A multi dataset validation model for hybrid feature selection in wind energy maximum power point tracking systems.

Area of Science:

• Renewable Energy Systems
• Computational Intelligence
• Machine Learning for Energy

Background:

• Modern wind turbines generate high-dimensional sensor data, posing computational challenges for Maximum Power Point Tracking (MPPT) systems.
• Existing MPPT methods struggle with the scale of data, impacting efficiency and accuracy in real-time applications.
• Need for methods that balance predictive performance with computational feasibility in wind energy.

Purpose of the Study:

• To develop and validate a two-stage hybrid Feature Selection (FS) method for wind energy MPPT.
• To enhance computational efficiency and prediction accuracy in high-dimensional datasets.
• To provide adaptable solutions for varying accuracy-efficiency trade-offs in operational MPPT systems.

Main Methods:

• A two-stage hybrid approach combining mutual information (MI) for initial filtering and Adaptive Multi-Objective Binary Harmony Search (AMO-BHS) for optimization.
• AMO-BHS incorporates adaptive parameter control and multi-objective optimization to generate diverse feature subsets.
• Validation using Random Forest regression on three distinct datasets (Kelmarsh, laboratory, VV Wind Farms) for active power and rotor speed prediction.

Main Results:

• Achieved significant dimensionality reduction (76.9-87.5%), reducing feature sets from hundreds to tens.
• Improved prediction accuracy, with Root Mean Square Errors (RMSE) reduced by 9.4-14.7% compared to full-feature baselines.
• Demonstrated superior performance over standard FS methods (MI ranking, LASSO), reducing RMSE by 20.6-34.7%.

Conclusions:

• The proposed two-stage FS method offers a computationally efficient solution for real-time MPPT on resource-constrained hardware.
• Reduced feature sets enhance operational benefits, including lower sensor dependency, maintenance costs, and improved model resilience.
• Significant potential for energy production optimization in utility-scale wind farms through improved MPPT accuracy.