An efficient Robust Random Vector Functional Link network for Solar Irradiance, Power and Wind speed prediction

Abstract

This paper proposes an efficient technique for prediction of solar irradiance, solar power and wind speed at different time intervals (i.e. 5min, 10min and 60min). With the deliberation of historical solar irradiance, power and wind speed data, an ultra-short Prediction model has been established which is known as Robust Regularized Random Vector Functional link (RRVFL) network. This method utilizes a weighted factor in ridge regularized model, for training the samples to assess the weights in output layer. A Huber's cost function has been applied to gain the robustness here. To get the accuracy of the proposed methodology, the test has been carried out with solar and wind for various time intervals in different atmospheric condition. The result shows that the proposed RRVFL method is very superior as compared with other models (i.e. Random vector functional link (RVFL) and Robust Extreme learning machine(R-ELM), etc. Solar and wind data of California, USA has been taken here. The proposed model can be validated in real time scenario by using test bench application and in industries of solar and wind farm.