The accurate modeling and forecasting of the power output of photovoltaic (PV) systems are critical to efficiently managing their integration in smart grids, delivery, and storage. This paper intends to provide efficient short-term forecasting of solar power production using Variational AutoEncoder (VAE) model. Adopting the VAE-driven deep learning model is expected to improve forecasting accuracy because of its suitable performance in time-series modeling and flexible nonlinear approximation. Both single- and multi-step-ahead forecasts are investigated in this work. Data from two grid-connected plants (a 243 kW parking lot canopy array in the US and a 9 MW PV system in Algeria) are employed to show the investigated deep learning models’ performance. Specifically, the forecasting outputs of the proposed VAE-based forecasting method have been compared with seven deep learning methods, namely recurrent neural network, Long short-term memory (LSTM), Bidirectional LSTM, Convolutional LSTM network, Gated recurrent units, stacked autoencoder, and restricted Boltzmann machine, and two commonly used machine learning methods, namely logistic regression and support vector regression. The results of this investigation demonstrate the satisfying performance of deep learning techniques to forecast solar power and point out that the VAE consistently performed better than the other methods. Also, results confirmed the superior performance of deep learning models compared to the two considered baseline machine learning models.

In recent years, human health across the world is becoming concerned by a constant threat of air pollution, which causes many chronic diseases and premature mortalities. Poor air quality does not have only serious adverse effects on human health and vegetation but also some major negative political, societal, and economic impacts. Hence, it is essential to invest more effort on accurate forecasting of ambient air pollution to provide practical and relevant solutions, achieve acceptable air quality, and plan for prevention. In this work, we propose a flexible and efficient deep learning-driven model to forecast concentrations of ambient pollutants. This article introduces first the traditional variational autoencoder (VAE) and the attention mechanism to develop the forecasting modeling strategy based on the innovative integrated multiple directed attention (IMDA) deep learning architecture. To assess the performance of the proposed forecasting methodology, experimental validation is then performed using air pollution data from four US states. Six statistical indicators have been used to evaluate the forecasting accuracy. A discussion of the results obtained finally demonstrates the satisfying performance of integrated multiple directed attention variational autoencoder (IMDA-VAE) methods to forecast different pollutants in different locations. Furthermore, results indicate that the proposed IMDA-VAE model can effectively improve air pollution forecasting performance and outperforms the deep learning models, namely VAE, long short-term memory (LSTM), gated recurrent units (GRU), bidirectional LSTM, bidirectional GRU, and ConvLSTM. We also showed that the forecasting results of the proposed model surpass the performance of LSTM and GRU with the attention mechanism.

Obstacle detection is an essential element for the development of intelligent transportation systems so that accidents can be avoided. In this study, we propose a stereovisionbased method for detecting obstacles in urban environment. The proposed method uses a deep stacked auto-encoders (DSA) model that combines the greedy learning features with the dimensionality reduction capacity and employs an unsupervised k-nearest neighbors algorithm (KNN) to accurately and reliably detect the presence of obstacles. We consider obstacle detection as an anomaly detection problem. We evaluated the proposed method by using practical data from three publicly available datasets, the Malaga stereovision urban dataset (MSVUD), the Daimler urban segmentation dataset (DUSD), and Bahnhof dataset. Also, we compared the efficiency of DSA-KNN approach to the deep belief network (DBN)-based clustering schemes. Results show that the DSA-KNN is suitable to visually monitor urban scenes.