With the intensification of competition in electricity markets and the increase in renewable energy penetration, electricity retailers urgently need intelligent pricing strategies to address supply-demand fluctuations and competitive pressures. This drive for dynamic strategies is similarly seen in the textile industry, where intelligent pricing is vital for manufacturers to adapt quickly to fluctuating raw fiber costs and competitive market demands for finished goods. Traditional static pricing models cannot adapt to real-time market changes, and existing dynamic pricing studies often ignore the coupling effects of consumer price elasticity and competitors'behaviors. This study proposes an algorithm framework based on Multi-Agent Deep Deterministic Policy Gradient (MADDPG), modeling retailers, consumers, and energy suppliers as interactive agents. The framework uses a double-layer LSTM network to process historical load data (RMSE=0.12) and real-time market sales data (updated every 15 minutes), and designs a multi-dimensional reward function integrating dynamic price elasticity coefficients (ηt) and profit-risk equilibrium constraints. Simulation results show that the proposed strategy significantly outperforms both basic and state-of-the-art benchmarks. The results indicate that the reinforcement learning-driven dynamic pricing algorithm provides efficient decision support for electricity retail pricing through elastic and differentiated price responses. This approach to optimizing dynamic responses is equally applicable in the textile industry, where similar machine learning models can be used to set agile pricing for textiles based on real-time factors like inventory levels and immediate market demand.

dynamic pricing, electricity retail market, multi-agent reinforcement learning, demand response, textile industry