<p>The emergence of a new epidemic impacts overall aspects of human life, disrupting healthcare systems, and economies. The COVID-19 pandemic, caused by the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), exemplified this challenge on a global scale. In response, countries implemented various non-pharmaceutical interventions, such as mobility restrictions, along with extensive vaccination programs to mitigate the spread and severity of the disease. To address the complexity of modeling such a dynamic and multifaceted pandemic, this study proposes a novel ten-compartment epidemiological model named Susceptible-Exposed-Infected-Recovered-Deceased-Protected-Vaccinated (SEIRDPV). The proposed model extends the classical SEIRD model by introducing additional compartments to account for multi-dose vaccination schemes, protected individuals, and immunized susceptibles. Furthermore, the model integrates factors such as vaccine efficacy, demographic characteristics, mobility restrictions, geographical variations, and the emergence of new viral variants through time-varying parameters. To estimate these parameters efficiently, a swarm intelligence based optimization algorithm is employed, enabling the model to adapt to changing conditions while minimizing computational overhead. The SEIRDPV model is rigorously evaluated using COVID-19 confirmed cases and death cases from three regions: India and the USA at the national level, and California at the subnational level, focusing on the period following the initiation of vaccination rollouts. Extensive empirical analysis demonstrates that the proposed model closely aligns with real-world epidemic trends and significantly outperforms existing state-of-the-art forecasting methods across five standard performance metrics. The results affirm the potential of the proposed model as a robust and flexible framework for epidemic forecasting, with practical implications for policy-making in future outbreaks.</p>