Deterministic Prediction Research Articles

How much does a high-resolution global ensemble forecast improve upon deterministic prediction skill for the Indian summer monsoon?

How much does a high-resolution global ensemble forecast improve upon deterministic prediction skill for the Indian summer monsoon?

Experimental assessment of a nonlinear, deterministic sea wave prediction method using instantaneous velocity profiles

This paper presents the experimental proof-of-concept of a nonlinear, deterministic wave prediction method. The method is based on the adapted version of the HOS-NWT wave model and uses wave-induced velocity profiles as input information on the sea state. Unlike most HOS approaches, it does not need any optimization procedure or data assimilation step to initialize the simulation. Wave tank experiments were conducted to validate the method for irregular, long-crested waves with a low significant steepness. Data was collected using wave probes, an ADV and the fifth beams of two ADCP sensors set-up with a High-Resolution mode. Despite challenging experimental conditions for the ADCPs, the method proved able to reconstruct reliable instantaneous horizontal velocity profiles from the acoustic sensor measurements. These profiles were used as boundary conditions in the wave prediction model. The sea surface elevation predicted was compared to wave probe measurements and showed good agreement all over the theoretical prediction area.

Suggestion of new models for predicting performance of raise boring machines based on indentation tests

Raise Boring Machines (RBMs) as the fast, safe, and efficient machines are used for drilling/excavation of vertical and/or inclined shafts in the underground structures. Proper selection and accurate performance prediction of these machines are very important issues affecting the cost estimation and planning of the projects. However, due to lack of reliable models, performance prediction of these machines has been seriously limited. This study aims at proposing new models for performance prediction of RBMs based on indentation and full-scale linear cutting tests, as well as physical–mechanical property tests, performed on rock samples. Operational and performance data of RBMs and rock samples with different geological origins are collected from different raise boring projects. Then, indentation tests are performed on different core samples by using a real-life tungsten carbide insert of the button cutters used on the analyzed RBMs. Full-scale linear cutting tests were previously performed on the block samples by using a 305 mm (12 in.) diameter button (kerf) cutter. The study indicates that velocity of S-wave and Cerchar abrasivity index can be used to estimate brittleness index of rocks, which is also obtained from the indentation tests. It also shows the feasibility of using power function as relationship between the indentation force and penetration values for performance prediction of RBMs. It is shown that the results of indentation tests can be used in both empirical and deterministic predictions of RBM performance. However, the proposed models should be improved by additional rock types, operational-performance parameters of different RBMs, indentation and cutting test results of different rocks.

Comparison of the Stability and Accuracy of Deterministic Project Cost Prediction Methods in Earned Value Management

Completing a project on time and on budget are essential factors for the success of any project. One technique that allows predicting the final cost of a project is earned value management (EVM). In this technique, different mathematical methods for predicting the final project cost have been proposed over the last 30 years. These formulas make use of activities’ actual costs and durations as the project progresses. EVM is a technique widely used by many project management professionals. However, very few studies have compared the stability and accuracy of the multiple existing methods for predicting the final cost of the project (commonly abbreviated as estimated cost at completion, EAC). This study compares the stability and accuracy of 30 deterministic cost prediction methods (EAC) in EVM. For this purpose, a representative database of 4100 simulated projects of various topological structures is used. Our results suggest that the methods with the simplest mathematical configurations achieve better stability and accuracy performance. Knowing which EVM methods are the most stable and accurate for predicting the final cost of the project will help project practitioners choose the most reliable cost prediction techniques when they are managing their own projects in real contexts.

