In order to satisfy agro-industrial firm requirements, softwheat grains stored incountryelevators are often divided into batches characterized by different values of grain protein content. The decision to allocate farmer’s production to a high grain protein content batch or to a low grain protein content batch is often based on grain protein content measurements performed at harvest in trailers carrying wheat grains. This approach is quite expensive, does not always give a satisfactory result, and cannot be used to predict before harvest the quantity, and the quality of the production for crop collecting organization at a regional level. This paper presents a new method foroptimizing batches segregated incountryelevators. The general principle is to define and solve Linear Programming (LP) models including constraints on the batch weight and on the batch average grain protein content. The LP model coefficients are calculated by using field measurements or by using a crop model that predicts the values of yield and of grain protein content for the different fields of the collecting area. This method can be used to determine before harvest the optimal combination ofwheat fields that would give a batch with a satisfactory average grain protein content and with a maximal weight. A case study is presented in which several optimal batches are determined for various objectives and fora collecting area including 46 wheat fields. The characteristics of the batches obtained with our method are satisfactory when the grain protein content lower bound of the LP model is not higher than 11.5%.