Dynamic Prediction Method Research Articles

Knowledge fusion and spatio-temporal modeling for dynamic bottleneck prediction in sustainable production processes

ABSTRACT Sustainable production is for the manufacturing industry to improve production efficiency and reduce energy consumption. Production system complexities, which cause bottlenecks, greatly impact process efficiency. However, the challenge lies in effectively utilizing sustainable production process knowledge, which is often under-exploited, particularly when integrating this knowledge with knowledge graphs for production bottleneck prediction. To solve the problem, this study presents a method for dynamic bottleneck prediction by integrating knowledge graphs and spatio-temporal models, leveraging underused production process knowledge. Initially, developing a sustainable production process knowledge graph effectively captures correlations between workstation states and external factors at production bottlenecks. Subsequently, a knowledge cross-fertilization module merges this correlation knowledge with spatio-temporal features, enhancing bottleneck prediction accuracy and reliability. Therefore, it efficiently predicts each workstation’s blockage and starvation, which has significance in identifying accurately future bottleneck workstations. A case study demonstrates that this method significantly reduces root-mean-square error by 7% to 42% in predicting overall system blockage and starvation metrics. Moreover, the mean-absolute-error of this method is within 4.5% over various time scales of 30,60,90, and 120 minutes. Furthermore, these results confirm that the method introduced holds considerable practical significance in precisely identifying future workstation bottlenecks, offering novel tools and insights for enhancing production efficiency.

Prediction Accuracy of Stackelberg Game Model of Electricity Price in Smart Grid Power Market Environment

With the deepening of power market reform and the increasingly fierce competition in the power market, the accurate prediction of electricity price has become an important demand for power market participants to make scientific decisions, optimize resource allocation, and reduce risks. Electricity price forecast can provide a key reference for the power market, help market participants make wise decisions, promote competition and efficient operation and cope with complex market fluctuations, provide a scientific basis for various entities to optimize resource allocation, reduce risks and improve benefits, and promote the sustainable development of the power industry. This study presents a dynamic retail price prediction method for smart grid based on the Stackelberg game model. Firstly, the correlation test is used to verify the strong correlation between electric load and electricity price. Secondly, the parameters of the Stackelberg model are determined, and the load and electricity price are tested using the white noise test. Finally, by comparing the BP neural network model and quantifying the model parameters, the superiority of the model is verified. The results show that the Stackelberg game model has higher prediction accuracy than the BP neural network model in electricity price prediction.

Joint quantile regression of longitudinal continuous proportions and time-to-event data: Application in medication adherence and persistence.

This study introduces a novel joint modeling framework integrating quantile regression for longitudinal continuous proportions data with Cox regression for time-to-event analysis, employing integrated nested Laplace approximation for Bayesian inference. Our approach facilitates an examination across the entire distribution of patient health metrics over time, including the occurrence of key health events and their impact on patient outcomes, particularly in the context of medication adherence and persistence. Integrated nested Laplace approximation's fast computational speed significantly enhances the efficiency of this process, making the model particularly suitable for applications requiring rapid data analysis and updates. Applying this model to a dataset of patients who underwent treatment with atorvastatin, we demonstrate the significant impact of targeted interventions on improving medication adherence and persistence across various patient subgroups. Furthermore, we have developed a dynamic prediction method within this framework that rapidly estimates persistence probabilities based on the latest medication adherence data, demonstrating integrated nested Laplace approximation's quick updates and prediction capability. The simulation study validates the reliability of our modeling approach, evidenced by minimal bias and appropriate credible interval coverage probabilities across different quantile levels.

Structural Online Damage Identification and Dynamic Reliability Prediction Method Based on Unscented Kalman Filter.

As sensor monitoring technology continues to evolve, structural online monitoring and health management have found numerous applications across various fields. However, challenges remain concerning the real-time diagnosis of structural damage and the accuracy of dynamic reliability predictions. In this paper, a structural online damage identification and dynamic reliability prediction method based on Unscented Kalman Filter (UKF) is presented. Specifically, in the Wiener degradation process with random effects on structural performance, the structural damage identification is initially realized using UKF. Following that, the EM algorithm is employed for estimating the performance model parameters. Eventually, dynamic reliability prediction is realized based on conditional probability. The simulation results indicate that the method effectively estimates the damage state during the structure's use while providing accurate, real-time, and dynamic reliability predictions for the system.

