A self-organized criticality model is given to simulate power system blackout process when the power flow limits of all system components are relatively fixed. It is used to investigate the faults of the IEEE 39-bus power system with different load rates. The influences of the distribution of load rates on power system self-organized criticality are analyzed. The results indicate that the dispatchers should pay attention to both the degree of overall system load rate and the slope for each component load rate in double logarithmic plot at the same time. Under the same system load rate， larger absolute value of the slope of distribution curve indicates higher probability that a large -scale cascading failure may happen in the power system.