Agent based simulation of centralized electricity transaction market using bi-level and Q-learning algorithm approach. (January 2022)

• Record Type: Journal Article
• Authors: Namalomba, Ellen; Feihu, Hu; Shi, Haijie
• Is Part Of: International journal of electrical power & energy systems. Volume 134(2022)
• Publication Date: 2022-01
• Subjects: Electricity markets -- Double side auction -- Bidding strategies -- Agent simulation -- Bi-level mathematical programming; Electrical engineering -- Periodicals; Electric power systems -- Periodicals; Électrotechnique -- Périodiques; Réseaux électriques (Énergie) -- Périodiques; Electric power systems; Electrical engineering; Periodicals; 621.3
• DOI: 10.1016/j.ijepes.2021.107415
• Languages: English
• ISSNs: 0142-0615

Abstract

Highlights: A centralized electricity market of double sided auction with elastic demand is proposed. Flexible consumers participate in the electricity auction and respond to market prices using Q learning method. The participation of consumers with elastic demand in electricity auction offers significant reduction in energy transaction prices. Bi-level mathematical optimization and Q-learning method is used to analyze the strategic bidding of market players at Nash Equilibrium. Abstract: The market participants face risks in market price fluctuations and uncertainties in the demand behaviour, regardless of the series of restructuring reforms of China power industry. To address this gap, the strategic pricing and bidding behaviour of generation and electricity selling enterprises in double-sided auctions of centralized electricity transaction market with elastic demand is examined. The market is modelled with two-level (Bi-level) mathematical optimization problem and Q-Learning algorithm. First, the upper level solves the profit maximization of the individual market players. Second, the lower level represents the market clearing at uniform transaction price by a market auctioneer using Lagrange relaxation method. Agent learning approach of Q-Learning algorithm is used to solve the two level mathematical problem. Both generation and electricity selling enterprises are modelled as Q-Learning agents with incomplete information about their counterparts. The simulation results show significant reduction in energy transaction price with the participation of flexible consumers in electricity market auctions, hence boosting the consumer savings. This study offers positive effects in reducing market prices in double-sided auction of electricity markets with elastic consumers using Q-learning algorithm as compared to a market with inelastic consumers.