Three-step Prediction Research Articles

Dynamic predictions for the composition and efficiency of heating, ventilation and air conditioning systems in urban building energy modeling

The composition and efficiency of heating, ventilation and air conditioning (HVAC) systems are crucial inputs for urban building energy modeling (UBEM). However, in current studies, the heterogeneity of HVAC systems is often ignored, leading to huge modeling errors. To address these issues, this study developed a three-step prediction approach that can dynamically forecast the composition of HVAC systems (SystemID), the efficiency of heat sources and the efficiency of entire systems at the urban scale. The performance of the prediction models was evaluated through the ten-fold cross-validation. The results showed that the accuracy in predicting the composition could reach 0.87, and the coefficient of determination (R2) was greater than 0.93 in efficiency forecasts. The accuracy of the developed approach was further evaluated through the testing set in two situations, i.e., with the predicted and actual SystemID data. In the first situation, the overall accuracy in predicting the composition reached 78.2 %, and the R2 for forecasting the system efficiencies was over 0.6. Then, with the SystemID assumed to be accurate, the R2 increased to more than 0.87. To analyze the performance of the developed method in energy use predictions, two case regions in Changzhou and Nanjing were used, respectively. The simulation results were compared with the scenario in which the system efficiencies were determined based on the national standard. The results showed that the developed method could enhance the accuracy of total energy use predictions by approximately 10 % and HVAC energy use predictions by roughly 40 % at both urban and building scales. This helps policy-makers craft energy-saving strategies more reasonably.

A traffic-weather generative adversarial network for traffic flow prediction for road networks under bad weather

Traffic flow prediction is pivotal in providing reliable information for intelligent traffic systems. Unexpected events, such as bad weather, unavoidably impact the precision of traffic flow prediction. Therefore, to achieve accurate traffic flow prediction results in road networks under bad weather, a novel Traffic-Weather Generative Adversarial Network (TWeather-GAN model) is developed. This model comprises a Generator and a Discriminator. The Generator incorporates both the traffic and weather modules to extract the spatiotemporal patterns hidden in traffic flow and weather data. In the traffic and weather modules, the gated convolutional layer, Encoder-Decoder architecture, and attention mechanism are established. In the Discriminator, the gated convolutional layer and bidirectional long short-term memory neural network are introduced. Traffic flow data under fog, strong wind, and heavy rain are selected to test the seven baseline models and the proposed model, and the ablation experiments are conducted to analyze the mechanism of the proposed model. The experiments demonstrate that the TWeather-GAN model outperforms the baseline models under bad weather, and makes the prediction error have an average reduction of 0.20%–34.81%, 3.21%–35.22%, and 9.46%–39.10%, respectively, under one-step prediction, three-step prediction, and six-step prediction. Furthermore, establishing the gated convolutional layer and the weather module enhances the accuracy of traffic flow prediction under bad weather. Results show that traffic flow fluctuations and distributions differ under fog and strong wind, and heavy rain affects the trend of traffic flow over one day.

An adaptive variational mode decomposition for wind power prediction using convolutional block attention deep learning network

Random intermittency and high fluctuation limit the wind power prediction accuracy. Although current studies offer various solutions, the prediction accuracy and data fitting performance are not satisfactory during violently fluctuating periods. To address the issue, a novel hybrid model is proposed in this paper, which combines adaptive variational mode decomposition (VMD), temporal convolution network with convolutional block attention module (CBTCN), and gated recurrent unit (GRU). First, VMD is used for data decomposition. Due to the high fluctuation of wind power data, it is difficult to optimize parameters for VMD. Thus grey wolf crossover optimization algorithm (GSCSO) is proposed, which combines four optimization algorithms. Then, an important index, i.e., dynamic error entropy (DEE) is proposed as the fitness function to ensure decomposition integrity, mode complexity, and predictability of sub-sequences for the first time. Thereafter, by extracting deep temporal features with CBTCN, GRU is cascaded to further mine the temporal correlation of these features and predict the wind power. Multiple experiments are conducted, and the results demonstrate that the proposed hybrid model can track the peaks and troughs satisfactorily, especially during violently fluctuating periods. For instance, in three-step prediction, the RMSE reduces by over 40% compared with other advanced models.

