An efficient uncertainty analysis of performance of hydrogen storage systems

The metal hydride hydrogen storage systems are gaining popularity due to their high volumetric capacity, safety and stability. The designing of these systems is complex due to many reasons, including input uncertainties. The performances of these systems are often evaluated in deterministic framework, ignoring uncertainties. This includes over-conservative safety factors in the design process increasing time and costs involved in designs. The uncertainty analysis could be a better alternative to assess system performance under such scenarios. This study investigates- i) the effect of input uncertainties on uncertainties of multiple and implicit system outputs, i.e., reaction fraction and bed temperature, ii) application of response surface and Borgonovo's global sensitivity analysis for efficient analysis, and iii) a comparative assessment between different uncertainty methods. The methodology is demonstrated for a space heating system. Initially, a surrogate relationship is constructed between inputs-outputs using moving least square response surface, based on known input-output data estimated using COMSOL for random input realizations. Next, the sensitive inputs were identified using Monte-Carlo simulations based Borgonovo's analysis. Finally, the effect of uncertainties of sensitive inputs on outputs were estimated using different uncertainty methods. Harr's and Hong's (2n) point estimate methods were observed to be highly accurate, mathematically simpler and efficient, as compared to other methods. The uncertainties of outputs were directly dependent on uncertainties of sensitive inputs. The probabilistic safety measure, reliability index, estimated using output statistics was of significant practical utility for industries to avoid deterministic safety factors based over-conservative and costly designs of storage systems.