We report a multi-harmonic histogram method for extracting and analyzing electric network frequency (ENF) signals to identify power grids. Given a voltage-time measurement of a power grid with a base frequency f0, we compute the ENF signals at multiple harmonic locations f_0 and extract (i) a histogram of the magnitudes of the ENF; (ii) a histogram of the signal power and noise power surrounding the ENF; (iii) a histogram of the signal-to noise-ratio (SNR) of the ENF. Different from existing methods which use spectral-combining techniques to compress multiple harmonics into one single ENF, the proposed multi-harmonic histogram method retains significantly richer information which allows one to perform classification. At the classification stage, we propose a histogram matching method with a multi-layer decision rule for identification. On the practice data set consisting of 50 testing samples, we achieve 92% overall accuracy.