Search

As more and more photovoltaic (PV) systems are integrated into the grid, the intelligent operation of the grid system is facing significant challenges. Therefore, accurately forecasting PV power output at various time scales is particularly urgent. To meet this demand, this paper proposes an LSTM-Informer model based on an improved Stacking ensemble algorithm (ISt-LSTM-Informer). The proposed model improves the k-fold cross validation in the traditional Stacking algorithm to a time-series cross validation for integrating time-series forecasting models. Simultaneously, it utilizes long short-term memory (LSTM) and Informer as the base models. By integrating the advantages of the two base models, the ISt-LSTM-Informer achieves accurate short and medium-term PV power forecasting. To validate the effectiveness of the model, a historical dataset from a PV system located in Uluru, Australia, is used for various types of experiments. Among them, comparative experiments validate the superiority of the model. Compared with five other methods, the ISt-LSTM-Informer obtains 21 optimal results for the four evaluation metrics of RMSE, MAE, MAPE, and R 2 across eight forecasting time scales. In addition, different combinations of base models are conducted to verify the advantages of the Stacking ensemble algorithm and the two base models, respectively. • A novel multi-timescale photovoltaic power forecasting model is proposed. • Time-series cross validation is introduced into the Stacking algorithm. • LSTM and Informer are utilized as the base models of the Stacking algorithm. • Various methods are compared to verify the proposed model's effectiveness. • The model's predictive accuracy is illustrated through various visualization methods. [ABSTRACT FROM AUTHOR]

An accurate predictive model is essential for monitoring the energy produced by a solar system based on different meteorological parameters. In the present paper, a novel machine-learning model named DNN-RODDPSO is proposed to improve the real-time prediction accuracy of the hourly energy produced by four dual-axis solar trackers. This model integrates a new deep neural network (DNN) model with a recent variant of PSO algorithm referred to as a randomly occurring distributed delayed particle swarm optimization (RODDPSO) algorithm. This algorithm is adopted to enhance the training process of the DNN model by reducing the risk of being trapped into local optima and for the search space diversification. Furthermore, to develop the DNN-RODDPSO model, the hourly observations of seven meteorological parameters including time variable measured during 2014–2015 in Alice Springs city, Australia, are used. This model integrates two novel hidden layers, the first one is a selective layer based on daytime/nighttime data selection. The second one named automatic inputs relevance determination to point out the most relevant inputs for an accurate prediction. The obtained results demonstrate the high performance of the two novel hidden layers and the RODDPSO algorithm to improve significantly the prediction accuracy compared to the actual literature standards. • A novel hybrid ANN based on RODDPSO algorithm is developed. • A novel selective layer based on daytime/nighttime data selection is introduced. • A novel automatic inputs relevance determination layer is proposed and validated. • The energy potential produced by four solar trackers is monitored. [ABSTRACT FROM AUTHOR]

Modelling workflow for BRF-LSTM and BRF-GRU predictive models. [Display omitted] • Deep learning hybrid predictive model (BRF-LSTM) is built for streamflow forecasting. • Climate indices, rainfall and periodicity are factored for accurate predictions. • Adopts feature algorithm (i.e. Boruta) with long short-term memory and gated recurrent unit. • Six gauging sites tested; BRF-LSTM yields > 98% errors in ± 0.015 m, ~1.30% relative error. • The proposed model is useful for hydrological and strategic water resources planning. Long-term forecasting of any hydrologic phenomena is essential for strategic environmental planning, hydrologic and other forms of structural design, agriculture, and water resources management. Climate mode indices, utilising machine learning methods, are frequently considered as predictor variables in order to forecast several different hydrological variables. In this study, a feature selection algorithm based on two different deep learning models, i.e., long short-term memory and a gated recurrent unit, is applied to improve the forecasting capability of streamflow water levels at six gauging stations in the Murray Darling Basin of Australia. This paper therefore aggregates the significant antecedent lag memory of climate mode indices, rainfall, and the monthly factor based on the periodicity as the predictor variables to attain significantly accurate stream water level forecasts. This novel method identifies an improved relationship between the stream water level and climate mode indices through the aggregation of the significant lagged datasets capturing the historical features to predict the future streamflow water level. The boruta feature selection algorithm (BRF) was then applied in a two phase process before and after attaining the significant lagged inputs to screen the optimum predictor variables. The merits of the forecast models were evaluated through different performance evaluation criteria. The results show that the accumulated significant lagged inputs based on climate mode indices, along with the rainfall and periodicity factors are seen to provide improved forecasting of the SWL over the non-BRF deep learning approaches where no prior feature selection was applied. The hybrid LSTM method (i.e., BRF-LSTM model) achieved a unique advantage in terms of SWL forecasting, particularly attaining over 98% of the predictive errors lying within a band of +/-0.015 m with relatively low relative errors (RRMSE ≈1.30% and RMAE ≈ 0.882%), outperforming all of the benchmark models. It is also found that the periodicity factor has a potential influence on the accuracy of the forecast models for the four monitored study stations. This study concludes that the newly developed hybrid deep learning approaches, coupled with the BRF feature selection, provide improved forecasting performance. The hybrid approach developed in this paper can therefore be used to provide a strong provide predictive response algorithm for the hydrological variables that were influenced by the low-frequency variability of the climate model indices in respect to streamflow water level. [ABSTRACT FROM AUTHOR]

