Prediction intervals for electricity load forecasting using neural networks

Most of the research in time series is concerned with point forecasting. In this paper we focus on interval forecasting and its application for electricity load prediction. We extend the LUBE method, a neural network-based method for computing prediction intervals. The extended method, called LUBEX, includes an advanced feature selector and an ensemble of neural networks. Its performance is evaluated using Australian electricity load data for one year. The results showed that LUBEX is able to generate high quality prediction intervals, using a very small number of previous lag variables and having acceptable training time requirements. The use of ensemble is shown to be critical for the accuracy of the results.

Effects of mechanical deformation on electric performance of rechargeable batteries embedded in load carrying composite structures

Integrating rechargeable battery cells with fibre reinforced polymer matrix composites is a promising technology to enable composite structures to concurrently carry load and store electric energy, thus significantly reducing weight at the system level. To develop a design criterion for structural battery composites, rechargeable lithium polymer battery cells were embedded into carbon fibre/epoxy matrix composite laminates, which were then subjected to tensile, flexural and compressive loading. The electric charging/discharging properties were measured at varying levels of applied loads. The results showed that degradation in battery performance, such as voltagea and energy storage capacity, correlated well with the applied strain under three different loading conditions. Under compressive loading, battery cells, due to their multilayer construction, were unable to prevent buckling of composite face sheets due to the low lateral stiffness, leading to lower compressive strength that sandwich panels with foam core.

Peak oxygen uptake of 12-18-year-old boys living in a densely populated urban environment

The purposes of this study were to provide data on the peak Vo 2 of 12-18-year-old boys from Hong Kong, a densely populated urban environment; to compare these data with those for other similarly aged populations; and to examine the correlations between peak Vo 2 and various anthropometric parameters of this group. A stratified, random sample of 86 ethnic Chinese boys had their peak Vo 2 determined using an on-line gas analysis system during incremental, treadmill running. The mean peak 17o2 of the boys was 2.7 SD 0.44 1- rain - 1 or, when expressed in relation to body mass, 52.0 SD 5-8 ml- kg- 1. min - 1. Peak Vo 2 (1. min - 1) was significantly correlated with body mass (r = 0.72, p < 0.001, age (r = 0.49, p < 0.001) and height (r = 0.71, p = 0-001). Peak Vo 2 (ml 'kg- 1. min- 1) showed no correlation with age or height. These data suggest that this population group has peak Vo 2 values very similar to those observed in boys from most other population groups.

A preliminary study of the 20-m multistage shuttle run as a predictor of peak VO2 in Hong Kong Chinese students

The purpose of the present study was to determine the validity of the 20-meter multistage shuttle run (MSR) for predicting peak VO2 in Hong Kong Chinese students, ages 12-15 years. Fifty-five subjects, 27 boys and 28 girls, performed the MSR in the school environment and had peak VO2 determined in the laboratory. A correlation of 0.72 (p is less than 0.001) was found between peak VO2 and predicted peak VO2 using an equation previously developed with Canadian children (6). However, maximal shuttle run speed alone was a better predictor in this group (r=0.74, SEE=4.6 ml.kg-1.min-1, p is less than 0.001). Multiple-regression analysis (best-subsets) was performed and the best predictor variables were maximal speed and sex with either triceps skinfold or weight. For practical application in the school setting, the equation peak VO2=24.2-5.0(sex) minus 0.8(age) plus 3.4(maximal speed) (r=0.82, SEE=4.0), where for sex, male = 0 and female = 1, is suggested.

Bayesian model averaging of load demand forecasts from neural network models

Creating a set of a number of neural network (NN) models in an ensemble and accumulating them can achieve better overview capability as compared to single neural network. Neural network ensembles are designed to provide solutions to particular problems. Many researchers and academicians have adopted this NN ensemble technique, especially in machine learning, and has been applied in various fields of engineering, medicine and information technology. This paper present a robust aggregation methodology for load demand forecasting based on Bayesian Model Averaging of a set of neural network models in an ensemble. This paper estimate a vector of coefficient for individual NN models' forecasts using validation data-set. These coefficients, also known as weights, are equal to posterior probabilities of the models generating the forecasts. These BMA weights are then used in combining forecasts generated from NN models with test data-set. By comparing the Bayesian results with the Simple Averaging method, it was observed that benefits are obtained by utilizing an advanced method like BMA for forecast combinations.

The impact of working memory load on task execution and online plan adjustment during multitasking in a virtual environment

Three experiments investigated the impact of working memory load on online plan adjustment during a test of multitasking in young, nonexpert, adult participants. Multitasking was assessed using the Edinburgh Virtual Errands Test (EVET). Participants were asked to memorize either good or poor plans for performing multiple errands and were assessed both on task completion and on the extent to which they modified their plans during EVET performance. EVET was performed twice, with and without a secondary task loading a component of working memory. In Experiment 1, articulatory suppression was used to load the phonological loop. In Experiment 2, oral random generation was used to load executive functions. In Experiment 3, spatial working memory was loaded with an auditory spatial localization task. EVET performance for both good- and poor-planning groups was disrupted by random generation and sound localization, but not by articulatory suppression. Additionally, people given a poor plan were able to overcome this initial disadvantage by modifying their plans online. It was concluded that, in addition to executive functions, multiple errands performance draws heavily on spatial, but not verbal, working memory resources but can be successfully completed on the basis of modifying plans online, despite a secondary task load.

