I have a time series dataset that has yearly data points, and we can see that there is recurring pattern every few years
My question is : can I use SARIMA with this model, by changing my dataset as a time series dataset that has a frequency? as technically speaking there are no seasonality components, would that be frowned upon?
Certainly you can do that, if you know the frequency of your seasonality. After all, there are annual time series that do exhibit cyclical behavior, e.g., sunspots have an 11-year cyclicality and could thus be modeled using SARIMA.
That said, I do not see a recurring pattern in your time series, other than "sometimes it's high, and then it usually comes down again, perhaps after a while, except when it doesn't, and then it goes up again, usually at least". May I suggest that you are at least open to the possibility that you are reading more into your data than is actually there? Unless there is some theory that underpins a five-year cycle (say), but not a seven-year one, I would be very careful indeed. (And note that humans are very good at post hoc explanations of stuff, even if those explanations do not help in predicting.)
Building on Stephan's answer, if you do think there is a seasonal pattern in your data, I would certainly apply some tests for seasonality to your data instead of just eyeballing it. I think this thread is a great place to start.