A dynamic strain prediction method for malfunction of sensors in buildings subjected to seismic loads using CWT and CNN

Structural health monitoring (SHM) is a technique used to evaluate the safety of a structure based on the structural response. In the SHM, sensors may fail to measure the structural response, or data loss can occur during the data transmission and reception. In this regard, this paper proposes a method of predicting the strain response when a sensor installed in a structural member suffers a malfunction. To this end, a convolutional neural network (CNN) is introduced to predict the strain response. Time-history strain data for seismic loads measured in adjacent structural members are set as the input layer of the CNN, and continuous wavelet transform (CWT) results are additionally set in the input layer of the CNN to reflect the dynamic structural behavior during earthquakes in the CNN predictions. Time-history strain data for a member, the target of prediction, are set as the output layer of the CNN. Then, the trained CNN uses the time-history strain data of adjacent members with no sensor failure and the CWT result of the data to predict the strain response of a structural member with sensor failure. A numerical study on the seismic load of a single-span three-story steel moment frame and an experimental study on a single-span three-story reinforced concrete frame were carried out to verify the validity of the proposed method. This study presents a method of applying CWT data to the input layer of the CNN and discusses the effectiveness of CWT data for predicting the nonlinear response of a structural member with damaged sensors.

Refined subsidence monitoring and dynamic prediction in narrow and long mining areas based on InSAR and probabilistic integral method

Continuous exploitation in mining areas damages the surrounding environment and has various severe geological impacts. Hence, long-term monitoring of mining areas is crucial to reducing these impacts. Differential interferometric synthetic aperture radar (D-InSAR) is widely applied to monitor the subsidence in mining areas, but it cannot obtain accurate large-gradient subsidence result in the centre of the subsidence basin in mining areas due to the de-coherence phenomenon. The probability integral method (PIM) is a prediction method that can cooperate with D-InSAR (D-InSAR PIM, DPIM) to solve the problem. However, with this method, there is early convergence of the edge subsidence in narrow and long mining areas. Moreover, the PIM can only predict the spatial domain; it cannot achieve dynamic prediction. To address the above problems in the traditional DPIM data processing process, in this study, firstly, the traditional PIM was improved by adjusting the radius of the parameter and constructed an improved DPIM (IDPIM) method. The hybrid algorithm was applied to solve the parameters of the IDPIM method and then acquire subsidence results, thus solving the early convergence of edge subsidence problem characteristic of traditional PIM prediction in mining. Additionally, an area-weighting based fusion method was proposed to integrate the IDPIM results and the D-InSAR results (Area-weighting based fusion of the IDPIM and D-InSAR results, AFID) achieving whole-basin refined subsidence in mining areas. Secondly, based on a summary of subsidence laws in mining areas, the Hossfeld model was introduced and combined with the IDPIM method (IDPIM Hossfeld, IDPIM-H) to construct a subsidence dynamic prediction method. This achieved dynamic prediction of the subsidence in mining areas. A coal mine in Ordos was used as the study area, and the feasibility of the IDPIM method, the AFID method and the DPIM-H method was verified through a comparative analysis of leveling data. The results demonstrated that: (1) The results of the IDPIM method showed 8% and 66% improvement in RMSE along the striking and dip lines, respectively, over the D-InSAR results, improving the early convergence of the DPIM along the dip direction of the mine. (2) The results of the AFID method provide a 69% improvement in whole-basin RMSE over the D-InSAR results, which improves the lack of monitoring capacity in the D-InSAR technology center. (3) The results of the DPIM-H method provide a 35% improvement in basin-wide RMSE over the D-InSAR results, solving the problem of low temporal resolution of the D-InSAR technology and realizing the dynamic prediction with high temporal resolution. These findings provide a theoretical basis for future refined exploration of dynamic subsidence in mining areas

Bidirectional dynamic frame prediction network for total-body [68Ga]Ga-PSMA-11 and [68Ga]Ga-FAPI-04 PET images