Developing machine learning approaches to identify candidate persistent, mobile and toxic (PMT) and very persistent and very mobile (vPvM) substances based on molecular structure

Determining which substances on the global market could be classified as persistent, mobile and toxic (PMT) substances or very persistent, very mobile (vPvM) substances is essential to prevent or reduce drinking water contamination from them. This study developed machine learning models based on different molecular descriptors (MDs) and defined applicability domains for the screening of PMT/vPvM substances. The models were trained with 3111 substances with expert weight-of-evidence based PMT/vPvM hazard classifications that considered the highest quality data available. The model was based on the hypothesis that PMT/vPvM substances contain similar MDs, representative of chemical structures resistant to degradation, be associated with low sorption (or high-water solubility) and in some cases be associated with known toxic mechanisms. All possible model combinations were tested by integrating different molecular description methods, data balancing strategies and machine learning algorithms. Our model allows one-step prediction of candidate PMT/vPvM substances, and our method was compared with the approach predicting P, M and T separately (i.e. three-step prediction). The results showed that the one-step model achieved a higher accuracy of 92% for PMT/vPvM identification (i.e. positive samples) for an internal test set, and also resulted in a higher accuracy of 90% for an external test set of chemical pollutants detected in Taihu Lake, China. Furthermore, prediction mechanism of the model was interpreted by Shapley additive explanations (SHAP). This work presents an advance of big data in silico screening models for the identification of substances that potentially meet the PMT/vPvM criteria.

A novel network training approach for solving sample imbalance problem in wind power prediction

Randomness and intermittency are common challenges in wind power prediction. Most studies focus on randomness but usually ignore the intermittency of wind power that leads to sample imbalance and impairs prediction accuracy. To address the sample imbalance problem, a segment imbalance regression (SIR) method with crisscross optimization (CSO) is proposed to proactively dig and utilize the imbalance nature of samples. By investigating the interactions among adjacent samples, SIR is employed to adaptively assign different learning weights to each sample. SIR focuses on the training samples within the segment while retaining useful information outside, and thus can facilitate the gradient descend process and achieve better training performance. On this basis, aiming to extract more feature information, a novel combination network is constructed by integrating two networks, i.e., Temporal self-attention network (TSA) with temporal self-correction and bidirectional gate recurrent unit (BiGRU) with bi-direction temporal memory respectively. The sequence features of samples are first obtained by ensemble empirical mode decomposition (EEMD) as the input of the combination network, and then the SIR-CSO method is used to train the TSA-BiGRU network and enhance the adaptive learning ability. Massive experiments are conducted, and the results demonstrate the excellent performance of SIR and the superiority of the combination network. Especially in three-step prediction, the root mean square errors of the proposed model reduce 38.97%∼47.04% compared with other state-of-the-art methods.

Research on Multi-Step Prediction of Short-Term Wind Power Based on Combination Model and Error Correction

The instability of wind power poses a great threat to the security of the power system, and accurate wind power prediction is beneficial to the large-scale entry of wind power into the grid. To improve the accuracy of wind power prediction, a short-term multi-step wind power prediction model with error correction is proposed, which includes complete ensemble empirical mode decomposition adaptive noise (CEEMDAN), sample entropy (SE), improved beetle antennae search (IBAS) and kernel extreme learning machine (KELM). First, CEEMDAN decomposes the original wind power sequences into a set of stationary sequence components. Then, a set of new sequence components is reconstructed according to the SE value of each sequence component to reduce the workload of subsequent prediction. The new sequence components are respectively sent to the IBAS-KELM model for prediction, and the wind power prediction value and error prediction value of each component are obtained, and the predicted values of each component are obtained by adding the two. Finally, the predicted values of each component are added to obtain the final predicted value. The prediction results of the actual wind farm data show that the model has outstanding advantages in high-precision wind power prediction, and the error evaluation indexes of the combined model constructed in this paper are at least 34.29% lower in MAE, 34.53% lower in RMSE, and 36.36% lower in MAPE compared with other models. prediction decreased by 30.43%, RMSE decreased by 29.67%, and MAPE decreased by 28.57%, and the error-corrected three-step prediction decreased by 55.60%, RMSE decreased by 50.00%, and MAPE decreased by 54.17% compared with the uncorrected three-step prediction, and the method significantly improved the prediction accuracy.

Deep Learning Method on Deformation Prediction for Large-Section Tunnels

With the continuous development of engineering construction in China, more and more large-section highway tunnels have emerged. Different geological engineering environments determine the diversity of construction plans. The determination of construction plans and the prediction of tunnel deformations have always been the key points of engineering construction. In this paper, we use numerical simulations to determine specific construction parameters in the context of actual highway tunnel projects, and then use deep learning methods to predict deformation during tunnel construction, thus providing guidance for construction. We have found that: (i) Different excavation sequences and excavation depths have different effects on the surrounding rock deformation around the tunnel. The optimal excavation sequence through numerical simulation in this study is symmetrical excavation, and the excavation depth is 2 m. (ii) Numerical simulation based on Long Short-Term Memory (LSTM) algorithm is used to predict the tunnel deformation. It is found that the prediction results of the LSTM algorithm are more consistent with the actual monitoring data. (iii) Multi-step prediction is more important for engineering guidance, and three-step prediction can be considered during the process of engineering construction. Therefore, the machine learning algorithm provides a new method for engineering prediction.

