An Eco-Industrial Park (EIP) is composed of a number of Industrial Symbiosis (IS) instances, which allow energy/material exchanges among the different industrial enterprises (Individual Actors, IA) therein. By so doing, IA's economic and environmental performances can be improved. Despite recent methodological advancements, the existing approaches to EIP design optimization are still suffering from several major problems: (i) dominance of the global EIP optimum over the local IA optimum, (ii) limited number of optimization objectives falling into the categories of economic and environmental objectives and (iii) EIP optimization performed without considering possible operational uncertainties. In addition, the adoption of the bio-inspired concept of IS makes EIP evolve from classical engineered systems towards complex engineered systems, with associated static and dynamic complexity characteristics. To highlight and understand the complexity features of EIP, in this paper we present them as intelligent networks for multiple energy and material exchanges, drawing a parallel with another typical complex system, that of the electric power network, in its modern Smart Grid (SG) concept, conceived to improve and optimize the distribution and use of electrical energy. Then, the modelling and optimization framework proposed in this paper adopts a more systematic methodology for accounting of the EIP complexity characteristics and addressing the associated optimization challenges. The proposed approach allows ensuring a sustainable and robust EIP design, thanks to the due account given to the related uncertainties and risks, e.g., due to major changes in the regulatory context and IA operational strategies, failures of interconnections among IA, interruption or shutdown of IA operation.