This　paper　proposes　a　comprehensive　method　for　the　electromechanical　probabilistic　analysis　of　piezoelectric　energy　harvesters　subjected　to　modulated　and　filtered　white　Gaussian　noise　(WGN)　at　the　base.　Specifically,　the　dynamic　excitation　is　simulated　by　means　of　an　amplitude-modulated　WGN,　which　is　filtered　through　the　Clough-Penzien　filter.　The　considered　piezoelectric　harvester　is　a　cantilever　bimorph　modeled　as　Euler-Bernoulli　beam　with　a　concentrated　mass　at　the　free-end,　and　its　global　behavior　is　approximated　by　the　fundamental　vibration　mode　(which　is　tuned　with　the　dominant　frequency　of　the　dynamic　input).　A　resistive　electrical　load　is　considered　in　the　circuit.　Once　the　Lyapunov　equation　of　the　coupled　electromechanical　problem　has　been　formulated,　an　original　and　efficient　semi-analytical　procedure　is　proposed　to　estimate　mean　and　standard　deviation　of　the　electrical　energy　extracted　from　the　piezoelectric　layers.　©　2017　Elsevier　Ltd