Short-term scheduling for hydroelectric power plants in a deregulated power market by means of a deep deterministic policy gradient algorithm

The main task of short-term scheduling in a free market is to provide decision support for market bidding and unit commitment for the coming hours and days. In such a realm, an optimal bidding strategy maximizes the profits while taking all essential aspects regarding the system operation into account. In today's hydropower industry, the bidding process often rests on the skills of the operating engineers. This paper investigates how a reinforcement learning agent can provide decision support or even fully alleviate the human operator of the short-term scheduling. The work also captures the interplay between long or medium-term scheduling and short-term scheduling ensuring a more seamless connection between the two horizons.