From entry power analysis in glossary Statistics for HCI: Making Sense of Quantitative Data
Power analysis attempts to calculate the statistical power of a proposed experiment or study ahead of time. Given an effect size (the real difference you'd like to be able to detect) and the number of participants / trials you can calculate the probability of a Type II error (false negative). It can be used to assess the size of the experiment needed for a desired sensitivity.
Power analysis is based on (i) the experimental design, (ii) the planned statistical tests to be used, and (iii) knowledge of the distribution and size of the noise. However, the last of these will usually be estimates based on past studies in the area, so a power analysis, while useful, always comes with its own uncertainty.