Development of a short-term operational planning tool for geothermal plants with heat and power production connected to large district heating systems

A short-term operational planning tool for geothermal plants with heat and power production connected to large district heating systems is developed. The software tool contains among other features a heat demand forecasting model for district heating systems. Several options, such as linear regression and artificial neural networks, are compared. As the result shows, artificial neural networks with the Bayesian Regularization Backpropagation Algorithm have a high generalization capability and are suitable to forecast the heat demand of large district heating systems with high accuracy. Data from a district heating system with about 70 MWth load supplied by a geothermal plant in the south of Munich (Germany) are used for the comparison and assessment of all methods. After developing a suitable heat forecast, the heat and power production site is modeled by using mixed integer linear programming, which has been solved in MATLAB®. Mixed-integer linear programming has proven to be a suitable method to model the operation of geothermal plants with heat and power production as well as to solve the planning optimization problem. As the result shows, the short-term operational planning tool can optimize the operation of single components as well as of the overall geothermal plant with regard to various objective functions. The tool maximizes the revenues from the sold heat and electricity minus the costs for the boiler fuel and the heat purchased from a connected adjacent geothermal plant. A retro perspective operation investigation has proven that the profitability of the considered geothermal plant could be significantly increased by using the developed software.     «

A short-term operational planning tool for geothermal plants with heat and power production connected to large district heating systems is developed. The software tool contains among other features a heat demand forecasting model for district heating systems. Several options, such as linear regression and artificial neural networks, are compared. As the result shows, artificial neural networks with the Bayesian Regularization Backpropagation Algorithm have a high generalization capability and ar...     »