Alkaline Liquid Hydrazine Fuel Cell: from Catalyst Development to AI-based Performance Prediction

Abstract

Although vehicles that are equipped with proton exchange membrane fuel cells (PEMFCs) are beginning to be seen on the streets, there are still limits to their widespread use. This is because hydrogen is intractable and there is a price burden due to the large use of platinum group metal (PGM) catalysts. Therefore, the use of non-PGM catalysts and liquid fuels could be a suitable way to reduce the hurdle in terms of the high price of fuel cells and the supply of fuel cells to the public. In this regard, the hydrazine fuel cell (HzFC), which is one of the alkaline liquid fuel cells, can be a good candidate because it does not need PGM catalysts and produces no carbon-based gases during fuel cell operation. Consequently in this dissertation, we focus on (1) the catalyst development for hydrazine oxidation, (2) control of the key factors for hydrazine fuel cell stack-making, and (3) the operational analysis via artificial intelligence to enable the commercialization of hydrazine fuel cells. In the process of catalyst development, we systematically investigated the electrocatalytic activity of hydrazine oxidation by carefully adjusting the ratio of three different non-PGM catalysts (Ni, Co, and Cu). We successfully controlled that nickel, cobalt, and copper serve as the main electrocatalyst, the promoter (lowering onset potential), and the electron-transfer helper, respectively. Secondly, to extend HzFC to the stack, we conducted a lot of HzFC tests by controlling various factors. Because these operational factors are inter-dependent variables rather than independent variables, the optimized condition in single-cell could not be adapted to the stack. We confirmed the difference in the optimized condition between the single-cell and HzFC stack. So we experimented under more conditions, accumulated the fuel cell results, and pre-treated features to establish the database for artificial intelligence (AI). The operating factors of the fuel cell were prioritized through AI, which helps to understand problems that occur when the reaction area becomes larger and the internal structure becomes complicated, rather than verifying the performance of the prepared catalyst. Without human judgment, the algorithm prioritized the operational factor of the cathode part, which is a more sluggish reaction than the anode side.