Identification method of market power abuse of generators based on lasso-logit model in spot market

Abstract Accurate identification of violations of generators' market power abuse strongly guarantees that the electricity spot market operates smoothly. To reduce the collinearity of indicators and improve the model's performance, the least absolute shrinkage and selection operator (Lasso) is introduced into the binary logit model, and a method for identifying the market power abuse of generators based on the Lasso-logit model is proposed. First, an identification indicator system is constructed using three aspects: structural, behavioral, and impact indicators. Then, it is screened using the Lasso variable selection for identifying the market power abuse of generators. Based on these results, the Lasso-logit model for identifying the market power abuse of generators is established, and its performance is evaluated using the errors of two kinds and receiver operating characteristic (ROC) curve. Finally, the model is applied to an area's electricity spot market. The results show that the Lasso-logit model has a 96.6% correct rate and can be used to identify illegal generators of the market power abuse in the region. Indicators such as the high-price declaration rate, out-of-merit capacity index, marginal generator reaching the limit rate, market-clearing price and TOP-4 index have practical significance for identifying the market power abuse of generators.