Multi-agent based experimental analysis on bidding mechanism in electricity auction markets

Abstract

Determining the best market pricing rule is a controversial issue in the design of electricity auction markets. Pay-as-bid (PAB) has been proposed to replace the market-clearing price (MCP) in deregulated electricity markets, with the expectation that it would lower market prices and reduce price volatility. A multi-agent based experiment was constructed to compare and analyze the balanced process of the two auction mechanisms. Each adaptive agent represents a generator who develops a bid price and quantity based on a reinforcement learning algorithm. In the simulation experiments, the experimental results are not as expected. Before the market reaches equilibrium, the market price with the PAB mechanism is lower than it is with the MCP mechanism. However, the market price in the PAB mechanism is higher than it is with the MCP when the market achieves a state of equilibrium. With the PAB mechanism, the volatility of volume and price of each generator is less than it is with the MCP system, as expected. The experimental results also show that unconscious collusion behaviors by generators during the game process are a key reason for producing generator market power in the two auction mechanisms.