论文标题
部分可观测时空混沌系统的无模型预测
Causal impact of severe events on electricity demand: The case of COVID-19 in Japan
论文作者
论文摘要
储层计算是预测湍流的有力工具，其简单的架构具有处理大型系统的计算效率。然而，其实现通常需要完整的状态向量测量和系统非线性知识。我们使用非线性投影函数将系统测量扩展到高维空间，然后将其输入到储层中以获得预测。我们展示了这种储层计算网络在时空混沌系统上的应用，该系统模拟了湍流的若干特征。我们表明，使用径向基函数作为非线性投影器，即使只有部分观测并且不知道控制方程，也能稳健地捕捉复杂的系统非线性。最后，我们表明，当测量稀疏、不完整且带有噪声，甚至控制方程变得不准确时，我们的网络仍然可以产生相当准确的预测，从而为实际湍流系统的无模型预测铺平了道路。
As of May 2022, the coronavirus disease 2019 (COVID-19) still has a severe global impact on people's lives. Previous studies have reported that COVID-19 decreased the electricity demand in early 2020. However, our study found that the electricity demand increased in summer and winter even when the infection was widespread. The fact that the event has continued over two years suggests that it is essential to introduce the method which can estimate the impact of the event for long period considering seasonal fluctuations. We employed the Bayesian structural time-series model to estimate the causal impact of COVID-19 on electricity demand in Japan. The results indicate that behavioral restrictions due to COVID-19 decreased the daily electricity demand (-5.1% in weekdays, -6.1% in holidays) in April and May 2020 as indicated by previous studies. However, even in 2020, the results show that the demand increases in the hot summer and cold winter (the increasing rate is +14% in the period from 1st August to 15th September 2020, and +7.6% from 16th December 2020 to 15th January 2021). This study shows that the significant decrease in electricity demand for the business sector exceeded the increase in demand for the household sector in April and May 2020; however, the increase in demand for the households exceeded the decrease in demand for the business in hot summer and cold winter periods. Our result also implies that it is possible to run out of electricity when people's behavior changes even if they are less active.
