Search

Solar energy is one of the most common and promising sources of renewable energy. In photovoltaic (PV) systems, operators can benefit from future solar irradiance predictions for efficient load balancing and grid stability. Therefore, short-term solar irradiance forecasting plays a crucial role in the transition to renewable energy. Modern PV grids collect large volumes of data that provide valuable information for forecasting models. Although the nature of these data presents an ideal setting for online learning methodologies, research to date has mainly focused on offline approaches. Hence, this work proposes a novel data streaming method for real-time solar irradiance forecasting on days with variable weather conditions and cloud coverage. Our method operates under an asynchronous dual-pipeline framework using deep learning models. For the experimental study, two datasets from a Canadian PV solar plant have been simulated as streams at different data frequencies. The experiments involve an exhaustive parameter grid search to evaluate four state-of-the-art deep learning architectures: multilayer percep tron (MLP), long-short term memory network (LSTM), convolutional network (CNN), and Transformer network. The obtained results demonstrate the suitability of deep learning models for this problem. In particular, MLP and CNN achieved the best accuracy, with a high capacity to adapt to the evolving data stream

The progressive substitution of conventional synchronous generation for renewable-based generation imposes a series of challenges in many aspects of modern power systems, among which are the issues related to the low rotational inertia systems. Rotational inertia and the kinetic energy stored in the rotating masses in the power system play a fundamental role in the operation of power systems as it represents in some sort the ability of the system to withstand imbalances between generation and demand. Therefore, transmission system operators (TSOs) need tools to forecast the inertia or the kinetic energy available in the systems in the very short term (from minutes to hours) in order to take appropriate actions if the values fall below the one that ensures secure operation. This paper proposes a methodology based on machine learning (ML) techniques for short-term kinetic energy forecasting available in power systems; it focuses on the length of the moving window, which allows for obtaining a balance between the historical information needed and the horizon of forecasting. The proposed methodology aims to be as flexible as possible to apply to any power system, regardless of the data available and the software used. To illustrate the proposed methodology, time series of the kinetic energy recorded in the Nordic Power System (NPS) has been used as a case study. The results show that Linear Regression (LR) is the most suitable method for a time horizon of one hour due to its high accuracyto-simplicity ratio, while Long Short-Term Memory (LSTM) is the most accurate for a forecasting horizon of four hours. Experimental assessment has been carried out using Typhoon HIL-404 simulator, verifying that both algorithms are suitable for real-time simulation.

Processing data streams arriving at high speed requires the development of models that can provide fast and accurate predictions. Although deep neural networks are the state-of-the-art for many machine learning tasks, their performance in real-time data streaming scenarios is a research area that has not yet been fully addressed. Nevertheless, much effort has been put into the adaption of complex deep learning (DL) models to streaming tasks by reducing the processing time. The design of the asynchronous dual-pipeline DL framework allows making predictions of incoming instances and updating the model simultaneously, using two separate layers. The aim of this work is to assess the performance of different types of DL architectures for data streaming classification using this framework. We evaluate models such as multi-layer perceptrons, recurrent, convolutional and temporal convolutional neural networks over several time series datasets that are simulated as streams at different speeds. In addition, we evaluate how the different architectures react to concept drifts typically found in evolving data streams. The obtained results indicate that convolutional architectures achieve a higher performance in terms of accuracy and efficiency, but are also the most sensitive to concept drifts.

