Evaluation of a Stochastic Vehicle Travel Pattern Generation Model with Real-World Travel Data

Higher electric vehicle penetration and associated grid charging will have significant reliability impacts on electric power systems. Therefore, distribution network planning studies need suitable models to capture spatial and temporal distribution of vehicle charging demand. When making expensive planning decisions, it is important to have high confidence in the models used. This paper proposes a systematic method and metrics to evaluate a stochastic model that generate random vehicle travel patterns in an urban area. The particular model considered consists of a collection of Hidden Markov Models, and its efficacy is evaluated using a database of actual vehicle travel data from Winnipeg, Manitoba, Canada. The probabilities associated with the model were extracted from the actual data. The evaluation approach involves comparison of actual and simulated values of model outputs representing the spatial-temporal distributions and use of Euclidean distances to measure the similarity of probability distributions produced by actual and simulated data.