Bolanča Tomislav, Strahovnik Tomislav, Ukić Šime, Stankov Mirjana Novak, Rogošić Marko
Faculty of Chemical Engineering and Technology, University of Zagreb, Marulićev trg 19, Zagreb, Croatia.
Croatian Energy Regulatory Agency, HERA, Ulica grada Vukovara 14, Zagreb, Croatia.
Environ Sci Pollut Res Int. 2017 Jul;24(19):16172-16185. doi: 10.1007/s11356-017-9216-x. Epub 2017 May 24.
This study describes the development of tool for testing different policies for reduction of greenhouse gas (GHG) emissions in energy sector using artificial neural networks (ANNs). The case study of Croatia was elaborated. Two different energy consumption scenarios were used as a base for calculations and predictions of GHG emissions: the business as usual (BAU) scenario and sustainable scenario. Both of them are based on predicted energy consumption using different growth rates; the growth rates within the second scenario resulted from the implementation of corresponding energy efficiency measures in final energy consumption and increasing share of renewable energy sources. Both ANN architecture and training methodology were optimized to produce network that was able to successfully describe the existing data and to achieve reliable prediction of emissions in a forward time sense. The BAU scenario was found to produce continuously increasing emissions of all GHGs. The sustainable scenario was found to decrease the GHG emission levels of all gases with respect to BAU. The observed decrease was attributed to the group of measures termed the reduction of final energy consumption through energy efficiency measures.
本研究描述了一种利用人工神经网络（ANN）测试能源部门减少温室气体（GHG）排放不同政策的工具的开发。阐述了克罗地亚的案例研究。使用两种不同的能源消耗情景作为温室气体排放计算和预测的基础：照常营业（BAU）情景和可持续情景。它们都基于使用不同增长率预测的能源消耗；第二种情景中的增长率源于最终能源消耗中相应能源效率措施的实施以及可再生能源份额的增加。对人工神经网络架构和训练方法进行了优化，以生成能够成功描述现有数据并在时间向前推移时实现可靠排放预测的网络。发现BAU情景会导致所有温室气体排放持续增加。发现可持续情景相对于BAU情景会降低所有气体的温室气体排放水平。观察到的减少归因于通过能源效率措施降低最终能源消耗这一组措施。
