Optimizing　power　purchase　model　is　an　effective　way　to　reduce　cost　of　power　purchase　to　the　minimum,　which　can　meet　the　purchased　profits　to　the　greatest　degree.　Therefore　research　of　power　purchase　model　and　algorithm　optimization　is　of　vital　practical　significance　for　power　purchaser.　This　paper　first　takes　total　expense　from　many　power　suppliers　as　an　objective　function,　the　model　of　power　purchase　for　power　companies　is　established,　which　synthetically　considers　supply　and　demand,　and　transfer　capability　constraints.　It　is　very　complex　to　deal　with　these　problems　by　classical　optimization　method　such　as　the　simplex　method　and　the　Lagrangian　relaxation　method.　Their　basic　shortcomings　are　long　searching　time　and　hard　finding　globe　optimal　solution.　The　particle　swarm　optimization　(PSO)　algorithm　is　applied　to　solve　the　model　which　makes　sure　that　the　solution　is　global　optimal,　then　introduces　a　novel　PSO　algorithm　which　is　modified　by　means　of　restricting　the　search　tactics,　simultaneously　considering　the　non-continual　variables,　and　through　ordering　the　power　suppliers　in　advance.　The　improved　particle　swarm　optimization　algorithm　makes　the　search　more　direct　and　faster.　Numerical　simulation　results　show　that　the　improved　PSO　algorithm　has　advantages　both　in　effectiveness　and　efficiency.　It　is　capable　of　obtaining　higher　quality　solution　efficiently　and　also　saves　considerable　costs,　furthermore,　the　algorithm　is　versatile　and　robust,　and　has　merits　of　higher　reliability　and　speed　of　convergence　than　other　methods.　Therefore,　it　is　concluded　that　the　algorithm　is　supposed　to　be　an　effective　way　to　deal　with　the　optimized　issue　in　the　power　market,　and　has　a　wide　potential　application　in　power　market.