The statistical properties of electricity price are important information for generator investment and evaluating the market risk. The statistical properties of day-ahead prices in Pennsylvania-New Jersey-Maryland（PJM） market are discussed by Kolmogorov-Smirnove method in this paper. The conclusions are as follows: First of all， the probability density function on day-ahead locational marginal prices（LMPs） in PJM market in every year obeys log-normal distribution; Second， by classifying the electricity prices in different levels according to its load， the LMPs show the log-normal distribution when load is in middle level or in high level， but if the load is in low level， the LMPs show the normal distribution; At last， the relations of load with average and root mean square of corresponding LMPs are approximately linearity in different load situations.