Search

Estimating depth from RGB images is a long-standing ill-posed problem, which has been explored for decades by the computer vision, graphics, and machine learning communities. Among the existing techniques, stereo matching remains one of the most widely used in the literature due to its strong connection to the human binocular system. Traditionally, stereo-based depth estimation has been addressed through matching hand-crafted features across multiple images. Despite the extensive amount of research, these traditional techniques still suffer in the presence of highly textured areas, large uniform regions, and occlusions. Motivated by their growing success in solving various 2D and 3D vision problems, deep learning for stereo-based depth estimation has attracted a growing interest from the community, with more than 150 papers published in this area between 2014 and 2019. This new generation of methods has demonstrated a significant leap in performance, enabling applications such as autonomous driving and augmented reality. In this paper, we provide a comprehensive survey of this new and continuously growing field of research, summarize the most commonly used pipelines, and discuss their benefits and limitations. In retrospect of what has been achieved so far, we also conjecture what the future may hold for deep learning-based stereo for depth estimation research. [ABSTRACT FROM AUTHOR]

Objective: To determine the incidence of patients presenting in pain to a large Australian inner-city emergency department (ED) using a clinical text deep learning algorithm., Materials and Methods: A fine-tuned, domain-specific, transformer-based clinical text deep learning model was used to interpret free-text nursing assessments in the electronic medical records of 235,789 adult presentations to the ED over a three-year period. The model classified presentations according to whether the patient had pain on arrival at the ED. Interrupted time series analysis was used to determine the incidence of pain in patients on arrival over time. We described the changes in the population characteristics and incidence of patients with pain on arrival occurring with the start of the Covid-19 pandemic., Results: 55.16% (95%CI 54.95%-55.36%) of all patients presenting to this ED had pain on arrival. There were differences in demographics and arrival and departure patterns between patients with and without pain. The Covid-19 pandemic initially precipitated a decrease followed by a sharp, sustained rise in pain on arrival, with concurrent changes to the population arriving in pain and their treatment., Discussion: Applying a clinical text deep learning model has successfully identified the incidence of pain on arrival. It represents an automated, reproducible mechanism to identify pain from routinely collected medical records. The description of this population and their treatment forms the basis of intervention to improve care for patients with pain. The combination of the clinical text deep learning models and interrupted time series analysis has reported on the effects of the Covid-19 pandemic on pain care in the ED, outlining a methodology to assess the impact of significant events or interventions on pain care in the ED., Conclusion: Applying a novel deep learning approach to identifying pain guides methodological approaches to evaluating pain care interventions in the ED, giving previously unavailable population-level insights., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 The Authors. Published by Elsevier B.V. All rights reserved.)

A large proportion of photovoltaic (PV) power generation is connected to the power grid, and its volatility and stochasticity have significant impacts on the power system. Accurate PV power forecasting is of great significance in optimizing the safe operation of the power grid and power market transactions. In this paper, a novel dual-channel PV power forecasting method based on a temporal convolutional network (TCN) is proposed. The method deeply integrates the PV station feature data with the model computing mechanism through the dual-channel model architecture; utilizes the combination of multihead attention (MHA) and TCN to extract the multidimensional spatio-temporal features between other meteorological variables and the PV power; and utilizes a single TCN to fully extract the temporal constraints of the power sequence elements. The weighted fusion of the dual-channel feature data ultimately yields the ideal forecasting results. The experimental data in this study are from a 26.52 kW PV power plant in central Australia. The experiments were carried out over seven different input window widths, and the two models that currently show superior performance within the field of PV power forecasting: the convolutional neural network (CNN), and the convolutional neural network combined with a long and short-term memory network (CNN_LSTM), are used as the baseline models. The experimental results show that the proposed model and the baseline models both obtained the best forecasting performance over a 1-day input window width, while the proposed model exhibited superior forecasting performance compared to the baseline model. It also shows that designing model architectures that deeply integrate the data input method with the model mechanism has research potential in the field of PV power forecasting. [ABSTRACT FROM AUTHOR]

This paper introduces an innovative framework for wind power prediction that focuses on the future of energy forecasting utilizing intelligent deep learning and strategic feature engineering. This research investigates the application of a state-of-the-art deep learning model for wind energy prediction to make extremely short-term forecasts using real-time data on wind generation from New South Wales, Australia. In contrast with typical approaches to wind energy forecasting, this model relies entirely on historical data and strategic feature engineering to make predictions, rather than relying on meteorological parameters. A hybrid feature engineering strategy that integrates features from several feature generation techniques to obtain the optimal input parameters is a significant contribution to this work. The model's performance is assessed using key metrics, yielding optimal results with a Mean Absolute Error (MAE) of 8.76, Mean Squared Error (MSE) of 139.49, Root Mean Squared Error (RMSE) of 11.81, R-squared score of 0.997, and Mean Absolute Percentage Error (MAPE) of 4.85%. Additionally, the proposed framework outperforms six other deep learning and hybrid deep learning models in terms of wind energy prediction accuracy. These findings highlight the importance of advanced data analysis for feature generation in data processing, pointing to its key role in boosting the precision of forecasting applications. [ABSTRACT FROM AUTHOR]

Simple Summary: Breast cancer is one of the leading causes of cancer-related death in women. The early detection of breast cancer with screening mammograms plays a pivotal role in reducing mortality rates. Although population-based double-reading screening mammograms have reduced mortality by over 31% in women with breast cancer in Europe, continuing this program is difficult due to the shortage of radiologists. Artificial intelligence (AI) is an emerging technology that has provided promising results in medical imaging for disease detection. This study investigates the performance of AI models on an Australian mammographic database, demonstrating how transfer learning from a USA mammographic database to an Australian one, contrast enhancement on mammographic images and the quality of training data according to radiologists' concordance can improve breast cancer diagnosis. Our proposed methodology offers a more efficacious approach for AI to contribute to radiologists' decision making when interpreting mammography images. This paper investigates the adaptability of four state-of-the-art artificial intelligence (AI) models to the Australian mammographic context through transfer learning, explores the impact of image enhancement on model performance and analyses the relationship between AI outputs and histopathological features for clinical relevance and accuracy assessment. A total of 1712 screening mammograms (n = 856 cancer cases and n = 856 matched normal cases) were used in this study. The 856 cases with cancer lesions were annotated by two expert radiologists and the level of concordance between their annotations was used to establish two sets: a 'high-concordances subset' with 99% agreement of cancer location and an 'entire dataset' with all cases included. The area under the receiver operating characteristic curve (AUC) was used to evaluate the performance of Globally aware Multiple Instance Classifier (GMIC), Global-Local Activation Maps (GLAM), I&H and End2End AI models, both in the pretrained and transfer learning modes, with and without applying the Contrast Limited Adaptive Histogram Equalization (CLAHE) algorithm. The four AI models with and without transfer learning in the high-concordance subset outperformed those in the entire dataset. Applying the CLAHE algorithm to mammograms improved the performance of the AI models. In the high-concordance subset with the transfer learning and CLAHE algorithm applied, the AUC of the GMIC model was highest (0.912), followed by the GLAM model (0.909), I&H (0.893) and End2End (0.875). There were significant differences (p < 0.05) in the performances of the four AI models between the high-concordance subset and the entire dataset. The AI models demonstrated significant differences in malignancy probability concerning different tumour size categories in mammograms. The performance of AI models was affected by several factors such as concordance classification, image enhancement and transfer learning. Mammograms with a strong concordance with radiologists' annotations, applying image enhancement and transfer learning could enhance the accuracy of AI models. [ABSTRACT FROM AUTHOR]

