Industry　power　customers　occupy　most　of　the　national　electricity　consumption,　and　analyzing　the　electricity　consumption　patterns　of　industry　power　users　to　understand　their　electricity　consumption　characteristics　is　an　important　foundation　for　fine-grained　response　to　the　shortage　of　electric　energy　supply.　To　address　the　problem　of　high　dimensionality　of　electric　load　data　and　difficulty　in　extracting　temporal　features　at　the　present　stage,　this　paper　proposes　a　Deep　Embedding　Clustering　method　based　on　Long　Short-Term　Memory　Auto-Encoder　(DEC-LSTM-AE).　First,　LSTM-AE　is　used　to　extract　the　temporal　features　embedded　in　the　load　data.　Then,　the　extracted　load　feature　vectors　are　soft-segmented　using　a　custom　clustering　layer.　Finally,　Kullback-Leibler　Divergence　(KLD)　is　used　as　the　loss　function　to　jointly　optimize　LSTM-AE　and　the　clustering　layer　to　obtain　the　clustering　results.　The　model　analysis　shows　that　the　method　is　superior　in　Davies-Bouldin　Index,　Silhouette　Coefficient　Index　(SCI)　and　Classification　Validity　Index　(CVI)　indexes,　and　its　method　can　effectively　improve　the　accuracy　of　industry　user　load　clustering.　©　2022,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.