Wind forecasting-driven profit optimization in deregulated energy markets using long short-term memory and American...

Area of Science:

• Electrical Engineering
• Renewable Energy Systems
• Artificial Intelligence in Energy

Background:

• Renewable energy sources (RES), especially wind power (WP), introduce operational uncertainties in deregulated electricity markets due to wind velocity variations.
• Discrepancies between actual wind velocity (AWV) and forecasted wind velocity (FWV) lead to significant price imbalances (PIMB), impacting market scheduling, reliability, and profitability.

Purpose of the Study:

• To develop a comprehensive techno-economic model for profit maximization in wind-integrated deregulated power systems.
• To enhance wind forecasting accuracy and optimize operational decisions under market uncertainties.

Main Methods:

• Utilized Long Short-Term Memory (LSTM) networks for accurate wind speed forecasting, capturing nonlinear temporal dependencies.
• Integrated LSTM-based forecasting with an American Zebra Optimization (AZO) algorithm within an Optimal Power Flow (OPF) market framework.
• Validated the model using real wind datasets from Oman and the IEEE 14-bus test system, considering Locational Marginal Pricing (LMP) and grid constraints.

Main Results:

• LSTM demonstrated superior forecasting accuracy compared to traditional methods and Random Forest (RF).
• The proposed framework reduced imbalance pricing by 28-35% and improved system profitability by approximately 6% (5.9-6.4%).
• AZO algorithm outperformed Sequential Quadratic Programming (SQP) in convergence and profitability.

Conclusions:

• The LSTM-based framework effectively models wind variability, reduces market uncertainty, and enhances renewable energy integration.
• The developed model provides a robust decision-support tool for competitive electricity market operations under high RES penetration.