This is probably not what you want to do. We are in an age where AI is becoming much more intelligent in helping us find what we want when we search, but classic keyword searches do not use AI and they do not work the way you seem to think. My comments that follow apply to classic keyword searching, not to rapidly evolving AI-based searching (for which I have no general comments, other than that its potential is very exciting).
It is a common misunderstanding that with a keyword search, we tell the computer what we want it to find for us. No, a keyword search tells the computer what we do NOT want to see. This makes more sense if you think of something like this: suppose we have 1000 articles in our database. A keyword search of "apple* OR banana*" does not tell the computer to give us all articles about apples or bananas. Rather, it tells the computer to show us everything it has except for articles that do not mention apples or bananas. Suppose the computer comes up with 500 articles that mention apples or bananas. But 400 of these talk about pears. I'm not interested in pears, only apples or bananas. If I then search for "apple* OR banana* NOT pear*", I probably would not get what I want. I might end up with 25 articles that indeed talk about apples and bananas with no mention of pears, but I would exclude 375 articles that mention all three. Most likely some of those excluded articles would be very interesting to me, especially those that talk about "why apples and bananas are great but pears are not".
Please excuse my simplistic example, but I hope it makes it clear that by searching for "NOT optim*" (or something like that), you would exclude articles that cover "the energy topic you really want, but not focusing on optimization".
Keyword searches are best used as an initial broad net that should catch many articles. The NOT keyword should be used very sparingly. The appropriate approach would be to search for the keywords that you want and then read the titles and abstracts of all the articles you identify and then manually exclude the ones that are overly focused on optimization. I hope that makes sense.