Hybrid Model of Artificial Neural Network-Cuckoo Search for Irradiance and Load Forecasting

It is worth noting the importance of irradiance and load time series forecasting, in order to contribute to the planning and operation of the smart-grids, due to the increasing Distributed Energy Resource (DER) integration in the electricity network. Thus, Artificial Neural Network (ANN) has been widely adopted, in recent studies, to solve this forecasting problem. However, there is not a consolidated methodology which defines the better combination of some ANN parameters, such as the number of historical data in the input and neurons in the hidden layer. Therefore, in this paper, the authors proposed an optimization methodology which minimizes the Mean Absolute percentage error (MAPE), considering the lowest amount of historical data in the input and neurons in the hidden layer. Thus, this study compared a modified Cuckoo Search (CS) and Particle Swarm Optimization (PSO) performance with the exhaustive search, in order to validate the optimization methodology results. It was possible to notice, from the results, that the CS presented lower MAPE than the PSO, and both presented a considerably lower processing time than trial-and-error.