Hybrid LSTM-PSO optimization techniques for enhancing wind power bidding efficiency in electricity markets

Abstract

Past research has predominantly focused on utilizing meta-heuristic algorithms to optimize neural network structures, while the exploration of deep learning in optimization has remained relatively limited. The proposed hybrid approach seeks to enhance wind power bidding strategies, improving profitability by predicting optimal output power for day-ahead electricity markets. This method integrates Long Short-Term Memory (LSTM) with Particle Swarm Optimization (PSO), leveraging LSTM’s ability to predict the active movement tendencies of particles for more efficient and faster optimization. Experiments conducted on the IEEE 30-bus power system show that the LSTM-PSO hybrid outperforms mathematical models and standalone PSO algorithms. It also delivers an optimal wind power bidding strategy, yielding peak annual revenue, while recommending a 16 % reduction in bidding output power variance in models that integrate wind power with thermal power and energy storage systems (ESS). Ultimately, this approach fosters confidence in wind energy investment, contributing to sustainable development.