Solar radiation prediction based on recurrent neural networks trained by Levenberg-Marquardt backpropagation learning algorithm

Zhang, Nian; Behera, Pradeep K.

doi:10.1109/isgt.2012.6175757

Cited by 42 publications

(17 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Derivative-based methods are widely used in this move. Although the second derivative methods, such as Hessian optimization [28] and Levenberg-Marquardt algorithm [29], can get optimum estimation accurately, the first derivative methods are more efficient in practice because of the balance between calculation complication and iteration effect.…”

Section: Gradient Problemmentioning

confidence: 99%

Reconstruct Recurrent Neural Networks via Flexible Sub-Models for Time Series Classification

Chang

Lü

et al. 2018

Applied Sciences

View full text Add to dashboard Cite

Recurrent neural networks (RNNs) remain challenging, and there is still a lack of long-term memory or learning ability in sequential data classification and prediction. In this paper, we propose a flexible recurrent model, BIdirectional COnvolutional RaNdom RNNs (BICORN-RNNs), incorporating a series of sub-models: random projection, convolutional operation, and bidirectional transmission. These subcategories advance classification accuracy, which was limited by the gradient vanishing and the exploding problem. Experiments on public time series datasets demonstrate that our proposed method substantially outperforms a variety of existing models. Furthermore, the coordination of the accuracy and efficiency concerning a variety of factors, including SNR, length, data missing, and overlapping, is also discussed.

show abstract

Section: Gradient Problemmentioning

confidence: 99%

Reconstruct Recurrent Neural Networks via Flexible Sub-Models for Time Series Classification

Chang

Lü

et al. 2018

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…It is a network training function that updates weight and bias values according to Levenberg-Marquardt optimization. It is often the fastest backpropagation algorithm for training moderate-sized feedforward neural networks (up to several hundred weights), although it does require more memory than other algorithms [13].…”

Section: Learning Algorithmmentioning

confidence: 99%

“…The error is the difference between the output and the target. Validation vectors are used to stop training early if the network performance on the validation vectors fails to improve or substantially remains the same [13]; the best validation performance is 2.221e-07 at epoch 31. As showed in Figure 6.…”

Section: Mean Squared Errormentioning

confidence: 99%

Estimation of the Solar Power Tower Heliostat Position using Neural Network

Zeghoui¹,

Chermitti²

2014

IJCA

View full text Add to dashboard Cite

The alignment and control of heliostats have been one of most issues in solar tower power. In this paper, we review our work to estimate the alignment of solar tower heliostat field based on the combination of two control systems (open loop and closed loop system) using a neural network approach. Several factors influence the path of each heliostat such as the azimuth, the elevation time, the date, the position of the sun, the plant location, the height of the tower and the heliostat size and slope. Firstly, we have modeled the heliostat position using astronomical formulas. Secondly, the neural network model is trained using data from January of three heliostats for both paths. Then, it was tested with another heliostat position for the month of March. The accuracy of the model was evaluated using the mean absolute error (MAE) and absolute percentage error (MAPE). The simulation results show the accuracy of the proposed for estimating the heliostat position without detecting the position of the sun.

show abstract

“…Several papers used statistical time series methods such as auto-regressive (AR) [4] and auto-regressive with moving average (ARMA) [5]. Others use artificial intelligence [6] or neural networks (NN) [7], [8]. In [9] wavelet recurrent neural networks are used.…”

Section: Introductionmentioning

confidence: 99%

Solar radiation forecasting using zenith angle

Fatemi

Kuh

2013

2013 IEEE Global Conference on Signal and Information Processing

View full text Add to dashboard Cite

This paper discusses simple methods for forecasting solar irradiation. We use the zenith angle (the angle between sun beam and perpendicular line on horizontal surface) to remove both seasonal and time of day effects. Then we forecast by using least-squares (LS), time-varying least squares (TVLS), exponentially weighted recursive least squares (EWRLS) and one step estimation of second order statistics. For comparison we also consider a standard method of normalizing data by subtracting mean and dividing by deviation (at that time of day, sixty days moving average) and then applying the same signal processing algorithms. Finally, we compare simulation results for different sites and various time of prediction. Simulation results shows that using zenith angle in signal processing methods improves the performance.

show abstract

Solar radiation prediction based on recurrent neural networks trained by Levenberg-Marquardt backpropagation learning algorithm

Cited by 42 publications

References 19 publications

Reconstruct Recurrent Neural Networks via Flexible Sub-Models for Time Series Classification

Reconstruct Recurrent Neural Networks via Flexible Sub-Models for Time Series Classification

Estimation of the Solar Power Tower Heliostat Position using Neural Network

Solar radiation forecasting using zenith angle

Contact Info

Product

Resources

About