An Overview of Deep Learning in Smart Grids

📘 Authors :

Dr. Rudra Kalyan Nayak
Dr. Kolla Bhanu Prakash

📘 Conference Name :

<aside> 💡 IEEE India Info. Vol. 14 No. 3 Jul – Sep 2019

</aside>

📘 Publication Year:

📘 Web Address:

📘 Main Idea :

In recent years, with the growth of the computational methods particularly in case of data management and analysis several machine learning (ML) approaches has been implemented in various industries. According to the current situation most of the researchers have concentrated their studies on deep learning (DL) because deep learning has been treated as one of the emerging areas for feature extraction and handling huge amount of data where machine learning methods fail. Altogether, artificial intelligence encloses with numerous subfields, counting as machine learning, deep learning, computer vision, neural network and natural language processing etc. Deep learning utilizes massive neural networks with a lot of layers of processing units, for advances in computing power and enhanced training techniques to learn versatile patterns from huge quantity of data. Common applications include image and speech recognition. Deep learning is implemented through neural network. The motivation behind neural network is the biological neuron. Generally, deep learning is a sub-branch of machine learning and machine learning is also another sub-branch of artificial intelligence as given in the figure1.

📘 The biggest advantages of deep learning are:

Deep learning gives better performance on dissimilar troubles that significantly outperforms other solutions in diverse domains. This take in speech recognition, language, computer vision, playing games etc. • Deep learning overcomes the limitation of machine learning methods in case of feature extraction that means it takes less time. • For adopting new type of problems in upcoming days deep learning architecture performs well. • Robustness to natural variations in the data is automatically learned.

📘 Importance of deep learning over machine learning:

Machine learning is not enough capable for managing high dimensional data wherever input & output is relatively large. • Deep learning has sufficient power for managing the high dimensional data as well as to focus on the exact features by themselves. This procedure is known as feature mining. • Deep learning can do better than traditional methods. For example, in image classification, deep learning algorithms gave 41percent more correct result than machine learning algorithm and gave 27 percent more accurate in facial recognition and lastly 25 percent in case of voice recognition.