scholarly journals Multi-Step Solar Irradiance Forecasting and Domain Adaptation of Deep Neural Networks

Energies ◽  
2020 ◽  
Vol 13 (15) ◽  
pp. 3987
Author(s):  
Giorgio Guariso ◽  
Giuseppe Nunnari ◽  
Matteo Sangiorgio
Keyword(s):  
Neural Networks ◽  
Time Series ◽  
Solar Irradiance ◽  
Short Term Memory ◽  
Domain Adaptation ◽  
Model Parameters ◽  
Network Architectures ◽  
Long Time ◽  
Unique Predictor ◽  
Hidden Neurons

The problem of forecasting hourly solar irradiance over a multi-step horizon is dealt with by using three kinds of predictor structures. Two approaches are introduced: Multi-Model (MM) and Multi-Output (MO). Model parameters are identified for two kinds of neural networks, namely the traditional feed-forward (FF) and a class of recurrent networks, those with long short-term memory (LSTM) hidden neurons, which is relatively new for solar radiation forecasting. The performances of the considered approaches are rigorously assessed by appropriate indices and compared with standard benchmarks: the clear sky irradiance and two persistent predictors. Experimental results on a relatively long time series of global solar irradiance show that all the networks architectures perform in a similar way, guaranteeing a slower decrease of forecasting ability on horizons up to several hours, in comparison to the benchmark predictors. The domain adaptation of the neural predictors is investigated evaluating their accuracy on other irradiance time series, with different geographical conditions. The performances of FF and LSTM models are still good and similar between them, suggesting the possibility of adopting a unique predictor at the regional level. Some conceptual and computational differences between the network architectures are also discussed.

Selective Linear Segmentation for Detecting Relevant Parameter Changes*

2020 ◽  
Author(s):  
Arnaud Dufays ◽  
Elysee Aristide Houndetoungan ◽  
Alain Coën
Keyword(s):  
Time Series ◽  
Hedge Fund ◽  
Expectation Maximization Algorithm ◽  
Model Parameters ◽  
Relevant Parameter ◽  
Time Varying ◽  
Long Time ◽  
Model Selection Uncertainty ◽  
Penalized Likelihood Approach ◽  
Changes Over Time

Abstract Change-point (CP) processes are one flexible approach to model long time series. We propose a method to uncover which model parameters truly vary when a CP is detected. Given a set of breakpoints, we use a penalized likelihood approach to select the best set of parameters that changes over time and we prove that the penalty function leads to a consistent selection of the true model. Estimation is carried out via the deterministic annealing expectation-maximization algorithm. Our method accounts for model selection uncertainty and associates a probability to all the possible time-varying parameter specifications. Monte Carlo simulations highlight that the method works well for many time series models including heteroskedastic processes. For a sample of fourteen hedge fund (HF) strategies, using an asset-based style pricing model, we shed light on the promising ability of our method to detect the time-varying dynamics of risk exposures as well as to forecast HF returns.

Cosinor Analysis for Temperature Time Series Data of Long Duration

2007 ◽  
Vol 9 (1) ◽  
pp. 30-41 ◽  
Author(s):  
Nikhil S. Padhye ◽  
Sandra K. Hanneman
Keyword(s):  
Time Series ◽  
Time Series Data ◽  
Model Fit ◽  
Series Data ◽  
Model Parameters ◽  
Cosinor Analysis ◽  
Long Time ◽  
Improved Model ◽  
Data Length ◽  
Cosinor Models

The application of cosinor models to long time series requires special attention. With increasing length of the time series, the presence of noise and drifts in rhythm parameters from cycle to cycle lead to rapid deterioration of cosinor models. The sensitivity of amplitude and model-fit to the data length is demonstrated for body temperature data from ambulatory menstrual cycling and menopausal women and from ambulatory male swine. It follows that amplitude comparisons between studies cannot be made independent of consideration of the data length. Cosinor analysis may be carried out on serial-sections of the series for improved model-fit and for tracking changes in rhythm parameters. Noise and drift reduction can also be achieved by folding the series onto a single cycle, which leads to substantial gains in the model-fit but lowers the amplitude. Central values of model parameters are negligibly changed by consideration of the autoregressive nature of residuals.

scholarly journals Minutely Active Power Forecasting Models Using Neural Networks

