The COVID-19 pandemic, caused by the SARS-CoV-2 virus, has been a global health crisis since its emergence in December 2019. South Korea, one of the first countries after China to be affected, has faced multiple waves of infection and the emergence of new variants of the virus. This study aims to understand the spread dynamics of COVID-19 in South Korea using a Physics Informed Neural Networks (PINNs) framework. Using the susceptible-infected-recovered-dead (SIsRD) model as the mathematical basis for PINNs, we analyzed the infection, recovery and mortality rates from the COVID-19 virus and predict the number of actively infected, recovering, susceptible and deceased individuals at any given time. Data from four South Korean cities for two time periods (December 7, 2020~December 6, 2021 and March 5, 2022~March 4, 2023) were used to train the neural network. Our results contribute to the understanding of the epidemiological impact of the simultaneous spread of different COVID-19 variants and have the potential to develop future public health strategies.