In the context of the ongoing COVID-19 pandemic, while millions of people await the administration of a vaccine, social distancing remains the leading approach towards the effect commonly known as "flattening the curve" of infections. Over the last year, governmental administrations throughout the globe have implemented various lockdown policies in hopes of slowing down the transmission of the disease. However, the current lack of consensus on when and how these policies should be implemented reflects the need for further studies regarding these questions. In this paper, we tackle the issue of lockdown policy management, in particular in terms of lockdown placement (how often, when, and how long these periods should be), in order to minimize the peak of infections in a specific population. We introduce a novel combination of classic mathematical disease modelling using the equation-based SEIR model, and Evolutionary Strategies (ES) for optimizing the peak of infections. The method is evaluated using data collected in different countries, and a particular focus is placed on the study of the effect of specific model parameters on lockdown optimization, such as the transmission rate ($\beta$), of which 4 alternative modelling functions have been proposed and analyzed. Our results indicate that this transmission rate parameter significantly influences the resulting optimal strategies. In particular, the presence of a gradual decay of the rate of transmission during lockdown leads to longer, more sparsely placed confinement periods while an abrupt, instantaneous drop in the amount of contacts per person favors shorter but more frequent lockdowns. Although these results are limited by the scope of action provided by the simplicity of the SEIR model, they suggest that the influence of the evolution of the rate of transmission along the disease should be assessed in further studies with alternative optimization strategies (agent-based) and models (SEIRSHUD).