This paper presents the development and evaluation of short-term traffic prediction models using unidirectional and bidirectional deep learning long short-term memory (LSTM) neural networks. The unidirectional LSTM (Uni-LSTM) model provides high performance through its ability to recognize longer sequences of traffic time series data. In this work, Uni-LSTM is extended to bidirectional LSTM (BiLSTM) networks which train the input data twice through forward and backward directions. The paper presents a comparative evaluation of the two models for short-term speed and traffic flow prediction using a common dataset of field observations collected from multiple freeways in Australia. The results showed BiLSTM performed better for variable prediction horizons for both speed and flow. Stacked and mixed Uni-LSTM and BiLSTM models were also investigated for 15-minute prediction horizons resulting in improved accuracy when using 4-layer BiLSTM networks. The optimized 4-layer BiLSTM model was then calibrated and validated for multiple prediction horizons using data from three different freeways. The validation results showed a high degree of prediction accuracy exceeding 90% for speeds up to 60-minute prediction horizons. For flow, the model achieved accuracies above 90% for 5- and 10-minute prediction horizons and more than 80% accuracy for 15- and 30-minute prediction horizons. These findings extend the set of AI models available for road operators and provide them with confidence in applying robust models that have been tested and evaluated on different freeways in Australia. [ABSTRACT FROM AUTHOR]

This paper proposes a novel data driven framework for detecting topology transitions in a distribution network. The framework analyzes data from phasor measurement units (PMUs) and relies on the fact that changes in network topology results in changes in the structure and admittance of the network. Using voltage and current phasors recorded by PMUs, the proposed method approximates network parameters using an ensemble-based deep learning model and thus, it does not require any knowledge of network parameters and load models. Using the prediction error of the proposed model, a connectivity matrix which shows the status of switches is constructed. In contrast to other methods, this proposed framework does not require a library of voltage and current transients associated with possible network transitions. It can also detect simultaneous switching actions and is robust to noise and load variations. The model yields a lower error detection rate, and its performance is validated using a modified version of the IEEE 33 bus network and a real feeder located in Queensland, Australia, under full and partial observability conditions. The proposed model has also been compared with another data driven method in terms of inference time and error detection rates. [ABSTRACT FROM AUTHOR]

Data pre-processing for developing a generalised land use and land cover (LULC) deep learning model using earth observation data is important for the classification of a different date and/or sensor. However, it is unclear how to approach deep learning segmentation problems in earth observation data. In this paper, we trialled different methods of data preparation for Convolutional Neural Network (CNN) training and semantic segmentation of LULC features within aerial photography over the Wet Tropics and Atherton Tablelands, Queensland, Australia. This was conducted by trialling and ranking various training patch selection sampling strategies, patch and batch sizes, data augmentations and scaling and inference strategies. Our results showed: a stratified random sampling approach for producing training patches counteracted class imbalances; a smaller number of larger patches (small batch size) improves model accuracy; data augmentations and scaling are imperative in creating a generalised model able to accurately classify LULC features in imagery from a different date and sensor; and producing the output classification by averaging multiple grids of patches and three rotated versions of each patch produced a more accurate and aesthetic result. Combining the findings from the trials, we fully trained five models on the 2018 training image and applied the model to the 2015 test image. The output LULC classifications achieved an average kappa of 0.84, user accuracy of 0.81, and producer accuracy of 0.87. Future research using CNNs and earth observation data should implement the findings of this project to increase LULC model accuracy and transferability. [ABSTRACT FROM AUTHOR]

Video-based human action recognition is currently one of the most active research areas in computer vision. Various research studies indicate that the performance of action recognition is highly dependent on the type of features being extracted and how the actions are represented. Since the release of the Kinect camera, a large number of Kinect-based human action recognition techniques have been proposed in the literature. However, there still does not exist a thorough comparison of these Kinect-based techniques under the grouping of feature types, such as handcrafted versus deep learning features and depth-based versus skeleton-based features. In this paper, we analyze and compare 10 recent Kinect-based algorithms for both cross-subject action recognition and cross-view action recognition using six benchmark datasets. In addition, we have implemented and improved some of these techniques and included their variants in the comparison. Our experiments show that the majority of methods perform better on cross-subject action recognition than cross-view action recognition, that the skeleton-based features are more robust for cross-view recognition than the depth-based features, and that the deep learning features are suitable for large datasets. [ABSTRACT FROM AUTHOR]

Electronic Medical Records (EMRs) contain clinical narrative text that is of great potential value to medical researchers. However, this information is mixed with Personally Identifiable Information (PII) that presents risks to patient and clinician confidentiality. This paper presents an end-to-end de-identification framework to automatically remove PII from Australian hospital discharge summaries. Our corpus included 600 hospital discharge summaries which were extracted from the EMRs of two principal referral hospitals in Sydney, Australia. Our end-to-end de-identification framework consists of three components: (1) Annotation: labelling of PII in the 600 hospital discharge summaries using five pre-defined categories: person, address, date of birth, individual identification number, phone/fax number; (2) Modelling: training six named entity recognition (NER) deep learning base-models on balanced and imbalanced datasets; and evaluating ensembles that combine all six base-models, the three base-models with the best F1 scores and the three base-models with the best recall scores respectively, using token-level majority voting and stacking methods; and (3) De-identification: removing PII from the hospital discharge summaries. Our results showed that the ensemble model combined using the stacking Support Vector Machine (SVM) method on the three base-models with the best F1 scores achieved excellent results with a F1 score of 99.16% on the test set of our corpus. We also evaluated the robustness of our modelling component on the 2014 i2b2 de-identification dataset. Our ensemble model, which uses the token-level majority voting method on all six base-models, achieved the highest F1 score of 96.24% at strict entity matching and the highest F1 score of 98.64% at binary token-level matching compared to two state-of-the-art methods. The end-to-end framework provides a robust solution to de-identifying clinical narrative corpuses safely. It can easily be applied to any kind of clinical narrative documents., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2022 The Author(s). Published by Elsevier Inc. All rights reserved.)

Energy forecasting plays a vital role in mitigating challenges in data rich smart grid (SG) systems involving various applications such as demand‐side management, load shedding, and optimum dispatch. Managing efficient forecasting while ensuring the least possible prediction error is one of the main challenges posed in the grid today, considering the uncertainty in SG data. This paper presents a comprehensive and application‐oriented review of state‐of‐the‐art forecasting methods for SG systems along with recent developments in probabilistic deep learning (PDL). Traditional point forecasting methods including statistical, machine learning (ML), and deep learning (DL) are extensively investigated in terms of their applicability to energy forecasting. In addition, the significance of hybrid and data pre‐processing techniques to support forecasting performance is also studied. A comparative case study using the Victorian electricity consumption in Australia and American electric power (AEP) datasets is conducted to analyze the performance of deterministic and probabilistic forecasting methods. The analysis demonstrates higher efficacy of DL methods with appropriate hyper‐parameter tuning when sample sizes are larger and involve nonlinear patterns. Furthermore, PDL methods are found to achieve at least 60% lower prediction errors in comparison to other benchmark DL methods. However, the execution time increases significantly for PDL methods due to large sample space and a tradeoff between computational performance and forecasting accuracy needs to be maintained. [ABSTRACT FROM AUTHOR]

High precision short-term photovoltaic (PV) power prediction can reduce the damage associated with large-scale photovoltaic grid-connection to the power system. In this paper, a combination deep learning forecasting method based on variational mode decomposition (VMD), a fast correlation-based filter (FCBF) and bidirectional long short-term memory (BiLSTM) network is developed to minimize PV power forecasting error. In this model, VMD is used to extract the trend feature of PV power, then FCBF is adopted to select the optimal input-set to reduce the forecasting error caused by the redundant feature. Finally, the input-set is put into the BiLSTM network for training and testing. The performance of this model is tested by a case study using the public data-set provided by a PV station in Australia. Comparisons with common short-term PV power forecasting models are also presented. The results show that under the processing of trend feature extraction and feature selection, the proposed methodology provides a more stable and accurate forecasting effect than other forecasting models. [ABSTRACT FROM AUTHOR]

