Wind Speed Intervals Prediction using Meta-cognitive Approach

Abstract

In this paper, an interval type-2 neural fuzzy inference system and its meta-cognitive learning algorithm for wind speed prediction is proposed. Interval type-2 neuro-fuzzy system is capable of handling uncertainty associated with the data and meta-cognition employs self-regulation mechanism for learning. The proposed system realizes Takagi-Sugeno-Kang inference mechanism and adopts a fast data-driven interval-reduction method. Meta-cognitive learning enables the network structure to evolve automatically based on the knowledge in data. The parameters are updated based on an extended Kalman filter algorithm. In addition, the proposed network is able to construct prediction intervals to quantify uncertainty associated with forecasts. For performance evaluation, a real-world wind speed prediction problem is utilized. Using historical data, the model provides short-term prediction intervals of wind speed. The performance of proposed algorithm is compared with existing state-of-the art fuzzy inference system approaches and the results clearly indicate its advantages in forecasting problems.