LSTM, WaveNet, and 2D CNN for nonlinear time history prediction of seismic responses

Predicting the nonlinear time-history responses of civil engineering structures under seismic loading remains an essential task in earthquake engineering. This paper explores the promise of developing three deep learning (DL) models, i.e., long short-term memory (LSTM), WaveNet, and 2D convolution neural network (CNN), to predict seismic response time histories of three benchmark structures, including a single degree-of-freedom (SDOF) system, a steel building frame, and a multi-component bridge structure. LSTM has been previously developed and is utilized to serve as a reference model, while WaveNet and 2D CNN (i.e., it deals with the data in coupled time–frequency dimensions) are newly developed in the current study. One other novel contribution is to replace the final layer of the WaveNet with an LSTM layer, which significantly improves the model performance. Methodological backgrounds of these DL models are introduced, followed by discussions on model architectures and hyperparameters, the list of evaluation metrics, and the ground motion (GM) suite selected for data generation. High-fidelity numerical models are developed for conducting nonlinear time history analyses that generate numerous motion-response pairs for training, validating, and testing the DL models. The LSTM and WaveNet directly use time series GM inputs and seismic response outputs to train the models, whereas the CNN makes inferences on time–frequency spectrogram images converted through the short-time Fourier transform (STFT). These three DL models are investigated by comparing deterministic predictions under one testing GM and probabilistic distributions of six evaluation metrics. The models’ accuracy, efficiency, and robustness are further examined using a sensitivity study under different training samples and model architectures. Research findings from this study provide a sound reference for the community to leverage these three DL models to achieve reliable and efficient time history response predictions, which are crucial for fulfilling cutting-edge research/practical tasks such as regional seismic risk assessment and performance-based seismic design.

Probabilistic Forecasting of Residential Energy Consumption Based on SWT-QRTCN-ADSC-NLSTM Model

Residential electricity consumption forecasting plays a crucial role in the rational allocation of resources reducing energy waste and enhancing the grid-connected operation of power systems. Probabilistic forecasting can provide more comprehensive information for the decision-making and dispatching process by quantifying the uncertainty of electricity load. In this study, we propose a method based on stationary wavelet transform (SWT), quantile regression (QR), Bidirectional nested long short-term memory (BiNLSTM), and Depthwise separable convolution (DSC) combined with attention mechanism for electricity consumption probability prediction methods. First, the data sequence is decomposed using SWT to reduce the complexity of the sequence; then, the combined neural network model with attention is used to obtain the prediction values under different quantile conditions. Finally, the probability density curve of electricity consumption is obtained by combining kernel density estimation (KDE). The model was tested using historical demand-side data from five UK households to achieve energy consumption predictions 5 min in advance. It is demonstrated that the model can achieve both reliable probabilistic prediction and accurate deterministic prediction.

Probing asymmetric uncertain effect on symmetric three-lobe journal bearing

The conventional design process of journal bearings is completely based on deterministic theoretical predictions. The deterministic approach ignores the probabilistic and random nature of input variables and consequently, poses an arbitrary tolerance or safety factor on the performance parameters. However, such arbitrary safety factors result in performance loss when the factor is very high or failure of the system when the factor is low. This paper presents a probabilistic approach for the design of a three-lobe journal bearing considering the uncertainty in geometric parameters such as clearance, preload, and offset factor. The Monte Carlo simulation (MCS) algorithm is used for the propagation of input uncertainties to the systems’ output. To increase the efficiency of computationally expensive MCS, the moving least square (MLS) method is used as a surrogate model. The training data of the surrogate model is obtained by solving the Reynolds equation, at selective design points, using the finite difference method. To explain the probabilistic response of the bearing for various applications and different operating conditions, the results are presented for long, short, and finite bearings at three different supply pressures considering three different eccentricity ratios. The results of the probabilistic analysis show a significant deviation of the system response from its deterministically predicted values and suggest safe design values for the reliable operation of a three-lobe bearing.