2020 ◽  
Vol 12 (8) ◽  
pp. 3177 ◽  
Author(s):  
Dimitrios Kontogiannis ◽  
Dimitrios Bargiotas ◽  
Aspassia Daskalopulu
Keyword(s):  
Neural Networks ◽  
Time Series ◽  
Power Systems ◽  
Demand Response ◽  
Short Term Memory ◽  
Electricity Consumption ◽  
Absolute Error ◽  
Real World Data ◽  
Training Time ◽  
Power Forecasting

Power forecasting is an integral part of the Demand Response design philosophy for power systems, enabling utility companies to understand the electricity consumption patterns of their customers and adjust price signals accordingly, in order to handle load demand more effectively. Since there is an increasing interest in real-time automation and more flexible Demand Response programs that monitor changes in the residential load profiles and reflect them according to changes in energy pricing schemes, high granularity time series forecasting is at the forefront of energy and artificial intelligence research, aimed at developing machine learning models that can produce accurate time series predictions. In this study we compared the baseline performance and structure of different types of neural networks on residential energy data by formulating a suitable supervised learning problem, based on real world data. After training and testing long short-term memory (LSTM) network variants, a convolutional neural network (CNN), and a multi-layer perceptron (MLP), we observed that the latter performed better on the given problem, yielding the lowest mean absolute error and achieving the fastest training time.

scholarly journals Long Short-Term Memory Neural Networks for Online Disturbance Detection in Satellite Image Time Series

2018 ◽  
Vol 10 (3) ◽  
pp. 452 ◽  
Author(s):  
Yun-Long Kong ◽  
Qingqing Huang ◽  
Chengyi Wang ◽  
Jingbo Chen ◽  
Jiansheng Chen ◽  
...  
Keyword(s):  
Neural Networks ◽  
Time Series ◽  
Short Term Memory ◽  
Satellite Image ◽  
Short Term ◽  
Term Memory ◽  
Long Short Term Memory ◽  
Disturbance Detection

scholarly journals Predicting Solar Irradiance Using Time Series Neural Networks

2014 ◽  
Vol 36 ◽  
pp. 623-628 ◽  
Author(s):  
A. Alzahrani ◽  
J.W. Kimball ◽  
C. Dagli
Keyword(s):  
Neural Networks ◽  
Time Series ◽  
Solar Irradiance

scholarly journals Neural Networks Applied to the Wave-Induced Fatigue Analysis of Steel Risers

2018 ◽  
Vol 2018 ◽  
pp. 1-16 ◽  
Author(s):  
João P. R. Cortina ◽  
Fernando J. M. de Sousa ◽  
Luis V. S. Sagrilo
Keyword(s):  
Neural Networks ◽  
Time Series ◽  
Cross Sections ◽  
Computational Cost ◽  
Offshore Structures ◽  
Fatigue Analysis ◽  
Element Analysis ◽  
Long Time ◽  
Wave Induced ◽  
Hybrid Methodology

Time domain stochastic wave dynamic analyses of offshore structures are computationally expensive. Considering the wave-induced fatigue assessment for such structures, the combination of many environmental loading cases and the need of long time-series responses make the computational cost even more critical. In order to reduce the computational burden related to the wave-induced fatigue analysis of Steel Catenary Risers (SCRs), this work presents the application of a recently developed hybrid methodology that combines dynamic Finite Element Analysis (FEA) and Artificial Neural Networks (ANN). The methodology is named hybrid once it requires short time series of structure responses (obtained by FEA) and imposed motions (evaluated analytically) to train an ANN. Subsequently, the ANN is employed to predict the remaining response time series using the prescribed motions imposed at the top of the structure by the floater unit. In this particular work, the methodology is applied aiming to predict the tension and bending moments’ time series at structural elements located at the top region and at the touchdown zone (TDZ) of a metallic riser. With the predicted responses (tensions and moments), the stress time series are determined for eight points along the pipe cross sections, and stress cycles are identified using a Rainflow algorithm. Fatigue damage is then evaluated using SN curves and the Miner-Palmgren damage accumulation rule. The methodology is applied to a SCR connected to a semisubmersible platform in a water depth of 910 m. The obtained results are compared to those from a full FEA in order to evaluate the accuracy and computer efficiency of the hybrid methodology.

