Tacit Collusion with Imperfect Information: Ex-Ante Detection

The liberalization of electricity markets had a significant impact on the whole process of optimal dispatch. The positive effect of competition is partly canceled out by profit-maximizing behavior of market participants. Strategic generators can exercise market power or collude tacitly with other participants to ensure higher electricity price and therefore higher profits. In this paper we study the generators' possibility to come to a tacit collusion by distributed computations. Due to the strategic nature of market interactions, the information corresponding to the power production by generating companies is often fully confidential outside the generating unit. Such information includes the data about planned and emergency outages, capacity constraints, ramping rates specific to the units, probability of failures, etc. Under certain assumptions the constraint set of a generating unit can be represented as a convex set. Using a novel multi-agent distributed constrained optimization algorithm we show how market participants with confidential constraint sets can reach the collusion on the production share, while maximizing their own profit. We compare this outcome to the outcome of Nash equilibrium over specified time horizon and conclude, if both generators find it profitable to collude. This modeling is important for early detection and prevention of tacit collusion.