Neural Network-Based Surrogates for Alkaline Electrolyser System Modeling in Smart Grid Model

Abstract

Modeling of a complex detailed system-of-systems (SoS) can require substantial computational power. In some cases, it can be more beneficial to use neural network (NN)-based models as surrogates to model the subsystems. Surrogate modeling technique have been used in various engineering disciplines, providing approximation of complex system behaviors with reduced computational cost. An example of such complex modeling problem is a modern electrical grid that consist of multiple functions and subsystems, e.g. flexible loads, that interact with each other. This paper focuses on data-based modeling of a controlled alkaline electrolyser (AEL) system as part of a smart grid model. In this paper, the presented case-study describes the replacing of the modeled AEL system with NN surrogate and its further verification through tests in time domain under certain grid conditions.