Emission Stations Location Selection Based on Conditional Measurement GAN Data

Publication date: Available online 13 January 2020Source: NeurocomputingAuthor(s): Zhenyi Xu, Yu Kang, Yang CaoAbstractUrban vehicle emission monitoring can help make suggestions for the pollution emission control, and can protect public health. However, it is hard to get an overview of the vehicle emission in the city scale due to the sparse emission remote sensing stations in city, and selecting the appropriate stations locations which can reflect the emission variation in the given region mostly is another challenge. The existing methods solve the problem by spatial interpolation based on geographical statistics methods, without considering that urban vehicle emission varies by locations non-linearly and depends on many complex external factors. To tackle the spatial sparsity of vehicle remote sensing data, we design a data augmentation strategy based on Generative Adversarial Networks (GAN), which leverages prior model COPERT with conditional measurement data to filling the missing entries at other locations. With this strategy, we can generate realistic emission data while accelerating the training process. In addition, to address the emission stations location selection problem, we design a novel location selection strategy based on Spearman’s rank correlation coefficients, which leverages the realistic data generated to discover the grids with maximum link correlation for the pre-deployed station. Finally, we present experiments with the remote emission sensing data ...

Source: Neurocomputing - January 13, 2020 Category: Neuroscience Source Type: research