Improvement potential for fuel planning optimization using BoFiT

Abstract: In 2018, Stockholm Exergi, the largest district heating producer in Stockholm, used around 36% of imported fuel from other countries. With a large amount of imported fuel, fuel management is taking a vital part for Stockholm Exergi to ensure district heating supply stability. To manage fuel planning, Stockholm Exergi will calculate fuel demand forecasting for the next three years by using an optimization tool called “BoFiT”. But high optimization time is the main issue for the current midterm model. More than six hours are spent on regular model optimization. Therefore, midterm model optimization is typically run overnight. If there is an error occurring while the model is running, more time might be lost. From a preliminary study on the midterm model optimization time, the time spent in the calculation process is accounted for 80% of the total optimization time. In BoFiT, the midterm visual model is transformed into the mathematical model to solve for optimum results. The idea of decreasing the midterm model optimization time is the activation of LP relaxation in BoFiT. With LP relaxation activation, the computation time in the calculation process will be minimized. However, the usage of LP relaxation is giving some of the consequences to the optimization results.  Based on the study of the midterm model optimization time, the usage of LP relaxation can decrease the time spent in the calculation process for 67.72%. When considering the overall optimization time, 48.60% of the optimization time was reduced by LP relaxation activation. The statistical analysis in 2018 BoFiT optimization showed that results from relaxed optimization have slightly lower accuracy than the results from normal optimization in heat and electricity production. However, the relaxed optimization results in fuel consumption are considered comparable for most of the fuel except by fossil oils. For midterm optimization, heat and electricity production forecasting from normal optimization and relaxed optimization are comparable. For fuel demand forecasting, fossil oils give the most significant different results in terms of forecasting analysis. However, the wood chip has the most considerable difference in the demand for midterm fuel planning. The result from relaxed optimization showed that it requires 305.89 GWh more than that from normal optimization. With a vast amount of wood chip consumption, it could have high effects on the wood chip preparation and storage for Stockholm Exergi.