Day-ahead solar forecasting using time series stationarization and feed-forward neural network

Alanazi, Mohana; Khodaei, Amin

doi:10.1109/naps.2016.7747994

Cited by 7 publications

(10 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It can be observed in Table 3 that the residues that are given by the difference between the current irradiance curve and the detrending models were stationary series, since all pValues were less than 0.001. Wu et al (2011), Benmouiza et al (2015) and Alanazi et al (2016) selected the Al-Sadah model, and it can be seen in Fig. 6 that this model of 6º order presented greater similarity of the January real curve and the same occurred for the other months.…”

Section: Ann Resultsmentioning

confidence: 87%

“…However, SVM algorithm is difficult to achieve largescale training samples, which is difficult to solve a variety of problems, while ANN can solve the complex problem of uncertainty (Gu et al, 2006). Wu et al (2011), Benmouiza et al (2015) and Alanazi et al (2016) emphasized the data preprocessing importance by removing the nighttime data, applying detrending models and normalizing the data for neural network application.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Twenty-four Hours Ahead Solar Irradiance Forecast Based on Artificial Neural Network, ARIMA and SARIMA

Ferraz

Souza

et al. 2019

Anais Do 14º Simpósio Brasileiro De Automação Inteligente

View full text Add to dashboard Cite

Over the past years, the Brazilian electricity matrix has shown a tendency to increase the diversification with the insertion of renewable energy as solar and wind. However, the change in the electric power scenario leads to greater complexity in the planning and operation of the system due to its intermittency and the impossibility of storing both natural resources. In order to mitigate this problem, predictive systems have been sought to accurately forecast short-term irradiance and wind speed. The purpose of this paper was to predict irradiance 24 hours ahead from a method based on artificial intelligence, neural network and applying a data pre-processing stage which consisted of obtaining the stochastic components of the daytime data. Thus, the nighttime data were eliminated and the value of each daytime sample was subtracted from the calculated daily trend curve value at its corresponding hour, and finally, the data were normalized. In addition, statistical methods, such as integrated auto-regressive of moving average and integrated auto-regressive of seasonal moving average, were used in order to compare the results achieved, and the artificial neural network presented better results. Resumo: Nos últimos anos, a matriz elétrica brasileira tem apresentado uma tendência de maior diversificação com a inserção de fontes renováveis como solar e eólica. Porém, essa mudança do cenário elétrico tem resultado em uma maior complexidade no planejamento e operação do sistema em função da intermitência dessas fontes e a impossibilidade de armazenamento desses recursos naturais. Com o intuito de mitigar esse problema, tem-se buscado a melhor técnica de previsão dessas fontes a curto prazo. Para esse trabalho, realizou-se a predição da irradiância no horizonte de 24 horas, a partir do método baseado em inteligência artificial, a rede neural. Vale ressaltar que aplicou-se uma etapa de pré-processamento dos dados que consistiu na obtenção dos componentes estocásticos dos dados diurnos, com isso, dados noturnos foram eliminados, e o valor de cada amostra do período diurno é descontado o valor correspondente ao ponto da curva de tendência diária naquele instante, e por fim, os dados foram normalizados. Ademais, utilizou-se métodos estatísticos como auto-regressivo integrado de média móvel e auto-regressivo integrado de média móvel sazonal a fim de comparar os resultados alcançados, sendo que a rede neural apresentou melhor resultado.

show abstract

Section: Ann Resultsmentioning

confidence: 87%

Section: Introductionmentioning

confidence: 99%

Twenty-four Hours Ahead Solar Irradiance Forecast Based on Artificial Neural Network, ARIMA and SARIMA

Ferraz

Souza

et al. 2019

Anais Do 14º Simpósio Brasileiro De Automação Inteligente

View full text Add to dashboard Cite

show abstract

“…Output gate activation vector at time t P fore(i) Forecasted time-series value P true(i) True time-series value r A vector containing the radius of each element in the normalized time-series vector r 2 Coefficient of determination The capital cost for a renewable energy system has been falling over recent years. In fact, the cost of solar PV systems has declined by 75% since 2010, causing a significant increase in investment in the field [1].…”

