Extrapolation of stack load spectrum for fuel cell bus based on kernel density estimation

ObjectiveThe key bottleneck, i.e., lack of an effective full-lifecycle load spectrum, is in current studies on mechanical durability of automotive fuel cell stack system. This study takes a demonstration fuel cell bus as the specific research object. It aims to construct a high-precision and high-reliability stack load spectrum covering the target mileage of 500 000 km for vehicles; provide core data support for the durability design, service life prediction, and performance optimization on fuel cell stacks; as wells as provide necessary supplements to the industry in the methodology for constructing load spectra over the entire lifecycle.MethodThe study adopted two-dimensional nonparametric kernel density estimation method as the core technical approach. First, it collected the original data of stack load-time history in the complete operating cycle conditions of the demonstration bus, ensuring the data covered typical driving conditions, e.g., starting, accelerating, constant speed, and decelerating. Gaussian kernel function combined with an adaptive bandwidth factor was selected in the data processing phase. Monte Carlo method was used to accurately fit the probability density function of two-dimensional random loads, constructing a full-process technical framework of data collection, kernel function selection, probability fitting, and load extrapolation. The collected limited-load data were extended at extrapolation rate of 15 times to form the preliminary result of a complete full-lifecycle load spectrum.ResultThrough multi-dimensional comparative verification of the extrapolated load spectrum with original measured data and traditional linear extrapolation results, the accuracy and reliability of the proposed extrapolation method are fully verified. Finally, the target full-lifecycle load spectrum of fuel cell stack for 500 000 km is successfully obtained. The data integrity and precision meet the requirements of durability research.ConclusionThis study effectively solves the industry-wide problem of lack of full-lifecycle load spectra for automotive fuel cell stacks through two-dimensional nonparametric kernel density estimation method combined with Monte Carlo fitting technology. It provides a standardized technical reference for subsequent related study and holds significant practical significance for promoting the industrialization development of fuel cell vehicles.