Accurate and reliable low-voltage load forecasting is critical to optimal operation and control of distribution network and smart grid. However, compared to traditional regional load forecasting at high-voltage level, it faces tough challenges due to the inherent high uncertainty of the low-capacity load and distributed renewable energy integrated in the demand side. This paper proposes a novel hybrid ensemble deep learning (HEDL) approach for deterministic and probabilistic low-voltage load forecasting. The deep belief network (DBN) is applied to low-voltage load point prediction with the strong ability of approximating nonlinear mapping. A series of ensemble learning methods including bagging and boosting variants are introduced to improve the regression ability of DBN. In addition, the differencing transformation technique is utilized to ensure the stationarity of load time series for the application bagging and boosting methods. On the basis of the integrated thought of ensemble learning, a new hybrid ensemble algorithm is developed via integrating multiple separate ensemble methods. Considering the diversity in various ensemble algorithms, an effective K nearest neighbor classification method is utilized to adaptively determine the weights of sub-models. Furthermore, HEDL based probabilistic forecasting is proposed by taking advantage of the inherent resample idea in bagging and boosting. The effectiveness of the HEDL method for both deterministic and probabilistic forecasting has been systematically verified based on realistic load data from East China and Australia, indicating its promising prospective for practical applications in distribution networks. [ABSTRACT FROM AUTHOR]

Across the globe, remote image data is rapidly being collected for the assessment of benthic communities from shallow to extremely deep waters on continental slopes to the abyssal seas. Exploiting this data is presently limited by the time it takes for experts to identify organisms found in these images. With this limitation in mind, a large effort has been made globally to introduce automation and machine learning algorithms to accelerate both classification and assessment of marine benthic biota. One major issue lies with organisms that move with swell and currents, such as kelps. This paper presents an automatic hierarchical classification method local binary classification as opposed to the conventional flat classification to classify kelps in images collected by autonomous underwater vehicles. The proposed kelp classification approach exploits learned feature representations extracted from deep residual networks. We show that these generic features outperform the traditional off-the-shelf CNN features and the conventional hand-crafted features. Experiments also demonstrate that the hierarchical classification method outperforms the traditional parallel multi-class classifications by a significant margin (90.0% vs. 57.6% and 77.2% vs. 59.0%) on Benthoz15 and Rottnest datasets respectively. Furthermore, we compare different hierarchical classification approaches and experimentally show that the sibling hierarchical training approach outperforms the inclusive hierarchical approach by a significant margin. We also report an application of our proposed method to study the change in kelp cover over time for annually repeated AUV surveys.

Background: The paper introduces a deep learning-based approach for real-time detection and insights generation about one of the most prevalent chronic conditions in Australia - Pollen allergy. The popular social media platform is used for data collection as cost-effective and unobtrusive alternative for public health monitoring to complement the traditional survey-based approaches., Methods: The data was extracted from Twitter based on pre-defined keywords (i.e. 'hayfever' OR 'hay fever') throughout the period of 6 months, covering the high pollen season in Australia. The following deep learning architectures were adopted in the experiments: CNN, RNN, LSTM and GRU. Both default (GloVe) and domain-specific (HF) word embeddings were used in training the classifiers. Standard evaluation metrics (i.e. Accuracy, Precision and Recall) were calculated for the results validation. Finally, visual correlation with weather variables was performed., Results: The neural networks-based approach was able to correctly identify the implicit mentions of the symptoms and treatments, even unseen previously (accuracy up to 87.9% for GRU with GloVe embeddings of 300 dimensions)., Conclusions: The system addresses the shortcomings of the conventional machine learning techniques with manual feature-engineering that prove limiting when exposed to a wide range of non-standard expressions relating to medical concepts. The case-study presented demonstrates an application of 'black-box' approach to the real-world problem, along with its internal workings demonstration towards more transparent, interpretable and reproducible decision-making in health informatics domain.

Power grid damage and blackouts are increasing with climate change. Load forecasting methods that integrate climate resilience are therefore essential to facilitate timely and accurate network reconfiguration during periods of extreme stress. Our paper proposes a generalised Wildfire Resilient Load Forecasting Model (WRLFM) to predict electricity load based on operational data of a Distribution Network (DN) in Australia during wildfire seasons in 2015–2020. We demonstrate that load forecasting during wildfire seasons is more challenging than during non-wildfire seasons, motivating an imperative need to improve forecast performance during wildfire seasons. To develop the robust WRLFM, comprehensive comparative analyses were conducted to determine proper Machine Learning (ML) forecast structures and methods for incorporating multiple factors. Bi-directional Gated Recurrent Unit (Bi-GRU) and Vision Transformer (ViT) were selected as they performed the best among all 13 recently trending ML methods. Multi-factors were incorporated to contribute to forecast performance, including input sequence structures, calendar information, flexible correlation-based temperature conditions, and categorical Fire Weather Index (FWI). High-resolution categorical FWI was used to build a forecasting model with climate resilience for the first time, significantly enhancing the average stability of forecast performances by 42%. A sensitivity analysis compared data set patterns and model performances during wildfire and non-wildfire seasons. The improvement rate of load forecasting performance during wildfire seasons was more than two times greater than in non-wildfire seasons. This indicates the significance and effectiveness of applying the WRLFM to improve forecast accuracy under extreme weather risks. Overall, the WRLFM reduces the Mean Absolute Percentage Error (MAPE) of the forecast by 14.37% and 20.86% for Bi-GRU and ViT-based models, respectively, achieving an average forecast MAPE of around 3%. • Novel climate-resilient load forecast model for use during extreme wildfire seasons. • Bi-GRU and ViT outperform 11 recently trending machine learning models. • Categorical FWI with high resolution enhances forecast stability by 42%. • Optimised calendar, temperature and FWI input reduce errors by 17.6%. • Forecast improvement during wildfires is twice that in non-wildfire seasons. [ABSTRACT FROM AUTHOR]

The recent technological advances and their applications to agriculture provide leverage for the new paradigm of smart agriculture. Remote sensing applications can help optimise resources, making agriculture more ecological, increasing productivity and helping farmers to anticipate events that could not otherwise be avoided. Considering that losses caused by anomalies such as diseases, weeds and pests account for 20–40 % of overall agricultural productivity, a successful research effort in this area would be a breakthrough for agriculture. In this paper, we propose a methodology with which to discover and classify anomalies in images of crops, taken from a wide range of distances, using different Convolutional Neural Network architectures. This methodology also deals with several difficulties that usually appear in this kind of problems, such as class imbalance, the insufficient and small variety of images, overtraining or lack of models generalisation. We have implemented four convolutional neural network architectures in a high-performance computing environment, and propose a methodology based on data augmentation with the addition of Gaussian noise to the images to solve the above problems. Our approach was tested using two well-established open datasets that are unalike: DeepWeeds, which provides a classification of 8 weed species native to Australia using images that were taken at a distance of 1 m, and Agriculture-Vision, which classifies 6 types of crop anomalies using multispectral satellite imagery. Our methodology attained accuracies of 98 % and 95.3% respectively, improving the state-of-the-art by several points. In order to ease reproducibility and model selection, we have provided a comparison in terms of computational time and other metrics, thus enabling the choice between architectures to be made according to the resources available. The complete code is available in an open repository in order to encourage reproducibility and promote scientific advances in sustainable agriculture. • A convolutional deep learning model was developed for anomaly detection in agricultural images. • High and low-resolution imagery can be used as input for the model. • Data transformation and augmentation are part of our methodology. • Extensive experiments comparing 4 architectures and 2 datasets are shown. [ABSTRACT FROM AUTHOR]

Although traditional physical models have been used to analyse groundwater systems, the emergence of novel machine learning models can improve the accuracy of the predictions. Deep learning has been prominent in environmental and climate change problems. In this paper, we present a framework for utilising deep learning models to predict groundwater levels based on nearby streamflow and rainfall data. We address the missing data problem using a Bayesian linear regression model within the deep learning framework. Our deep learning framework utilises models such as long-short term memory (LSTM) networks and convolutional neural networks (CNN) for multi-step ahead time series prediction. We examine the fluctuations in groundwater levels at various boreholes located near Middle Creek in New South Wales, Australia. We use the National Collaborative Research Infrastructure Strategy (NCRIS) groundwater database and utilise Bayesian linear regression to impute missing data. We investigate the accuracy of the selected models for individual and regional basins and univariate and multivariate strategies. Our results show that the LSTM-based regional model with multivariate strategy using rainfall data provided the best accuracy. • We predict groundwater levels based on stream flow and rainfall data. • Our framework uses Bayesian data imputation and deep learning for multi-step prediction. • We examine the groundwater fluctuations at various boreholes located in Australia. • Results show a fine-tuned regional model with rainfall data yielded the best accuracy. • We find that LSTM models outperformed CNNs using the multivariate approach. [ABSTRACT FROM AUTHOR]

