Optimization-Based Synthesis of Large-Scale Energy Systems by Time-Series Aggregation

Global greenhouse gas emissions significantly accelerate climate change and adversely affect life on earth. To mitigate these effects, greenhouse gas emissions need to be cut-down by significant reductions in primary energy consumption. Primary energy consumption can be reduced by increasing the efficiency of energy supply, which is mainly fixed during the synthesis of energy systems. Optimal synthesis of energy systems can be realized by mathematical optimization, however the solvable problem complexity is limited. In contrast, energy systems encompass highly complex energy conversion technologies and an increasing number of time-varying operation conditions. Thus, the synthesis of real-world energy systems usually result in large-scale optimization problems, which are computationally prohibitive. In this thesis, a solution framework is proposed to enable large-scale synthesis of energy systems. The framework exploits the two-stage character of synthesis problems: the decision stage of optimal investment for technologies and the operation stage. Thus, the framework solves two consecutive subproblems with reduced complexity by applying time-series aggregation methods. Time-series aggregation is performed by systematic clustering methods. Using these clustering methods, the framework simultaneously identifies the appropriate period length, required number of aggregated periods and number of aggregated time steps per period for accurate synthesis. Based on aggregated synthesis problems, the proposed framework calculates feasible solutions with known accuracy. In addition, the framework measures the accuracy as error of the objective function and iteratively refines the aggregation until an accuracy criterion is satisfied. The proposed solution framework for synthesis of energy systems is applied to two case studies motivated by real-world applications from industry. All results indicate that few time steps and few typical periods are sufficient to identify feasible near-optimal solutions of large-scale synthesis problems. Moreover, results show that intuitive aggregation methods (e.g., monthly averages) are not beneficial - only sophisticated clustering methods allow a strong time-series aggregation with high accuracy for synthesis of energy systems. Overall, the proposed framework outperforms available solution software in both solution time and solution quality.