The　distribution　network　environment　is　complex　and　has　serious　harmonic　problems.　The　large-scale　integration　of　new　energy　has　exacerbated　the　diversification　of　harmonic　sources　and　voltage　distortion.　Aiming　at　the　poor　estimation　accuracy　of　traditional　dynamic　harmonic　state　estimation　algorithm　when　there　are　time-varying　noise　and　abrupt　data,　an　adaptive　harmonic　state　estimation　algorithm　for　distribution　networks　considering　new　energy　pseudo-measurement　is　proposed.　Firstly,　the　Sage-Husa　noise　estimation　method　based　on　the　forgetting　factor　is　introduced　to　estimate　the　time-varying　noise　interference　when　the　noise　covariance　is　reduced　in　real-time.　Then,　an　adaptive　factor　is　constructed　to　online　correct　the　error　variance　matrix,　thereby　reducing　the　prediction　error　caused　by　sudden　load　changes.　Secondly,　a　confidence　interval　pseudo　measurement　generation　method　for　quantile　regression　Bayesian　gated　recurrent　neural　network　is　proposed　to　solve　the　problem　of　missing　short-term　harmonic　data　in　new　energy　and　improve　effect　of　the　real-time　harmonic　state　estimation.　Finally,　the　effectiveness　of　the　algorithm　proposed　in　this　paper　is　verified　by　an　modified　distribution　network　system.