Wind Speed Prediction for Offshore Sites Using a Clockwork Recurrent Network

Knowledge Graph (KG), which commonly consists of fruitful connected facts about items, presents an unprecedented opportunity to alleviate the sparsity problem in recommender system. However, existing KG based recommendation methods mainly rely on handcrafted meta-path features or simple triple-level entity embedding, which cannot automatically capture entities' long-term relational dependencies for the recommendation. In this paper, a two-channel neural interaction method named Knowledge Graph enhanced Neural Collaborative Filtering with Residual Recurrent Network (KGNCF-RRN) is proposed, which leverages both long-term relational dependencies KG context and user-item interaction for recommendation. (1) For the KG context interaction channel, we propose a Residual Recurrent Network (RRN) to construct context-based path embedding, which incorporates residual learning into traditional recurrent neural networks (RNNs) to efficiently encode the long-term relational dependencies of KG. The self-attention network is then applied to the path embedding to capture the polysemy of various user interaction behaviours. (2) For the user-item interaction channel, the user and item embeddings are fed into a newly designed two-dimensional interaction map. (3) Finally, above the two-channel neural interaction matrix, we employ a convolutional neural network to learn complex correlations between user and item. Extensive experimental results on three benchmark datasets show that our proposed approach outperforms existing state-of-the-art approaches for knowledge graph based recommendation.

Computational intelligence (CI) based prediction models increase the efficient and effective utilization of resources for wind prediction. However, the traditional recurrent neural networks (RNN) are difficult to train on data having long-term temporal dependencies, thus susceptible to an inherent problem of vanishing gradient. This work proposed a method based on an advanced version of RNN known as long short-term memory (LSTM) architecture, which updates recurrent weights to overcome the vanishing gradient problem. This, in turn, improves training performance. The RNN model is developed based on stack LSTM and bidirectional LSTM. The parameters like mean absolute error (MAE), standard deviation error (SDE), and root mean squared error (RMSE) are utilized as performance measures for comparison with recent state-of-the-art techniques. Results showed that the proposed technique outperformed the existing techniques in terms of RMSE and MAE against all the used wind farm datasets. Whereas, a reduction in SDE is observed for larger wind farm datasets. The proposed RNN approach performed better than the existing models despite fewer parameters. In addition, the approach requires minimum processing power to achieve compatible results.

It is a general trend for new energy vehicles to replace traditional fuel vehicles, while the number of electric vehicles has increased dramatically, the pressure of charging on the power grid has also increased. Therefore, from the angle of wind power absorption of electric vehicles, this paper mainly studies the time series of wind speed power. The time series of wind speed is decomposed into the sum of 15-minute average wind speed signal and turbulent wind speed signal. The relationship between the magnitude of turbulent signal and average wind speed is studied. Based on the discrete Markov chain, the battery state transition probability of electric vehicle is established according to the results of Monte Carlo simulation. The state transition matrix varies with time, and the stationary distribution is obtained to obtain the charging potential of the electric vehicle in a certain area, that is, the charging demand of the electric vehicle with time. The main factors affecting the state transition matrix are analyzed, and the results of each factor affecting the state transition matrix are given. Thus, a more accurate charging potential model can be obtained. By predicting the average wind speed, the number of electric vehicles participating in the regulation can be obtained. The results of simulation indicate that wind fluctuation can be effectively reduced by controlling charging power of electric vehicle.

Simplified non-homogeneous Markov Chain (MC) models for generation of synthetic time series of wind speed (WS) and wind direction (WD) are proposed. The proposed models consider daily, monthly, and seasonal WS and WD characteristics. The accuracy and computational burden of the proposed models are compared with those of a complete model (CM), proposed by the same authors in a previous paper, trough numerical experiments based on field data.

A deep learning framework for constitutive modeling based on temporal convolutional network

