The daily food intake for dairy cows has to be effectively controlled to breed a sound group of cows as well as to enhance the productivity of the cows. But, feed stuffs are fed in the common bulk for a group of cows in most cases despite that the individual food intake has to be varied. To obtain the feed for each cow, both the nutrient requirements and the nutrient composition of fred have to be provided in advance, which are based on the status of cows such as weigh marginal weight amount of milk, fat concentration in milk, growth and milking stages, and rough feed ratio, etc. Then, the mixed ration fur diet would be computed by the nutrient requirements constraints. However, when TMR (Total Mixed Ration) is conventionally supplied for a group of cows, it is almost impossible to get an optimal feed mixed ration meeting the nutrient requirements of each individual cow since the constraints are usually conflicting and over-constrained although they are linear. Hence, addressed in this paper is a simulated annealing (SA) technique to find the food intake for dairy cows, considering the characteristics of individual or grouped cows. Appropriate parameters fur the successful working of SA are determined through preliminary experiments. The parameters include initial temperature, epoch length. cooling scheduling, and stopping criteria. In addition, a neighborhood solution generation method for the effective improvement of solutions is presented. Experimental results show that the final solution for the mixture of feed fits the rough feed ratio and some other nutrient requirements such as rough fiber, acid detergent fiber, and neutral detergent fiber, with 100 percent, while fulfilling net energy for lactating, metabolic energy, total digestible nutrients, crude protein, and undegraded intake protein within average five percent.