attention module is applied to simultaneously extract both the temporal and spatial correlations of the historical traffic flow data in order to generate the output prediction. Experiments have been conducted on NYCTaxi and NYCBike datasets with abnormal events, and the results indicate that the developed model can efficiently predict traffic flow in abnormal events, especially short-term traffic disruptions, outperforming the baseline methods under both abnormal and normal conditions.