Section: N Gaf Number Of Gaf Images After Transformation O Tmentioning

confidence: 99%

“…The power generated by PV plants depends on the intermittent energy that is provided by the sun. The variability caused by the daily sun cycle and other meteorological factors gives rise to uncertainties in the determination of this power generation [2]. Power forecasting is critically important in the operation of power systems, as it helps operators dispatch/schedule the power generation from traditional fossil fuels [3].…”

Section: N Gaf Number Of Gaf Images After Transformation O Tmentioning

confidence: 99%

Day-Ahead Solar Irradiation Forecasting Utilizing Gramian Angular Field and Convolutional Long Short-Term Memory

2020

View full text Add to dashboard Cite

The operations of power systems are becoming more challenging on account of the high penetration of renewable power generation, including photovoltaic systems. One method for improving the power system operation involves making accurate forecasts of day-ahead solar irradiation, enabling operators to minimize uncertainty in managing the balance between generation and load. To overcome the limitations of Long Short-term Memory (LSTM) in a one-dimensional forecasting problem, this work proposes a novel method in forecasting solar irradiation by encoding time-series data into images using the Gramian Angular Field and the Convolutional LSTM (ConvLSTM) network. The pre-processed data become a five-dimensional input tensor that is perfectly suitable for ConvLSTM. The ConvLSTM network uses convolution operations in its input-to-state transition and state-to-state transition. The network thus enables time-series forecasting by a feature-rich approach, which ultimately provides competitive forecasting performance despite the use of a small dataset. The proposed method was evaluated in day-ahead solar irradiation forecasting using a univariate dataset of Global Horizontal Irradiation (GHI) data from Fuhai in Taiwan. The proposed method was resampled using 5×2-fold cross-validation, and assessed using the Wilcoxon signed-rank test to determine the statistical significance of the result. It outperformed benchmark methods such as Autoregressive Integrated Moving Average (ARIMA), Convolutional Neural Network cascaded with Long Short-term Memory (CNN-LSTM), and LSTM cascaded with a fully-connected (FC) network.

show abstract

“…Normalization is an important step to ensure all data sets are under same reference scale, and to eliminate any variability due to the changes in the peak of the clear sky irradiance. More detail on GHI data preprocessing can be found in [28].…”

Section: A Data Prepration and Preprocessingmentioning

confidence: 99%

Two-stage hybrid day-ahead solar forecasting

Alanazi

Mahoor

Khodaei

2017

2017 North American Power Symposium (NAPS)

Self Cite

View full text Add to dashboard Cite

Abstract-Power supply from renewable resources is on a global rise where it is forecasted that renewable generation will surpass other types of generation in a foreseeable future. Increased generation from renewable resources, mainly solar and wind, exposes the power grid to more vulnerabilities, conceivably due to their variable generation, thus highlighting the importance of accurate forecasting methods. This paper proposes a two-stage day-ahead solar forecasting method that breaks down the forecasting into linear and nonlinear parts, determines subsequent forecasts, and accordingly, improves accuracy of the obtained results. To further reduce the error resulted from nonstationarity of the historical solar radiation data, a data processing approach, including pre-process and post-process levels, is integrated with the proposed method. Numerical simulations on three test days with different weather conditions exhibit the effectiveness of the proposed two-stage model.

show abstract

Day-ahead solar forecasting using time series stationarization and feed-forward neural network

Cited by 7 publications

References 12 publications

Twenty-four Hours Ahead Solar Irradiance Forecast Based on Artificial Neural Network, ARIMA and SARIMA

Twenty-four Hours Ahead Solar Irradiance Forecast Based on Artificial Neural Network, ARIMA and SARIMA

Day-Ahead Solar Irradiation Forecasting Utilizing Gramian Angular Field and Convolutional Long Short-Term Memory

Two-stage hybrid day-ahead solar forecasting

Contact Info

Product

Resources

About