TRSWA-BP Neural Network for Dynamic Wind Power Forecasting Based on Entropy Evaluation.

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence
• Time Series Analysis

Background:

• Accurate wind power forecasting is crucial for grid management and economic decisions.
• The stochastic and volatile nature of wind power presents significant challenges for traditional forecasting methods.

Purpose of the Study:

• To develop and evaluate novel forecasting models for wind power.
• To introduce an effective criterion for assessing the dynamic accuracy of wind power predictions.

Main Methods:

• A novel Tabu, Real-coded, Small-world Optimization Algorithm (TRSWA) integrated with a Backpropagation (BP) neural network (TRSWA-BP).
• Three forecasting models combining TRSWA-BP with Empirical Mode Decomposition (EMD), Phase Space Reconstruction (PSR), and EMD-based PSR.
• Development of a Normalized Renyi's Quadratic Entropy (NRQE) criterion to evaluate prediction accuracy.

Main Results:

• The proposed TRSWA-BP models effectively mitigate errors caused by wind power fluctuations and multi-fractal properties.
• Error sequences from the forecasting methods exhibit non-Gaussian characteristics.
• The NRQE criterion demonstrates feasibility in assessing stochastic prediction errors.

Conclusions:

• The novel TRSWA-BP models offer improved performance for wind power forecasting.
• The NRQE criterion provides a robust method for evaluating dynamic prediction accuracy in stochastic environments.
• The findings contribute to more reliable wind energy integration and management.