Performance of the Canadian Arctic Prediction System during the YOPP Special Observing Periods

ABSTRACT As a contribution to the Year of Polar Prediction (YOPP), Environment and Climate Change Canada (ECCC) developed the Canadian Arctic Prediction System (CAPS), a high-resolution (3-km horizontal grid-spacing) deterministic Numerical Weather Prediction (NWP) system that ran in real-time from February 2018 to November 2021. During YOPP, ECCC was also running two other operational systems that cover the Arctic: the 10-km Regional Deterministic Prediction System (RDPS) and the 25-km Global Deterministic Prediction System (GDPS). The performance of these three systems over the Arctic was monitored and routinely compared during 2018, both subjectively and with objective verification scores. This work provides a description of CAPS and compares the surface variable objective verification for the Canadian deterministic NWP systems operational during YOPP, focusing on the Arctic winter and summer Special Observing Periods (Feb-March and July-Aug-Sept, 2018). CAPS outperforms RDPS and GDPS in predicting near-surface temperature, dew-point temperature, wind and precipitation, in both seasons and domains. All three systems exhibit a diurnal cycle in the near-surface temperature biases, with maxima at night and minima in day-time. In order to mitigate representativeness issues associated with complex topography, model tile temperatures are adjusted to the station elevation by applying a standard atmosphere lapse-rate: especially for the coarse-resolution models, the lapse-rate adjustment reduces the temperature cold biases characterising mountain terrains. Verification of winter precipitation is performed by adjusting solid precipitation measurement errors from the undercatch in windy conditions: the Canadian models’ systematic positive bias, which was artificially inflated by the undercatch, is reduced by the adjustment, to attain neutral bias. These YOPP dedicated intense verification activities have identified some strengths, weaknesses and systematic behaviours of the Canadian deterministic prediction systems at high latitudes: these results can serve as a benchmark, for comparison and further development. Moreover, this YOPP verification exercise has revealed some issues related to the verification of surface variables and has led to the development of better verification practices for the polar regions (and beyond).

Seasonal Variability of the Ocean Circulation in Queen Charlotte Strait, British Columbia

ABSTRACT Queen Charlotte Strait (QCST) is a large marine area separating northern Vancouver Island from the British Columbia mainland. Although it is an important waterway connecting Queen Charlotte Sound with Johnstone Strait and eventually the Strait of Georgia, complex dynamics in QCST, including tides, surface winds, river runoffs, and coastal currents, are yet understudied. In this numerical study, a high-resolution (up to approximately 10 m in the horizontal) model was developed for QCST and adjacent areas to investigate the circulation in the strait, using the unstructured grid, Finite-Volume, primitive equation Community Ocean Model (FVCOM). Two operational larger-scale models force the FVCOM model: the Coastal Ice Ocean Prediction System for the West coast (CIOPS-W) at the open ocean (one-way nesting), and the High-Resolution Deterministic Prediction System (HRDPS) at surface boundaries. Prevailing winds over QCST are southeasterly in winter and northwesterly in summer and the model demonstrated the capacity to reproduce the ocean circulation regimes in both seasons of 2019. Complex empirical orthogonal function (CEOF) and correlation analyses determined that surface winds play a dominant role on the variability of residual flow in QCST in the winter, while river runoffs strongly influence the variability during the summer. The analyses highlight that QCST is a complex estuarine system with significant seasonal variability rather than just a simple conduit between Queen Charlotte Sound and the Strait of Georgia.

Probabilistic prediction model of concrete carbonation depth considering the influence of multiple factors

AbstractIn order to realize the probability prediction of concrete carbonation depth, this paper proposed a probabilistic prediction model of carbonation depth based on 433 sets of existing experimental data. First, considering the influence of water‐binder ratio, fly ash content, stress level, and carbonation time, a deterministic prediction model of carbonation depth was established based on stepwise regression method. Then, based on the established deterministic prediction model, considering the influence of aleatory and epistemic uncertainties, a probabilistic prediction model of carbonation depth was established, and the posterior distributions of the parameters in the probabilistic prediction model were updated iteratively by using Bayesian theory and Markov chain Monte Carlo method, and the influence of standard deviation of prior distribution of the parameters in probabilistic prediction model on their posterior distributions was discussed. Finally, the prediction accuracy of the probabilistic prediction model was verified by comparing with another 127 sets of existing experimental data and the predicted values of the deterministic prediction model, and the effect of the standard deviations of prior distribution of the parameters on the prediction accuracy of the probabilistic prediction model was analyzed. The results showed that the prediction accuracy of the proposed probabilistic prediction model of carbonation depth after three updates was higher than that of the deterministic prediction model. When the standard deviation of prior distributions of the parameters in probabilistic prediction model was 100, the proposed probabilistic prediction model after three updates had the highest prediction accuracy. Under the tensile stress state of concrete, the prediction accuracy of the probabilistic prediction model was improved by 26.76% compared with the deterministic prediction model, and when the concrete under compressive stress state, the prediction accuracy of the probabilistic prediction model was 44.07% higher than that of the deterministic prediction model.