Recurrent Neural Networks (RNNs), which are a powerful scheme for modeling temporal and sequential data need to capture long-term dependencies on datasets and represent them in hidden layers with a powerful model to capture more information from inputs. For modeling long-term dependencies in a dataset, the gating mechanism concept can help RNNs remember and forget previous information. Representing the hidden layers of an RNN with more expressive operations (i.e., tensor products) helps it learn a more complex relationship between the current input and the previous hidden layer information. These ideas can generally improve RNN performances. In this paper, we proposed a novel RNN architecture that combine the concepts of gating mechanism and the tensor product into a single model. By combining these two concepts into a single RNN, our proposed models learn long-term dependencies by modeling with gating units and obtain more expressive and direct interaction between input and hidden layers using a tensor product on 3-dimensional array (tensor) weight parameters. We use Long Short Term Memory (LSTM) RNN and Gated Recurrent Unit (GRU) RNN and combine them with a tensor product inside their formulations. Our proposed RNNs, which are called a Long-Short Term Memory Recurrent Neural Tensor Network (LSTMRNTN) and Gated Recurrent Unit Recurrent Neural Tensor Network (GRURNTN), are made by combining the LSTM and GRU RNN models with the tensor product. We conducted experiments with our proposed models on word-level and character-level language modeling tasks and revealed that our proposed models significantly improved their performance compared to our baseline models.

The main neural network architectures applied for product demand forecasting are examined. Particular attention is given to Recurrent Neural Networks (RNN), which demonstrate high efficiency when working with time series and detecting dependencies within data. A detailed analysis is provided for the Long Short-Term Memory (LSTM) architecture, an advanced variant of RNN that effectively addresses the vanishing gradient problem, a common issue in traditional RNN. Due to its capability to retain information over extended periods, LSTM is ideal for tasks requiring the analysis of long-term dependencies, such as seasonal fluctuations in product demand. The application of Convolutional Neural Networks (CNN) is analyzed, which show strong performance when processing structured data, such as images or matrices. With their ability to effectively detect spatial dependencies, CNN can be useful for demand forecasting when complex relationships between various data features need to be considered, particularly for demand evaluations dependent on multiple factors. The architectures Feedforward Neural Networks (FNN), Gated Recurrent Units (GRU), Attention-based models (ABM) and Autoencoders (AE) are considered, which can also be applied to solve demand forecasting problems. A comparative analysis of the architectures is made in terms of forecast accuracy and ability to work with time series. By evaluating the training speed, data processing time, and reliability of these architectures in terms of the stability of results and forecast reliability, it has been determined that LSTM architectures are the most effective for product demand forecasting. These networks provide high accuracy and reliability of results and perform well with time series, which is crucial for forecasting demand under dynamic conditions.

The article described the relevant knowledge of wind speed prediction and the verification method of the forecast. First, the characteristics of wind speed are general analyzed. Through the form of wind, speed frequency of wind, shear speed and turbulence characteristics of wind, it was found that the wind speed had instability and non-linear characteristics. By analysing the basic principles of the three methods included the time series, neural network method and empirical mode decomposition, the wind speed time series prediction method was proposed. On the basis of this method, it was applied to a raw wind factory speed time series and simulated reality by Matlab. Compared the predictive conclusion with original data, it gained a higher accuracy and verified the efficiency of the method.

In the traditional wind speed forecast of a general wind farm, this farm is often regarded as a whole for point forecasting, and only a single wind speed forecast result is given in a large spatial scale. However, wind turbines are distributed in diverse locations in the wind farm, and wind speeds are often different. Therefore, this paper comes up with a model to forecast the wind speeds of multiple wind turbines by learning the potential relationship between wind speed time series and wind direction time series. First, wind direction feature is preprocessed by trigonometric function. Subsequently Simple Recurrent Unit (SRU) is designed to learn the temporal correlation information. Wind speed and trigonometric wind direction form a three-dimensional feature matrix. Compared with other Recurrent Neural Network (RNN), SRU can greatly reduce the computational cost. Therefore, SRU is used to extract the dynamic information of characteristic data of multiple wind turbines in a wind farm over time. The simulation results with the real data prove that compared with the existing forecast methods using temporal correlation feature, the method not only significantly reduces the computational cost, but also improves the accuracy of forecast.

A novel hybrid algorithm based on Empirical Fourier decomposition and deep learning for wind speed forecasting

Wind has been one of the popular renewable energy generation methods in the last decades. Foreknowledge of power to be generated from wind is crucial especially for planning and storing the power. It is evident in various experimental data that wind speed time series has non‐linear characteristics. It has been reported in the literature that nonlinear prediction methods such as artificial neural network (ANN) and adaptive neuro fuzzy inference system (ANFIS) perform better than linear autoregressive (AR) and AR moving average models. Polynomial AR (PAR) models, despite being non‐linear, are simpler to implement when compared with other non‐linear AR models due to their linear‐in‐the‐parameters property. In this study, a PAR model is used for one‐day ahead wind speed prediction by using the past hourly average wind speed measurements of Çeşme and Bandon and performance comparison studies between PAR and ANN‐ANFIS models are performed. In addition, wind power data which was published for Global Energy Forecasting Competition 2012 has been used to make power predictions. Despite having lower number of model parameters, PAR models outperform all other models for both of the locations in speed predictions as well as in power predictions when the prediction horizon is longer than 12 h.