PurposeTotal-body dynamic positron emission tomography (PET) imaging with total-body coverage and ultrahigh sensitivity has played an important role in accurate tracer kinetic analyses in physiology, biochemistry, and pharmacology. However, dynamic PET scans typically entail prolonged durations (\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\:\\ge\\:$$\\end{document}60 minutes), potentially causing patient discomfort and resulting in artifacts in the final images. Therefore, we propose a dynamic frame prediction method for total-body PET imaging via deep learning technology to reduce the required scanning time.MethodsOn the basis of total-body dynamic PET data acquired from 13 subjects who received [68Ga]Ga-FAPI-04 (68Ga-FAPI) and 24 subjects who received [68Ga]Ga-PSMA-11 (68Ga-PSMA), we propose a bidirectional dynamic frame prediction network that uses the initial and final 10 min of PET imaging data (frames 1–6 and frames 25–30, respectively) as inputs. The peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) were employed as evaluation metrics for an image quality assessment. Moreover, we calculated parametric images (68Ga-FAPI: \\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\:{K}_{1}$$\\end{document}, 68Ga-PSMA: \\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\:{K}_{i}$$\\end{document}) based on the supplemented sequence data to observe the quantitative accuracy of our approach. Regions of interest (ROIs) and statistical analyses were utilized to evaluate the performance of the model.ResultsBoth the visual and quantitative results illustrate the effectiveness of our approach. The generated dynamic PET images yielded PSNRs of 36.056 ± 0.709 dB for the 68Ga-PSMA group and 33.779 ± 0.760 dB for the 68Ga-FAPI group. Additionally, the SSIM reached 0.935 ± 0.006 for the 68Ga-FAPI group and 0.922 ± 0.009 for the 68Ga-PSMA group. By conducting a quantitative analysis on the parametric images, we obtained PSNRs of 36.155 ± 4.813 dB (68Ga-PSMA, \\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\:{K}_{1}$$\\end{document}) and 43.150 ± 4.102 dB (68Ga-FAPI, \\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\:{K}_{i}$$\\end{document}). The obtained SSIM values were 0.932 ± 0.041 (68Ga-PSMA) and 0.980 ± 0.011 (68Ga-FAPI). The ROI analysis conducted on our generated dynamic PET sequences also revealed that our method can accurately predict temporal voxel intensity changes, maintaining overall visual consistency with the ground truth.ConclusionIn this work, we propose a bidirectional dynamic frame prediction network for total-body 68Ga-PSMA and 68Ga-FAPI PET imaging with a reduced scan duration. Visual and quantitative analyses demonstrated that our approach performed well when it was used to predict one-hour dynamic PET images. https://github.com/OPMZZZ/BDF-NET.

Rotor dynamic response prediction using physics-informed multi-LSTM networks

The deep learning method provides an effective alternative to numerical simulations for establishing the nonlinear input-output relationship and calculating dynamic responses of rotor systems. To overcome the low generalization capability of pure data-driven long short-term memory (LSTM) networks when predicting dynamic responses to out-of-distribution inputs, a dynamic response prediction method using physics-informed multi-LSTM networks is proposed. This approach incorporates required physical constraints into the deep LSTM network, allowing the model training process to optimize the network parameters within the feasible solution space that adheres to physical laws. Consequently, this enhances the physical interpretability of the deep learning model. Specifically, two physics-informed multi-LSTM network architectures are introduced, and physical laws of equation of motion, state dependency and hysteretic constitutive relationship are considered to construct the physics loss. The feasibility of the proposed method is verified by a Bouc-Wen hysteresis model and a simulated gas generator rotor. The response prediction performance of the two networks is validated on a constructed fault rotor dataset with significant sample differences, along with cross-speed and cross-node prediction validation for the rotor system. The results demonstrate that the trained networks exhibit strong robustness and generalization capabilities, making them suitable as surrogate models for rotor systems.

Dynamic output feedback robust MPC hybrid fault-tolerant control for discrete uncertain systems: a switching control strategy

To guarantee the control performance of the discrete system under unmeasurable state and partial actuator failure, a robust dynamic output feedback predictive fault-tolerant control method derived from switching strategy is developed. A new iterative error switching model is established, which both ensures tracking performance and lays the foundation for the development of the next switching controller. Based on this model, output feedback switching fault-tolerant controller is developed. Compared with conventional fault-tolerant control methods, the designed controller can be used under the case of unmeasurable states, and it is more fault tolerant and less conservative. According to the Lyapunov stability theory, the robust stability analysis is carried out under normal and fault conditions, respectively. Then, depending on the findings of the stability analysis, sufficient conditions to guarantee the stability of the closed-loop normal and faulty system are derived. Finally, the suggested method's efficacy is confirmed using the injection moulding process as an example.

