To　determine　the　population　size　of　multi-objective　particle　swarm　optimization　algorithm　(MOPSO),　an　improved　MOPSO,　based　on　the　convergence　speed　and　diversity,　named　CD-MOPSO,　is　proposed.　Firstly,　the　fitness　function　of　population　size,　which　is　developed　by　the　convergence　speed　and　diversity　during　the　evolutionary　process,　is　used　to　describe　the　relationship　between　the　population　size　and　the　performance　of　MOPSO.　Secondly,　according　to　the　fitness　function,　an　adaptive　adjustment　method　is　designed　to　update　the　population　size　of　MOPSO　dynamically.　Finally,　the　proposed　CD-MOPSO　is　tested　on　the　ZDT　benchmark　optimization　problems　and　applied　to　a　real　optimization　problem　of　urban　pipe　networks.　The　experimental　results　show　that　the　proposed　CD-MOPSO　can　adjust　the　population　size　automatically　according　to　the　problem,　compared　with　the　performance　of　NSGA,　MOPSO,　SPEA2　and　EMDS-MOPSO,　CD-MOPSO　has　faster　convergence　speed　with　better　optimization　results.　©　2018,　Chinese　Institute　of　Electronics.　All　right　reserved.