A methodological approach for modeling the spread of disease using geographical discrete-event spatial models

The study of infectious disease models has become increasingly important during the COVID-19 pandemic. The forecasting of disease spread using mathematical models has become a common practice by public health authorities, assisting in creating policies to combat the spread of the virus. Common approaches to the modeling of infectious diseases include compartmental differential equations and cellular automata, both of which do not describe the spatial dynamics of disease spread over unique geographical regions. We introduce a new methodology for modeling disease spread within a pandemic using geographical models. We demonstrate how geography-based Cell-Discrete-Event Systems Specification (DEVS) and the Cadmium JavaScript Object Notation (JSON) library can be used to develop geographical cellular models. We exemplify the use of these methodologies by developing different versions of a compartmental model that considers geographical-level transmission dynamics (e.g. movement restriction or population disobedience to public health guidelines), the effect of asymptomatic population, and vaccination stages with a varying immunity rate. Our approach provides an easily adaptable framework that allows rapid prototyping and modifications. In addition, it offers deterministic predictions for any number of regions simulated simultaneously and can be easily adapted to unique geographical areas. While the baseline model has been calibrated using real data from Ontario, we can update and/or add different infection profiles as soon as new information about the spread of the disease become available.

Omnivision forecasting: Combining satellite and sky images for improved deterministic and probabilistic intra-hour solar energy predictions

Integrating large proportions of intermittent renewable energy sources into electric grids is challenging. A well-established approach aimed at addressing this difficulty involves the anticipation of the upcoming energy supply variability to adapt the response of the grid. In solar energy, short-term changes in electricity production caused by occluding clouds can be predicted at different time scales from all-sky cameras (up to 30-min ahead) and satellite observations (up to 6 h ahead). In this study, we integrate these two complementary points of view on the cloud cover in a single machine learning framework to improve intra-hour (up to 60-min ahead) irradiance forecasting. Both deterministic and probabilistic predictions are evaluated in different weather conditions (clear-sky, cloudy, overcast) and with different input configurations (sky images, satellite observations and/or past irradiance values). Our results show that the hybrid model benefits predictions in clear-sky conditions and improves longer-term forecasting. This study lays the groundwork for future innovative approaches of combining sky images and satellite observations in a single learning framework to advance solar nowcasting.

UAV's air combat decision-making based on deep deterministic policy gradient and prediction

To solve the enemy uncertain manipulation problem during a UAV's autonomous air combat maneuver decision-making, this paper proposes an autonomous air combat maneuver decision-making method that combines target maneuver command prediction with the deep deterministic policy algorithm. The situation data of both sides of air combat are effectively fused and processed, the UAV's six-degree-of-freedom model and maneuver library are built. In air combat, the target generates its corresponding maneuver library instructions through the deep Q network algorithm; at the same time, the UAV on our side gives the target maneuver prediction results through the probabilistic neural network. A deep deterministic policy gradient reinforcement learning method that considers both the situation information of two aircraft and the prediction results of enemy aircraft is proposed, so that the UAV can choose the appropriate maneuver decision according to the current air combat situation. The simulation results show that the method can effectively use the air combat situation information and target maneuver prediction information so that it can improve the effectiveness of the reinforcement learning method for UAV's autonomous air combat decision-making on the premise of ensuring convergence.