Accurate and comparable annual mapping is critical to understanding changing vegetation distribution and informing land use planning and management. A U-Net convolutional neural network (CNN) model was used to map natural vegetation and forest types based on annual Landsat geomedian reflectance composite images for a 500 km × 500 km study area in southeastern Australia. The CNN was developed using 2018 imagery. Label data were a ten-class natural vegetation and forest classification (i.e., Acacia, Callitris, Casuarina, Eucalyptus, Grassland, Mangrove, Melaleuca, Plantation, Rainforest and Non-Forest) derived by combining current best-available regional-scale maps of Australian forest types, natural vegetation and land use. The best CNN generated using six Landsat geomedian bands as input produced better results than a pixel-based random forest algorithm, with higher overall accuracy (OA) and weighted mean F1 score for all vegetation classes (93 vs. 87% in both cases) and a higher Kappa score (86 vs. 74%). The trained CNN was used to generate annual vegetation maps for 2000–2019 and evaluated for an independent test area of 100 km × 100 km using statistics describing accuracy regarding the label data and temporal stability. Seventy-six percent of pixels did not change over the 20 years (2000–2019), and year-on-year results were highly correlated (94–97% OA). The accuracy of the CNN model was further verified for the study area using 3456 independent vegetation survey plots where the species of interest had ≥ 50% crown cover. The CNN showed an 81% OA compared with the plot data. The model accuracy was also higher than the label data (76%), which suggests that imperfect training data may not be a major obstacle to CNN-based mapping. Applying the CNN to other regions would help to test the spatial transferability of these techniques and whether they can support the automated production of accurate and comparable annual maps of natural vegetation and forest types required for national reporting. [ABSTRACT FROM AUTHOR]

A deep learning model designed for time series predictions, the long short-term memory (LSTM) architecture, is regularly producing reliable results in local and regional rainfall–runoff applications around the world. Recent large-sample hydrology studies in North America and Europe have shown the LSTM model to successfully match conceptual model performance at a daily time step over hundreds of catchments. Here we investigate how these models perform in producing monthly runoff predictions in the relatively dry and variable conditions of the Australian continent. The monthly time step matches historic data availability and is also important for future water resources planning; however, it provides significantly smaller training datasets than daily time series. In this study, a continental-scale comparison of monthly deep learning (LSTM) predictions to conceptual rainfall–runoff (WAPABA model) predictions is performed on almost 500 catchments across Australia with performance results aggregated over a variety of catchment sizes, flow conditions, and hydrological record lengths. The study period covers a wet phase followed by a prolonged drought, introducing challenges for making predictions outside of known conditions – challenges that will intensify as climate change progresses. The results show that LSTM models matched or exceeded WAPABA prediction performance for more than two-thirds of the study catchments, the largest performance gains of LSTM versus WAPABA occurred in large catchments, the LSTMs struggled less to generalise than the WAPABA models (e.g. making predictions under new conditions), and catchments with few training observations due to the monthly time step did not demonstrate a clear benefit with either WAPABA or LSTM. [ABSTRACT FROM AUTHOR]

Accurately predicting estuaries water quality is essential to support immediate intervention to water quality problem management. Deep learning is used to improve forecasting of water quality parameters in many aquatic systems. However, it is frequently constrained by low data frequency and quality. High-frequency, continuous monitoring using integrated in-situ water quality and environmental sensors can be an input to deep learning models resulting in highly accurate water quality and environmental predictions. This paper proposes a novel approach to improve forecasting of water quality and environmental variables. The results of the real-world data from the Swan Canning Estuary sites show how well the suggested model works. With different sizes of training and testing sets, the model can still predict the increased number of hours in high scores. Eliminating highly correlated variables impact the model's performance, emphasising the usefulness of strongly correlated variables in scarce data scenarios. • This article proposes a framework to enhance water quality variable forecasting. • The framework reframes the time series problem as a supervised learning problem. • The results on the set of data from the estuary show how well the framework works. • The approach is envisaged to support management in detecting water quality issues. [ABSTRACT FROM AUTHOR]

As the global population rises, there is an ever-increasing demand for food, in terms of volume, quality and sustainable production. Protected Cropping Structures (PCS) provide controlled farming environments that support the optimum use of crop inputs for plant growth, faster production cycles, multiple growing seasons per annum and increased yield, while offering greater control of pests, disease and adverse weather. Globally, there has been a rapid increase in the adoption of PCS. However, there remains a concerning knowledge gap in the availability of accurate and up-to-date spatial information that defines the extent (location and area) of PCS. This data is fundamental for providing metrics that inform decision making around forward selling, labour, processing and infrastructure requirements, traceability, biosecurity and natural disaster preparedness and response. This project addresses this need, by developing a national map of PCS for Australia using remotely sensed imagery and deep learning analytics, ancillary data, field validation and industry engagement. The resulting map presents the location and extent of all commercial glasshouses, polyhouses, polytunnels, shadehouses and permanent nets with an area of >0.2 ha. The outcomes of the project revealed deep learning techniques can accurately map PCS with models achieving F-Scores > 0.9 and accelerate the mapping where suitable imagery is available. Location-based tools supported by web mapping applications were critical for the validation of PCS locations and for building industry awareness and engagement. The final national PCS map is publicly available through an online dashboard which summarises the area of PCS structures at a range of scales including state/territory, local government area and individual structure. The outcomes of this project have set a global standard on how this level of mapping can be achieved through a collaborative, multifaceted approach. [ABSTRACT FROM AUTHOR]

Despite the considerable success of deep learning methods in modelling physical processes, they suffer from a variety of issues such as overfitting and lack of interpretability. In hydrology, conceptual rainfall-runoff models are simple yet fast and effective tools to represent the underlying physical processes through lumped storage components. Although conceptual rainfall-runoff models play a vital role in supporting decision-making in water resources management and urban planning, they have limited flexibility to take data into account for the development of robust region-wide models. The combination of deep learning and conceptual models has the potential to address some of the aforementioned limitations. This paper presents a sub-model hybridization of the GR4J rainfall-runoff model with deep learning architectures such as convolutional neural networks (CNN) and long short-term memory (LSTM) networks. The results show that the hybrid models outperform both the base conceptual model as well as the canonical deep neural network architectures in terms of the Nash–Sutcliffe Efficiency (NSE) score across 223 catchments in Australia. We show that our hybrid model provides a significant improvement in predictive performance, particularly in arid catchments, and generalizing better across catchments. • Hybrid modelling features parsimony of data-driven and conceptual hydrologic models. • We present a deep learning-based hybridization for the GR4J rainfall-runoff model. • We present a hierarchical training approach consisting of gradient-based and evolutionary optimization of the respective models. • The results show that hybrid models outperform the canonical (base) hydrologic and deep learning model. • The addition of surface temperature and antecedent streamflow further improves the predictive performance. [ABSTRACT FROM AUTHOR]

The effective management of surface water bodies, such as rivers, lakes, and reservoirs, necessitates a comprehensive understanding of water quality status. Altered precipitation patterns due to climate change may significantly affect the water quality and influence treatment procedures. This study aims to identify the most suitable water quality prediction models for the assessment of the water quality status for three water supply reservoirs in Toowoomba, Australia. It employed four machine learning and two deep learning models for determining the Water Quality Index (WQI) based on five parameters sensitive to rainfall impact. Temporal WQI variations over a period of 22 years (2000–2022) are scrutinised across 4 seasons and 12 months. Through regression analysis, both machine learning and deep learning models anticipate WQI gauged by seven accuracy metrics. Notably, XGBoost and GRU yielded exceptional outcomes, showcasing an R2 value of 0.99. Conversely, Bidirectional LSTM (BiLSTM) demonstrated moderate accuracy with results hovering at 88% to 90% for water quality prediction across all reservoirs. The Coefficient of Efficiency (CE) and Willmott Index (d) showed that the models capture patterns well, while MAE, MAPE and RMSE provided good performance metrics for the RFR, XGBoost and GRU models. These models have provided valuable knowledge that can be utilised to assess the adverse consequences of extreme climate events such as shifts in rainfall patterns. These insights can be used to improve strategies for managing water bodies more effectively. [ABSTRACT FROM AUTHOR]