In contrast to the recent proliferation of studies incorporating ordinal methods to generate health state values from adults, to date relatively few studies have utilised ordinal methods to generate health state values from adolescents. This paper reports upon a study to apply profile case best worst scaling methods to derive a new adolescent specific scoring algorithm for the Child Health Utility 9D (CHU9D), a generic preference based instrument that has been specifically designed for the estimation of quality adjusted life years for the economic evaluation of health care treatment and preventive programs targeted at young people. A survey was developed for administration in an on-line format in which consenting community based Australian adolescents aged 11–17 years (N = 1982) indicated the best and worst features of a series of 10 health states derived from the CHU9D descriptive system. The data were analyzed using latent class conditional logit models to estimate values (part worth utilities) for each level of the nine attributes relating to the CHU9D. A marginal utility matrix was then estimated to generate an adolescent-specific scoring algorithm on the full health = 1 and dead = 0 scale required for the calculation of QALYs. It was evident that different decision processes were being used in the best and worst choices. Whilst respondents appeared readily able to choose ‘best’ attribute levels for the CHU9D health states, a large amount of random variability and indeed different decision rules were evident for the choice of ‘worst’ attribute levels, to the extent that the best and worst data should not be pooled from the statistical perspective. The optimal adolescent-specific scoring algorithm was therefore derived using data obtained from the best choices only. The study provides important insights into the use of profile case best worst scaling methods to generate health state values with adolescent populations. [ABSTRACT FROM AUTHOR]

The prediction of future drought is an effective mitigation tool for assessing adverse consequences of drought events on vital water resources, agriculture, ecosystems and hydrology. Data-driven model predictions using machine learning algorithms are promising tenets for these purposes as they require less developmental time, minimal inputs and are relatively less complex than the dynamic or physical model. This paper authenticates a computationally simple, fast and efficient non-linear algorithm known as extreme learning machine (ELM) for the prediction of Effective Drought Index (EDI) in eastern Australia using input data trained from 1957–2008 and the monthly EDI predicted over the period 2009–2011. The predictive variables for the ELM model were the rainfall and mean, minimum and maximum air temperatures, supplemented by the large-scale climate mode indices of interest as regression covariates, namely the Southern Oscillation Index, Pacific Decadal Oscillation, Southern Annular Mode and the Indian Ocean Dipole moment. To demonstrate the effectiveness of the proposed data-driven model a performance comparison in terms of the prediction capabilities and learning speeds was conducted between the proposed ELM algorithm and the conventional artificial neural network (ANN) algorithm trained with Levenberg–Marquardt back propagation. The prediction metrics certified an excellent performance of the ELM over the ANN model for the overall test sites, thus yielding Mean Absolute Errors, Root-Mean Square Errors, Coefficients of Determination and Willmott's Indices of Agreement of 0.277, 0.008, 0.892 and 0.93 (for ELM) and 0.602, 0.172, 0.578 and 0.92 (for ANN) models. Moreover, the ELM model was executed with learning speed 32 times faster and training speed 6.1 times faster than the ANN model. An improvement in the prediction capability of the drought duration and severity by the ELM model was achieved. Based on these results we aver that out of the two machine learning algorithms tested, the ELM was the more expeditious tool for prediction of drought and its related properties. [ABSTRACT FROM AUTHOR]

