top of page


Mr. John Smith

Job title



As the hydrogen economy begins in earnest and the construction of hydrogen refueling stations increases, issues regarding the operation of hydrogen refueling stations are emerging. Typically, hydrogen tube trailers are essential for transporting high-pressure and large amounts of hydrogen gas to hydrogen refueling stations and demand sources. Since the amount of hydrogen in each trailer varies slightly, station managers are trying to document the amounts of gas in the tube trailer. If the gas pressure falls below a certain level, it could put a load on refueling station facilities, so it is good to replace the trailer on time to avoid safety risks. Since the charge of one-time transportation of tube trailers is quite expensive, replacing tube trailers too quickly for refueling station operators causes costly and inefficient operation. However, it is difficult for the station manager to predict every situation and request a replacement of the tube trailer at an appropriate time. It could be possible that the amount of hydrogen in the supply chain is insufficient, So the tube trailer delivery sometimes takes a long time. Therefore, an accurate forecast of the amount of residual gas can optimize the replacement of the tube trailer and reduce the risk of accidents caused by excessive use or damage to the trailer. This study proposes an operation solution for hydrogen refueling stations by forecasting the pressure of hydrogen tube trailers based on excellent time series forecasting models such as SCINet. This study creates a dataset based on refueling station real-time operation data and refueling station infrastructure data collected for about a year and performs time-series stationarity verification. Time series stationarity refers to a property of a time series where statistical properties such as the mean, variance, and covariance of the series do not change over time. In other words, the time series data shouldn't have any trend, seasonal effects, or systematic patterns that change over time. But stationarity is not satisfied when analyzing the original data. Differencing data makes itself have time series stationarity. In addition, statistical features of pressure over time add to the dataset. Finally, complete a custom dataset that has undergone a series of analyses and processing like normalization. Several state-of-art (SOTA) models that show excellent performance in time series forecasts are modified and optimized. And these models are trained with custom datasets to generate a refueling station-specific pressure forecast model. As a result of evaluating the performance by putting a verification dataset into several model algorithms learned based on custom data, the model that shows the best performance has a tube trailer pressure error(MAE-unit bar) of around 5.0. As a result of comparing the performance of each station with the same time series forecast algorithm, more errors occurred in refueling stations where store the inaccurate station operation data due to failure or omission of refueling station facility data. A future study plans to conduct additional research to supplement the results of refueling stations where inaccurate data are collected. In this work, we propose a method to forecast the pressure of hydrogen tube trailers at refueling stations using various time series forecast models such as SCINet. The solution shows significant results that can help reduce transportation costs that have a large portion of the budget for operation by forecasting pressure accurately and helping to replace tube trailers on time. It can also provide demand forecast monitoring services to hydrogen manufacturers and refueling station operators to help them plan for hydrogen supply. And it gives a guideline for refueling station managers on when to replace tube trailers and adjust the amount of hydrogen refueling. It will no longer happen that hydrogen car drivers are unable to refuel their vehicles due to sudden trailer replacement. Furthermore, this forecast result can be used as data to design hydrogen supply chains and predict facility failures in the future. Positive expectations are that this will significantly improve the efficiency and reliability of hydrogen refueling station operations and that it also causes the extension of the hydrogen fuel cell market.

Ms. Yeonjoon Han, sw engineer, HL Mando

A Study on the Forecasting Method of Pressure of Hydrogen Tube Trailer Using Deep-Learning-Based Time Series Forecast Model

FWC2023-REI-001 • FISITA World Congress 2023 • Road & energy infrastructure


Sign up or login to the ICC to download this item and access the entire FISITA library.

Upgrade your ICC subscription to access all Library items.

Congratulations! Your ICC subscription gives you complete access to the FISITA Library.


Retrieving info...

Available for purchase on the FISITA Store


bottom of page