Esta tesis, presentada como un compendio de artículos de investigación, aborda la predicción de series temporales en un entorno de streaming mediante técnicas de deep learning. En primer lugar, se aporta un innovador framework asíncrono para la aplicación de modelos deep learning a datos provenientes de un stream a gran velocidad. Además, se realiza un exhaustivo estudio sobre la aplicabilidad de los métodos de deep learning para el problema de predicción de series temporales. La minería de datos en streaming es un problema fundamental aplicable a una gran multitud de campos donde los datos son generados secuencialmente a gran velocidad. Los requisitos de velocidad que caracterizan este escenario no permiten el uso de técnicas de deep learning, las cuales presentan un gran coste computacional. En esta tesis presentamos una solución a este problema: un framework asíncrono (ADLStream) que separa las fases de entrenamiento y predicción de los modelos de entrenamiento, aliviando así el coste computacional de los modelos y permitiéndoles adaptarse a la evolución de los datos a lo largo del tiempo. Esta propuesta ha sido evaluada experimentalmente usando diversos conjuntos de datos de clasificación de series temporales y comparándolo con los modelos del estado del arte para la minería de datos en streaming, como son los árboles de Hoeffding, detectores de deriva o modelos ensembles. Los resultados demostraron la mejora en el rendimiento conseguida con nuestra propuesta. La predicción de series temporales es uno de los problemas de aprendizaje automático y estadísticos más comunes, engloba todas los datos que tienen una componente temporal, presentes en problemas meteorológicos, energéticos, médicos, logísticos o financiarios. La irrupción del aprendizaje profundo (deep learning) como estado del arte en la minería de datos, ha beneficiado a la investigación relacionada con la predicción de series temporales. Por ello, en esta tesis presentamos el estudio experimental más c, This thesis addresses the problem of time series forecasting in a streaming scenario using deep learning techniques. Firstly, it provides an innovative asynchronous framework for the application of deep learning models to data from a high-speed stream. In addition, an extensive study on the applicability of deep learning methods to the time series prediction problem is carried out. Data stream mining is a fundamental problem applicable to a multitude of fields where data is generated sequentially at high speed. The speed requirements that characterise this scenario do not allow the use of deep learning techniques, which are computationally expensive. In this work we present a solution to this problem - an asynchronous framework (ADLStream) that separates the training and prediction phases of the training models, thus alleviating the computational cost of the models and allowing them to adapt to the evolution of the data distribution. This proposal has been experimentally evaluated using different time series classification datasets and compared with state-of-the-art models for data stream mining, such as Hoeffding trees, drift detectors or ensemble models. The results demonstrated the performance improvement achieved with our proposal. Time series forecasting is one of the most common statistical and machine learning problems, covering all data with a temporal component, such as meteorological, energy, medical, logistical or financial data. The emergence of deep learning as the state-of-the-art in data mining has benefited research related to time series forecasting. Therefore, we present the most comprehensive experimental study on the applicability of deep learning for time series forecasting. More than 50,000 time series were used for this study. We trained and evaluated a total of 3,800 models of different architectures: multilayer perceptron (MLP), Elman recurrent neural network (ERNN), long-short term memory network (LSTM), gated recurrent unit network (GRU)

Object detection has been one of the most active topics in computer vision for the past years. Recent works have mainly focused on pushing the state-of-the-art in the general-purpose COCO benchmark. However, the use of such detection frameworks in specific applications such as autonomous driving is yet an area to be addressed. This study presents an enhanced 2D object detector based on Faster RCNN that is better suited for the context of autonomous vehicles. Two main aspects are improved: the anchor generation procedure and the performance drop in minority classes. The default uniform anchor configuration is not suitable in this scenario due to the perspective projection of the vehicle cameras. Therefore, we propose a perspective-aware methodology that divides the image into key regions via clustering and uses evolutionary algorithms to optimize the base anchors for each of them. Furthermore, we add a module that enhances the precision of the second-stage header network by including the spatial information of the candidate regions proposed in the first stage. We also explore different reweighting strategies to address the foreground-foreground class imbalance, showing that the use of a reduced version of focal loss can significantly improve the detection of difficult and underrepresented objects in two-stage detectors. Finally, we design an ensemble model to combine the strengths of the different learning strategies. Our proposal is evaluated with the Waymo Open Dataset, which is the most extensive and diverse up to date. The results demonstrate an average accuracy improvement of 6.13% mAP when using the best single model, and of 9.69% mAP with the ensemble. The proposed modifications over the Faster R-CNN do not increase computational cost and can easily be extended to optimize other anchor-based detection frameworks.

The attention-based Transformer architecture is earning in- creasing popularity for many machine learning tasks. In this study, we aim to explore the suitability of Transformers for time series forecasting, which is a crucial problem in di erent domains. We perform an extensive experimental study of the Transformer with di erent architecture and hyper-parameter con gurations over 12 datasets with more than 50,000 time series. The forecasting accuracy and computational e ciency of Transformers are compared with state-of-the-art deep learning networks such as LSTM and CNN. The obtained results demonstrate that Trans- formers can outperform traditional recurrent or convolutional models due to their capacity to capture long-term dependencies, obtaining the most accurate forecasts in ve out of twelve datasets. However, Transformers are generally more di cult to parametrize and show higher variability of results. In terms of e ciency, Transformer models proved to be less competitive in inference time and similar to the LSTM in training time.

