A Hybrid Forecasting Model Based on Modified Bat Algorithm and ELM: A Case Study for Wind Speed Forecasting

Accurate wind speed forecasting is essential to the dispatch and management of wind power systems for the improvement in operation reliability of wind power plants. However, the wind velocity is the most volatile and random nonlinear series which has difficulty to obtain satisfactory prediction values. Currently to achieve higher forecasting accuracy, some numerical optimization algorithms have been employed in computational methods to remedy the deficiencies. In this paper, we propose a hybrid forecasting model based on a modified bat algorithm and extreme learning machine (ELM). The process consists of two layers: a bat algorithm (BA) is improved by conjugate gradient method to optimize the parameter of ELM, which proves to converge faster than using bat algorithm purely. The second layer is the training of ELM network to obtain the final forecasting results. To verify the effectiveness of proposed novel model, we collect data from wind power stations in Penglai, China, with the comparison indicating that MBA-ELM algorithm gains more accurate forecasting results than purely applying ELM for forecasting.