—The　electrical　load　of　the　iron　and　steel　industry　is　significantly　influenced　by　physical　parameters　and　production　flow.　However,　the　current　research　neglects　the　interconnected　nature　of　consecutive　processes,　resulting　in　imprecise　simulations.　To　address　these　issues,　a　hybrid　mechanism-　and　data-driven　load　forecasting　model　for　the　iron　and　steel　industry　is　proposed.　Firstly,　the　attenuation　law　of　physical　variables　between　successive　processes　is　established　based　on　the　coupling　characteristics　of　material　and　energy　flow　during　transportation.　Then,　a　general　expression　of　power　function　is　developed　considering　the　equipment＇s　operational　characteristic.　Additionally,　physical　quantities　and　characteristic　parameters　of　the　power　function　are　fitted　using　Kernel-Based　Extreme　Learning　Machine　(KELM).　Finally,　the　predictive　power　of　different　processes　is　combined　in　time　domain　to　derive　the　total　power　curve　of　the　iron　and　steel　industry.　Through　case　studies　on　the　iron　and　steel　industry,　the　accuracy　and　superiority　of　the　model　proposed　in　this　paper　have　been　verified.　The　error　of　the　hybrid-driven　model　is　reduced　by　24.01　%　and　16.55　%　respectively　compared　with　the　mechanism-driven　model　and　the　data-driven　model,　highlighting　its　improved　accuracy　in　capturing　the　enterprise＇s　load　dynamics.　©　2025　Elsevier　Ltd