Implementing a grid state indicator for responsive retail demand

Historically, wholesale and retail electricity markets have been designed and maintained as separate markets, without any mechanisms for interactions between participants in the distinct marketplaces. Growing interest in developing responsive demand brings into question the logic of this separation, as retail customers could begin to play a role in maintaining a stable power grid and market price. With this in mind, the California ISO has proposed a price-based signal - the grid state indicator - that would be initiated by the CAISO, updated by distribution system operators, and sent to responsive demand. Thus the signal will allow retail customers to respond to the state of the wholesale market and high voltage grid. In this paper, an electricity price signal model consistent with the proposed CAISO electricity grid state indicator is developed. A customer-agent model applying Q-learning, a form of reinforcement learning, is designed to predict electricity load reduction based on price-responsive demand. Results based on data from the New York ISO assess potential savings, as well as load reduction to smooth out demand, from implementing the proposed grid state signal.