Abstract
•A novel hMPGA-WTD framework was proposed for representative driving cycle construction.•Real-world telematics data from 15,367 km was used to develop the driving cycle.•The proposed method was compared with micro-trip concatenation and Markov chain methods.•Cycle feasibility and fuel consumption rates were analyzed through vehicle and dynamometer tests.
Driving cycles are essential for vehicle energy and emission standards, but standard cycles often overlook regional variations. This study proposes a novel framework for constructing representative driving cycles for heavy-duty vehicles by optimizing a Markov chain with a hybrid Multi-Population Genetic Algorithm and Wavelet Threshold Denoising (hMPGA-WTD). The methodology involves: (1) spectral clustering of 1.27 million telematics data points to identify operational scenarios; (2) generating scenario-specific cycles using a two-dimensional Markov chain; and (3) enhancing driving cycle fidelity with hMPGA-WTD to reduce feature distortions. The method achieves a mean relative error of 4.61% across key parameters, with velocity standard deviation error below 1%. It outperforms micro-trip concatenation and conventional Markov approaches. Fuel consumption is validated via real-world and chassis dynamometer tests under identical loads, showing a 4.26% deviation, better than the 7.71% under CHTC-TT. The framework offers a robust basis for efficiency assessment and standard development.