Production optimization for district heating : Short-term planning of district heating grid in Gävle, Sweden

Abstract: Energy systems with a high portion of renewable energy from wind and solar power can suffer from fluctuations in production due to weak winds or cloudy weather, which may affect the electricity price. When producing heat and power in a combined heat and power plant, an additional heat storage tank can be used to store the heat surplus which is obtained when the power production is high, and the heat demand is low. To optimize heat and power production economically, short-term planning can be applied. Short-term planning covers the production in the near future of 1-3 days. The optimization in this degree project is based on the district heating production, which means that the heating demand always needs to be fulfilled. The district heating production is based on the weather. Therefore a suitable period for simulation is three days due to the accuracy of the weather forecasts are reasonable. The optimization is performed on the district heat system in Gävle, Sweden. The system comprises several different production units, such as combined heat and power plants, backup plants, and industrial waste heat recovery. Two different models are made, one using linear programming and one using mixed integer non-linear programming. The model stated as a linear programming problem is not as accurate as of the one stated as a mixed integer non-linear programming problem which uses binary variables. Historical input data from Bomhus Energi AB, a company owned together by the local heat and power supplier Gävle Energi AB and the pulp and paper manufacturer BillerudKorsnäs AB, was given to simulate different scenarios. The different scenarios have various average temperatures and in some scenarios are there some issues with the pulp and paper industry affecting the waste heat recovery. In all scenarios is the heat storage tank charged when the demand is low and then discharged when the demand increases to avoid starting some of the more expensive backup plants if possible. The simulation time varies a lot between the two approaches, from a couple of seconds to several hours. Particularly when observing scenarios with a rather high demand since the backup generators use binary variables which take a lot of time to solve.