A　full-scale　measurement　of　wind　pressure　loadings　was　conducted　on　photovoltaic　(PV)　solar　panels　mounted　on　the　flat　roof　of　a　multipurpose　low-rise　building.　According　to　the　characteristics　of　wind　pressure　time　series,　several　probability　density　functions　were　properly　selected　for　nonlinear　regression　analysis　of　wind　pressure　coefficients　of　each　gauging　point.　Numerical　fittings　of　mean　and　peak　(positive　maxima　and　negative　minima)　and　area-averaged　peak　pressure　coefficients　were　performed　and　compared　by　using　six　probability　distributions.　The　analysis　results　show　that　the　probability　distributions　of　negative　minima　and　positive　maxima　of　pressure　coefficients　are　left-skewed　and　right-skewed,　respectively.　The　mean　pressure　coefficient　time　series　comply　with　unbiased　distribution　and　the　t　distribution　fits　the　measured　data　best,　followed　by　Logistic,　Normal,　Type　I　EDV　distribution,　respectively;　Type　I　EVD　has　the　best　fitting　effect　for　peak　values,　and　then　the　Lognormal　distribution,　Gamma　distribution　and　Normal　distribution　in　order;　The　most　unfavorable　pressure　coefficient　is　dominated　by　negative　peak　pressure,　that　is,　suction　forces　are　predominant　over　wind　forces　on　PV　panels;　The　tributary　area-averaged　unfavorable　pressure　coefficient　is　recommended　to　take　-2.3.　©　2016,　Science　Press.　All　right　reserved.