Benchmark calculations of Canada deuterium uranium (CANDU) reactor physics design and analysis codes have been performed for a lattice code WIMS-AECL, a supercell code DRAGON, and a core analysis code RFSP by using the physics measurement data of Wolsong nuclear power plants. In this study, the lattice and reactivity device models were examined based on Wolsong-2 measurement data for the criticality and reactivity device worth. Sensitivity calculations were also performed for the number of energy groups and the cross-section library. Using the lattice and reactivity device models obtained from the Wolsong-2 calculation, the benchmark calculations were extended to the Wolsong-3 and Wolsong-4 plants. Compared to a previous study, this study showed that the results of the criticality and reactivity device worth calculations were improved when the material data were updated and the exact two-group cross sections were used. For the three nuclear power plants, the calculated core reactivity was within 0.2% k of criticality. The zone controller unit reactivity worth was estimated to have a maximum error of ~8%. The total reactivity worth of other reactivity control devices was consistent with the measurement data within 13%. The root-mean-square error of the flux distribution calculation was <12% when compared with flux scans performed during Phase B physics tests. In conclusion, the CANDU physics design and analysis codes used in this benchmark study predicted the physics parameters within the allowed uncertainty level of the measurement data.