This paper proposes an automated active fire detection framework using Sentinel-2 imagery. The framework is made up of three basic parts including data collection and preprocessing, deep-learning-based active fire detection, and final product generation modules. The active fire detection module is developed on a specifically designed dual-domain channel-position attention (DCPA)+HRNetV2 model and a dataset with semi-manually annotated active fire samples is constructed over wildfires that commenced on the east coast of Australia and the west coast of the United States in 2019–2020 for the training process. This dataset can be used as a benchmark for other deep-learning-based algorithms to improve active fire detection accuracy. The performance of active fire detection is evaluated regarding the detection accuracy of deep-learning-based models and the processing efficiency of the whole framework. Results indicate that the DCPA and HRNetV2 combination surpasses DeepLabV3 and HRNetV2 models for active fire detection. In addition, the automated framework can deliver active fire detection results of Sentinel-2 inputs with coverage of about 12,000 km2 (including data download) in less than 6 min, where average intersections over union (IoUs) of 70.4% and 71.9% were achieved in tests over Australia and the United States, respectively. Concepts in this framework can be further applied to other remote sensing sensors with data acquisitions in SWIR-NIR-Red ranges and can serve as a powerful tool to deal with large volumes of high-resolution data used in future fire monitoring systems and as a cost-efficient resource in support of governments and fire service agencies that need timely, optimized firefighting plans. [ABSTRACT FROM AUTHOR]

In this paper, a deep learning long-short-term memory (LSTM) method is applied to the forecasting of the critical frequency of the ionosphere F2 layer (foF2). Hourly values of foF2 from 10 ionospheric stations in China and Australia (based on availability) from 2006 to 2019 are used for training and verifying. While 2015 and 2019 are exclusive for verifying the forecasting accuracy. The inputs of the LSTM model are sequential data for the previous values, which include local time (LT), day number, solar zenith angle, the sunspot number (SSN), the daily F10.7 solar flux, geomagnetic the Ap and Kp indices, geographic coordinates, neutral winds, and the observed value of foF2 at the previous moment. To evaluate the forecasting ability of the deep learning LSTM model, two different neural network forecasting models: a back-propagation neural network (BPNN) and a genetic algorithm optimized backpropagation neural network (GABP) were established for comparative analysis. The foF2 parameters were forecasted under geomagnetic quiet and geomagnetic disturbed conditions during solar activity maximum (2015) and minimum (2019), respectively. The forecasting results of these models are compared with those of the international reference ionosphere model (IRI2016) and the measurements. The diurnal and seasonal variations of foF2 for the 4 models were compared and analyzed from 8 selected verification stations. The forecasting results reveal that the deep learning LSTM model presents the optimal performance of all models in forecasting the time series of foF2, while the IRI2016 model has the poorest forecasting performance, and the BPNN model and GABP model are between two of them. [ABSTRACT FROM AUTHOR]

The cyanobacterium Trichodesmium is responsible for approximately half of the ocean's nitrogen input through nitrogen fixation. Although it was first recorded near Australia in the 18th century, the knowledge of where and when large quantity of Trichodesmium around Australia could be found is still lacking. Here, using multi‐band satellite imagery acquired between 2012 and 2021, we fill this knowledge gap through the use of deep learning, designed to recognize both the spectral shapes of individual pixels and spatial morphology of surface aggregations (scums) of Trichodesmium. Trichodesmium scums were found nearly everywhere around Australia, with a cumulative footprint (i.e., where the 10‐year average density is >0.001‰) exceeding 4.6 million km2. Strong seasonality was found, with peak months between September and November. Furthermore, temperature, iron‐rich dust and black carbon aerosols, with the latter being a result of frequent bushfires, play major roles in determining the spatial distributions and seasonality of Trichodesmium. Plain Language Summary: Responsible for half of the ocean's nitrogen input through nitrogen fixation, the saltwater cyanobacterium Trichodesmium is ubiquitous in global tropical and subtropical oceans but particularly abundant around Australia. However, although the earliest report goes back to the 18th century, the knowledge of where and when large quantities of Trichodesmium can be found around Australia is still incomplete. Based on satellite imagery and deep learning, we quantified relative abundance of Trichodesmium around Australia for the period of 2012–2021. Surface aggregations of Trichodesmium were found almost everywhere except the southern coast, with a cumulative footprint exceeding 4.6 million km2. Strong seasonality was found, with peak months between September and November. The spatial distributions and seasonality were found to correlate well with water temperature, iron‐rich dust from Australian desert, and black carbon aerosols from frequent bushfires. With the projected ocean warming in the coming century, Trichodesmium may expand further south, making the cumulative footprint even larger. Key Points: Deep learning was applied to multi‐band satellite images to detect and quantify Trichodesmium surface scums around AustraliaTrichodesmium scums were found nearly everywhere around Australia with a seasonality and a cumulative footprint exceeding 4.6 million km2Distribution and seasonality of Trichodesmium were driven by temperature, iron‐rich dust and black carbon from the mainland bushfires [ABSTRACT FROM AUTHOR]

The prediction of sea level rise is extremely important for improved future climate change mitigation and adaptation strategies. This study uses a hybrid convolutional neural Network (CNN) and a bidirectional long short-term (BiLSTM) model with successive variational mode decomposition (SVMD) to predict the absolute sea level for two study sites in Australia (Port Kembla and Milner Bay). More importantly, the sea level measurements using a tide gauge were corrected using Global Navigation Satellite System (GNSS) measurements of the vertical land movement (VLM). The SVMD-CNN-BiLSTM model was benchmarked by a multi-layer perceptron (MLP), support vector regression (SVR) and gradient boosting (GB). The SVMD-CNN-BiLSTM model outperformed all the comparative models with high correlation values of more than 0.95 for Port Kembla and Milner Bay. Similarly, the SVMD-CNN-BiLSTM model achieved the highest values for the Willmott index, the Nash–Sutcliffe index and the Legates and McCabe index for both study sites. The projected linear trend showed the expected annual mean sea rise for 2030. Using the current trend, Port Kembla was projected to have an MSL value of 1.03 m with a rate rise of approx. 4.5 mm/year. The rate of the MSL for Milner Bay was comparatively lower with a value of approx. 2.75 mm/year and an expected MSL value of 1.27 m for the year 2030. [ABSTRACT FROM AUTHOR]

Satellite imagery is the only feasible approach to annual monitoring and reporting on land cover change. Unfortunately, conventional pixel-based classification methods based on spectral response only (e.g., using random forests algorithms) have shown a lack of spatial and temporal stability due, for instance, to variability between individual pixels and changes in vegetation condition, respectively. Machine learning methods that consider spatial patterns in addition to reflectance can address some of these issues. In this study, a convolutional neural network (CNN) model, U-Net, was trained for a 500 km × 500 km region in southeast Australia using annual Landsat geomedian data for the relatively dry and wet years of 2018 and 2020, respectively. The label data for model training was an eight-class classification inferred from a static land-use map, enhanced using forest-extent mapping. Here, we wished to analyse the benefits of CNN-based land cover mapping and reporting over 34 years (1987–2020). We used the trained model to generate annual land cover maps for a 100 km × 100 km tile near the Australian Capital Territory. We developed innovative diagnostic methods to assess spatial and temporal stability, analysed how the CNN method differs from pixel-based mapping and compared it with two reference land cover products available for some years. Our U-Net CNN results showed better spatial and temporal stability with, respectively, overall accuracy of 89% verses 82% for reference pixel-based mapping, and 76% of pixels unchanged over 33 years. This gave a clearer insight into where and when land cover change occurred compared to reference mapping, where only 30% of pixels were conserved. Remaining issues include edge effects associated with the CNN method and a limited ability to distinguish some land cover types (e.g., broadacre crops vs. pasture). We conclude that the CNN model was better for understanding broad-scale land cover change, use in environmental accounting and natural resource management, whereas pixel-based approaches sometimes more accurately represented small-scale changes in land cover. [ABSTRACT FROM AUTHOR]