Sensitivity analysis of the system of wind speed distributions

Asset integrity management is essential in many industries, especially oil and gas, to ensure assets operate efficiently and safely. Traditionally, maintenance monitoring has relied on manual inspections. However, this approach is time-consuming, labor-intensive, inefficient, error-prone, ineffective for real-time tracking, and incurs high operational costs. This study seeks to leverage artificial intelligence (AI) to minimize reliance on manual inspections and maintenance, thereby reducing human error and enhancing efficiency. This paper proposes a hybrid machine learning model utilizing a recurrent neural network (RNN) and long short-term memory (LSTM) for asset integrity management using oil and gas plant data to detect potential issues before they occur. The hybrid model combining LSTM and RNN enhances predictive performance and analysis in dynamic contexts by leveraging the advantages of both algorithms. Using a Monte Carlo simulation approach with 100 iterations, we demonstrate that the LSTM-RNN model performs better than traditional RNN, LSTM, and hybrid RNN-LSTM models for time-series data. Subsequently, we implement the models using oil and gas industry data through transfer learning. Our results show that the LSTM-RNN model can serve as a baseline for an autonomous system that maintains assets in the oil and gas industry.

This paper provides a comparison between two methods for time series forecasting. The first method is based on traditional recurrent neural networks (RNNs) while the second method is based in Reservoir Computing (RC). Reservoir Computing is a new paradigm that offers an intuitive methodology for using the temporal processing power of RNNs without the inconvenience of training them. So we decided to compare the advantages / disadvantages of using Reservoir Computing and RNNs in the problem of time series forecasting. The first method uses a Nonlinear Autoregressive Network with exogenous inputs (NARX). Optimization was carried out on the NARX architecture through an optimization procedure focused on the best mean squared error (MSE) metrics in the training set. The second method, called RCDESIGN, combines an evolutionary algorithm with Reservoir Computing and simultaneously looks for the best values of parameters, topology and weight matrices without rescaling the reservoir by the spectral radius. Nevertheless RCDESIGN has yielded fast tracking and excellent performance in some benchmark problems including the Narma and Mackey-Glass time-series.

<p>Dust devils are convective vortices able to lift sand and dust grains from the surface. They are common not only in the terrestrial deserts, but also on Mars, where they give a substantial contribution to the planetary dust budget (∼50%). Dust devils are characterized by a central pressure drop that generates the rotation and translate advected by the wind background.<br>One of the problems related to their monitoring is due to the impossibility to directly separate the translational and rotational motion components from the study of the wind speed and direction time series. This means that it is not possible to directly retrieve information on their translational speed and direction and on the maximum rotational wind speed using only a meteorological station.  <br>In addition, there are other fundamental parameters that are not directly measurable, such as the distance of passage of the vortex from the station, its sense of rotation and its diameter.<br>These limitations lead in general to a size/distance degeneracy of the results, i.e. the acquired meteorological signatures of a smaller dust devils passing near the station could not be distinguished from the ones of a bigger and farther event. This, in turn, leads to several problems in the study of their physics.<br>To avoid these issues, the monitoring meteorological station can be equipped with an imaging camera system with a sufficient acquisition rate, resolution and field of view. However, this is not always possible, in particular for the planetary space missions.<br>Here, we want to present two simple methods for the characterization of the wind speed and direction time series of dust devils that allow an easy solution for the measure of the vortex translational wind speed and direction, distance of passage and sense of rotation.<br>The knowledge of these parameters allows to completely characterize the measured dust devils encounter just using the meteorological station acquisition.<br>In order to test the methods, we performed a field campaign in the Sahara desert, deploying a fully equipped meteorological station coupled with a camera system. We compared the results of the meteorological analysis with the ones obtained from the images, confirming the effectiveness of our methodology. <br>This methodology can give a substantial improvement in the interpretation of the past and next martian dust devils data. For example, the ESA/Roscosmos ExoMars 2020 mission will host on its lander a meteorological station (METEO package) and the Dust Complex, a suite of specific instruments devoted to the study of the primary airborne dust. These instruments can be used in tandem for the characterization of the local dust devils activity.</p>