Reliability-based assessment of scour at pier in gravel beds

ABSTRACT Estimation of local scour depth at a bridge pier must be accurate and precise for the safety assessment of pier foundation. Many deterministic scour depth prediction models for bridge piers in sands and gravels have already been reported in the literature. However, the various types of uncertainties may be involved in the deterministic scour depth prediction models and thus affect the accuracy of the scour depth. To account for such uncertainties, a reliability analysis method is generally adopted. A sizable amount of laboratory and field data on scour has been procured in the present study to assess the performance of existing scour prediction models for gravel beds and develop a probabilistic scour prediction model using a First-Order Reliability Method (FORM) based on the object-oriented constraints optimization spreadsheet algorithm for the most appropriate selected deterministic scour prediction model. A new relationship of the silt factor has also been developed to estimate the scour depth in gravel bed as per Indian Practice. A safety factor based on the target reliability index has also been proposed in this study.

Spatially correlated Vs30 estimation in the Beijing area

Beijing is an international metropolis, that is also an earthquake-prone city. The aims of this study are detailed quantifying and qualifying soil layer properties for an accurate seismic safety evaluation in the Beijing area. The time average shear-wave velocity in the first 30 m of subsoil, Vs30, is an important site parameter used in site response analysis, site classification, and seismic loss estimation. Mapping of Vs30over a city-scaled region is commonly done through proxy-based methods by correlating Vs30with geological or topographic information. In this paper, a geostatistical-based random field model is presented and applied to mapping Vs30over extended areas. This random field model is then coupled with Monte Carlo simulations to obtain an averaged Vs30map and its associated uncertainties. Unlike the traditional deterministic prediction model, this framework accounts for spatial variations of Vs30values and uncertainties, which makes the prediction more reliable. A total of 388 shear wave velocity measurements in the Beijing area are used to calculate Vs30values, from which the statistical and spatial properties for the random field realizations are inferred. New spatially correlated probabilistic Vs30maps for the Beijing area are then represented, and the effect of the maximum number of previously generated elements to correlate to in estimating Vs30maps is tested.

Multi-step ahead probabilistic forecasting of multiple hydrological variables for multiple stations

The demand for more accurate simulation of physical river basin puts forward higher requirements on the number of hydrological forecast stations, types of hydrological variables, forecast accuracy, forecast period and quantitative uncertainty. Therefore, how to simultaneously obtain multi-step ahead probabilistic forecasting of multiple hydrological variables for multiple stations is a key issue to be solved in this study. Firstly, the background of River Basin Digital Twin is introduced and the new problems faced by hydrological forecasting in this background are analyzed. Then, the input and output of hydrological forecasting are reconstructed into 4-D tensors. Next, a new hybrid deep learning model (B-CM-C3D) based on 3-D Convolutional Neural Network, Convolutional Minimum Gate Memory Neural Network and Variational Bayesian Neural Network is proposed to obtain multi-step ahead probabilistic forecasting of multiple hydrological variables for multiple stations. Finally, in order to verify the performance of this model, it was compared with four state-of-the-art models in the Yangtze River Basin. The experimental results show that: (1) the deterministic prediction accuracy and probabilistic forecasting comprehensive performance of B-CM-C3D are better than other comparison models in 80% of the results. (2) B-CM-C3D shortens the training time of B-CL-C3D by 43% and improves the prediction accuracy.

Dual Accuracy-Quality-Driven Neural Network for Prediction Interval Generation.