Modern power systems are experiencing the challenge of high uncertainty with the increasing penetration of renewable energy resources and the electrification of heating systems. In this paradigm shift, understanding electricity users' demand is of utmost value to the retailers, aggregators, and policymakers. However, behind-the-meter (BTM) equipment and appliances at the household level are unknown to the other stakeholders mainly due to privacy concerns and tight regulations. In this paper, we seek to identify residential consumers based on their BTM equipment, mainly rooftop photovoltaic (PV) systems and electric heating, using imported/purchased energy data from utility meters. To solve this problem with an interpretable, fast, secure, and maintainable solution, we propose an integrated method called Interpretable Refined Motifs And binary Classification (IRMAC). The proposed method comprises a novel shape-based pattern extraction technique, called Refined Motif (RM) discovery, and a single-neuron classifier. The first part extracts a sub-pattern from the long time series considering the frequency of occurrences, average dissimilarity, and time dynamics while emphasising specific times with annotated distances. The second part identifies users' types with linear complexity while preserving the transparency of the algorithms. With the real data from Australia and Denmark, the proposed method is tested and verified in identifying PV owners and electrical heating system users. The performance of the IRMAC is studied and compared with various state-of-the-art methods. The proposed method reached an accuracy of 96% in identifying rooftop PV users and 94.4% in identifying electric heating users, which is comparable to the best solution based on deep learning, while the speed of the inference model is a thousand times faster. Last but not least, the proposed method is a transparent algorithm, which can tackle the concerns regarding the agnostic decision-making process when policies prohibit some machine learning methods. [ABSTRACT FROM AUTHOR]

Accurate measurements of the structural characteristics of trees such as height, diameter, sweep and taper are an important part of forest inventories in managed forests and commercial plantations. Both terrestrial and aerial LiDAR are currently employed to produce pointcloud data from which inventory metrics can be determined. Terrestrial/ground-based scanning typically provides pointclouds resolutions of many thousands of points per m 2 from which tree stems can be observed and inventory measurements made directly, whereas typical resolutions from aerial scanning (tens of points per m 2 ) require inventory metrics to be regressed from LiDAR variables using inventory reference data collected from the ground. Recent developments in miniaturised LiDAR sensors are enabling aerial capture of pointclouds from low-flying aircraft at high-resolutions (hundreds of points per m 2 ) from which tree stem information starts to become directly visible, enabling the possibility for plot-scale inventories that do not require access to the ground. In this paper, we develop new approaches to automated tree detection, segmentation and stem reconstruction using algorithms based on deep supervised machine learning which are designed for use with aerially acquired high-resolution LiDAR pointclouds. Our approach is able to isolate individual trees, determine tree stem points and further build a segmented model of the main tree stem that encompasses tree height, diameter, taper, and sweep. Through the use of deep learning models, our approach is able to adapt to variations in pointcloud densities and partial occlusions that are particularly prevalent when data is captured from the air. We present results of our algorithms using high-resolution LiDAR pointclouds captured from a helicopter over two Radiata pine forests in NSW, Australia. [ABSTRACT FROM AUTHOR]

Particulate matter (PM) or particle pollution include the tiny particles of dust and fly ash particles are expelled from coal-burning power plants. Coal combustion is an extremely prevalent source of air pollution, and resulting PM has substantial impacts on human health, especially in industrial zones. This paper aims to design a novel hybrid deep learning framework based on long short-term memory (LSTM) integrated with a two-stage data filtering technique to accurately predict the air quality indices (i.e., PM 2.5 and PM 10) in a chosen study region 'Miles Airport, Queensland,' that meets the needs of the coal seam gas industry in Australia. The data used to construct the novel hybrid two-stage data filtering technique based on LSTM comprising of six meteorological parameters (i.e., wind direction, wind speed, air temperature, relative humidity, solar radiation, and rainfall) and two environmental factors (i.e., ozone and total suspended particulate). In the first stage, two robust feature selection methods, namely, extreme gradient boosting (XGBoost) and the classification and regression tree (CART) approach, were adopted to explore the most significant predictors. Then in the second stage, the best subset regression (BSR) technique is utilized to determine the best subsets input combinations i.e., C1, C2, and C3 based on several particular metrics. The three BSR-based input combinations were employed in the LSTM model to estimate the PM 2.5 and PM 10. Furthermore, to validate the main hybrid framework, two advanced machine learning (ML) methods (i.e., LightGBM and ridge kernel regression (KRR)) and two traditional ML methods (i.e., Adaptive neuro-fuzzy inference system (ANFIS) and multilayer perceptron neural network (MLP)) were hybridized with the multi-level data filtering strategy by examining the optimal input combinations. Several statistical metrics, graphical tools, and diagnostic analyses evaluated the hybrid models. The outcomes of the PM 2.5 simulation based on 2375 data samples showed that the LSTM-C3 containing all the selected predictors yielded the most promising accuracy, followed by the LightGBM-C3 and MLP-C3 models. On the other hand, the simulation of PM 10 distribution demonstrated that the LSTM-C3 was superior to other models, followed by the KRR-C3 and LightGBM-C3 models. [ABSTRACT FROM AUTHOR]

Electricity price forecasting (EPF) is important for energy system operations and management which include strategic bidding, generation scheduling, optimum storage reserves scheduling and systems analysis. Moreover, accurate EPF is crucial for the purpose of bidding strategies and minimizing the risk for market participants in the competitive electricity market. Nevertheless, accurate time-series prediction of electricity price is very challenging due to complex nonlinearity in the trend of electricity price. This work proposes a mid-term forecasting model based on the demand and price data, renewable and non-renewable energy supplies, the seasonality and peak and off-peak hours of working and nonworking days. An optimized Gated Recurrent Unit (GRU) which incorporates Bagged Regression Tree (BTE) is developed in the Recurrent Neural Network (RNN) architecture for the mid-term EPF. Tanh layer is employed to optimize the hyperparameters of the heterogeneous GRU with the aim to improve the model's performance, error reduction and predict the spikes. In this work, the proposed framework is assessed using electricity market data of five major economical states in Australia by using electricity market data from August 2020 to May 2021. The results showed significant improvement when adopting the proposed prediction framework compared to previous works in forecasting the electricity price. [ABSTRACT FROM AUTHOR]

• The LSTM method is proposed to construct the LUBE approach-based prediction interval for wind power forecast. • An innovative fuzzy approach is suggested to adjust the prediction interval coverage probability. • CDOA is applied to achieve the most optimal prediction intervals. This paper proposes a novel deep learning method based on lower-upper-bound-estimation (LUBE) and long short-term memory (LSTM) models to capture the uncertainty effects in the power generation of wind turbines. The LUBE model deploys the prediction interval concept as well as the LSTM approach to make a robust and resilient model. The LSTM approach is applied to construct the optimal prediction intervals with an appropriate upper bound and lower bound with a certain confidence level. In addition, the fuzzy set theory is proposed in the model to let adjust LSTM parameters based on the decision maker's ideas. This approach empowers operators to satisfy both coverage probability and width indicators of prediction intervals to achieve the optimal solution. The collective decision optimization algorithm is introduced to provide more flexibility in tuning the LSTM parameters. The efficiency and quality of the proposed scheme are studied using some datasets gathered from the Australia wind farms. [Display omitted] [ABSTRACT FROM AUTHOR]

More timely and accurate air quality forecasting could contribute to better public health protection and air pollution prevention. Particulates are a significant indicator for measuring the degree of air pollution. This paper reports on research to model an early warning tool for coarse particulates when assessing the impact of the 12 satellite-derived and ground-based meteorological pollutants out of 30 pollutants considered using hourly Australian data from January 2018–December 2020. A one-dimensional convolutional neural network (CNN) was integrated with a one-directional fully gated recurrent unit (GRU) to forecast consecutive hours' air quality. The CNN model acts as a spatial feature extractor, whereas the new generation GRU makes it computationally efficient. The resultant hybrid ' CNN-GRU' is then comprehensively benchmarked outperforming an ensemble of six other deep learning models. The proposed model's efficacy is indicated at the four most air polluted Australian postcodes in the testing phase. A detailed error analysis with visual and statistical metrics for air quality forecasting ascertains the proposed model's countermeasure to reduce harm and loss. The practical tool is immensely beneficial and can be widely deployed to the regions of public health concern where air pollution is a significant hazard. [Display omitted] • A novel early warning hybrid PM10 forecasting framework designed for Australia. • 30 satellite and ground-based meteorological pollutants were considered with 7 modeling approaches. • CNN integrated with GRU outperformed the other models. • Early warning AI model show potential in the health, environment sector, and decision-making. [ABSTRACT FROM AUTHOR]

