A shift from fossil-based energy and products to more sustainable alternatives is essential to reduce greenhouse gas emissions and associated climate change impacts. Biomass represents a promising alternative for providing fuels and carbon-based products with reduced environmental impact. Combined with power-to-X technologies, it holds the potential to store energy from excess renewable electricity and valorize it in a multitude of low-impact energy carriers and chemical products. Biomass is a limited resource, creating the need for its efficient valorization. Due to the complexity induced by different processing pathways and products to which biomass can be converted, advanced computational methods offer great potential to support decision-making when designing biorefineries. This thesis aims to provide comprehensive decision-support tools that assist the development of integrated biorefinery designs for converting biomass to value-added products. Process modeling and integration are combined with rigorous multi-objective optimization approaches, including economic and environmental indicators. Multi-criteria decision analysis is explored with the acknowledgment of uncertainty in modeling parameters, providing robust solutions adapted to decision-makers. Furthermore, excessive computational loads evoked by simulation-based superstructure components and model complexity are addressed using surrogate modeling and machine learning. The developed methods are demonstrated on a Kraft pulp mill integrated with fuel production opportunities from biogenic residue streams. Results of the initial analysis of combined pulp and fuel production indicate that the carbon efficiency of the mill can be increased by up to 10% while offering economic and environmental benefits compared to conventional operation. Moreover, multi-criteria decision analysis coupled with interactive optimization illustrates the promising potential for generating and identifying solutions tailored to the needs and preferences of decision-makers. The consideration of uncertainties makes it possible to investigate the influence of model parameters on the optimization results and to leverage the obtained information for decision-making. Extending superstructure formulations by simulation-based model components increases model capability but introduces computational complexity. Therefore, surrogate modeling approaches are coupled with active learning, reducing the computational time of solution generation significantly. Furthermore, machine learning is used to efficiently generate Pareto frontiers for non-linear optimization problems and reduce computational time by up to 60%. When analyzing the integrated biorefinery in consideration of temporal variability of energy demands and resource availability, carbon capture, storage and utilization, and power-to-X technologies enable an increase of the mill's carbon efficiency to over 90%, compared to 50% for conventional operation. The integration of the mill with a nearby residential district reveals valuable synergies for both economic and environmental indicators. Depending on the costs of internal exchanges and the degree of self-sufficiency reached, emission reduction on the system level can lead to economic benefits for the actors of the considered system. Overall, it is shown that the suggested methods considerably enhance decision-making capabilities for integrated biorefinery design, enabling efficient valorization of biomass.