Usually we take absorption maxima as excitation wavelength but you can also excite at other absorption bands and check which excitation wavelength gives you maximum emission intensity.

the excitation energy should be higher than the expected emission energy

In the case where the investigated structure has a band gap the excitation energy should be higher than the band gap

Usually, wavelength for excitation should be in correspondence with absorption spectrum: it can be absorption maximum (preferentially) or another wavelength around the maximum (in this case the luminescence intensity normally should be lower).

The question is: what do you want to investigate? If you want to see integral spectra, Bouraoui Ilahi is completely right. If you want to investigate connections between high and low energy emission peaks, it might be helpul to excite resonantly. If you decide to use the absorbance maximum wavelength for excitation, take care not to dissociate the material or to introduce structural defects which might affect your spectra as well.

In PL excitation, the detection energy is fixed at the energy of the investigated level and the excitation enrgy, which should be near and upper of the energy level, is varied. The excitation energy could be selected from a fluorescence band of a dye laser with appropriate colorant, a lamp emitting white light or or other devices.

Excuse me, you talk about PL emission and not PL excitation, so the excitation energy should be higher than the investigated energy domain.

The answer of Gopal Singth is the best for undestanding. You should know absorption (or PL excitation)  band range.