Numerical method for dynamic prediction of temperature field in aviation gear profile grinding

Numerical method for dynamic prediction of temperature field in aviation gear profile grinding

Research progress on quantitative evaluation methods of urban ecological carrying capacity

Quantitative evaluation of urban ecological carrying capacity is a critical foundation for measuring urban sustainable development in the new era. This review would enrich the concept and connotation of urban ecological carrying capacity by sorting out its components and characteristics. We categorized the methods for quantifying urban ecological carrying capacity into static evaluation methods, including ecological footprint method, comprehensive evaluation method, state space method, net primary productivity method, and carbon-oxygen balance method, as well as dynamic simulation prediction methods, including system dynamics models, BP neural network prediction models, and grey prediction models. We further summarized both the advantages and disadvantages of these quantitative evaluation methods. Finally, we predicted the development trends of quantitative evaluation methods for urban ecological carrying capacity, and put forward further in-depth research from aspects such as strengthening theoretical foundations, constructing indicator systems, establishing evaluation principles, and optimizing methods.

Prediction of Progressive Erosion Characteristics of Centrifugal Pump Blades Based on Dynamic Boundaries

Abstract The conventional method of fixed boundaries for predicting erosion inadequately reflects the erosion characteristics of centrifugal pumps under realistic operating conditions. In this research, we employ the dynamic boundary method, considering the influence of changes in wall morphology due to erosion on the flow field. We utilize the Euler-Lagrange approach and the E/CRC erosion model, incorporating a dynamic boundary geometry reconstruction strategy, to predict the erosion characteristics of the IS80-50-315 single-stage single-suction centrifugal pump. Solid-liquid two-phase flow calculations are conducted on the pump to predict erosion characteristics under varied operating conditions. The results show that the dynamic boundary-based progressive erosion prediction method realistically forecasts the erosion characteristics of the overflow wall and captures the evolution of blade morphology with erosion time. Erosion intensity on the blade suction surface significantly surpasses that on the pressure surface, guided by particle distribution characteristics in the centrifugal pump flow channel. Particle diameter primarily influences the erosion position of centrifugal pump blades; smaller particles induce erosion in the middle region of the blade, while an increase in particle diameter shifts the severe erosion position towards the impeller outlet direction. At a consistent particle diameter, higher particle concentrations correspond to elevated erosion rates in the blade area. We establish a relationship between the blade mass loss rate and erosion time during 2500 hours of pump operation at the design condition, introducing the blade mass loss rate for convenient assessment and prediction of blade damage in the centrifugal pump.

The Dynamic Prediction Method for Aircraft Cabin Temperatures Based on Flight Test Data

The Dynamic Prediction Method for Aircraft Cabin Temperatures Based on Flight Test Data

E-PINN: A fast physics-informed neural network based on explicit time-domain method for dynamic response prediction of nonlinear structures

The physics-informed neural networks (PINNs), serving as the surrogate models, have emerged in the dynamic response prediction of nonlinear systems. However, in the conventional PINN model, the global differential equation of motion of the nonlinear system is directly integrated into the loss function, which requires the network to capture the intricate global dynamic evolution mechanism of the system and thus poses significant challenges for network learning. In this study, inspired by the explicit time-domain method (ETDM), a novel PINN based on ETDM, called E-PINN, is proposed to address the challenges arising from the network learning of the existing PINNs. A lightweight long short-term memory (LSTM) module is trained to learn the nonlinear evolution mechanism of restoring forces, while a single convolutional layer (SCL) module is utilized to reflect the linear evolution mechanism of the primary structure under the combined action of the external excitations and the nonlinear restoring forces. The weights of the SCL module can be directly retrieved from the discrete convolutional formulation of ETDM, and only the parameters in the LSTM module need to be optimized through a loss function solely based on the nonlinear restoring forces, thereby enabling easy network learning of the proposed E-PINN by decoupling the linear and nonlinear evolution mechanisms embedded in the nonlinear system. Three numerical examples are investigated by the present approach, and the results demonstrate the superior training efficiency and prediction accuracy of E-PINN in dynamic response prediction of nonlinear systems.