Ultra-Short-Term Photovoltaic Power Prediction Model Based on the Localized Emotion Reconstruction Emotional Neural Network

Due to the intermittency and randomness of photovoltaic (PV) power, the PV power prediction accuracy of the traditional data-driven prediction models is difficult to improve. A prediction model based on the localized emotion reconstruction emotional neural network (LERENN) is proposed, which is motivated by chaos theory and the neuropsychological theory of emotion. Firstly, the chaotic nonlinear dynamics approach is used to draw the hidden characteristics of PV power time series, and the single-step cyclic rolling localized prediction mechanism is derived. Secondly, in order to establish the correlation between the prediction model and the specific characteristics of PV power time series, the extended signal and emotional parameters are reconstructed with a relatively certain local basis. Finally, the proposed prediction model is trained and tested for single-step and three-step prediction using the actual measured data. Compared with the prediction model based on the long short-term memory (LSTM) neural network, limbic-based artificial emotional neural network (LiAENN), the back propagation neural network (BPNN), and the persistence model (PM), numerical results show that the proposed prediction model achieves better accuracy and better detection of ramp events for different weather conditions when only using PV power data.

A Novel High-Performance Predictive Control Formulation for Multilevel Inverters

This article proposes a novel high-performance predictive control with long prediction horizons to improve the control performance of the multilevel inverters. The finite control set model predictive control (FCS-MPC) has obtained a lot of attention for power converters due to its advantages of high dynamic performance, multi objective capability, no need for PI regulators and PWM modulators. However, the MPC method requires a high number of computations especially for higher-level power converter topologies due to the existence of a huge amount of switching combinations and redundancies. Real-time searching for the optimal switching state among a large candidate pool at a high sampling rate is sometimes impossible with the standard commercial processors. This limitation also prevents real-time implementation of an MPC for multilevel converters with step prediction more than one. To solve the aforementioned issues, this paper presents a novel high-performance FCS-MPC scheme. The proposed FCS-MPC is reformulated mathematically MPC approach to an l 2 norm optimization problem, which can be solved on-line through matrix theory. Compared with the existing MPC optimization algorithms, the proposed MPC formulation has the advantage of a substantial reduction in computational burden, no need for the weighting factors or cost functions, and thus can operate at long horizon prediction length. The proposed method is verified experimentally on a seven-level CHB inverter prototype with a three-step prediction length to demonstrate the ability and performance of the proposed method for multilevel inverters.

An inverse modeling approach for predicting filled rubber performance

In this paper, a computational procedure combining experimental data and interphase inverse modeling is presented to predict filled rubber compound properties. The Fast Fourier Transformation (FFT) based numerical homogenization scheme is applied on the high quality filled rubber 3D Transmission Electron Microscope (TEM) image to compute its complex shear moduli. The 3D TEM filled rubber image is then compressed into a material microstructure database using a novel Reduced Order Modeling (ROM) technique, namely Self-consistent Clustering Analysis (a two-stage offline database creation from training and learning, followed by data compression via unsupervised learning, and online prediction approach), for improved efficiency and accuracy. An inverse modeling approach is formulated for quantitatively computing interphase complex shear moduli in order to understand the interphase behaviors. The two-stage SCA and the inverse modeling approach formulate a three-step prediction scheme for studying filled rubber, whose loss tangent curve can be computed in agreement with test data.

Air PM2.5 concentration multi-step forecasting using a new hybrid modeling method: Comparing cases for four cities in China

Exposure to particulate matter (PM2.5) with high concentrations can increase the risk of human illness and mortality. Consequently, it is meaningful to build an accurate model for PM2.5 forecasting and provide reference for air pollution management and short-term warning. This paper develops a novel hybrid model called WPD-PSO-BP-Adaboost, based on WPD (Wavelet Packet Decomposition), the PSO (Particle Swarm Optimization) algorithm, BPNN (Back Propagation Neural Network) and Adaboost algorithm. In the proposed structure, to obtain better performance of PM2.5 forecasting, the novel hybrid model can be describe as: the WPD is utilized to decompose the raw PM2.5 data into several sub-layers with low frequency and high frequency; optimized by PSO and Adaboost algorithm, the BPNN is employed to compete the three-step prediction for every single subseries. To investigate the three-step forecasting performance of the proposed models, there are three experiments involving eleven models for the comparisons, including the BP model, BP-Adaboost model, WPD-BP model, PSO-BP model, WPD-BP-Adaboost model, WPD-PSO-BP model, PSO-BP-Adaboost model, WPD-PSO -BP-Adaboost model, EEMD-GRNN model, CEEMDAN-ICA-ELM model and WPD-CEEMD-PSOGSA -SVM model. The experiments results show that: (1) the WPD is useful in improving the forecasting performance; (2) the PSO and Adaboost algorithm can enhance the precision of forecasting significantly; (3) in all models, the WPD-PSO-BP-Adaboost model performs best in multi-step forecasting.

