Optimal selection of raw grain for ethanol production at Lantmännen Agroetanol

Sammanfattning: Lantmännen is a Swedish agricultural cooperative in agriculture, machinery, bioenergy and food. With over 10,000 employees and 5 divisions, Lantmännen is the Nordic region’s largest ethanol manufacturer. In the Energy division business area are Agroetanol, which produces sustainable ethanol that is used for, among other things, biofuels. To produce ethanol, grains such as wheat, barley, triticale, corn and rye are mixed. The grain is purchased through forward contracts for up to a year before it is used in production. Depending on which mixture of grain is used, the amount of ethanol that is produced differs. Therefore, is it of great importance which mixture is used. Agroetanol currently use mixtures based on what is available in the procured contracts and production stocks, without regard to ethanol yields. The purpose of the master thesis is to provide a basis for Lantmännen Agroetanol that can help the company choose which mixture of raw grain is most suitable in relation to the ethanol yield. The main goal of the thesis is to create a model for calculating the expected ethanol yield in the production of specific mixtures, as well as an optimization model to optimize which mixtures are to be used from available stocks at each quarter. To achieve the goal, a statistical model was created based on data from Agroetanol. The model was created using a regression analysis in several steps. The steps began with a literature search on similar studies to determine what form the predictable model would take. At the same time as the literature search was ongoing, several interviews were conducted with staff from Agroetanol. The interviews aimed to collect historical data regarding ethanol production at the plant, as well as a survey of how the production was carried out and how the plant functioned. Based on the historical data collected, a statistical model could be produced. The model can be used to estimate the expected ethanol yield of different combinations of raw grains. The standard error of the regression was 1.3898 percentage points. The model can be used to determine the amount of each grain to be procured, to have a high ethanol yield. Based on the regression model, a mathematical mixing model for grains (BMFS) was created with the goal to plan how the available grain should be distributed in different mixtures. BMFS was based on the information about Agroetanol’s processes and production facility obtained through interviews. The model’s solution consists of which grains are to be included in each mixture for each week, how much of each grain is to be included in the mixture, from which contract the grain is to be procured and in which silo at Agroetanol it is to be stored before it is used in production. BMFS was optimized in AMPL as an optimization model, with the BARON solver. The solution is time consuming and therefore EpsR is used to make the solution easier. The EpsR value means that Baron stops looking for better solutions if the objective functioan value does not become better than the EpsR value. With the help of the regression model and BMFS, Agroetanol can improve its long-term production planning. With the regression model, they can already, when buying grain, plan according to which mixtures give the highest ethanol content. The solution from BMFS means that Agroetanol can plan the ethanol production for the entire quarter before the quarter begins.