Wind power is a sustainable and renewable energy source that has garnered significant attention as the world shifts toward cleaner energy solutions to meet its growing power demands while reducing environmental impact. Accurately predicting wind power generation is crucial for effective energy management and grid stability, as it allows for better integrating wind energy into current power systems. This paper presents a novel approach integrating Partial Least Squares (PLS) with LSTM neural network to predict wind power generation based on historical and high-dimensional weather and power-related data. Using PLS for dimensionality reduction and feature selection helps transform the high-dimensional feature space into a compact set of relevant components, minimizing redundancy and preserving the most informative features for prediction. The proposed PLS-enhanced LSTM model is evaluated using real-world wind power datasets, where the input features include weather conditions, turbine characteristics, and operational data. The methodology involves extensive data preprocessing, including normalization, handling missing values, and constructing PLS components to serve as inputs to the LSTM network. These findings suggest that the PLS-LSTM approach improves prediction accuracy and offers a scalable and robust solution for real-time wind power generation forecasting. Such a model can facilitate better decision-making in energy management, contributing to the optimized utilization of renewable energy resources and enhancing the reliability of power systems.