Hybrid Fuzzy-Neuro System for Electrical Load Forecasting

The adaptability of the CANFIS architecture to diverse data patterns and complexities is facilitated by the application of Bell and Gaussian functions. These membership functions play a crucial role in enabling the model to accurately capture the inherent uncertainties and nonlinearities present in real-world datasets. The utilization of the Delta Bar Delta algorithm for weight adaptation further enhances the model's capacity to refine its parameters based on input data, contributing to improved accuracy in load prediction tasks. Additionally, the integration of Momentum learning mechanisms accelerates the learning process and fosters faster convergence, thereby enhancing the overall efficiency of the model. This paper delves into a thorough examination of how different configurations within the CANFIS framework influence key performance metrics such as prediction accuracy, generalization capacity, and learning efficiency. Through extensive experimentation conducted using datasets specifically tailored for load prediction, we aim to elucidate the nuanced effects of varying configurations on the model's performance characteristics. By systematically analyzing the outcomes of these experiments, valuable insights can be gleaned to guide the selection of optimal configurations for practical implementation in real-world load forecasting scenarios.