Abstract: Fire danger indices are used by fire management agencies to assess fire weather conditions and issue public warnings. The most widely used fire danger indices in Australia are the McArthur Fire Forest Danger Index and the Grassland Fire Danger Index. These indices are calculated at weather stations using measurements of weather variables and fuel information. For a vast country like Australia when assessing the risk of severe fire weather events, it is also important to calculate the spatial distribution of these indices considering the extreme tail of the distribution. The spatial distribution of one of the fire weather danger indices regularly used in Australia is presented in this paper. In particular, we present the spatial distribution of the long-term tendency of extreme values of the McArthur Forest Fire Danger Index (FFDI). This indicator of fire weather conditions was assessed by calculating the return period of its extreme values by fitting extreme value distributions to data sets of FFDI at 78 recording stations around Australia. The spatial distribution of these return periods was obtained by using spatial interpolation algorithms with the recording stations measurements. Two conventional and two new algorithms based on machine-learning techniques were tested. This study shows that the best interpolation results for the FFDI can be obtained by using a combination of random forest and inverse distance weighting interpolation algorithms. The spatial distribution of the seasonal FFDI return period shows that the highest FFDI over large parts of southern Australia occurs during the summer months whilst in northern Australia it occurs in spring. The results also show that the FFDI in eastern Australia, the most populated region of the country, is higher inland than in the coastal areas particularly during spring and summer. [Copyright &y& Elsevier]

Summary: This paper describes an algorithm for disaggregating daily rainfall into sub-daily rainfall ‘fragments’ (fine-resolution rainfall sequences) under a future, warmer climate. The algorithm uses a combined generalised additive model (GAM) and method of fragments (MoFs) framework to resample sub-daily rainfall fragments from the historical record conditional on daily rainfall amount and a range of atmospheric covariates. The rationale is that as the atmosphere warms, future rainfall patterns will be more reflective of historical rainfall patterns corresponding to warmer days at the same location, or to locations which have an atmospheric profile more reflective of expected future climate. It was found that the daily to sub-daily scaling relationship varied significantly by season and by location, with rainfall patterns on warmer seasons or at warmer locations typically showing more intense rainfall occurring over shorter periods compared with cooler seasons and stations. Importantly, by regressing against atmospheric covariates such as temperature, this effect was substantially reduced, suggesting that the approach may also be valid when extrapolating to a future climate. The GAM–MoF algorithm was then applied to nine stations around Australia, with the results showing that relative to the daily rainfall amount, the maximum intensity of short duration rainfall increased by between 4.1% and 13.4% per degree change in temperature for the maximum six minute burst, and by between 3.1% and 6.8% for the maximum 1h burst. The fraction of each wet day with no rainfall also increased by between 1.5% and 3.5%. This highlights that a significant proportion of the change to the distribution of rainfall is likely to occur at sub-daily timescales, with important implications for many hydrological systems. [Copyright &y& Elsevier]

Abstract: Fire detection using satellites is an important source of information for fire management, ecological studies and emission estimates. However, little is known about the minimum sizes of fires that are being detected. This paper presents an approach using fire radiative power estimated from MODIS satellite data to determine the detection threshold for fire-prone savannas in Northern Australia. The results indicate that fires with an active flaming area 100–300m2 can be detected in the study region. It is also shown that the algorithm is slightly more sensitive at night. As expected the detection threshold shows strong view angle dependence. While this study has been undertaken in the savannas of Northern Australia, the results should be transferable to other savanna regions worldwide and other areas where fires are not obscured by a dense tree canopy. [Copyright &y& Elsevier]

Abstract: This paper provides a novel Exceptional Object Analysis for Finding Rare Environmental Events (EOAFREE). The major contribution of our EOAFREE method is that it proposes a general Improved Exceptional Object Analysis based on Noises (IEOAN) algorithm to efficiently detect and rank exceptional objects. Our IEOAN algorithm is more general than already known outlier detection algorithms to find exceptional objects that may be not on the border; and experimental study shows that our IEOAN algorithm is far more efficient than directly recursively using already known clustering algorithms that may not force every data instance to belong to a cluster to detect rare events. Another contribution is that it provides an approach to preprocess heterogeneous real world data through exploring domain knowledge, based on which it defines changes instead of the water data value itself as the input of the IEOAN algorithm to remove the geographical differences between any two sites and the temporal differences between any two years. The effectiveness of our EOAFREE method is demonstrated by a real world application – that is, to detect water pollution events from the water quality datasets of 93 sites distributed in 10 river basins in Victoria, Australia between 1975 and 2010. [Copyright &y& Elsevier]

