as per mr kai observations its true,actually the load will not be pure resistive or inductive,it will be mixed load and can`t predict as this supply is of multi utility>>>now let me tel u what exactally i am doing>>>suppose at given instance the pulse width is say 40%,what i did i samples the previousely mentioned bridge>>resistor devider derived output of winding for 10 milliseconds,totalling 64 samples and then adding all these to divide by 64 to get an average.then i checks whether the o/p is above a given threshold is its above that threshold then pulse width is reduced by some 3 or 4 percent in following cycle and at the same time if the o/p is less then threshold-20 then so pulse width of following cycle is increased bysome 2 or 3 %,the threshold-20 is just a hysterises provide.though the same scheme works well when the output is sampled directly from the output winding taps where load is actually connected,but dont know why its not working when sensed from a additional winding of o/p transformer.one more thing...suppose if the no load o/p voltages are 230 vac and if some 100 watt load is applied then o/p becomes say some thing 215 vac,though these 215 now remains stable till battery exhausts,and againg if the noload o/p volts are 230 vac and a 200 watt of load is applied o/p becomes 200 vac,though as same in above case now these 200 vac o/p remains stable till battery ends....so what i observed from all this is that at any given load there is fixed drop in o/p voltage,but i need some thing where the o/p should remain constant,i know its all due to some improper sampling.

AND AS PER MR ERIC`S COMMENT>>>
i had already tried the scheme proposed by mr eric in past,and offcourse this was the first thing i did,but this failed,may be due to some ringing etc ,i can not evaluate this on scope as i m not having one at hand,hope this all details will help u all guys....a great thanks again for all ur suggestions