ON-BOARD PREDICTION OF POWER CONSUMPTION IN AUTOMOBILE ACTIVE SUSPENSION SYSTEMS—II: VALIDATION AND PERFORMANCE EVALUATION

✍ Scribed by R. Ben Mrad; S.D. Fassois; J.A. Levitt; B.I. Bachrach

• Publisher: Elsevier Science
• Year: 1996
• Tongue: English
• Weight: 449 KB
• Volume: 10
• Category: Article
• ISSN: 0888-3270
• DOI: 10.1006/mssp.1996.0011

✦ Synopsis

The focus of this part of the paper is on validation and performance evaluation. The indirect (standard) and novel direct predictors of part I, which use time-recursive realisations and no leading indicators, are critically compared by using the non-linear active suspension system model. The results, constituting the first known comparison between indirect and direct schemes, show similar performance with a slight superiority of the former.

Experimental validation is based on an especially developed active suspension vehicle. The power consumption non-stationarity is, in this case, shown to be of the homogeneous type, and completely ''masking'' the signal's second-order characteristics, which are revealed only after the non-stationarity's effective removal. The analysis leads to two distinct types of indirect predictors: An explicit type, based on non-stationary integrated autoregressive moving average models, and an implicit type, based on stationary autoregressive moving average models. The explicit predictor is shown to be uniformly better than the implicit, although the difference is small for short prediction horizons. The experimental results indicate that accurate power consumption prediction is possible, with errors ranging from 2.22% for a prediction horizon of 0.156 s, to still less than 10% for horizons that are up to 0.470 s long, and about 25% for 1.563 s long horizons.