Model　predictive　control　(MPC)　algorithms　are　popularly　studied　for　modular　multilevel　converters　(MMCs)　because　of　the　multi-objective　regulation　capability　and　fast　dynamic　response.　However,　they　have　some　inherent　drawbacks,　for　example,　calculation　complexity,　non-fixed　switching　frequency,　complex　weighting　factor　determination　as　well　as　weak　steady　state　performance.　This　article　proposes　a　deadbeat　model　predictive　control　algorithm　for　MMCs.　The　proposed　algorithm　can　accurately　track　the　reference　of　output　ac　and　circulating　currents　in　one　control　period,　therefore　can　provide　a　fast-dynamic　performance.　In　addition,　switching　state　and　cost　function　calculation　and　weighting　factor　determination　are　removed.　Hence,　its　calculation　burden　is　very　low,　and　not　related　to　the　submodules　(SMs)　number.　Since　a　modulator　is　adopted,　a　fixed　switching　frequency　and　thus　a　good　steady-state　performance　are　achieved.　The　impacts　of　circuit　parameter　mismatches　as　well　as　SM　capacitor　voltage　ripples　on　the　control　algorithm　are　analyzed.　Improvement　measures　are　proposed　to　enhance　the　steady-state　performance　and　system　stability.　The　effectiveness　of　the　proposed　control　algorithm　is　verified　by　experiments.　©　2021　IEEE.