Similarity Grouping-Guided Neural Network Modeling for Maritime Time Series Prediction

Reliable and accurate prediction of time series plays a crucial role in maritime industry, such as economic investment, transportation planning, port planning and design, etc. The dynamic growth of maritime time series has the predominantly complex, nonlinear and non-stationary properties. To guarantee high-quality prediction performance, we propose to first adopt the empirical mode decomposition (EMD) and ensemble EMD (EEMD) methods to decompose the original time series into high- and low-frequency components. The low-frequency components can be easily predicted directly through traditional neural network (NN) methods. It is more difficult to predict high-frequency components due to their properties of weak mathematical regularity. To take advantage of the inherent self-similarities within high-frequency components, these components will be divided into several continuous small (overlapping) segments. The grouped segments with high similarities are then selected to form more proper training datasets for traditional NN methods. This regrouping strategy can assist in enhancing the prediction accuracy of high-frequency components. The final prediction result is obtained by integrating the predicted high- and low-frequency components. Our proposed three-step prediction frameworks benefit from the time series decomposition and similar segments grouping. Experiments on both port cargo throughput and vessel traffic flow have illustrated its superior performance in terms of prediction accuracy and robustness.

Inferring tag co-occurrence relationship across heterogeneous social networks

Predicting the occurrence of links or interactions between objects in a network is a fundamental problem in network analysis. In this work, we address a novel problem about tag co-occurrence relationship prediction across heterogeneous networks. Although tag co-occurrence has recently become a hot research topic, many studies mainly focus on how to produce the personalized recommendation leveraging the tag co-occurrence relationship and most of them are considered in a homogeneous network. So far, few studies pay attention to how to predict tag co-occurrence relationship across heterogeneous networks. In order to solve this novel problem mentioned previously, we propose a novel three-step prediction approach. First, image-tag bins are generated by utilizing the TF-IDF like method, which help reduce the search space. And then, weight path-based topological features are systematically extracted from the network. At last, a supervised model is used to learn the best weights associated with different topological features in deciding the co-occurrence relationships. Experiments are performed on real-world dataset, the Flickr network, with comprehensive measurements. Experimental results demonstrate that weight path-based heterogeneous topological features have substantial advantages over commonly used link prediction approaches in predicting co-occurrence relations in Flickr networks.

Predicting hospitality firm failure: mixed sample modelling

PurposeAvailable information for evaluating the possibility of hospitality firm failure in emerging countries is often deficient. Oversampling can compensate for this but can also yield mixed samples, which limit prediction models’ effectiveness. This research aims to provide a feasible approach to handle possible mixed information caused by oversampling.Design/methodology/approachThis paper uses mixed sample modelling (MSM) when evaluating the possibility of firm failure on enlarged hospitality firms. The mixed sample is filtered out with a mixed sample index through control of the noisy parameter and outliner parameter and meta-models are used to build MSM models for hospitality firm failure prediction, with performances compared to traditional models.FindingsThe proposed models are helpful in predicting hospitality firm failure in the mixed information situation caused by oversampling, whereas MSM significantly improves the performance of traditional models. Meanwhile, only partial mixed hospitality samples matter in predicting firm failure in both rich- and poor-information situations.Practical implicationsThis research is helpful for managers, investors, employees and customers to reduce their hospitality-related risk in the emerging Chinese market. The two-dimensional sample collection strategies, three-step prediction process and five MSM modelling principles are helpful for practice of hospitality firm failure prediction.Originality/valueThis research provides a means of processing mixed hospitality firm samples through the early definition and proposal of MSM, which addresses the ranking information within samples in deficient information environments and improves forecasting accuracy of traditional models. Moreover, it provides empirical evidence for the validation of sample selection and sample pairing strategy in evaluating the possibility of hospitality firm failure.