Prediction of resilient modulus and critical dynamic stress of recycled aggregates: Experimental study and machine learning methods

This paper aimed to investigate the feasibility of partially or completely replacing natural aggregates with recycled aggregates from construction and demolition wastes for low-carbon-emission use as coarse-grained embankment fill materials. The laboratory specimens were prepared by blending natural and recycled aggregates at varying proportions, and a series of laboratory repeated load triaxial compression tests were carried out to study the effects of material index properties and dynamic stress states on the resilient modulus and permanent strain characteristics. Based on the experimental results and by considering the main influencing parameters of the resilient modulus and permanent deformation, an artificial neural network (ANN) prediction model with optimal architecture was developed and optimized by the particle swarm optimization (PSO) algorithm, and its performance and accuracy were verified by supplementary analyses. A shakedown state classification method was proposed based on the unsupervised clustering algorithm, and a prediction model of critical dynamic stress was established based on the machine learning (ML) method and the shakedown state classification results. The research results indicate that the stress state has a greater influence on the resilient modulus and permanent deformation characteristics than other factors, and the shear stress ratio has a significant effect on the shakedown state. The resilient modulus and critical dynamic stress of such specimens vary linearly with confining pressure. The improved PSO-ANN prediction model exhibits high prediction accuracy and robustness, superior to several other commonly used ML regression prediction algorithms. The resilient modulus and critical dynamic stress prediction methods based on ML algorithms can provide technical guidance and theoretical basis for the design and in-service maintenance of similar unbound granular materials.

Research on dynamic bottleneck prediction methods in manufacturing workshops

Abstract To address the challenge of identifying bottleneck responsibilities when multiple bottlenecks arise in process workshops and predicting bottleneck trends before their occurrence, a novel dynamic prediction method based on the Markov chain is introduced. This approach effectively identifies all bottlenecks in a multi-bottleneck process shop and identifies the primary bottleneck unit with the highest level of responsibility, enabling anticipation of bottleneck shifts prior to their occurrence. Initially, the concept of bottleneck is elaborated, accounting for the intricate interplay among manufacturing processes, leading to the development of a static mathematical model for multi-bottleneck prediction incorporating parameters such as time, cost, and quality. This model evaluates the bottleneck severity of each manufacturing unit, facilitating the identification of all bottleneck units and the assignment of bottleneck responsibilities, thereby establishing primary and secondary bottlenecks. Subsequently, a bottleneck state transition matrix is formed based on historical bottleneck processes, initializing the state matrix through the process with the highest bottleneck responsibility in the static setting. Leveraging the Markov chain prediction technique, the method anticipates the location of bottleneck transitions preceding the occurrence of bottleneck drift. Finally, simulations were conducted in an authentic production line environment to validate the method’s efficacy. Results indicate a 9.1% increase in the main bottleneck coincidence rate compared to single-bottleneck prediction models. Moreover, the Markov chain prediction accurately forecasts bottleneck trends prior to drift occurrences, boasting an 86.4% accuracy rate. These findings underscore the applicability and guidance potential of the proposed method.

Remaining useful life prediction for a cracked rotor system via moving feature fusion based deep learning approach

This study proposes a novel approach for predicting the remaining useful life (RUL) of a cracked rotor system by using a moving fusion gated recurrent unit (MFGRU) model. Firstly, 12 multi-domain features were extracted from the raw collected vibration signals, which were then fused and distributed adaptively using bidirectional exponential moving average (EMA) and multi-head attention (MHA). Then a bidirectional gated recurrent unit network combined with moving feature fusion method was proposed to capture the long-term dependence relationship in the monitoring data of the cracked rotor for RUL prediction. Finally, the model’s performance was validated through accelerated life experiments, with the measured values of root mean square error (RMSE) and mean absolute error (MAE) below 3.17 and 2.60, respectively. This study offers a dynamic RUL prediction method with certain significance and valuable reference for designing deep learning models.

