This thesis investigates a set of advanced topics inspired from the heat integration and the pinch technology, all tied around the challenge of identifying energy savings in large-scale chemical industrial plants. In search for a general-purpose methodology for energy efficiency, it has been observed that some well established methods exist already, but either they are case specific or relatively short handed when it comes to dealing with the large size plants. A general multi-step methodology is proposed in this thesis and then continuously improved, with a particular focus on the industrial realism and the usability for large-scale problems. The general methodology has been first put in practice to identify energy savings options in real case applications with the purpose of identifying the challenges to be tackled. Different aspects of large scale systems integration are being tackled: we start by studying issues experienced during the data gathering stage and come up with a multilevel data extraction approach. It is based on the different representation of the process unit heat transfer interfaces that satisfy the same heat requirement and let us tune the required level of detail with respect to how detailed the energy requirement definition needs to be for the site scale integration. The combination of those heat requirement levels is then applied to the Total Site Integration (TSI), alleviating the difficulties of the data extraction for large-scale problems. The heat recovery targets of the total site are then explored to identify further energy savings by the introduction of advanced technologies such as the heat pump and the Mechanical Vapor Recompression (MVR). Following the heat recovery improvement target, the study investigates the integration of one of the major heat consumers in the chemical processes : the distillation columns. The thermodynamic optimization of distillation columns has been examined together with the TSI and the integration of advanced separation concepts like Dividing-Wall Columns has been investigated considering a holistic site scale approach. The retrofit of the Heat Exchanger Network (HEN) of large-scale plants has been studied next. The definition of the same heat requirement presenting different heat transfer units has been used to introduce the existing network into the optimization model. The purpose is to evaluate and to minimize the required modifications to the existing system. We have proposed a multi-objective optimization with the evolutionary algorithm and a Mixed Integer Linear Programming (MILP) model. The optimal Pareto-frontier has been generated with respect to the minimum operating cost, the investment cost, the number of required modifications and the thermal exergy destruction. This approach offers two main advantages: (i) it is helpful to predict the region of good solutions without going through numerous iterations between targeting levels, usually required in the classical heat integration methodology for the evaluation of the retrofit of existing facilities and (ii) it helps considerably reduce the size of the posterior HEN re-design problem by minimizing the number of streams to be considered. The concluding chapters consider ways to reduce the size and the computational complexities of the HEN design for industrial size problems, with a main focus on the Heat Load Distribution (HLD) subproblem.