Abstract: The prospect that decision support systems (DSS) can help farmers adjust their management to suit seasonal conditions by putting scientific knowledge and rational risk management algorithms at farmers’ fingertips continues to challenge the science and extension community. A number of reviews of agricultural DSS have called for a re-appraisal of the field and for the need to reflect on past mistakes and to learn from social and management theory. The objective of this paper was to investigate whether there is an emerging consensus, among stakeholders in DSS for Australian agriculture, about the lessons learned from past experience with DSS tools. This investigation was conducted in three parts. The first part was a distillation of suggestions for best practice from the relevant literature. The second part was a reflection on what the champions of five current DSS development and delivery efforts in Australia learned from their recent efforts. The third part tested the level of support for the combined findings from the first and second approaches by surveying 23 stakeholders in the research, development, delivery and funding of DSS. The key propositions relating to best practice that were supported by the survey, listed according to the strength of support, were: 1. It is essential to have a plan for delivery of the DSS beyond the initial funding period. 2. DSS need to be embedded in a support network consisting of farmers, consultants and researchers. 3. DSS development requires the commitment of a critical mass of appropriately skilled people. 4. A DSS should aim to educate farmers’ intuition rather than replace it with optimised recommendations. 5. A DSS should enable users to experiment with options that satisfy their needs rather than attempt to present ‘optimised’ solutions. 6. DSS tools stand on the quality and authority of their underlying science and require ongoing improvement, testing and validation. 7. DSS development should not commence unless it is backed by marketing information and a plan for delivery of the DSS beyond the initial funding period. While the DSS stakeholders supported the proposition that it is essential to have a plan for delivery of a DSS beyond the funding period, the majority resisted the notion of DSS development being market-driven and especially commercial delivery of DSS. We argue that since public funding of the delivery of DSS for farmers’ management of climate risk is highly unlikely, reaping the benefits of lessons learned from past efforts will require that DSS stakeholders change their perception of the commercial delivery model or find an alternative way to fund the delivery of DSS beyond the R&D phase. [Copyright &y& Elsevier]

Abstract: Accurate high-resolution soil moisture data are needed for a range of agricultural and hydrologic activities. To improve the spatial resolution of ∼40km resolution passive microwave-derived soil moisture, a methodology based on 1km resolution MODIS (MODerate resolution Imaging Spectroradiometer) red, near-infrared and thermal-infrared data has been implemented at 4km resolution. The three components of that method are (i) fractional vegetation cover, (ii) soil evaporative efficiency (defined as the ratio of actual to potential evaporation) and (iii) a downscaling relationship. In this paper, 36 different disaggregation algorithms are built from 3 fractional vegetation cover formulations, 3 soil evaporative efficiency models, and 4 downscaling relationships. All algorithms differ with regard to the representation of the nonlinear relationship between microwave-derived soil moisture and optical-derived soil evaporative efficiency. Airborne L-band data collected over an Australian agricultural area are used to both generate ∼40km resolution microwave pixels and verify disaggregation results at 4km resolution. Among the 36 disaggregation algorithms, one is identified as being more robust (insensitive to soil, vegetation and atmospheric variables) than the others with a mean slope between MODIS-disaggregated and L-band derived soil moisture of 0.94. The robustness of that algorithm is notably assessed by comparing the disaggregation results obtained using composited (averaged) Terra and Aqua MODIS data, and using data from Terra and Aqua separately. The error on disaggregated soil moisture is systematically reduced by compositing daily Terra and Aqua data with an error of 0.012vol./vol. [ABSTRACT FROM AUTHOR]

Abstract: With the introduction of restructuring into the electric power industry, the price of electricity has become the focus of all activities in the power market. Electricity price forecast is key information for electricity market managers and participants. However, electricity price is a complex signal due to its non-linear, non-stationary, and time variant behavior. In spite of performed research in this area, more accurate and robust price forecast methods are still required. In this paper, a new forecast strategy is proposed for day-ahead price forecasting of electricity markets. Our forecast strategy is composed of a new two stage feature selection technique and cascaded neural networks. The proposed feature selection technique comprises modified Relief algorithm for the first stage and correlation analysis for the second stage. The modified Relief algorithm selects candidate inputs with maximum relevancy with the target variable. Then among the selected candidates, the correlation analysis eliminates redundant inputs. Selected features by the two stage feature selection technique are used for the forecast engine, which is composed of 24 consecutive forecasters. Each of these 24 forecasters is a neural network allocated to predict the price of 1h of the next day. The whole proposed forecast strategy is examined on the Spanish and Australia’s National Electricity Markets Management Company (NEMMCO) and compared with some of the most recent price forecast methods. [Copyright &y& Elsevier]