Fast dynamic prediction of consequences of heavy gas leakage accidents based on machine learning

The field of emergency risk management in chemical parks has been characterized by a lack of fast, precise and dynamic prediction methods. The application of computational fluid dynamics (CFD) models, which offer the potential for dynamic and precise prediction, has been hindered by high computational costs. Therefore, taking liquid benzene as a case study, this paper combined machine learning (ML) algorithms with a CFD-based precise prediction model, to develop an ML model for fast dynamic prediction of heavy gas leakage consequences in chemical parks. Employing the CFD data as the input, the prediction models were developed using ML algorithms, refined with Bayesian optimization for parameter tuning. This study utilized PHOENICS software to establish a dynamic prediction model for the diffusion of liquid benzene leakage, validated by Burro nine experiment data. Comparative analyses of models based on five ML algorithms were conducted to evaluate the reliability of their predictions using both CFD standard and noisy data. The results indicated that temperature had the most significant effect on the consequences of the leakage accidents among four key factors (wind speed, temperature, leakage aperture and atmospheric stability), followed by wind speed. These factors served as input variables for ML model training. Among the models evaluated, the eXtreme Gradient Boosting (XGBoost) model showed superior performance, irrespective of the presence of noise in the data. An XGBoost-based fast prediction model was ultimately developed for predicting the consequences of liquid benzene leakage. A case analysis was conducted to validate the feasibility of the model prediction. The relative errors between the predicted and actual values of the model for acute exposure guideline level-1 (AEGL-1), AEGL-2, and AEGL-3 distances were 2.70%, 2.58%, and 0.23%, respectively. Furthermore, the XGBoost model completed the prediction in only 0.218 s, a stark contrast to the hours necessitated by the CFD model, thus offering substantial computational time savings while maintaining high accuracy levels. This paper introduces an ML model for fast dynamic prediction of heavy gas leakage, enabling chemical parks to make more timely and accurate decisions in emergency risk management.

A binary-domain recurrent-like architecture-based dynamic graph neural network

The integration of Dynamic Graph Neural Networks (DGNNs) with Smart Manufacturing is crucial as it enables real-time, adaptive analysis of complex data, leading to enhanced predictive accuracy and operational efficiency in industrial environments. To address the problem of poor combination effect and low prediction accuracy of current dynamic graph neural networks in spatial and temporal domains, and over-smoothing caused by traditional graph neural networks, a dynamic graph prediction method based on spatiotemporal binary-domain recurrent-like architecture is proposed: Binary Domain Graph Neural Network (BDGNN). The proposed model begins by utilizing a modified Graph Convolutional Network (GCN) without an activation function to extract meaningful graph topology information, ensuring non-redundant embeddings. In the temporal domain, Recurrent Neural Network (RNN) and residual systems are employed to facilitate the transfer of dynamic graph node information between learner weights, aiming to mitigate the impact of noise within the graph sequence. In the spatial domain, the AdaBoost (Adaptive Boosting) algorithm is applied to replace the traditional approach of stacking layers in a graph neural network. This allows for the utilization of multiple independent graph learners, enabling the extraction of higher-order neighborhood information and alleviating the issue of over-smoothing. The efficacy of BDGNN is evaluated through a series of experiments, with performance metrics including Mean Average Precision (MAP) and Mean Reciprocal Rank (MRR) for link prediction tasks, as well as metrics for traffic speed regression tasks across diverse test sets. Compared with other models, the better experiments results demonstrate that BDGNN model can not only better integrate the connection between time and space information, but also extract higher-order neighbor information to alleviate the over-smoothing phenomenon of the original GCN.

Deviation entropy-based dynamic multi-model ensemble interval prediction method for quantifying uncertainty of building cooling load

Interval prediction is a promising method that can reveal the uncertainty of building load and has been shown to effectively manage building energy systems. Previous studies focused on improving the performance of three types of generic single models (quantile regression, bootstrap, and lower upper bound estimation models). Multi-model ensemble methods have the advantages of reducing overfitting risk, improving stability. However, in the field of building cooling load interval prediction, there is a research gap that the performance of the multi-model ensemble method has not been conducted extensively. To compensate for this gap, firstly, referring to the information entropy in information theory, we proposed deviation entropy, which is a new concept that can be used to calculate the weight of a generic single model. Based on this, a static multi-model ensemble interval prediction method was developed. Then, we introduced a sliding time window to further upgrade the static method into a dynamic method. Finally, we used actual data to evaluate the performance of the proposed dynamic multi-model ensemble interval prediction method. The study results show that the multi-model ensemble method outperforms traditional generic single model, and the dynamic method can improve the prediction reliability by 8.5% with only 0.76% loss of prediction accuracy.