Renewable GenCo bidding strategy using newsvendor-based neural networks: An example from Turkish electricity market

Liberalization policies in the electricity markets have led to an increase in the number of electricity generation companies (GenCos). GenCos develop competitive strategies to thrive in the market and ensure long-term sustainability. In the day-ahead market (DAM), GenCos submit bids for the electricity quantity they plan to generate the next day. Deviating from the committed electricity quantity leads to either selling surplus electricity at a positive imbalance price (lower than the market price) or incurring a penalty for missing electricity at a negative imbalance price (higher than the market price). In this study, we develop a bid strategy for renewable GenCos participating in the Turkish DAM. The proposed approach involves utilizing a neural network (NN) to determine the bid quantity. The proposed NN aims to minimize a cost function considering both negative and positive imbalance prices, as opposed to the classical NN of minimizing the mean squared error. The proposed approach outperforms the classical NN, yielding a 2.38% (potentially reaching up to 7.43%) improvement in revenue generation. Moreover, we perform a comprehensive sensitivity analysis, specifically focusing on changes in penalty coefficients. Notably, as penalty coefficient ranges widen, our approach refines bidding strategy more effectively than alternatives.