Collusion Strategy Investigation and Detection for Generation Units in Electricity Market Using Supervised Learning Paradigm
- 28 May 2020
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Systems Journal
- Vol. 15 (1) , 146-157
Abstract
In an oligopoly market, producers compete together to seize the electricity market share. Since they cannot obtain their desired profits through fair competition, they may collude to set their bid prices illegally higher than the oligopoly level. Manipulation and increasing market price decrease social welfare and then market efficiency. This article intends to provide independent system operators (ISOs) with a tool to analyze day-ahead market data so as to identify generator units who intend to exercise collusion and raise the market prices. Toward this goal, all possible collusion and competition scenarios are simulated and then the generated data are used to train a supervised learning algorithm. By applying the proposed approach to the IEEE 57 and 30 bus test systems, the efficiency of the proposed approach was assessed. Furthermore, it is demonstrated how colluding generators choose between maximizing their colluded profit and reducing the risk of being detected by ISO. The results show machine learning is capable of identifying colluding companies with accuracy of 95%. Also, it was rightly obvious that the closer the bidding price of companies is to competitive level, the more downward the efficiency of the machine is.Keywords
This publication has 25 references indexed in Scilit:
- Tacit collusion games in pool-based electricity markets under transmission constraintsMathematical Programming, 2013
- Green Certificates and Market Power on the Nordic Power MarketThe Energy Journal, 2012
- Tacit Collusion in Real-Time U.S. Electricity AuctionsSSRN Electronic Journal, 2011
- Anomaly detectionACM Computing Surveys, 2009
- Market power monitoring and mitigation in the US wholesale power marketsEnergy, 2006
- Price wars and collusion in the Spanish electricity marketInternational Journal of Industrial Organization, 2005
- A comparison of methods for multiclass support vector machinesIEEE Transactions on Neural Networks, 2002
- Strategic gaming analysis for electric power systems: an MPEC approachIEEE Transactions on Power Systems, 2000
- Bagging predictorsMachine Learning, 1996
- Support-vector networksMachine Learning, 1995