The invention provides a recursive optimized least mean square beam forming method applied to a high throughput satellite, relates to the field of satellite communications, and specifically relates to an adaptive beam forming method for the high throughput satellite. The method comprises the following steps: firstly, constructing a Ka band channel model based on a rain attenuation prediction model of the ITU-R; secondly, initializing an array element weight vector based on evenly arranged phased array antennas with N array elements; and finally, at the kth sampling time, correspondingly inputting a signal x(k), an expected signal d(k) and a forgetting factor alpha, and a correlation matrix Rxx(k)=alphaRxx (k-1)+x(k)xH(k), calculating a gain vector, calculating an antenna output signal according to the array element weight vector, and meanwhile, obtaining an output error value according to the expected signal and the antenna output signal, and updating the weight vector and judging whether the iteration is convergent according to an error squared value, as the following formula. The invention is applicable to a satellite-borne beam forming process.