Towards Smart Energy Forecasting Using Deep Neural Networks Integrated with Optimization Techniques

Abstract

The rise of the global population and rapid industrialization trigger an uncontrollable need for energy. A steady uninterrupted power supply helps in streamlining this process. However, the biggest challenge associated with power supply is the need to eliminate outdated methods and techniques that result in power wastage. Short-term and very short-term load forecasting are very crucial for power system operations and the reliability of any system. Using different techniques to guarantee a system with higher accuracy. Deep learning methods provide high accuracy when using hybrid systems. Deep learning also depends mainly on deep neural networks (DNNs) which have a higher ability to train data and to introduce higher accuracy, whereas DNNs are suitable for industrial loads. Moreover, optimization techniques improve the accuracy of deep neural networks when hybridized with them. This paper introduces a model that depends on the DNNs hybrid with optimization techniques like Genetic Algorithm (GA) or Ant Lion Optimizer (ALO) to enhance the accuracy of the model. Using DNNs hybrid with ALO in the yearly model gave an accuracy of about 98.7% while Feed-forward Neural Network (FFNN) hybrid with ALO gave an accuracy of about 97.44%. This means the superiority of DNNs with ALO rather than the others in load forecasting.