Monitoring of surface soil moisture (SM) through microwave radiometry typically relies on the inversion of a radiative transfer model. Conventional inversion algorithms require proper calibration of surface roughness and radiometric parameters of the overlying canopy, including the scattering albedo and optical depth. However, uncertainty in global characterization of these parameters is one of the main sources of error in satellite SM retrievals. To cope with this uncertainty, this paper presents a new algorithm, called "temporal polarization ratio algorithm" (TPRA), that enables retrieval of SM independent of surface roughness and vegetation parameters. This approach uses the temporal differences of polarized emissivity observations assuming that the surface roughness and vegetation parameters are invariant over a window of time. Unlike the classical dual channel algorithms (DCA), TPRA is not only free of surface calibration but also robust to systematical errors arising from surface soil temperature errors. One caveat is that the algorithm is unable to retrieve SM when surface emissivity does not change appreciably in time. The performance of TPRA is evaluated through several controlled numerical experiments and validated using the Soil Moisture Active Passive (SMAP) satellite retrievals as well as in-situ SM measurements over Australia. The results show improved SM retrievals compared with the DCA and SMAP official products as long as the surface emissivity changes sufficiently over the retrieval time period. • A new algorithm is designed for the retrieval of soil moisture at L-band. • Soil moisture retrievals are independent of surface roughness. • Soil moisture retrievals are independent of vegetation parameters. • Soil moisture retrievals are robust to systematical errors in soil temperature. [ABSTRACT FROM AUTHOR]

Study Objective: Continuous Positive Airway Pressure (CPAP) is considered to be the gold standard treatment for obstructive sleep apnoea (OSA). CPAP monitoring systems allow tracking of patient CPAP adherence and treatment efficacy, by measuring residual sleep-disordered breathing, hours of CPAP use, and mask leak etc. The American Thoracic Society (ATS) published a position paper in 2013 highlighting issues of interpreting CPAP data such as a lack of consistency between CPAP manufacturers data algorithms, legal implications of CPAP data and implications for CPAP adherence. This paper extends on this work by investigating these issues in an Australasian context.Method: A review of current literature on CPAP monitoring systems, privacy and security of CPAP data for major Australasian CPAP providers, and CPAP adherence was undertaken. A legal review was also commissioned for issues related to privacy and security of CPAP data.Results: CPAP manufacturers' utilize different algorithms for respiratory event detection and clinicians need to be aware the implications for interpreting CPAP data. Australasian CPAP manufacturers have created security/privacy policies with the intent to follow relevant legislation to protect patients' CPAP data, however they do need to be constantly reviewed and updated to avoid data breaches and changes to agreements. No guarantees can be provided by the Australasian Sleep Association on CPAP manufacturers' compliance with these policies and there is the potential for some degree of liability for physicians and CPAP providers associated with CPAP data. Lastly, providing patients with feedback on their CPAP usage and OSA management appears to have positive influence CPAP adherence.Conclusions: CPAP data provides many opportunities to increase OSA patient care and to help patients self-manage this chronic condition. However, issues relating to lack of standardization of CPAP parameters, privacy, security, and legal implications will need to be managed in this changing technologic and clinical environment. [ABSTRACT FROM AUTHOR]

This paper aims to assess the performance of generalized aboveground biomass (AGB) allometric models in the Australian context by investigating the correlation between the AGB estimates and the lidar-based individual tree parameters. A hybrid tree segmentation algorithm is proposed to segment an airborne lidar point cloud into individual trees. Although the diameter at breast height (DBH) of a tree is a crucial parameter for the AGB estimation, a typical airborne lidar data contains only a few points representing partial DBH, hence a localized DBH regression model is proposed. Principal component analysis is applied to examine the multicollinearity in the input variables and ridge regression is applied to remove the less important variables. Four machine learning techniques, namely random forest, support vector regression, multilayer perceptron and radial basis function, are applied to generate AGB regression models. The qualities of the calibrated AGB models are assessed by calculating the adjusted-coefficient-of-determination, leave-one-out cross-validation, the Akaike information criterion, normalized-mean-square-error and the model efficiency index. The test results indicate that the random forest-based AGB model outperforms other machine learning techniques. It is concluded that, if the environmental conditions of tree samples resemble the study region, the generalized AGB allometric model would perform well with the tree samples regardless of their geographical context. [ABSTRACT FROM AUTHOR]