An adaptive and flexible biomass power plant control system based on on-line fuel image analysis

•Utilization of image analysis for on-line fuel analysis, especially mixture ratios.•Development of an optimized biomass furnace controller based on fuel properties.•Implementation of the optimized controller in a full-scale environment in the megawatt range.•Evaluation of the controller in full-scale during a 12+ month period of time.•Discussion of the transferability and full description of the selected approach and potential improvements. Fuel costs are the determining factor for the operational expenditure of solid biomass furnaces. Consequently, there is an incentive to shift towards cheaper feedstock like forest residues, composting residues or waste wood. This comes at price, namely challenging properties during combustion as well as fluctuating fuel properties due to inhomogeneities or varying manual mixing. Together, this poses challenges for an industrial furnace’s control system. Currently relying on only few static control loops for thermal load or air supply and distribution, furnaces struggle under rapidly changing fuel properties. Briefly: Traditional control systems are running in a reactive way. In contrast to that, this work focuses on the development of a pro-active control system for industrial furnaces and power plants. This contribution describes the approach and the methodology for such a system as well as the industrial implementation and obtained results from a 12+ month evaluation period at a 6 MW biomass plant. The first necessary step is to determine relevant properties of incoming biomass fuels prior to the combustion itself. For this purpose, we captured fuel photographs of the utilised biomass feedstock of a 6 MW biomass power plant. Subsequently, using image processing techniques and a statistical/machine learning approach, we define a parameter for the current ‘fuel quality’. Data from the plant’s process control system is used to provide alternative furnace settings for the fuel stoking rate, grate speed and the air supply with respect to the fuel quality. Improved settings are derived by technical and operational considerations as well as combustion experiments in a full-scale environment. The final step is to manipulate the furnace’s set points of the existing control loops automatically continuously based on the controller’s suggestion. During the 12+ month demonstration period including tests at deliberate extreme conditions and edge cases, a faster adaption to changing fuel properties and more stable operation can be