Wind Speed Prediction Based on Error Compensation.

Area of Science:

• Renewable Energy
• Machine Learning
• Time Series Analysis

Background:

• Accurate wind speed prediction is crucial for optimizing wind power generation.
• Existing methods may struggle with the dynamic fluctuations of natural wind speed.
• Improving prediction accuracy enhances the efficiency and reliability of wind farms.

Purpose of the Study:

• To develop a hybrid model for accurate univariate wind speed time series prediction.
• To improve the balance between computational cost and predictive power.
• To compensate for time lags and reduce prediction deviations in wind speed forecasting.

Main Methods:

• A hybrid Autoregressive Moving Average-Support Vector Regression (ARMA-SVR) model was developed.
• ARMA characteristics determined the optimal number of historical data points for SVR training.
• An Extreme Learning Machine (ELM)-based error correction technique was employed for deviation reduction.

Main Results:

• The proposed ARMA-SVR model effectively utilized historical data for prediction.
• The ELM error correction significantly reduced deviations between predicted and real wind speeds.
• Verification studies using real wind farm data confirmed superior performance over traditional methods.

Conclusions:

• The hybrid ARMA-SVR with ELM error correction offers a more accurate wind speed prediction solution.
• This approach can lead to improved quantity and quality of wind power generation.
• The method provides a valuable tool for enhancing wind farm operational efficiency.