RPI-Bind: a structure-based method for accurate identification of RNA-protein binding sites

RNA and protein interactions play crucial roles in multiple biological processes, while these interactions are significantly influenced by the structures and sequences of protein and RNA molecules. In this study, we first performed an analysis of RNA-protein interacting complexes, and identified interface properties of sequences and structures, which reveal the diverse nature of the binding sites. With the observations, we built a three-step prediction model, namely RPI-Bind, for the identification of RNA-protein binding regions using the sequences and structures of both proteins and RNAs. The three steps include 1) the prediction of RNA binding regions on protein, 2) the prediction of protein binding regions on RNA, and 3) the prediction of interacting regions on both RNA and protein simultaneously, with the results from steps 1) and 2). Compared with existing methods, most of which employ only sequences, our model significantly improves the prediction accuracy at each of the three steps. Especially, our model outperforms the catRAPID by >20% at the 3rd step. All of these results indicate the importance of structures in RNA-protein interactions, and suggest that the RPI-Bind model is a powerful theoretical framework for studying RNA-protein interactions.

Performance Comparison of Four New ARIMA-ANN Prediction Models on Internet Traffic Data

Prediction of Internet traffic time series data (TSD) is a challenging research problem, owing to the complicated nature of TSD. In literature, many hybrids of auto-regressive integrated moving average (ARIMA) and artificial neural networks (ANN) models are devised for the TSD prediction. These hybrid models consider such TSD as a combination of linear and non-linear components, apply combination of ARIMA and ANN in some manner, to obtain the predictions. Out of the many available hybrid ARIMA-ANN models, this paper investigates as to which of them suits better for Internet traffic data. This suitability of hybrid ARIMAANN models is studied for both one-step ahead and multistep ahead prediction cases. For the purpose of the study, Internet traffic data is sampled at every 30 and 60 minutes. Model performances are evaluated using the mean absolute error and mean square error measurement. For one-step ahead prediction, with a forecast horizon of 10 points and for three-step prediction, with a forecast horizon of 12 points, the moving average filter based hybrid ARIMA-ANN model gave better forecast accuracy than the other compared models.

Photo Album Compression for Cloud Storage Using Local Features

This paper proposes a novel feature-based photo album compression scheme for cloud storage. The main contribution of this paper is the use of local features rather than pixel values for analyzing and exploring the correlation between images. Unlike previously established image set compression schemes, we adopt content-based feature matching, which is invariant to scale, rotation, and robust to illumination changes, in both correlation estimation and redundancy reduction. Based on feature-based correlation analysis, we organize correlated images as a pseudo sequence by minimizing the predictive cost. Since pseudo sequences have much more complicated correlations than natural videos, we further propose a three-step prediction to reduce inter-image redundancy based on local features. We first propose a feature-based multi-model approach to reduce local geometric deformations between images, resulting in multiple deformed predictions. We then decrease the illumination difference between each deformed prediction and the target image with a photometric transformation. Finally, we reduce redundancy between images using a block-based motion compensation, similar to video compression. Experimental results show that our proposed feature-based coding scheme can be 10 times more efficient than individual JPEG compression.

Prediction of Maternal Predisposition to Preeclampsia

Objective: To derive a prediction index based on the most salient patient history, laboratory, and clinical parameters for identifying women at high risk for developing preeclampsia (PE). Methods: Nonpregnant women with a history of PE (n = 101) were compared with nonpregnant parous women with a history of one or more successful normotensive pregnancies (n = 50) but with comparable age, gestation, and parity profiles. The parameters included a medical examination (demographics, patient history, family history, and clinical and obstetrical findings), laboratory investigations (hemostasis, coagulation, and vitamins), and morphological and functional tests (cardiovascular and renal functions). Stepwise logistic regression analysis was applied to develop a three-step PE prediction index based on the most discriminant parameters. Results: Patients with and without PE differed significantly (p < 0.05) with respect to 1) maternal history of chronic hypertension, body mass index, and blood pressure; 2) APTT, PT, activated factor VIII, homocystein, free protein S and vitamin B1; and 3) relative plasma volume. Based on these three sets of parameters, a three-step PE prediction index was developed. The likelihood ratio of a positive index score was equal to 3.4, 7.3, and 8.8, respectively. Thus, assuming a PE prevalence (or prior probability) of 5%, a patient's chances of developing PE when presenting with a positive score on the three-step prediction index were 15%, 28%, and 32%, respectively. Discussion: In the absence of welldefined pre-pregnancy screening guidelines for PE, the present study attempts to proceed in a stepwise fashion by looking at medical examination data first, requesting, if necessary, specific hemostasis and coagulation tests next, and finally measuring the relative plasma volume for confirmatory purposes. This approach offers a satisfactory positive predictive value and cost efficiency ratio.