Manta ray foraging optimization algorithm–based feedforward neural network for electric energy consumption forecasting

Summary As a consequence of the growing world population along with the rapid developments in technology, electric energy consumption is increasing. Considering the rate of electricity consumption, investment in electric energy generation continues to rapidly expand worldwide. In addition, because of increasing electric energy consumption, the problem of ensuring supply security is an issue that should be considered by all countries. As a result of this issue, it has become necessary to predict short‐term, mid‐term, and long‐term electric energy consumption rates in order to plan for future generation investments. In this study, a feedforward neural network (FFNN) model based on Manta Ray Foraging Optimizer algorithm was proposed to forecast the electric energy consumption rates of Bursa, an industrial city in Turkey, with a rapidly growing economy. The dataset for the proposed model consists of the average data for environmental conditions, the days of the week, and the electric energy consumption rates. Using this dataset, simulation trials were conducted to find the optimal values of weight and bias coefficients in different network structures. The simulation results obtained from the proposed approach were compared with the results from the neural network models trained using the Hierarchical Particle Swarm Optimizer with Time Varying Acceleration Coefficients, improved grey wolf optimization, gradient‐based optimizer, Symbiotic Organisms Search (SOS), Harris Hawks Optimization, Spotted Hyena Optimizer, Salp Swarm Algorithm, and Arithmetic Optimization Algorithm. In order to test the success of the proposed model, the results of both the training and the testing process were analyzed according to the mean absolute error, mean absolute percentage error, and root mean square error evaluation criteria. In addition, the proposed approach was tested using five classification problems of varying difficulty levels presented in the literature in recent years. The simulation results were evaluated statistically and compared to the results of the other algorithms. According to the simulation results from both datasets, in the five classification problems and in the prediction of electric energy consumption, the neural network model trained with the MRFO algorithm performed better than those trained with the other algorithms. It shows the application of the MRFO algorithm to the training of forward‐propagation artificial neural networks. Here, the input data set