Contract Settlements for Exchanging Utilities through Automated Negotiations between Prosumers in Eco-Industrial Parks using Reinforcement Learning

DAN KROHLING; ERNESTO MARTÍNEZ

Computer-Aided Chemical Engineering

Elsevier B.V.

Lugar: Amsterdam; Año: 2019; p. 1675 - 1680

Peer-to-peer trading of utilities (heating, cooling and electric power) in an eco-industrial park (EIP) is key to realize significant economic and environmental benefits by exploiting synergistic co-generation and surplus trading. A crucial aspect of this symbiotic scheme is that each selfish company (prosumer) will participate in exchanging utilities to increase its own profits depending on its internal load. In this paper, an automated negotiation approach based on utility tokens is proposed to incentivize participation in a market of prosumer peers for trading surpluses in an EIP. During each negotiation episode, a pair of prosumers engage in a bilateral negotiation and resort to a learned policy to bid, concede and accept/reject using both private information and environmental variables. Contract negotiation revolves around agreeing (or not) on the price expressed in tokens of a utility profile. The time-varying value of the utility token accounts for contextual variables beyond the control of each prosumer. Simulation results demonstrate that reinforcement learning allows finding Nash equilibrium policies for smart contract negotiation in a blockchain environment.