Zhiqing Zhou
Reinforcement learning; Smart grid; Economic dispatch; Demandresponse; Load
The study addresses smart grid challenges (diversity, randomness, flexibility from electric vehicles, and smart buildings) by construct- ing a two-layer optimization model for dynamic retail pricing and load-unit demand response. The modified model uses reinforcement learning to learn the optimal electricity price, thereby balancing grid stability and user energy consumption. Through numerical simula- tion verification experiments, it was found that the power company’s total revenue value and the load’s total cost value in the research model were 7424.6 and 4152.8, respectively. Compared to the ran- dom parameter method, the difference was the smallest and better than the other two algorithms. Experiments have shown that price- based demand response models based on reinforcement learning can effectively solve related problems in unknown electricity market envi- ronments, and have important application value in maximizing social welfare in unknown market environments.
Important Links:
Go Back