In recent years, deep learning techniques have outperformed traditional models in many machine learning tasks. Deep neural networks have successfully been applied to address time series forecasting problems, which is a very important topic in data mining. They have proved to be an effective solution given their capacity to automatically learn the temporal dependencies present in time series. However, selecting the most convenient type of deep neural network and its parametrization is a complex task that requires considerable expertise. Therefore, there is a need for deeper studies on the suitability of all existing architectures for different forecasting tasks. In this work, we face two main challenges: a comprehensive review of the latest works using deep learning for time series forecasting and an experimental study comparing the performance of the most popular architectures. The comparison involves a thorough analysis of seven types of deep learning models in terms of accuracy and efficiency. We evaluate the rankings and distribution of results obtained with the proposed models under many different architecture configurations and training hyperparameters. The datasets used comprise more than 50,000 time series divided into 12 different forecasting problems. By training more than 38,000 models on these data, we provide the most extensive deep learning study for time series forecasting. Among all studied models, the results show that long short-term memory (LSTM) and convolutional networks (CNN) are the best alternatives, with LSTMs obtaining the most accurate forecasts. CNNs achieve comparable performance with less variability of results under different parameter configurations, while also being more efficient.

Object detection using remote sensing data is a key task of the perception systems of self-driving vehicles. While many generic deep learning architectures have been proposed for this problem, there is little guidance on their suitability when using them in a particular scenario such as autonomous driving. In this work, we aim to assess the performance of existing 2D detection systems on a multi-class problem (vehicles, pedestrians, and cyclists) with images obtained from the on-board camera sensors of a car. We evaluate several one-stage (RetinaNet, FCOS, and YOLOv3) and two-stage (Faster R-CNN) deep learning meta-architectures under different image resolutions and feature extractors (ResNet, ResNeXt, Res2Net, DarkNet, and MobileNet). These models are trained using transfer learning and compared in terms of both precision and efficiency, with special attention to the real-time requirements of this context. For the experimental study, we use theWaymo Open Dataset, which is the largest existing benchmark. Despite the rising popularity of one-stage detectors, our findings show that two-stage detectors still provide the most robust performance. Faster R-CNN models outperform one-stage detectors in accuracy, being also more reliable in the detection of minority classes. Faster R-CNN Res2Net-101 achieves the best speed/accuracy tradeoff but needs lower resolution images to reach real-time speed. Furthermore, the anchor-free FCOS detector is a slightly faster alternative to RetinaNet, with similar precision and lower memory usage.

Modern energy systems collect high volumes of data that can provide valuable information about energy consumption. Electric companies can now use historical data to make informed decisions on energy production by forecasting the expected demand. Many deep learning models have been proposed to deal with these types of time series forecasting problems. Deep neural networks, such as recurrent or convolutional, can automatically capture complex patterns in time series data and provide accurate predictions. In particular, Temporal Convolutional Networks (TCN) are a specialised architecture that has advantages over recurrent networks for forecasting tasks. TCNs are able to extract long-term patterns using dilated causal convolutions and residual blocks, and can also be more efficient in terms of computation time. In this work, we propose a TCN-based deep learning model to improve the predictive performance in energy demand forecasting. Two energy-related time series with data from Spain have been studied: the national electric demand and the power demand at charging stations for electric vehicles. An extensive experimental study has been conducted, involving more than 1900 models with different architectures and parametrisations. The TCN proposal outperforms the forecasting accuracy of Long Short-Term Memory (LSTM) recurrent networks, which are considered the state-of-the-art in the field.

Processing data streams arriving at high speed requires the development of models that can provide fast and accurate predictions. Although deep neural networks are the state-of-the-art for many machine learning tasks, their performance in real-time data streaming scenarios is a research area that has not yet been fully addressed. Nevertheless, there have been recent efforts to adapt complex deep learning models for streaming tasks by reducing their processing rate. The design of the asynchronous dual-pipeline deep learning framework allows to predict over incoming instances and update the model simultaneously using two separate layers. The aim of this work is to assess the performance of different types of deep architectures for data streaming classification using this framework. We evaluate models such as multi-layer perceptrons, recurrent, convolutional and temporal convolutional neural networks over several time-series datasets that are simulated as streams. The obtained results indicate that convolutional architectures achieve a higher performance in terms of accuracy and efficiency.

Data streaming classification has become an essential task in many fields where real-time decisions have to be made based on incoming information. Neural networks are a particularly suitable technique for the streaming scenario due to their incremental learning nature. However, the high computation cost of deep architectures limits their applicability to high-velocity streams, hence they have not yet been fully explored in the literature. Therefore, in this work, we aim to evaluate the effectiveness of complex deep neural networks for supervised classification in the streaming context. We propose an asynchronous deep learning framework in which training and testing are performed simultaneously in two different processes. The data stream entering the system is dual fed into both layers in order to concurrently provide quick predictions and update the deep learning model. This separation reduces processing time while obtaining high accuracy on classification. Several time-series datasets from the UCR repository have been simulated as streams to evaluate our proposal, which has been compared to other methods such as Hoeffding trees, drift detectors, and ensemble models. The statistical analysis carried out verifies the improvement in performance achieved with our dual-pipeline deep learning framework, that is also competitive in terms of computation time.