Optimization of Neural Network-Based Load Forecasting by Means of Whale Optimization Algorithm

Electric load forecasting is of utmost importance for governments and power market participants for planning and monitoring load generation and consumption. Reliable Short-Term Load Forecasting (STLF) can guarantee market operators and participants to manage their operations correctly, securely, and effectively. This paper presents the optimization of neural networks for power forecasting by means of whale optimization algorithm: two types of artificial neural networks namely, Feed-Forward Neural Network (FNN) and Echo State Network (ESN) have been used for STLF. ESN’s simplicity and strength have room for improvement. Therefore, an optimization algorithm called the Whale Optimization Algorithm (WOA) has been used to improve ESN’s performance. WOA-ESN was used for STLF of the first case study, namely Puget power utility in North America. The considered forecasting error indicators showed significant accuracy and reliability. WOA-ESN model and recursive approach resulted in better accuracy measures in terms of standard performance metrics