Comparison of Three Methods for Constructing Real Driving Cycles

Abstract

This work compares the Micro-trips (MT), Markov chains–Monte Carlo (MCMC) and Fuel-based (FB) methods in their ability of constructing driving cycles (DC) that: (i) describe the real driving patterns of a given region and (ii) reproduce the real fuel consumption and emissions exhibited by the vehicles in that region. To that end, we selected four regions and monitored simultaneously the speed, fuel consumption and emissions of CO2, CO and NOx from a fleet of 15 buses of the same technology during eight months of normal operation. The driving patterns exhibited by drivers in each region were described in terms of 23 characteristic parameters (CPs) such as average speed and average positive kinetic energy. Then, for each region, we constructed their DC using the MT method and evaluated how close it describes the observed driving pattern in each region. We repeated the process using the MCMC and FB methods. Given the stochastic nature of MT and MCMC methods, the DCs obtained changed every time the methods were applied. Hence, we repeated the process of constructing the DCs up to 1000 times and reported their average relative differences and dispersion. We observed that the FB method exhibited the best performance producing DCs that describe the observed driving patterns. In all the regions considered in this study, the DCs produced by this method showed average relative differences smaller than 20% for all the CPs considered. A similar performance was observed for the case of fuel consumption and emission of pollutants.

Comparing Driving Cycle Development Methods Based on Markov Chains

“…As ( 21 ), we studied the evolution of the average RD for different numbers of iterations fixing the microtrip segmentation and clustering algorithm to the first method and fixing the number of clusters. We completed the entire process of building and selecting driving cycles several times for several iterations between 1 and 1,000.…”

“…Preliminary results have been executed in oder to choose the number of random walks. The relative difference (RD i ) between each of the assessment parameters of the obtained cycle (CP i ) and the targeted parameters (TP i ) is used to compare the cycles ( 21 ) and is calculated with Equation 1:…”

“…As brought up by ( 7 , 21 ), one of the limitations of the microtrip and Markov-chains approaches is that they produce a different cycle each time, because of their stochastic nature. To overcome this limitation, ( 22 , 23 ) and the Hong Kong Cycle ( 7 ) repeated the construction process until the cycle met the assessment measures within a threshold ( 23 ).…”

“…Others repeat the construction process several times to select later the best candidate among the generated cycles. ( 21 ) argue that 500 iterations allow the confidence intervals of the RDs to stabilize. The same authors used 1,000 iterations in their recent work ( 17 ).…”

Study on Driving Cycle Synthesis Method for City Buses considering Random Passenger Load

“…ere are various methodologies to develop a typical driving cycle, which can be substantially classified into four categories: microtrips, segment-based, pattern classification, and Markov chain method [32][33][34][35]. Microtrips are usually extracted from the measurement data sequence between two uninterrupted time points at which the vehicle is stopped [12].…”