As the penetration level of photovoltaic（PV） generation increases rapidly in recent years， PV capacity credit evaluation is becoming one of the most important issues in the power system planning. Considering the uncertainties of PV output and network loads， a probabilistic model based on improved-sequential hour-deterministic method is established. The deterministic model reflects the long-time scale correlation between the PV output and the network load and also between different PV outputs. The rank correlation coefficients reflect the short-time ones. Based on Latin hypercube sampling（LHS） technique， a novel evaluation method for the PV capacity credit is proposed by considering the correlations between random variables. The effectiveness and accuracy of the method are proved by simulation tests on IEEE-RTS 79 system.