Ning, Dejun; Chen, Xihui; Chen, Jiyan; Meng, Tao; Xu, Biao; Zhang, Huai

ENERGY REPORTS

2024, 12,

Ning, Dejun; Chen, Xihui; Chen, Jiyan; Meng, Tao; Xu, Biao; Zhang, Huai

With the rapid development of distributed energy, the uncertainty within the energy systems of low-carbon parks is becoming increasingly pronounced, making it difficult for the existing scheduling strategy to meet the requirements of EMS operation. This study addresses the challenges of energy scheduling in low-carbon parks, exacerbated by the uncertainty in distributed energy systems. The proposed PPO-MixClip algorithm, an enhancement of the proximal policy optimization (PPO) approach, incorporates a hybrid clipping mechanism for faster convergence and improved reward outcomes. The algorithm successfully reduces average daily electricity costs by 3.52% compared to the original PPO algorithm. Additionally, the PPO-MixClip achieves a 4.24% daily cost reduction with faster convergence rates. A backup safety control strategy is also implemented to ensure system reliability during training and real-time scheduling.