Human-Centric Predictive Maintenance Optimise Asset Performance - Hydrogen Plant
Abstract:
The use of hydrogen has been steadily increasing in popularity within the energy sector, providing a clean source of power for use within a number of different industries. With electrolysis proving a popular method of hydrogen production, the need to ensure the continued functionality and safety of electrolyser units is paramount. In this regard, thanks to the increased access to interconnected sensor data, the use of Predictive Maintenance is fast becoming the strategy of choice for ensuring the continued operational efficiency of industrial systems. However, with the increasing size of data collected comes the challenge of understanding what such data represents. This project marks the first stage of the development of a Predictive Maintenance system for use with such an electrolyser, with the aim of providing a strong foundational understanding of the data supplied by the industrial stakeholder. An initial exploration of the dataset reveals that its structure comprises sensor data from two Anion Exchange Membrane electrolyser units, one operating as expected and the other encountering faults throughout the operational period. The investigation transitioned to a more in-depth analysis of one of the discovered faults, relating to a temperature sensor. This included the use of visualisations to explore the sensor data during and around the fault, as well as an exploration of vailable sensors relating to temperature, such as other temperature sensors and cooling fan speeds. A statistical analysis was conducted, comparing the significance of the differences between sensor values during the start-up period prior to the error and the normal start-up periods, resulting in the identification of further possible connections, such as flow rate and voltage, providing a basis for further investigation. This investigation culminated in the initial stages of development for an interactive dashboard application upon which the Predictive Maintenance system would be built. At present, this dashboard acts only as an assistant tool for investigating the general operation of the system and the temperature sensor which triggered the fault.