The short answer is that there are many methods that can achieve this. If you are wanting to adopt an assumed model form e.g., mean structure with covariates and seasonality, then you can identify changes within the data and use the last segment to forecast the series forward (usually assuming no further changepoints). This paper (open access) discusses the pros and cons of adding changepoints at different model-building stages.

If you are wanting to detect departures away from an assumed model (either mean or variance/dependence structure) then this talk may be helpful.

The author of both of these, Rebecca Killick, has also published papers using wavelets to identify changepoints, a couple of links below. You could reach out to them for help with your question too. There are a small number of researchers looking at changepoints and wavelets, these include, from the UK, Haeran Cho, Piotr Fryzlewicz and Idris Eckley.