Validity of an upper-body-mounted accelerometer to measure peak vertical and resultant force during running and change-of-direction tasks

This study assessed the validity of a tri-axial accelerometer worn on the upper body to estimate peak forces during running and change-of-direction tasks. Seventeen participants completed four different running and change-of-direction tasks (0°, 45°, 90°, and 180°; five trials per condition). Peak crania-caudal and resultant acceleration was converted to force and compared against peak force plate ground reaction force (GRF) in two formats (raw and smoothed). The resultant smoothed (10 Hz) and crania-caudal raw (except 180°) accelerometer values were not significantly different to resultant and vertical GRF for all running and change-of-direction tasks, respectively. Resultant accelerometer measures showed no to strong significant correlations (r = 0.00–0.76) and moderate to large measurement errors (coefficient of variation [CV] = 11.7–23.9%). Crania-caudal accelerometer measures showed small to moderate correlations (r = − 0.26 to 0.39) and moderate to large measurement errors (CV = 15.0–20.6%). Accelerometers, within integrated micro-technology tracking devices and worn on the upper body, can provide a relative measure of peak impact force experienced during running and two change-of-direction tasks (45° and 90°) provided that resultant smoothed values are used.

Improving peak shapes with counter gradients in two-dimensional high performance liquid chromatography

To achieve the greatest peak capacity in two-dimensional high performance liquid chromatography (2D-HPLC) a gradient should be operated in both separation dimensions. However, it is known that when an injection solvent that is stronger than the initial mobile phase composition is deleterious to peak performance, thus causing problems when cutting a portion from one gradient into another. This was overcome when coupling hydrophilic interaction with reversed phase chromatography by introducing a counter gradient that changed the solvent strength of the second dimension injection. It was found that an injection solvent composition of 20% acetonitrile in water gave acceptable results in one-dimensional simulations with an initial composition of 5% acetonitrile. When this was transferred to a 2D-HPLC separation of standards it was found that a marked improvement in peak shape was gained for the moderately retained analytes (phenol and dimethyl phthalate), some improvement for the weakly retained caffeine and very little change for the strongly retained n-propylbenzene and anthracene which already displayed good chromatographic profiles. This effect was transferred when applied to a 2D-HPLC separation of a coffee extract where the indecipherable retention profile was transformed to a successful application multidimensional chromatography with peaks occupying 71% of the separation space according to the geometric approach to factor analysis.

Removing the ambiguity of data processing methods: Optimizing the location of peak boundaries for accurate moment calculations

The calculation of the first few moments of elution peaks is necessary to determine: the amount of component in the sample (peak area or zeroth moment), the retention factor (first moment), and the column efficiency (second moment). It is a time consuming and tedious task for the analyst to perform these calculations, thus data analysis is generally completed by data stations associated to modern chromatographs. However, data acquisition software is a black box which provides no information to chromatographers on how their data are treated. These results are too important to be accepted on blind faith. The location of the peak integration boundaries is most important. In this manuscript, we explore the relationships between the size of the integration area, the relative position of the peak maximum within this area, and the accuracy of the calculated moments. We found that relationships between these parameters do exist and that computers can be programmed with relatively simple routines to automatize the extraction of key peak parameters and to select acceptable integration boundaries. It was also found that the most accurate results are obtained when the S/N exceeds 200.

Short-term load and wind power forecasting using neural network-based prediction intervals

Electrical power systems are evolving from today's centralized bulk systems to more decentralized systems. Penetrations of renewable energies, such as wind and solar power, significantly increase the level of uncertainty in power systems. Accurate load forecasting becomes more complex, yet more important for management of power systems. Traditional methods for generating point forecasts of load demands cannot properly handle uncertainties in system operations. To quantify potential uncertainties associated with forecasts, this paper implements a neural network (NN)-based method for the construction of prediction intervals (PIs). A newly introduced method, called lower upper bound estimation (LUBE), is applied and extended to develop PIs using NN models. A new problem formulation is proposed, which translates the primary multiobjective problem into a constrained single-objective problem. Compared with the cost function, this new formulation is closer to the primary problem and has fewer parameters. Particle swarm optimization (PSO) integrated with the mutation operator is used to solve the problem. Electrical demands from Singapore and New South Wales (Australia), as well as wind power generation from Capital Wind Farm, are used to validate the PSO-based LUBE method. Comparative results show that the proposed method can construct higher quality PIs for load and wind power generation forecasts in a short time.