Accurate uncertainty quantification is necessary to enhance the reliability of deep learning (DL) models in real-world applications. In the case of regression tasks, prediction intervals (PIs) should be provided along with the deterministic predictions of DL models. Such PIs are useful or "high-quality (HQ)" as long as they are sufficiently narrow and capture most of the probability density. In this article, we present a method to learn PIs for regression-based neural networks (NNs) automatically in addition to the conventional target predictions. In particular, we train two companion NNs: one that uses one output, the target estimate, and another that uses two outputs, the upper and lower bounds of the corresponding PI. Our main contribution is the design of a novel loss function for the PI-generation network that takes into account the output of the target-estimation network and has two optimization objectives: minimizing the mean PI width and ensuring the PI integrity using constraints that maximize the PI probability coverage implicitly. Furthermore, we introduce a self-adaptive coefficient that balances both objectives within the loss function, which alleviates the task of fine-tuning. Experiments using a synthetic dataset, eight benchmark datasets, and a real-world crop yield prediction dataset showed that our method was able to maintain a nominal probability coverage and produce significantly narrower PIs without detriment to its target estimation accuracy when compared to those PIs generated by three state-of-the-art neural-network-based methods. In other words, our method was shown to produce higher quality PIs.

Probability forecasting method of Three Gorges Reservoir inflow flood based on error distribution estimation

准确、及时的入库洪水预报,对三峡水库综合效益的发挥和长江流域水旱灾害防御、水资源利用、流域综合管理等具有重要作用。基于预报误差的最优分布估计和分布函数动态参数假定,提出了一种三峡水库入库洪水概率预报方法,并进行了洪水概率预报业务试验。结果表明:本文所提方法科学可行,计算快捷,使用方便,便于在实时作业预报中应用推广;概率预报结果较确定性预报结果,在水量预报、预警效果等方面均有所改善,1~5 d预见期预报的确定性系数提高0.1%~3.4%,水量误差减少0.1%~4.8%,可为三峡水库实时调度提供更可靠的预报信息;所提出的三峡水库入库洪水概率预报业务化产品,可提供更多风险信息,为三峡水库的科学调度,尤其是洪水资源化利用提供更好的优化决策支撑。;Accurate and timely inflow flood forecasting plays an important role in the comprehensive benefit of the Three Gorges Reservoir, and the prevention of floods and droughts, the utilization of water resources, the comprehensive management for the Yangtze River basin. Based on the optimal distribution estimation of the prediction error and the assumption of the dynamic parameters for the distribution function, a probability prediction method for the inflow flood of the Three Gorges Reservoir is proposed, and the operational test of the probability prediction of the flood is carried out. The results show that the physical rational of the presented method is clear, and it is easy to be applied and popularized in real-time operation forecast. Compared with the deterministic prediction results, the probability forecast has improved in water volume prediction and early warning effect. For example, the certainty coefficient of forecast in 1-5 d is increased by 0.1%-3.4%, and the water volume error is reduced by 0.1%-4.8%, which can provide more reliable forecast information for the real-time operation of the Three Gorges Reservoir. The proposed operational product of probability forecast of inflow flood process can provide more risk information, and provide better support for scientific and fine operation of the Three Gorges Reservoir, especially for optimal decision-making of flood resource utilization.

Research of a combination system based on fuzzy sets and multi-objective marine predator algorithm for point and interval prediction of wind speed.

Short-term wind speed forecasting is fundamental to improving the stability of power grid operation and enhancing its transmission efficiency; thus, it has long been a research hotspot. Nonetheless, quantities of literature in this field only used the single prediction model and overemphasized deterministic prediction, which resulted in deficient forecasting performance. To address these issues, a novel point and interval combination prediction system was developed in this paper. Specifically, wind speed time series were reconstructed by dividing windows and fuzzification to input highly effective data; next, four single prediction models and a multi-objective weight-determining mechanism were integrated to obtain the point prediction results; and their distributions were assessed to implement interval prediction under distinct confidence levels. In the meantime, this study demonstrated that the proposed system reached the Pareto optimal by the theoretical proof, and empirical research was conducted based on 10-min real wind speed data from the wind farm in China. Judging from the experimental results, the combination prediction system was always capable of providing the most satisfactory forecasting performance by contrast with the comparative models. Consequently, it has broad application prospects in guiding the operation of wind farms and optimizing the power grid dispatching.

Computing the ensemble spread from deterministic weather predictions using conditional generative adversarial networks

Computing the ensemble spread from deterministic weather predictions using conditional generative adversarial networks