Neural network based fuzzy cognitive reliability and error analysis method for risk assessment of hydrogen valley design under uncertain data

Abstract

Hydrogen Valleys are sustainable ecosystems that combine hydrogen that is a cost-effective, efficient, and environmentally friendly clean energy alternative for production, storage, and distribution operations. Safe and sustainable operations of these ecosystems require proactive risk assessment. This paper aims to develop a standardized risk assessment methodology that provides an expert-validated decision-support tool for the design phase of Hydrogen Valleys by proposing an integrated methodology that incorporates Artificial Intelligence (AI), the Cognitive Reliability and Error Analysis Method (CREAM), and the fuzzy set theory (FST). The structural model is designed with FST based DEMATEL integrated with fuzzy Cognitive Mapping (FCM). The weights derived from expert insight are then embedded as initial weights in a multilayer-perceptron based on the structural model. Then, Artificial Neural Networks (ANNs) are used for risk classification probabilities based on the CREAM's common performance condition (CPC) concept. Finally, the ANN probabilistic outputs are converted into fuzzy membership grades for the fuzzy-inference system (FIS) to obtain overall risk index. The proposed methodology is applied for the case of T & uuml;rkiye, an emerging hydrogen economy with proximity to energy demand points in both Europe and Asia. As result, inflation & interest-rate changes, failure to plan procurement processes, and uncertainties in policy and legal regulations are determined as the most crucial risk related to design of hydrogen valleys. The sensitivity and comparative analyses are also carried out to validation and verify the results. Based on the results and analyses, this paper contributes to applying knowledge and experiences from current operational and engineering fields to the hydrogen energy environment, to streamline its assessment process, and accelerate hydrogen energy development by applying machine-learning based fuzzy CREAM risk assessment methodology.