Optimization of a Bayesian game for Peer-to-Peer trading among prosumers under incomplete information via a CNN-LSTM-ATT

Abstract

In modern low-carbon industrial parks, various distributed renewable energy resources are employed to fulfill production needs. Despite the growing capacity of renewable energy generation, a significant portion of the power produced by these renewable resources remains unconsumed, resulting in a waste of resources. Within an industrial park, microgrids that both generate and consume energy resources act as energy prosumers. Peer-to-peer (P2P) trading provides an efficient means of utilizing renewable energy among these energy prosumers, who possess both power generation and consumption capabilities. However, within the current market mechanism, each prosumer retains private information that is not disclosed on the network. To address the issue of incomplete information among multiple prosumers during the decision-making process, we develop a Bayesian game model based on the CNN-LSTM-ATT prediction method for P2P electricity transactions among multiple prosumers. The energy prosumers in each industrial park aim to minimize their energy consumption costs by adjusting strategies that include P2P energy trading and managing thermal loads. Prosumers make decisions on the basis of their own characteristics and estimates of other prosumer characteristics, which are obtained from the joint probability distribution predicted by the CNN-LSTM-ATT method. These decisions are aimed at minimizing each prosumer's electricity costs. The simulation results demonstrate the effectiveness of the Bayesian game model proposed in this study.

Keywords

• Bayesian game
• Electricity energy price forecasting
• Peer-to-peer transaction
• Renewable energy consumption
• Thermal dynamics