This paper presents a new hybrid approach for Global Solar Radiation (GSR) prediction problems, based on deep learning approaches. Predictive models are useful ploys in solar energy industries to optimize the performance of photovoltaic power systems. Specifically, in this work we develop a new 4-phase hybrid CXGBRFR framework, which includes a deep learning Convolutional Neural Network, an Extreme Gradient Boosting with Random Forest Regression, and a Harris Hawks Optimization for initial feature selection. The proposed system has been tested in a problem of daily GSR prediction at six solar farms in Australia. Data from three global climate models (GCM) (CSIRO-BOM ACCESS1-0, MOHC Hadley-GEM2-CC and MRI MRI-CGCM3) have been considered as predictive (input) variables for the proposed approach. The variables from these GCMs contain enough information to obtain an accurate prediction of the GSR at each solar farm. The performance of the proposed approach is compared against different deep and shallows learning approaches: Deep Belief Network, Deep Neural Network, Artificial Neural Network, Extreme Learning Machine and Multivariate Auto-Regressive Spline models. We show that the proposed approach exhibits an excellent performance in GSR prediction, against all alternative approaches in all solar farms considered. [Display omitted] • We develop hybrid deep learning model to predict solar radiation • We optimize the algorithm with Harris Hawks, gradient boosting and random forest • We adopt global climate model and observed predictors as learning model inputs • We build an expert system with relevant knowledge in context of solar application • Our methods benefit solar photovoltaic power monitoring in solar engineering problems. [ABSTRACT FROM AUTHOR]

This study proposes a new hybrid deep learning (DL) model, the called CSVR, for Global Solar Radiation (GSR) predictions by integrating Convolutional Neural Network (CNN) with Support Vector Regression (SVR) approach. First, the CNN algorithm is used to extract local patterns as well as common features that occur recurrently in time series data at different intervals. Then, the SVR is subsequently adopted to replace the fully connected CNN layers to predict the daily GSR time series data at six solar farms in Queensland, Australia. To develop the hybrid CSVR model, we adopt the most pertinent meteorological variables from Global Climate Model and Scientific Information for Landowners database. From a pool of Global Climate Models variables and ground-based observations, the optimal features are selected through a metaheuristic Feature Selection algorithm, an Atom Search Optimization method. The hyperparameters of the proposed CSVR are optimized by mean of the HyperOpt method, and the overall performance of the objective algorithm is benchmarked against eight alternative DL methods, and some of the other Machine Learning approaches (LSTM, DBN, RBF, BRF, MARS, WKNNR, GPML and M5TREE) methods. The results obtained shows that the proposed CSVR model can offer several predictive advantages over the alternative DL models, as well as the conventional ML models. Specifically, we note that the CSVR model recorded a root mean square error/mean absolute error ranging between ≈ 2.172–3.305 MJ m2/1.624–2.370 MJ m2 over the six tested solar farms compared to ≈ 2.514–3.879 MJ m2/1.939–2.866 MJ m2 from alternative ML and DL algorithms. Consistent with this predicted error, the correlation between the measured and the predicted GSR , including the Willmott's, Nash-Sutcliffe's coefficient and Legates & McCabe's Index was relatively higher for the proposed CSVR model compared to other DL and Machine Learning methods for all of the study sites. Accordingly, this study advocates the merits of CSVR model to provide a viable alternative to accurately predict GSR for renewable energy exploitation, energy demand or other forecasting-based applications. [Display omitted] • This paper presents a hybrid CSVR algorithm for global solar radiation prediction. • The proposed approach combines Convolutional Neural Network with a Support Vector Machine. • An Atom Search-based feature selection is applied to improve the performance of the system. • Experiments are carried out in data at six solar energy farms in Queensland, Australia. • The results show the efficacy of deep learning CSVR for solar energy prediction problems. [ABSTRACT FROM AUTHOR]

Solar energy predictive models designed to emulate the long-term (e.g., monthly) global solar radiation (GSR) trained with satellite-derived predictors can be employed as decision tenets in the exploration, installation and management of solar energy production systems in remote and inaccessible solar-powered sites. In spite of a plethora of models designed for GSR prediction, deep learning, representing a state-of-the-art intelligent tool, remains an attractive approach for renewable energy exploration, monitoring and forecasting. In this paper, algorithms based on deep belief networks and deep neural networks are designed to predict long-term GSR. Deep learning algorithms trained with publicly-accessible Moderate Resolution Imaging Spectroradiometer (MODIS) satellite data are tested in Australia's solar cities to predict the monthly GSR: single hidden layer and ensemble models. The monthly-scale MODIS-derived predictors (2003–2018) are adopted, with 15 diverse feature selection approaches including a Gaussian Emulation Machine for sensitivity analysis used to select optimal MODIS-predictor variables to simulate GSR against ground-truth values. Several statistical score metrics are adopted to comprehensively verify surface GSR simulations to ascertain the practicality of deep belief and deep neural networks. In the testing phase, deep learning models generate significantly lower absolute percentage bias (≤3%) and high Kling–Gupta efficiency (≥97.5%) values compared to the single hidden layer and ensemble model. This study ascertains that the optimal MODIS input variables employed in GSR prediction for solar energy applications can be relatively different for diverse sites, advocating a need for feature selection prior to the modelling of GSR. The proposed deep learning approach can be adopted to identify solar energy potential proactively in locations where it is impossible to install an environmental monitoring data acquisition instrument. Hence, MODIS and other related satellite-derived predictors can be incorporated for solar energy prediction as a strategy for long-term renewable energy exploration. [ABSTRACT FROM AUTHOR]

The energy generated by a solar photovoltaic (PV) system depends on uncontrollable factors, including weather conditions and solar irradiation, which leads to uncertainty in the power output. Forecast PV power generation is vital to improve grid stability and balance the energy supply and demand. This study aims to predict hourly day-ahead PV power generation by applying Temporal Fusion Transformer (TFT), a new attention-based architecture that incorporates an interpretable explanation of temporal dynamics and high-performance forecasting over multiple horizons. The proposed forecasting model has been trained and tested using data from six different facilities located in Germany and Australia. The results have been compared with other algorithms like Auto Regressive Integrated Moving Average (ARIMA), Long Short-Term Memory (LSTM), Multi-Layer Perceptron (MLP), and Extreme Gradient Boosting (XGBoost), using statistical error indicators. The use of TFT has been shown to be more accurate than the rest of the algorithms to forecast PV generation in the aforementioned facilities. [ABSTRACT FROM AUTHOR]

Land cover mapping from satellite images has progressed from visual and statistical approaches to Random Forests (RFs) and, more recently, advanced image recognition techniques such as convolutional neural networks (CNNs). CNNs have a conceptual benefit over RFs in recognising spatial feature context, but potentially at the cost of reduced spatial detail. We tested the use of CNNs for improved land cover mapping based on Landsat data, compared with RFs, for a study area of approximately 500 km × 500 km in southeastern Australia. Landsat 8 geomedian composite surface reflectances were available for 2018. Label data were a simple nine-member land cover classification derived from reference land use mapping (Catchment Scale Land Use of Australia—CLUM), and further enhanced by using custom forest extent mapping (Forests of Australia). Experiments were undertaken testing U-Net CNN for segmentation of Landsat 8 geomedian imagery to determine the optimal combination of input Landsat 8 bands. The results were compared with those from a simple autoencoder as well as an RF model. Segmentation test results for the best performing U-Net CNN models produced an overall accuracy of 79% and weighted-mean F1 score of 77% (9 band input) or 76% (6 band input) for a simple nine-member land cover classification, compared with 73% and 68% (6 band input), respectively, for the best RF model. We conclude that U-Net CNN models can generate annual land cover maps with good accuracy from proxy training data, and can also be used for quality control or improvement of existing land cover products. [ABSTRACT FROM AUTHOR]

