On the performance of artificial neural network with sine-cosine algorithm in forecasting electricity load demand

To achieve safe and optimal planning and operations of power systems there is a need for using more efficient methods for peak load forecasting. This paper demonstrates the application of sine-cosine algorithm (SCA) for training artificial neural network (ANN) in the problem of load forecasting. The data used in this study were collected over three years, i.e. 2014-2016, for the day, temperature, relative humidity, electricity load demand. The data set was split into weekdays and weekends and the last year data was kept for testing the trained weights. Genetic algorithm (GA) is used to benchmark the results of SCA. The analysis results show that SCA and GA have good performance and provide good fitting for the trained data and good forecast; however, GA still outperform SCA under the same test conditions. Further improvement could be obtained by combining SCA with another swarm technique to solve wider range of forecasting problems.