Simulation-based learning of strategies for automated negotiation of utility exchanges using smart contracts in eco-industrial parks

KRÖHLING, DAN EZEQUIEL; MARTÍNEZ, ERNESTO

Santa Fe

Congreso; X Congreso Argentino de Ingeniería Química; 2019

Asociación Argentina de Ingenieros Químicos

Peer-to-peer trading of utilities (heating, cooling, electric power, etc.) can become key cornesstone for companies (prosumers) in eco-industrial parks (EIP) to accomplish the utopy of optimal global solutions in a fully descentralized scheme. By the exploitation of synergistic co-generation of utilities and the trading of the surplus of those utilities, significant economic and environmental benefits can be reached through selfish collaboration. A crucial aspect of this symbiotic scheme is that each prosumer is considered as a selfish player in a game-theoretic manner, meaning that it will participate in this utility exchange only under the promise (and subsequent realization) of increasing its own profits. In this paper, an automated negotiation approach based on utility tokens and blockchain transactions is proposed as a posible mechanism design to incentivize participation in a market of prosumers for trading surpluses in an EIP. During each negotiation episode, prosumers engage in bilateral negotiations resorting to a previously learned policy to bid, concede and accept or reject, considering the context they are immersed in. Contract negotiation revolves around agreeing (or not) on the price expressed in tokens of a utility profile, given the private and public information (contextual variables) available to prosumers. Simulation results highlights that reinforcement learning allows finding policies that lead to near Nash equilibrium solutions for smart contract negotiation in an EIP.