Cerebellar ataxia (CA) is concerned with the incoordination of movement caused by cerebellar dysfunction. Movements of the eyes, speech, trunk, and limbs are affected. Conventional machine learning approaches utilizing centralised databases have been used to objectively diagnose and quantify the severity of CA. Although these approaches achieved high accuracy, large scale deployment will require large clinics and raises privacy concerns. In this study, we propose an image transformation-based approach to leverage the advantages of state-of-the-art deep learning with federated learning in diagnosing CA. We use motion capture sensors during the performance of a standard neurological balance test obtained from four geographically separated clinics. The recurrence plot, melspectrogram, and poincaré plot are three transformation techniques explored. Experimental results indicate that the recurrence plot yields the highest validation accuracy (86.69%) with MobileNetV2 model in diagnosing CA. The proposed scheme provides a practical solution with high diagnosis accuracy, removing the need for feature engineering and preserving data privacy for a large-scale deployment. [ABSTRACT FROM AUTHOR]

Wildfires not only severely damage the natural environment and global ecological balance but also cause substantial losses to global forest resources and human lives and property. Unprecedented fire events such as Australia's bushfires have alerted us to the fact that wildfire prediction is a critical scientific problem for fire management. Therefore, robust, long‐lead models and dynamic predictions of wildfire are valuable for global fire prevention. However, despite decades of effort, the dynamic, effective, and accurate prediction of wildfire remains problematic. There is great uncertainty in predicting the future based on historical and existing spatiotemporal sequence data, but with advances in deep learning algorithms, solutions to prediction problems are being developed. Here, we present a dynamic prediction model of global burned area of wildfire employing a deep neural network (DNN) approach that produces effective wildfire forecasts based on historical time series predictors and satellite‐based burned area products. A hybrid DNN that combines long short‐term memory and a two‐dimensional convolutional neural network (CNN2D‐LSTM) was proposed, and CNN2D‐LSTM model candidates with four different architectures were designed and compared to construct the optimal architecture for fire prediction. The proposed model was also shown to outperform convolutional neural networks (CNNs) and the fully connected long short‐term memory (FcLSTM) approach using the refined index of agreement and evaluation metrics. We produced monthly global burned area spatiotemporal prediction maps and adequately reflected the seasonal peak in fire activity and highly fire‐prone areas. Our combined CNN2D‐LSTM approach can effectively predict the global burned area of wildfires 1 month in advance and can be generalized to provide seasonal estimates of global fire risk. [ABSTRACT FROM AUTHOR]

Photovoltaic (PV) power forecasting can provide strong support for the safe operation of the power system. Existing forecasting methods are ineffective for grid scheduling decisions or risk analysis. The novel multicluster interval prediction method is proposed to consider the volatility and randomness of PV power output. First, this method utilizes the sparse autoencoder (SAE) and Bayesian regularized NARX network (BRNARX) for point forecasting of PV power. Second, density peak clustering improved by kernel Mahalanobis distance (KMDDPC) is applied to classify the dataset into multiple clusters, including forecasting error and meteorological factors. Finally, the joint probability density is established by multivariate kernel density estimation (MKDE) to accomplish the PV power interval prediction. The proposed hybrid method is applied for the interval prediction of PV power at Yulara, Australia. Comparative research of point forecasting is implemented to evaluate the machine learning and deep learning methods, with the proposed SAE-BRNARX under four different periods. Results shows that the average values of nRMSE, MRE, nMAE, and R2 for the four periods are 4.45%, 0.90%, −0.15%, 3.39%, and 95.93%, respectively. Moreover, the results of interval prediction obtained by the other interval prediction approaches are compared with the proposed KMDDPC-MKDE. It shows that the average values of PICP, PINAW, ACE, and nMPICD for four periods are 93.93%, 9.50%, 3.93%, and 7.10% at 90% confidence level, respectively. Outcomes demonstrate that the proposed method can obtain more accuracy, a higher coverage rate, narrower average bandwidth, and a closer distance between the middle of interval and actual value than other methods. [ABSTRACT FROM AUTHOR]

• A novel CTSGAN deep learning framework is first proposed and applied to model day-ahead electricity price scenarios. • The CTSGAN point forecasting model can eventually be transformed into a probabilistic forecasting model by enhancing the diversity of random input. • The CTSGAN probabilistic forecasting model can directly yield high-quality forecasting intervals with different coverage probabilities as a multi-objective forecasting model. • PSO-based optimal conditions selection method can increase the forecasting accuracy. Electricity prices in spot markets are volatile and can be affected by various factors, such as generation and demand, system contingencies, local weather patterns, bidding strategies of market participants, and uncertain renewable energy outputs. Because of these factors, electricity price forecasting is challenging. This paper proposes a scenario modeling approach to improve forecasting accuracy, conditioning time series generative adversarial networks on external factors. After data pre-processing and condition selection, a conditional TSGAN or CTSGAN is designed to forecast electricity prices. Wasserstein Distance, weights limitation, and RMSProp optimizer are used to ensure that the CTGAN training process is stable. By changing the dimensionality of random noise input, the point forecasting model can be transformed into a probabilistic forecasting model. For electricity price point forecasting, the proposed CTSGAN model has better accuracy and has better generalization ability than the TSGAN and other deep learning methods. For probabilistic forecasting, the proposed CTSGAN model can significantly improve the continuously ranked probability score and Winkler score. The effectiveness and superiority of the proposed CTSGAN forecasting model are verified by case studies. [ABSTRACT FROM AUTHOR]

Load forecasting is of crucial importance for operations of electric power systems. In recent years, deep learning based methods are emerging for load forecasting because their strong nonlinear approximation capabilities can provide more forecasting precision than conventional statistical methods. However, they usually suffer from some problems, e.g., the gradient vanishment and over-fitting. In order to address these problems, an unshared convolution based deep learning model with densely connected network is proposed. In this model, the backbone is the unshared convolutional neural network and a densely connected structure is adopted, which could alleviate the gradient vanishment. What is more, we use a regularization method named clipped $L_2$ -norm to overcome over-fitting, and design a trend decomposition strategy to address the possible distribution differences between the training and validation data. Finally, we conduct five case studies to verify the outperformance of our proposed deep learning model for deterministic and interval load forecasting. Two high-voltage and an medium-voltage real load datasets from Australia, Germany and America are used for model training and validation, respectively. Results show that the proposed model can achieve higher load forecasting accuracy, compared with other existing methods including the popular conventional methods such as naive forecast and generalized additive model, and deep learning methods, e.g., long short-term memory network, convolutional neural network, fully connected network, etc. [ABSTRACT FROM AUTHOR]

Accurate forecasting of wind power generation plays a key role in improving the operation and management of a power system network and thereby its reliability and security. However, predicting wind power is complex due to the existence of high non-linearity in wind speed that eventually relies on prevailing weather conditions. In this paper, a novel hybrid deep learning model is proposed to improve the prediction accuracy of very short-term wind power generation for the Bodangora wind farm located in New South Wales, Australia. The hybrid model consists of convolutional layers, gated recurrent unit (GRU) layers and a fully connected neural network. The convolutional layers have the ability to automatically learn complex features from raw data while the GRU layers are capable of directly learning multiple parallel sequences of input data. The data sets of 5-min intervals from the wind farm are used in case studies to demonstrate the effectiveness of the proposed model against other advanced existing models, including long short-term memory, GRU, autoregressive integrated moving average and support vector machine, which are tuned to optimise outcome. To further evaluate the efficacy of the proposed model, another data set taken from the Capital wind farm in Australia is used. It is observed that the hybrid deep learning model exhibits superior performance in both the data sets over other forecasting models to improve the accuracy of wind power forecasting, numerically for the Bodangora wind farm, up to 1.59 per cent in mean absolute error, 3.73 per cent in root mean square error and 8.13 per cent in mean absolute percentage error. [ABSTRACT FROM AUTHOR]