scholarly journals Speeding Up Back-Propagation Neural Networks

10.28945/2931 ◽  
2005 ◽  
Author(s):  
Mohammed A. Otair ◽  
Walid A. Salameh
Keyword(s):  
Neural Networks ◽  
Back Propagation ◽  
Network Size ◽  
Learning Rate ◽  
Network Architectures ◽  
Local Minima ◽  
Set Size ◽  
Long Time ◽  
Multilayer Neural Networks ◽  
Optical Time

There are many successful applications of Backpropagation (BP) for training multilayer neural networks. However, it has many shortcomings. Learning often takes long time to converge, and it may fall into local minima. One of the possible remedies to escape from local minima is by using a very small learning rate, which slows down the learning process. The proposed algorithm presented in this study used for training depends on a multilayer neural network with a very small learning rate, especially when using a large training set size. It can be applied in a generic manner for any network size that uses a backpropgation algorithm through an optical time (seen time). The paper describes the proposed algorithm, and how it can improve the performance of back-propagation (BP). The feasibility of proposed algorithm is shown through out number of experiments on different network architectures.

Deep machine learning for detection of acoustic wave reflections

2019 ◽  
Vol 19 (5) ◽  
pp. 1340-1350
Author(s):  
Mulugeta A Haile ◽  
Edward Zhu ◽  
Christopher Hsu ◽  
Natasha Bradley
Keyword(s):  
Neural Networks ◽  
Time Series ◽  
Acoustic Emission ◽  
Time Series Data ◽  
Short Term Memory ◽  
Series Data ◽  
Short Term ◽  
Term Memory ◽  
Propagating Waves ◽  
Long Short Term Memory

Acoustic emission signals are information rich and can be used to estimate the size and location of damage in structures. However, many existing algorithms may be deceived by indirectly propagated acoustic emission waves which are modulated by reflection boundaries within the structures. We propose two deep learning models to identify such waves such that existing algorithms for damage detection and localization may be used. The first approach uses long short-term memory recurrent neural networks to learn distinct patterns directly from the time-series data. In the second approach, we transform the time-series data into spectrograms and utilize convolutional neural networks to perform binary classification by leveraging spectro-temporal features. We achieved 80% classification accuracy using long short-term memory and near-perfect accuracy using convolutional neural networks on a dataset of acoustic emission signals generated by the Hsu-Nielsen sources. Both long short-term memory and convolutional neural network models were able to learn general and context-specific features of the direct and reflected acoustic emission waves. Once accurately identified, the indirectly propagating waves are filtered out while the directly propagating waves are used for source location using existing methods.

scholarly journals Single and Multi-Sequence Deep Learning Models for Short and Medium Term Electric Load Forecasting

Energies ◽  
2019 ◽  
Vol 12 (1) ◽  
pp. 149 ◽  
Author(s):  
Salah Bouktif ◽  
Ali Fiaz ◽  
Ali Ouni ◽  
Mohamed Adel Serhani
Keyword(s):  
Neural Networks ◽  
Time Series ◽  
Deep Learning ◽  
Short Term Memory ◽  
Load Forecasting ◽  
Forecasting Model ◽  
Time Lags ◽  
Electric Load ◽  
Learning Approaches ◽  
Electric Load Forecasting

Time series analysis using long short term memory (LSTM) deep learning is a very attractive strategy to achieve accurate electric load forecasting. Although it outperforms most machine learning approaches, the LSTM forecasting model still reveals a lack of validity because it neglects several characteristics of the electric load exhibited by time series. In this work, we propose a load-forecasting model based on enhanced-LSTM that explicitly considers the periodicity characteristic of the electric load by using multiple sequences of inputs time lags. An autoregressive model is developed together with an autocorrelation function (ACF) to regress consumption and identify the most relevant time lags to feed the multi-sequence LSTM. Two variations of deep neural networks, LSTM and gated recurrent unit (GRU) are developed for both single and multi-sequence time-lagged features. These models are compared to each other and to a spectrum of data mining benchmark techniques including artificial neural networks (ANN), boosting, and bagging ensemble trees. France Metropolitan’s electricity consumption data is used to train and validate our models. The obtained results show that GRU- and LSTM-based deep learning model with multi-sequence time lags achieve higher performance than other alternatives including the single-sequence LSTM. It is demonstrated that the new models can capture critical characteristics of complex time series (i.e., periodicity) by encompassing past information from multiple timescale sequences. These models subsequently achieve predictions that are more accurate.

Sign in / Sign up

Export Citation Format

Share Document